diff --git a/.gitignore b/.gitignore
index 7422cead3..2c298e3b4 100644
--- a/.gitignore
+++ b/.gitignore
@@ -15,6 +15,7 @@ lapack-netlib/make.inc
 lapack-netlib/lapacke/include/lapacke_mangling.h
 lapack-netlib/TESTING/testing_results.txt
 *.so
+*.so.*
 *.a
 .svn
 *~
@@ -65,3 +66,5 @@ test/sblat3
 test/zblat1
 test/zblat2
 test/zblat3
+build
+build.*
diff --git a/.travis.yml b/.travis.yml
index 7d625c9dc..806cb0046 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -1,4 +1,13 @@
 language: c
+
+notifications:
+  webhooks:
+    urls:
+      - https://webhooks.gitter.im/e/8a6e4470a0cebd090344
+    on_success: change  # options: [always|never|change] default: always
+    on_failure: always  # options: [always|never|change] default: always
+    on_start: never     # options: [always|never|change] default: always
+
 compiler:
   - gcc
 
diff --git a/CMakeLists.txt b/CMakeLists.txt
new file mode 100644
index 000000000..3b436dc13
--- /dev/null
+++ b/CMakeLists.txt
@@ -0,0 +1,190 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+##
+
+cmake_minimum_required(VERSION 2.8.4)
+project(OpenBLAS)
+set(OpenBLAS_MAJOR_VERSION 0)
+set(OpenBLAS_MINOR_VERSION 2)
+set(OpenBLAS_PATCH_VERSION 14)
+set(OpenBLAS_VERSION "${OpenBLAS_MAJOR_VERSION}.${OpenBLAS_MINOR_VERSION}.${OpenBLAS_PATCH_VERSION}")
+
+enable_language(ASM)
+enable_language(C)
+
+if(MSVC)
+set(OpenBLAS_LIBNAME libopenblas)
+else()
+set(OpenBLAS_LIBNAME openblas)
+endif()
+
+#######
+if(MSVC)
+option(BUILD_WITHOUT_LAPACK "Without LAPACK and LAPACKE (Only BLAS or CBLAS)" ON)
+endif()
+option(BUILD_WITHOUT_CBLAS "Without CBLAS" OFF)
+option(BUILD_DEBUG "Build Debug Version" OFF)
+#######
+if(BUILD_WITHOUT_LAPACK)
+set(NO_LAPACK 1)
+set(NO_LAPACKE 1)
+endif()
+
+if(BUILD_DEBUG)
+set(CMAKE_BUILD_TYPE Debug)
+else()
+set(CMAKE_BUILD_TYPE Release)
+endif()
+
+if(BUILD_WITHOUT_CBLAS)
+set(NO_CBLAS 1)
+endif()
+
+#######
+
+
+message(WARNING "CMake support is experimental. This will not produce the same Makefiles that OpenBLAS ships with. Only x86 support is currently available.")
+
+include("${CMAKE_SOURCE_DIR}/cmake/utils.cmake")
+include("${CMAKE_SOURCE_DIR}/cmake/system.cmake")
+
+set(BLASDIRS interface driver/level2 driver/level3 driver/others)
+
+if (NOT DYNAMIC_ARCH)
+  list(APPEND BLASDIRS kernel)
+endif ()
+
+if (DEFINED UTEST_CHECK)
+  set(SANITY_CHECK 1)
+endif ()
+
+if (DEFINED SANITY_CHECK)
+  list(APPEND BLASDIRS reference)
+endif ()
+
+set(SUBDIRS	${BLASDIRS})
+if (NOT NO_LAPACK)
+  list(APPEND SUBDIRS lapack)
+endif ()
+
+# set which float types we want to build for
+if (NOT DEFINED BUILD_SINGLE AND NOT DEFINED BUILD_DOUBLE AND NOT DEFINED BUILD_COMPLEX AND NOT DEFINED BUILD_COMPLEX16)
+  # if none are defined, build for all
+  set(BUILD_SINGLE true)
+  set(BUILD_DOUBLE true)
+  set(BUILD_COMPLEX true)
+  set(BUILD_COMPLEX16 true)
+endif ()
+
+set(FLOAT_TYPES "")
+if (BUILD_SINGLE)
+  message(STATUS "Building Single Precision")
+  list(APPEND FLOAT_TYPES "SINGLE") # defines nothing
+endif ()
+
+if (BUILD_DOUBLE)
+  message(STATUS "Building Double Precision")
+  list(APPEND FLOAT_TYPES "DOUBLE") # defines DOUBLE
+endif ()
+
+if (BUILD_COMPLEX)
+  message(STATUS "Building Complex Precision")
+  list(APPEND FLOAT_TYPES "COMPLEX") # defines COMPLEX
+endif ()
+
+if (BUILD_COMPLEX16)
+  message(STATUS "Building Double Complex Precision")
+  list(APPEND FLOAT_TYPES "ZCOMPLEX") # defines COMPLEX and DOUBLE
+endif ()
+
+set(SUBDIRS_ALL ${SUBDIRS} test ctest utest exports benchmark ../laswp ../bench)
+
+# all :: libs netlib tests shared
+
+# libs :
+if (NOT DEFINED CORE OR "${CORE}" STREQUAL "UNKNOWN")
+  message(FATAL_ERROR "Detecting CPU failed. Please set TARGET explicitly, e.g. make TARGET=your_cpu_target. Please read README for details.")
+endif ()
+
+if (${NO_STATIC} AND ${NO_SHARED})
+  message(FATAL_ERROR "Neither static nor shared are enabled.")
+endif ()
+
+# get obj vars into format that add_library likes: $<TARGET_OBJS:objlib> (see http://www.cmake.org/cmake/help/v3.0/command/add_library.html)
+set(TARGET_OBJS "")
+foreach (SUBDIR ${SUBDIRS})
+  add_subdirectory(${SUBDIR})
+  string(REPLACE "/" "_" subdir_obj ${SUBDIR})
+  list(APPEND TARGET_OBJS "$<TARGET_OBJECTS:${subdir_obj}>")
+endforeach ()
+
+# netlib:
+
+# Can't just use lapack-netlib's CMake files, since they are set up to search for BLAS, build and install a binary. We just want to build a couple of lib files out of lapack and lapacke.
+# Not using add_subdirectory here because lapack-netlib already has its own CMakeLists.txt. Instead include a cmake script with the sources we want.
+if (NOT NOFORTRAN AND NOT NO_LAPACK)
+  include("${CMAKE_SOURCE_DIR}/cmake/lapack.cmake")
+if (NOT NO_LAPACKE)
+  include("${CMAKE_SOURCE_DIR}/cmake/lapacke.cmake")
+endif ()
+endif ()
+
+#Only generate .def for dll on MSVC
+if(MSVC)
+set(OpenBLAS_DEF_FILE "${PROJECT_BINARY_DIR}/openblas.def")
+endif()
+
+# add objects to the openblas lib
+add_library(${OpenBLAS_LIBNAME} SHARED ${LA_SOURCES} ${LAPACKE_SOURCES} ${TARGET_OBJS} ${OpenBLAS_DEF_FILE})
+
+include("${CMAKE_SOURCE_DIR}/cmake/export.cmake")
+
+
+if(NOT MSVC)
+#only build shared library for MSVC
+add_library(${OpenBLAS_LIBNAME}_static STATIC ${LA_SOURCES} ${LAPACKE_SOURCES} ${TARGET_OBJS})
+set_target_properties(${OpenBLAS_LIBNAME}_static PROPERTIES OUTPUT_NAME ${OpenBLAS_LIBNAME})
+set_target_properties(${OpenBLAS_LIBNAME}_static PROPERTIES CLEAN_DIRECT_OUTPUT 1)
+
+if(SMP)
+target_link_libraries(${OpenBLAS_LIBNAME} pthread)
+target_link_libraries(${OpenBLAS_LIBNAME}_static pthread)
+endif()
+
+#build test and ctest
+enable_testing()
+add_subdirectory(test)
+if(NOT NO_CBLAS)
+add_subdirectory(ctest)
+endif()
+endif()
+
+set_target_properties(${OpenBLAS_LIBNAME} PROPERTIES 
+  VERSION ${OpenBLAS_MAJOR_VERSION}.${OpenBLAS_MINOR_VERSION}
+  SOVERSION ${OpenBLAS_MAJOR_VERSION}
+)
+
+
+# TODO: Why is the config saved here? Is this necessary with CMake?
+#Save the config files for installation
+#	@cp Makefile.conf Makefile.conf_last
+#	@cp config.h config_last.h
+#ifdef QUAD_PRECISION
+#	@echo "#define QUAD_PRECISION">> config_last.h
+#endif
+#ifeq ($(EXPRECISION), 1)
+#	@echo "#define EXPRECISION">> config_last.h
+#endif
+###
+#ifeq ($(DYNAMIC_ARCH), 1)
+#	@$(MAKE) -C kernel commonlibs || exit 1
+#	@for d in $(DYNAMIC_CORE) ; \
+#	do  $(MAKE) GOTOBLAS_MAKEFILE= -C kernel TARGET_CORE=$$d kernel || exit 1 ;\
+#	done
+#	@echo DYNAMIC_ARCH=1 >> Makefile.conf_last
+#endif
+#ifdef USE_THREAD
+#	@echo USE_THREAD=$(USE_THREAD) >>  Makefile.conf_last
+#endif
+#	@touch lib.grd
+
diff --git a/CONTRIBUTORS.md b/CONTRIBUTORS.md
index b88e3671b..88e461dc4 100644
--- a/CONTRIBUTORS.md
+++ b/CONTRIBUTORS.md
@@ -127,5 +127,8 @@ In chronological order:
 * Ton van den Heuvel <https://github.com/ton>
   * [2015-03-18] Fix race condition during shutdown causing a crash in gotoblas_set_affinity().
 
+* Martin Koehler <https://github.com/grisuthedragon/>
+  * [2015-09-07] Improved imatcopy
+
 * [Your name or handle] <[email or website]>
   * [Date] [Brief summary of your changes]
diff --git a/Changelog.txt b/Changelog.txt
index 6941a9f96..422b8b519 100644
--- a/Changelog.txt
+++ b/Changelog.txt
@@ -1,4 +1,57 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.2.15
+27-Oct-2015
+common:
+	* Support cmake on x86/x86-64. Natively compiling on MS Visual Studio.
+	  (experimental. Thank Hank Anderson for the initial cmake porting work.)
+	  
+	  On Linux and Mac OSX, OpenBLAS cmake supports assembly kernels.
+	  e.g. cmake .
+	       make
+	       make test (Optional)
+
+	  On Windows MS Visual Studio, OpenBLAS cmake only support C kernels.
+	  (OpenBLAS uses AT&T style assembly, which is not supported by MSVC.)
+	  e.g. cmake -G "Visual Studio 12 Win64" .
+	       Open OpenBLAS.sln and build.
+	  
+	* Enable MAX_STACK_ALLOC flags by default.
+	  Improve ger and gemv for small matrices.
+	* Improve gemv parallel with small m and large n case.
+	* Improve ?imatcopy when lda==ldb (#633. Thanks, Martin Koehler)
+	* Add vecLib benchmarks (#565. Thanks, Andreas Noack.)
+	* Fix LAPACK lantr for row major matrices (#634. Thanks, Dan Kortschak)
+	* Fix LAPACKE lansy (#640. Thanks, Dan Kortschak)
+	* Import bug fixes for LAPACKE s/dormlq, c/zunmlq 
+	* Raise the signal when pthread_create fails (#668. Thanks, James K. Lowden)
+	* Remove g77 from compiler list.
+	* Enable AppVeyor Windows CI.
+
+x86/x86-64:
+	* Support pure C generic kernels for x86/x86-64.
+	* Support Intel Boardwell and Skylake by Haswell kernels.
+	* Support AMD Excavator by Steamroller kernels.
+	* Optimize s/d/c/zdot for Intel SandyBridge and Haswell.
+	* Optimize s/d/c/zdot for AMD Piledriver and Steamroller.
+	* Optimize s/d/c/zapxy for Intel SandyBridge and Haswell.
+	* Optimize s/d/c/zapxy for AMD Piledriver and Steamroller.
+	* Optimize d/c/zscal for Intel Haswell, dscal for Intel SandyBridge.
+	* Optimize d/c/zscal for AMD Bulldozer, Piledriver and Steamroller.
+	* Optimize s/dger for Intel SandyBridge.
+	* Optimize s/dsymv for Intel SandyBridge.
+	* Optimize ssymv for Intel Haswell.
+	* Optimize dgemv for Intel Nehalem and Haswell.
+	* Optimize dtrmm for Intel Haswell.
+
+ARM:
+	* Support Android NDK armeabi-v7a-hard ABI (-mfloat-abi=hard)
+	  e.g. make HOSTCC=gcc CC=arm-linux-androideabi-gcc NO_LAPACK=1 TARGET=ARMV7
+	* Fix lock, rpcc bugs (#616, #617. Thanks, Grazvydas Ignotas)
+POWER:
+	* Support ppc64le platform (ELF ABI v2. #612. Thanks, Matthew Brandyberry.)
+	* Support POWER7/8 by POWER6 kernels. (#612. Thanks, Fábio Perez.)
+
 ====================================================================
 Version 0.2.14
 24-Mar-2015
diff --git a/Makefile b/Makefile
index 3aaf092fc..6ad87d802 100644
--- a/Makefile
+++ b/Makefile
@@ -20,6 +20,8 @@ ifneq ($(NO_LAPACK), 1)
 SUBDIRS	+= lapack
 endif
 
+LAPACK_NOOPT := $(filter-out -O0 -O1 -O2 -O3 -Ofast,$(LAPACK_FFLAGS))
+
 SUBDIRS_ALL = $(SUBDIRS) test ctest utest exports benchmark ../laswp ../bench
 
 .PHONY : all libs netlib test ctest shared install
@@ -131,7 +133,7 @@ ifeq ($(CORE), UNKOWN)
 	$(error OpenBLAS: Detecting CPU failed. Please set TARGET explicitly, e.g. make TARGET=your_cpu_target. Please read README for the detail.)
 endif
 ifeq ($(NOFORTRAN), 1)
-	$(error OpenBLAS: Detecting fortran compiler failed. Please install fortran compiler, e.g. gfortran, ifort, openf90.)
+	$(info OpenBLAS: Detecting fortran compiler failed. Cannot compile LAPACK. Only compile BLAS.)
 endif
 ifeq ($(NO_STATIC), 1)
 ifeq ($(NO_SHARED), 1)
@@ -231,7 +233,7 @@ ifndef NOFORTRAN
 	-@echo "FORTRAN     = $(FC)" > $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "OPTS        = $(LAPACK_FFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "POPTS       = $(LAPACK_FPFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
-	-@echo "NOOPT       = $(LAPACK_FFLAGS) -O0" >> $(NETLIB_LAPACK_DIR)/make.inc
+	-@echo "NOOPT       = -O0 $(LAPACK_NOOPT)" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "PNOOPT      = $(LAPACK_FPFLAGS) -O0" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "LOADOPTS    = $(FFLAGS) $(EXTRALIB)" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "CC          = $(CC)" >> $(NETLIB_LAPACK_DIR)/make.inc
diff --git a/Makefile.arm b/Makefile.arm
index 9978a672a..272220ca9 100644
--- a/Makefile.arm
+++ b/Makefile.arm
@@ -1,13 +1,23 @@
 # ifeq logical or
 ifeq ($(CORE), $(filter $(CORE),CORTEXA9 CORTEXA15))
+ifeq ($(OSNAME), Android)
+CCOMMON_OPT += -marm -mfpu=neon  -mfloat-abi=hard -march=armv7-a
+FCOMMON_OPT += -marm -mfpu=neon  -mfloat-abi=hard -march=armv7-a
+else
 CCOMMON_OPT += -marm -mfpu=vfpv3  -mfloat-abi=hard -march=armv7-a
 FCOMMON_OPT += -marm -mfpu=vfpv3  -mfloat-abi=hard -march=armv7-a
 endif
+endif
 
 ifeq ($(CORE), ARMV7)
+ifeq ($(OSNAME), Android)
+CCOMMON_OPT += -marm -mfpu=neon  -mfloat-abi=hard -march=armv7-a
+FCOMMON_OPT += -marm -mfpu=neon  -mfloat-abi=hard -march=armv7-a
+else
 CCOMMON_OPT += -marm -mfpu=vfpv3  -mfloat-abi=hard -march=armv7-a
 FCOMMON_OPT += -marm -mfpu=vfpv3  -mfloat-abi=hard -march=armv7-a
 endif
+endif
 
 ifeq ($(CORE), ARMV6)
 CCOMMON_OPT += -marm -mfpu=vfp -mfloat-abi=hard  -march=armv6
@@ -16,8 +26,8 @@ endif
 
 
 ifeq ($(CORE), ARMV5)
-CCOMMON_OPT += -marm -mfpu=vfp -mfloat-abi=hard  -march=armv6
-FCOMMON_OPT += -marm -mfpu=vfp -mfloat-abi=hard  -march=armv6
+CCOMMON_OPT += -marm -march=armv5
+FCOMMON_OPT += -marm -march=armv5
 endif
 
 
diff --git a/Makefile.install b/Makefile.install
index e1deaae3e..9814302b0 100644
--- a/Makefile.install
+++ b/Makefile.install
@@ -11,6 +11,7 @@ OPENBLAS_BINARY_DIR := $(PREFIX)/bin
 OPENBLAS_BUILD_DIR := $(CURDIR)
 OPENBLAS_CMAKE_DIR := $(OPENBLAS_LIBRARY_DIR)/cmake/openblas
 OPENBLAS_CMAKE_CONFIG := OpenBLASConfig.cmake
+OPENBLAS_CMAKE_CONFIG_VERSION := OpenBLASConfigVersion.cmake
 
 .PHONY : install
 .NOTPARALLEL : install
@@ -86,8 +87,8 @@ ifeq ($(OSNAME), Darwin)
 	ln -fs $(LIBDYNNAME) $(LIBPREFIX).dylib
 endif
 ifeq ($(OSNAME), WINNT)
-	@-cp $(LIBDLLNAME) $(OPENBLAS_BINARY_DIR)
-	@-cp $(LIBDLLNAME).a $(OPENBLAS_LIBRARY_DIR)
+	@-cp $(LIBDLLNAME) $(DESTDIR)$(OPENBLAS_BINARY_DIR)
+	@-cp $(LIBDLLNAME).a $(DESTDIR)$(OPENBLAS_LIBRARY_DIR)
 endif
 ifeq ($(OSNAME), CYGWIN_NT)
 	@-cp $(LIBDLLNAME) $(OPENBLAS_BINARY_DIR)
@@ -97,6 +98,7 @@ endif
 	@echo Generating $(OPENBLAS_CMAKE_CONFIG) in $(DESTDIR)$(OPENBLAS_CMAKE_DIR)
 	@echo "SET(OpenBLAS_VERSION \"${VERSION}\")" > $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG)	
 	@echo "SET(OpenBLAS_INCLUDE_DIRS ${OPENBLAS_INCLUDE_DIR})" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG)
+
 ifndef NO_SHARED
 #ifeq logical or
 ifeq ($(OSNAME), $(filter $(OSNAME),Linux FreeBSD NetBSD))
@@ -112,5 +114,16 @@ else
 #only static
 	@echo "SET(OpenBLAS_LIBRARIES ${OPENBLAS_LIBRARY_DIR}/$(LIBPREFIX).$(LIBSUFFIX))" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG)
 endif
+#Generating OpenBLASConfigVersion.cmake
+	@echo Generating $(OPENBLAS_CMAKE_CONFIG_VERSION) in $(DESTDIR)$(OPENBLAS_CMAKE_DIR)
+	@echo "set (PACKAGE_VERSION \"${VERSION}\")" > $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
+	@echo "if (PACKAGE_VERSION VERSION_LESS PACKAGE_FIND_VERSION)" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
+	@echo "  set (PACKAGE_VERSION_COMPATIBLE FALSE)" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
+	@echo "else ()" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
+	@echo "  set (PACKAGE_VERSION_COMPATIBLE TRUE)" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
+	@echo "  if (PACKAGE_FIND_VERSION STREQUAL PACKAGE_VERSION)" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
+	@echo "    set (PACKAGE_VERSION_EXACT TRUE)" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
+	@echo "  endif ()" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
+	@echo "endif ()" >> $(DESTDIR)$(OPENBLAS_CMAKE_DIR)/$(OPENBLAS_CMAKE_CONFIG_VERSION)
 	@echo Install OK!
 
diff --git a/Makefile.rule b/Makefile.rule
index 1479de660..459f79c26 100644
--- a/Makefile.rule
+++ b/Makefile.rule
@@ -3,7 +3,7 @@
 #
 
 # This library's version
-VERSION = 0.2.14
+VERSION = 0.2.15
 
 # If you set the suffix, the library name will be libopenblas_$(LIBNAMESUFFIX).a
 # and libopenblas_$(LIBNAMESUFFIX).so. Meanwhile, the soname in shared library
@@ -162,13 +162,16 @@ COMMON_PROF = -pg
 # Improve GEMV and GER for small matrices by stack allocation.
 # For details, https://github.com/xianyi/OpenBLAS/pull/482
 #
-# MAX_STACK_ALLOC=2048
+ MAX_STACK_ALLOC=2048
 
 # Add a prefix or suffix to all exported symbol names in the shared library.
 # Avoid conflicts with other BLAS libraries, especially when using
 # 64 bit integer interfaces in OpenBLAS.
 # For details, https://github.com/xianyi/OpenBLAS/pull/459
 #
+# The same prefix and suffix are also added to the library name,
+# i.e. you get lib$(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX) rather than libopenblas
+#
 # SYMBOLPREFIX=
 # SYMBOLSUFFIX=
 
diff --git a/Makefile.system b/Makefile.system
index 525daa41b..42ad49849 100644
--- a/Makefile.system
+++ b/Makefile.system
@@ -23,6 +23,7 @@ CC = gcc
 UNAME_S := $(shell uname -s)
 ifeq ($(UNAME_S),Darwin)
      CC = clang
+#     EXTRALIB += -Wl,-no_compact_unwind
 endif
 endif
 
@@ -64,6 +65,9 @@ endif
 ifeq ($(TARGET), STEAMROLLER)
 GETARCH_FLAGS := -DFORCE_BARCELONA
 endif
+ifeq ($(TARGET), EXCAVATOR)
+GETARCH_FLAGS := -DFORCE_BARCELONA
+endif
 endif
 
 
@@ -91,6 +95,9 @@ endif
 ifeq ($(TARGET_CORE), STEAMROLLER)
 GETARCH_FLAGS := -DFORCE_BARCELONA
 endif
+ifeq ($(TARGET_CORE), EXCAVATOR)
+GETARCH_FLAGS := -DFORCE_BARCELONA
+endif
 endif
 
 
@@ -195,12 +202,18 @@ DLLWRAP = $(CROSS_SUFFIX)dllwrap
 OBJCOPY = $(CROSS_SUFFIX)objcopy
 OBJCONV = $(CROSS_SUFFIX)objconv
 
+
+# For detect fortran failed, only build BLAS.
+ifeq ($(NOFORTRAN), 1)
+NO_LAPACK = 1
+endif
+
 #
 #  OS dependent settings
 #
 
 ifeq ($(OSNAME), Darwin)
-export MACOSX_DEPLOYMENT_TARGET=10.2
+export MACOSX_DEPLOYMENT_TARGET=10.6
 MD5SUM = md5 -r
 endif
 
@@ -323,6 +336,11 @@ ifeq ($(ARCH), x86)
 ifndef BINARY
 NO_BINARY_MODE	= 1
 endif
+
+ifeq ($(CORE), generic)
+NO_EXPRECISION = 1
+endif
+
 ifndef NO_EXPRECISION
 ifeq ($(F_COMPILER), GFORTRAN)
 # ifeq logical or. GCC or LSB
@@ -341,6 +359,11 @@ endif
 endif
 
 ifeq ($(ARCH), x86_64)
+
+ifeq ($(CORE), generic)
+NO_EXPRECISION = 1
+endif
+
 ifndef NO_EXPRECISION
 ifeq ($(F_COMPILER), GFORTRAN)
 # ifeq logical or. GCC or LSB
@@ -408,7 +431,7 @@ endif
 ifeq ($(ARCH), x86_64)
 DYNAMIC_CORE = PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO
 ifneq ($(NO_AVX), 1)
-DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER
+DYNAMIC_CORE += SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER EXCAVATOR
 endif
 ifneq ($(NO_AVX2), 1)
 DYNAMIC_CORE += HASWELL
@@ -578,7 +601,7 @@ else
 FCOMMON_OPT += -m32
 endif
 endif
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FCOMMON_OPT += -fopenmp
 endif
 endif
@@ -590,14 +613,14 @@ ifneq ($(INTERFACE64), 0)
 FCOMMON_OPT += -i8
 endif
 endif
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FCOMMON_OPT += -openmp
 endif
 endif
 
 ifeq ($(F_COMPILER), FUJITSU)
 CCOMMON_OPT += -DF_INTERFACE_FUJITSU
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FCOMMON_OPT += -openmp
 endif
 endif
@@ -615,7 +638,7 @@ endif
 else
 FCOMMON_OPT += -q32
 endif
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FCOMMON_OPT += -openmp
 endif
 endif
@@ -633,7 +656,7 @@ FCOMMON_OPT += -tp p7-64
 else
 FCOMMON_OPT += -tp p7
 endif
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FCOMMON_OPT += -mp
 endif
 endif
@@ -662,7 +685,7 @@ FCOMMON_OPT += -mabi=n32
 endif
 endif
 
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FCOMMON_OPT += -mp
 endif
 endif
@@ -699,7 +722,7 @@ FCOMMON_OPT += -m64
 endif
 endif
 
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FEXTRALIB   += -lstdc++
 FCOMMON_OPT += -mp
 endif
@@ -747,14 +770,14 @@ FCOMMON_OPT  += -m32
 else
 FCOMMON_OPT  += -m64
 endif
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FCOMMON_OPT += -xopenmp=parallel
 endif
 endif
 
 ifeq ($(F_COMPILER), COMPAQ)
 CCOMMON_OPT  += -DF_INTERFACE_COMPAQ
-ifdef USE_OPENMP
+ifeq ($(USE_OPENMP), 1)
 FCOMMON_OPT += -openmp
 endif
 endif
@@ -857,12 +880,6 @@ ifdef USE_SIMPLE_THREADED_LEVEL3
 CCOMMON_OPT	+= -DUSE_SIMPLE_THREADED_LEVEL3
 endif
 
-ifndef LIBNAMESUFFIX
-LIBPREFIX = libopenblas
-else
-LIBPREFIX = libopenblas_$(LIBNAMESUFFIX)
-endif
-
 ifndef SYMBOLPREFIX
 SYMBOLPREFIX =
 endif
@@ -871,6 +888,12 @@ ifndef SYMBOLSUFFIX
 SYMBOLSUFFIX =
 endif
 
+ifndef LIBNAMESUFFIX
+LIBPREFIX = lib$(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
+else
+LIBPREFIX = lib$(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
+endif
+
 KERNELDIR	= $(TOPDIR)/kernel/$(ARCH)
 
 include $(TOPDIR)/Makefile.$(ARCH)
diff --git a/README.md b/README.md
index cdacf9888..0ec86d362 100644
--- a/README.md
+++ b/README.md
@@ -1,7 +1,10 @@
 # OpenBLAS
 
-[![Build Status](https://travis-ci.org/xianyi/OpenBLAS.png?branch=develop)](https://travis-ci.org/xianyi/OpenBLAS)
+[![Join the chat at https://gitter.im/xianyi/OpenBLAS](https://badges.gitter.im/Join%20Chat.svg)](https://gitter.im/xianyi/OpenBLAS?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge)
 
+Travis CI: [![Build Status](https://travis-ci.org/xianyi/OpenBLAS.png?branch=develop)](https://travis-ci.org/xianyi/OpenBLAS)
+
+AppVeyor: [![Build status](https://ci.appveyor.com/api/projects/status/09sohd35n8nkkx64/branch/develop?svg=true)](https://ci.appveyor.com/project/xianyi/openblas/branch/develop)
 ## Introduction
 OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.
 
diff --git a/TargetList.txt b/TargetList.txt
index 1c985080b..b2878ba32 100644
--- a/TargetList.txt
+++ b/TargetList.txt
@@ -33,6 +33,7 @@ BOBCAT
 BULLDOZER
 PILEDRIVER
 STEAMROLLER
+EXCAVATOR
 
 c)VIA CPU:
 SSE_GENERIC
@@ -43,6 +44,8 @@ NANO
 POWER4
 POWER5
 POWER6
+POWER7
+POWER8
 PPCG4
 PPC970
 PPC970MP
diff --git a/appveyor.yml b/appveyor.yml
new file mode 100644
index 000000000..394e48854
--- /dev/null
+++ b/appveyor.yml
@@ -0,0 +1,42 @@
+version: 0.2.15.{build}
+
+#environment:
+
+platform: 
+  - x64
+
+configuration: Release
+
+clone_folder: c:\projects\OpenBLAS
+
+init:
+  - git config --global core.autocrlf input
+
+build:
+  project: OpenBLAS.sln
+
+clone_depth: 5
+
+#branches to build
+branches:
+  only:
+    - master
+    - develop
+    - cmake
+
+skip_tags: true
+
+matrix:
+  fast_finish: true
+
+skip_commits:
+# Add [av skip] to commit messages
+  message: /\[av skip\]/
+
+before_build:
+  - echo Running cmake...
+  - cd c:\projects\OpenBLAS
+  - cmake -G "Visual Studio 12 Win64" .
+
+test_script:
+  - echo Build OK!
diff --git a/benchmark/Make_exe.sh b/benchmark/Make_exe.sh
new file mode 100755
index 000000000..4304f6fb3
--- /dev/null
+++ b/benchmark/Make_exe.sh
@@ -0,0 +1,9 @@
+#!/bin/bash
+
+for f in *.goto *.acml *.mkl *.atlas
+do
+	if [ -f "$f" ]; then
+		mv $f `echo $f|tr '.' '_'`.exe
+	fi
+done
+
diff --git a/benchmark/Makefile b/benchmark/Makefile
index b5eaa9343..492d2617f 100644
--- a/benchmark/Makefile
+++ b/benchmark/Makefile
@@ -10,7 +10,7 @@ include $(TOPDIR)/Makefile.system
 #LIBACML = -fopenmp $(ACML)/libacml_mp.a -lgfortran -lm
 
 # ACML 6.1 custom
-ACML=/home/werner/project/acml6.1/gfortran64_mp/lib
+ACML=/home/saar/acml6.1/gfortran64_mp/lib
 LIBACML = -fopenmp $(ACML)/libacml_mp.so -lgfortran -lm
  
 
@@ -30,7 +30,10 @@ LIBATLAS = -fopenmp $(ATLAS)/liblapack.a  $(ATLAS)/libptcblas.a  $(ATLAS)/libptf
 MKL=/home/saar/intel_mkl
 LIBMKL = -L$(MKL) -lmkl_intel_lp64 -lmkl_gnu_thread -lmkl_core -lgomp -lpthread -lm
 
+# Apple vecLib
+LIBVECLIB = -framework Accelerate
 
+ifeq ($(OSNAME), WINNT)
 
 goto :: slinpack.goto dlinpack.goto clinpack.goto zlinpack.goto \
        scholesky.goto dcholesky.goto ccholesky.goto zcholesky.goto \
@@ -39,9 +42,13 @@ goto :: slinpack.goto dlinpack.goto clinpack.goto zlinpack.goto \
        strsm.goto dtrsm.goto ctrsm.goto ztrsm.goto \
        ssyrk.goto dsyrk.goto csyrk.goto zsyrk.goto \
        ssyr2k.goto dsyr2k.goto csyr2k.goto zsyr2k.goto \
-       sger.goto dger.goto \
+       sger.goto dger.goto cger.goto zger.goto \
        sdot.goto ddot.goto \
        saxpy.goto daxpy.goto caxpy.goto zaxpy.goto \
+       scopy.goto dcopy.goto ccopy.goto zcopy.goto \
+       sswap.goto dswap.goto cswap.goto zswap.goto \
+       sscal.goto dscal.goto cscal.goto zscal.goto \
+       sasum.goto dasum.goto casum.goto zasum.goto \
        ssymv.goto dsymv.goto csymv.goto zsymv.goto \
        chemv.goto zhemv.goto \
        chemm.goto zhemm.goto \
@@ -49,6 +56,114 @@ goto :: slinpack.goto dlinpack.goto clinpack.goto zlinpack.goto \
        cher2k.goto zher2k.goto \
        sgemv.goto dgemv.goto cgemv.goto zgemv.goto \
        sgeev.goto dgeev.goto cgeev.goto zgeev.goto \
+       sgesv.goto dgesv.goto cgesv.goto zgesv.goto \
+       sgetri.goto dgetri.goto cgetri.goto zgetri.goto \
+       spotrf.goto dpotrf.goto cpotrf.goto zpotrf.goto \
+       ssymm.goto dsymm.goto csymm.goto zsymm.goto
+
+acml :: slinpack.acml dlinpack.acml clinpack.acml zlinpack.acml \
+       scholesky.acml dcholesky.acml ccholesky.acml zcholesky.acml \
+       sgemm.acml dgemm.acml cgemm.acml zgemm.acml \
+       strmm.acml dtrmm.acml ctrmm.acml ztrmm.acml \
+       strsm.acml dtrsm.acml ctrsm.acml ztrsm.acml \
+       ssyrk.acml dsyrk.acml csyrk.acml zsyrk.acml \
+       ssyr2k.acml dsyr2k.acml csyr2k.acml zsyr2k.acml \
+       sger.acml dger.acml cger.acml zger.acml \
+       sdot.acml ddot.acml \
+       saxpy.acml daxpy.acml caxpy.acml zaxpy.acml \
+       scopy.acml dcopy.acml ccopy.acml zcopy.acml \
+       sswap.acml dswap.acml cswap.acml zswap.acml \
+       sscal.acml dscal.acml cscal.acml zscal.acml \
+       sasum.acml dasum.acml casum.acml zasum.acml \
+       ssymv.acml dsymv.acml csymv.acml zsymv.acml \
+       chemv.acml zhemv.acml \
+       chemm.acml zhemm.acml \
+       cherk.acml zherk.acml \
+       cher2k.acml zher2k.acml \
+       sgemv.acml dgemv.acml cgemv.acml zgemv.acml \
+       sgeev.acml dgeev.acml cgeev.acml zgeev.acml \
+       sgesv.acml dgesv.acml cgesv.acml zgesv.acml \
+       sgetri.acml dgetri.acml cgetri.acml zgetri.acml \
+       spotrf.acml dpotrf.acml cpotrf.acml zpotrf.acml \
+       ssymm.acml dsymm.acml csymm.acml zsymm.acml
+
+atlas :: slinpack.atlas dlinpack.atlas clinpack.atlas zlinpack.atlas \
+       scholesky.atlas dcholesky.atlas ccholesky.atlas zcholesky.atlas \
+       sgemm.atlas dgemm.atlas cgemm.atlas zgemm.atlas \
+       strmm.atlas dtrmm.atlas ctrmm.atlas ztrmm.atlas \
+       strsm.atlas dtrsm.atlas ctrsm.atlas ztrsm.atlas \
+       ssyrk.atlas dsyrk.atlas csyrk.atlas zsyrk.atlas \
+       ssyr2k.atlas dsyr2k.atlas csyr2k.atlas zsyr2k.atlas \
+       sger.atlas dger.atlas cger.atlas zger.atlas\
+       sdot.atlas ddot.atlas \
+       saxpy.atlas daxpy.atlas caxpy.atlas zaxpy.atlas \
+       scopy.atlas dcopy.atlas ccopy.atlas zcopy.atlas \
+       sswap.atlas dswap.atlas cswap.atlas zswap.atlas \
+       sscal.atlas dscal.atlas cscal.atlas zscal.atlas \
+       sasum.atlas dasum.atlas casum.atlas zasum.atlas \
+       ssymv.atlas dsymv.atlas csymv.atlas zsymv.atlas \
+       chemv.atlas zhemv.atlas \
+       chemm.acml zhemm.acml \
+       chemm.atlas zhemm.atlas \
+       cherk.atlas zherk.atlas \
+       cher2k.atlas zher2k.atlas \
+       sgemv.atlas dgemv.atlas cgemv.atlas zgemv.atlas \
+       sgeev.atlas dgeev.atlas cgeev.atlas zgeev.atlas \
+       sgesv.atlas dgesv.atlas cgesv.atlas zgesv.atlas \
+       sgetri.atlas dgetri.atlas cgetri.atlas zgetri.atlas \
+       spotrf.atlas dpotrf.atlas cpotrf.atlas zpotrf.atlas \
+       ssymm.atlas dsymm.atlas csymm.atlas zsymm.atlas
+
+mkl :: slinpack.mkl dlinpack.mkl clinpack.mkl zlinpack.mkl \
+       scholesky.mkl dcholesky.mkl ccholesky.mkl zcholesky.mkl \
+       sgemm.mkl dgemm.mkl cgemm.mkl zgemm.mkl \
+       strmm.mkl dtrmm.mkl ctrmm.mkl ztrmm.mkl \
+       strsm.mkl dtrsm.mkl ctrsm.mkl ztrsm.mkl \
+       ssyrk.mkl dsyrk.mkl csyrk.mkl zsyrk.mkl \
+       ssyr2k.mkl dsyr2k.mkl csyr2k.mkl zsyr2k.mkl \
+       sger.mkl dger.mkl cger.mkl zger.mkl \
+       sdot.mkl ddot.mkl \
+       saxpy.mkl daxpy.mkl caxpy.mkl zaxpy.mkl \
+       scopy.mkl dcopy.mkl ccopy.mkl zcopy.mkl \
+       sswap.mkl dswap.mkl cswap.mkl zswap.mkl \
+       sscal.mkl dscal.mkl cscal.mkl zscal.mkl \
+       sasum.mkl dasum.mkl casum.mkl zasum.mkl \
+       ssymv.mkl dsymv.mkl csymv.mkl zsymv.mkl \
+       chemv.mkl zhemv.mkl \
+       chemm.mkl zhemm.mkl \
+       cherk.mkl zherk.mkl \
+       cher2k.mkl zher2k.mkl \
+       sgemv.mkl dgemv.mkl cgemv.mkl zgemv.mkl \
+       sgeev.mkl dgeev.mkl cgeev.mkl zgeev.mkl \
+       sgesv.mkl dgesv.mkl cgesv.mkl zgesv.mkl \
+       sgetri.mkl dgetri.mkl cgetri.mkl zgetri.mkl \
+       spotrf.mkl dpotrf.mkl cpotrf.mkl zpotrf.mkl \
+       ssymm.mkl dsymm.mkl csymm.mkl zsymm.mkl
+
+else
+
+goto :: slinpack.goto dlinpack.goto clinpack.goto zlinpack.goto \
+       scholesky.goto dcholesky.goto ccholesky.goto zcholesky.goto \
+       sgemm.goto dgemm.goto cgemm.goto zgemm.goto \
+       strmm.goto dtrmm.goto ctrmm.goto ztrmm.goto \
+       strsm.goto dtrsm.goto ctrsm.goto ztrsm.goto \
+       ssyrk.goto dsyrk.goto csyrk.goto zsyrk.goto \
+       ssyr2k.goto dsyr2k.goto csyr2k.goto zsyr2k.goto \
+       sger.goto dger.goto cger.goto zger.goto \
+       sdot.goto ddot.goto cdot.goto zdot.goto \
+       saxpy.goto daxpy.goto caxpy.goto zaxpy.goto \
+       scopy.goto dcopy.goto ccopy.goto zcopy.goto \
+       sswap.goto dswap.goto cswap.goto zswap.goto \
+       sscal.goto dscal.goto cscal.goto zscal.goto \
+       sasum.goto dasum.goto casum.goto zasum.goto \
+       ssymv.goto dsymv.goto csymv.goto zsymv.goto \
+       chemv.goto zhemv.goto \
+       chemm.goto zhemm.goto \
+       cherk.goto zherk.goto \
+       cher2k.goto zher2k.goto \
+       sgemv.goto dgemv.goto cgemv.goto zgemv.goto \
+       sgesv.goto dgesv.goto cgesv.goto zgesv.goto \
+       sgeev.goto dgeev.goto cgeev.goto zgeev.goto \
        sgetri.goto dgetri.goto cgetri.goto zgetri.goto \
        spotrf.goto dpotrf.goto cpotrf.goto zpotrf.goto \
        ssymm.goto dsymm.goto csymm.goto zsymm.goto
@@ -60,9 +175,13 @@ acml :: slinpack.acml dlinpack.acml clinpack.acml zlinpack.acml \
        strsm.acml dtrsm.acml ctrsm.acml ztrsm.acml \
        ssyrk.acml dsyrk.acml csyrk.acml zsyrk.acml \
        ssyr2k.acml dsyr2k.acml csyr2k.acml zsyr2k.acml \
-       sger.acml dger.acml \
+       sger.acml dger.acml cger.acml zger.acml \
        sdot.acml ddot.acml \
        saxpy.acml daxpy.acml caxpy.acml zaxpy.acml \
+       scopy.acml dcopy.acml ccopy.acml zcopy.acml \
+       sswap.acml dswap.acml cswap.acml zswap.acml \
+       sscal.acml dscal.acml cscal.acml zscal.acml \
+       sasum.acml dasum.acml casum.acml zasum.acml \
        ssymv.acml dsymv.acml csymv.acml zsymv.acml \
        chemv.acml zhemv.acml \
        chemm.acml zhemm.acml \
@@ -70,6 +189,7 @@ acml :: slinpack.acml dlinpack.acml clinpack.acml zlinpack.acml \
        cher2k.acml zher2k.acml \
        sgemv.acml dgemv.acml cgemv.acml zgemv.acml \
        sgeev.acml dgeev.acml cgeev.acml zgeev.acml \
+       sgesv.acml dgesv.acml cgesv.acml zgesv.acml \
        sgetri.acml dgetri.acml cgetri.acml zgetri.acml \
        spotrf.acml dpotrf.acml cpotrf.acml zpotrf.acml \
        ssymm.acml dsymm.acml csymm.acml zsymm.acml
@@ -81,9 +201,13 @@ atlas :: slinpack.atlas dlinpack.atlas clinpack.atlas zlinpack.atlas \
        strsm.atlas dtrsm.atlas ctrsm.atlas ztrsm.atlas \
        ssyrk.atlas dsyrk.atlas csyrk.atlas zsyrk.atlas \
        ssyr2k.atlas dsyr2k.atlas csyr2k.atlas zsyr2k.atlas \
-       sger.atlas dger.atlas \
+       sger.atlas dger.atlas cger.atlas zger.atlas\
        sdot.atlas ddot.atlas \
        saxpy.atlas daxpy.atlas caxpy.atlas zaxpy.atlas \
+       scopy.atlas dcopy.atlas ccopy.atlas zcopy.atlas \
+       sswap.atlas dswap.atlas cswap.atlas zswap.atlas \
+       sscal.atlas dscal.atlas cscal.atlas zscal.atlas \
+       sasum.atlas dasum.atlas casum.atlas zasum.atlas \
        ssymv.atlas dsymv.atlas csymv.atlas zsymv.atlas \
        chemv.atlas zhemv.atlas \
        chemm.acml zhemm.acml \
@@ -92,6 +216,7 @@ atlas :: slinpack.atlas dlinpack.atlas clinpack.atlas zlinpack.atlas \
        cher2k.atlas zher2k.atlas \
        sgemv.atlas dgemv.atlas cgemv.atlas zgemv.atlas \
        sgeev.atlas dgeev.atlas cgeev.atlas zgeev.atlas \
+       sgesv.atlas dgesv.atlas cgesv.atlas zgesv.atlas \
        sgetri.atlas dgetri.atlas cgetri.atlas zgetri.atlas \
        spotrf.atlas dpotrf.atlas cpotrf.atlas zpotrf.atlas \
        ssymm.atlas dsymm.atlas csymm.atlas zsymm.atlas
@@ -103,9 +228,13 @@ mkl :: slinpack.mkl dlinpack.mkl clinpack.mkl zlinpack.mkl \
        strsm.mkl dtrsm.mkl ctrsm.mkl ztrsm.mkl \
        ssyrk.mkl dsyrk.mkl csyrk.mkl zsyrk.mkl \
        ssyr2k.mkl dsyr2k.mkl csyr2k.mkl zsyr2k.mkl \
-       sger.mkl dger.mkl \
-       sdot.mkl ddot.mkl \
+       sger.mkl dger.mkl cger.mkl zger.mkl \
+       sdot.mkl ddot.mkl cdot.mkl zdot.mkl \
        saxpy.mkl daxpy.mkl caxpy.mkl zaxpy.mkl \
+       scopy.mkl dcopy.mkl ccopy.mkl zcopy.mkl \
+       sswap.mkl dswap.mkl cswap.mkl zswap.mkl \
+       sscal.mkl dscal.mkl cscal.mkl zscal.mkl \
+       sasum.mkl dasum.mkl casum.mkl zasum.mkl \
        ssymv.mkl dsymv.mkl csymv.mkl zsymv.mkl \
        chemv.mkl zhemv.mkl \
        chemm.mkl zhemm.mkl \
@@ -113,20 +242,56 @@ mkl :: slinpack.mkl dlinpack.mkl clinpack.mkl zlinpack.mkl \
        cher2k.mkl zher2k.mkl \
        sgemv.mkl dgemv.mkl cgemv.mkl zgemv.mkl \
        sgeev.mkl dgeev.mkl cgeev.mkl zgeev.mkl \
+       sgesv.mkl dgesv.mkl cgesv.mkl zgesv.mkl \
        sgetri.mkl dgetri.mkl cgetri.mkl zgetri.mkl \
        spotrf.mkl dpotrf.mkl cpotrf.mkl zpotrf.mkl \
        ssymm.mkl dsymm.mkl csymm.mkl zsymm.mkl
 
 
+
+
+endif
+
+
+
+veclib :: slinpack.veclib dlinpack.veclib clinpack.veclib zlinpack.veclib \
+       scholesky.veclib dcholesky.veclib ccholesky.veclib zcholesky.veclib \
+       sgemm.veclib dgemm.veclib cgemm.veclib zgemm.veclib \
+       strmm.veclib dtrmm.veclib ctrmm.veclib ztrmm.veclib \
+       strsm.veclib dtrsm.veclib ctrsm.veclib ztrsm.veclib \
+       ssyrk.veclib dsyrk.veclib csyrk.veclib zsyrk.veclib \
+       ssyr2k.veclib dsyr2k.veclib csyr2k.veclib zsyr2k.veclib \
+       sger.veclib dger.veclib cger.veclib zger.veclib \
+       sdot.veclib ddot.veclib cdot.veclib zdot.veclib \
+       saxpy.veclib daxpy.veclib caxpy.veclib zaxpy.veclib \
+       scopy.veclib dcopy.veclib ccopy.veclib zcopy.veclib \
+       sswap.veclib dswap.veclib cswap.veclib zswap.veclib \
+       sscal.veclib dscal.veclib cscal.veclib zscal.veclib \
+       sasum.veclib dasum.veclib casum.veclib zasum.veclib \
+       ssymv.veclib dsymv.veclib csymv.veclib zsymv.veclib \
+       chemv.veclib zhemv.veclib \
+       chemm.veclib zhemm.veclib \
+       cherk.veclib zherk.veclib \
+       cher2k.veclib zher2k.veclib \
+       sgemv.veclib dgemv.veclib cgemv.veclib zgemv.veclib \
+       sgeev.veclib dgeev.veclib cgeev.veclib zgeev.veclib \
+       sgesv.veclib dgesv.veclib cgesv.veclib zgesv.veclib \
+       sgetri.veclib dgetri.veclib cgetri.veclib zgetri.veclib \
+       spotrf.veclib dpotrf.veclib cpotrf.veclib zpotrf.veclib \
+       ssymm.veclib dsymm.veclib csymm.veclib zsymm.veclib
+
 goto_3m :: cgemm3m.goto zgemm3m.goto
 
-mkl_3m  :: cgemm3m.mkl zgemm3m.mkl
+mkl_3m :: cgemm3m.mkl zgemm3m.mkl
 
-all :: goto mkl atlas acml
+all :: goto mkl atlas acml veclib
+
+exe :
+	@./Make_exe.sh
 
 ##################################### Slinpack ####################################################
 slinpack.goto : slinpack.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 slinpack.acml : slinpack.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -137,9 +302,12 @@ slinpack.atlas : slinpack.$(SUFFIX)
 slinpack.mkl : slinpack.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+slinpack.veclib : slinpack.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dlinpack ####################################################
 dlinpack.goto : dlinpack.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dlinpack.acml : dlinpack.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -150,10 +318,13 @@ dlinpack.atlas : dlinpack.$(SUFFIX)
 dlinpack.mkl : dlinpack.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dlinpack.veclib : dlinpack.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Clinpack ####################################################
 
 clinpack.goto : clinpack.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 clinpack.acml : clinpack.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -164,10 +335,13 @@ clinpack.atlas : clinpack.$(SUFFIX)
 clinpack.mkl : clinpack.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+clinpack.veclib : clinpack.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zlinpack ####################################################
 
 zlinpack.goto : zlinpack.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zlinpack.acml : zlinpack.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -178,10 +352,13 @@ zlinpack.atlas : zlinpack.$(SUFFIX)
 zlinpack.mkl : zlinpack.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zlinpack.veclib : zlinpack.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Scholesky ###################################################
 
 scholesky.goto : scholesky.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 scholesky.acml : scholesky.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -192,10 +369,13 @@ scholesky.atlas : scholesky.$(SUFFIX)
 scholesky.mkl : scholesky.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+scholesky.veclib : scholesky.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dcholesky ###################################################
 
 dcholesky.goto : dcholesky.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dcholesky.acml : dcholesky.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -206,10 +386,13 @@ dcholesky.atlas : dcholesky.$(SUFFIX)
 dcholesky.mkl : dcholesky.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dcholesky.veclib : dcholesky.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ccholesky ###################################################
 
 ccholesky.goto : ccholesky.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ccholesky.acml : ccholesky.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -220,13 +403,14 @@ ccholesky.atlas : ccholesky.$(SUFFIX)
 ccholesky.mkl : ccholesky.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
-zcholesky.goto : zcholesky.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+ccholesky.veclib : ccholesky.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 
 ##################################### Zcholesky ###################################################
 
-xcholesky.goto : xcholesky.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+zcholesky.goto : zcholesky.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zcholesky.acml : zcholesky.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -237,10 +421,12 @@ zcholesky.atlas : zcholesky.$(SUFFIX)
 zcholesky.mkl : zcholesky.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zcholesky.veclib : zcholesky.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
 ##################################### Sgemm ####################################################
 sgemm.goto : sgemm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 sgemm.acml : sgemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -251,9 +437,12 @@ sgemm.atlas : sgemm.$(SUFFIX)
 sgemm.mkl : sgemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+sgemm.veclib : sgemm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dgemm ####################################################
 dgemm.goto : dgemm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dgemm.acml : dgemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -264,10 +453,13 @@ dgemm.atlas : dgemm.$(SUFFIX)
 dgemm.mkl : dgemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dgemm.veclib : dgemm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Cgemm ####################################################
 
 cgemm.goto : cgemm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 cgemm.acml : cgemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -278,10 +470,13 @@ cgemm.atlas : cgemm.$(SUFFIX)
 cgemm.mkl : cgemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+cgemm.veclib : cgemm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zgemm ####################################################
 
 zgemm.goto : zgemm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zgemm.acml : zgemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -292,9 +487,12 @@ zgemm.atlas : zgemm.$(SUFFIX)
 zgemm.mkl : zgemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zgemm.veclib : zgemm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ssymm ####################################################
 ssymm.goto : ssymm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ssymm.acml : ssymm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -305,9 +503,12 @@ ssymm.atlas : ssymm.$(SUFFIX)
 ssymm.mkl : ssymm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ssymm.veclib : ssymm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dsymm ####################################################
 dsymm.goto : dsymm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dsymm.acml : dsymm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -318,10 +519,13 @@ dsymm.atlas : dsymm.$(SUFFIX)
 dsymm.mkl : dsymm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dsymm.veclib : dsymm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Csymm ####################################################
 
 csymm.goto : csymm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 csymm.acml : csymm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -332,10 +536,13 @@ csymm.atlas : csymm.$(SUFFIX)
 csymm.mkl : csymm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+csymm.veclib : csymm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zsymm ####################################################
 
 zsymm.goto : zsymm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zsymm.acml : zsymm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -346,9 +553,12 @@ zsymm.atlas : zsymm.$(SUFFIX)
 zsymm.mkl : zsymm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zsymm.veclib : zsymm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Strmm ####################################################
 strmm.goto : strmm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 strmm.acml : strmm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -359,9 +569,12 @@ strmm.atlas : strmm.$(SUFFIX)
 strmm.mkl : strmm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+strmm.veclib : strmm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dtrmm ####################################################
 dtrmm.goto : dtrmm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dtrmm.acml : dtrmm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -372,10 +585,13 @@ dtrmm.atlas : dtrmm.$(SUFFIX)
 dtrmm.mkl : dtrmm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dtrmm.veclib : dtrmm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ctrmm ####################################################
 
 ctrmm.goto : ctrmm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ctrmm.acml : ctrmm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -386,10 +602,13 @@ ctrmm.atlas : ctrmm.$(SUFFIX)
 ctrmm.mkl : ctrmm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ctrmm.veclib : ctrmm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ztrmm ####################################################
 
 ztrmm.goto : ztrmm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ztrmm.acml : ztrmm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -400,10 +619,12 @@ ztrmm.atlas : ztrmm.$(SUFFIX)
 ztrmm.mkl : ztrmm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ztrmm.veclib : ztrmm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
 ##################################### Strsm ####################################################
 strsm.goto : strsm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 strsm.acml : strsm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -414,9 +635,12 @@ strsm.atlas : strsm.$(SUFFIX)
 strsm.mkl : strsm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+strsm.veclib : strsm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dtrsm ####################################################
 dtrsm.goto : dtrsm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dtrsm.acml : dtrsm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -427,10 +651,13 @@ dtrsm.atlas : dtrsm.$(SUFFIX)
 dtrsm.mkl : dtrsm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dtrsm.veclib : dtrsm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ctrsm ####################################################
 
 ctrsm.goto : ctrsm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ctrsm.acml : ctrsm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -441,10 +668,13 @@ ctrsm.atlas : ctrsm.$(SUFFIX)
 ctrsm.mkl : ctrsm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ctrsm.veclib : ctrsm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ztrsm ####################################################
 
 ztrsm.goto : ztrsm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ztrsm.acml : ztrsm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -455,9 +685,12 @@ ztrsm.atlas : ztrsm.$(SUFFIX)
 ztrsm.mkl : ztrsm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ztrsm.veclib : ztrsm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ssyrk ####################################################
 ssyrk.goto : ssyrk.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ssyrk.acml : ssyrk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -468,9 +701,12 @@ ssyrk.atlas : ssyrk.$(SUFFIX)
 ssyrk.mkl : ssyrk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ssyrk.veclib : ssyrk.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dsyrk ####################################################
 dsyrk.goto : dsyrk.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dsyrk.acml : dsyrk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -481,10 +717,13 @@ dsyrk.atlas : dsyrk.$(SUFFIX)
 dsyrk.mkl : dsyrk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dsyrk.veclib : dsyrk.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Csyrk ####################################################
 
 csyrk.goto : csyrk.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 csyrk.acml : csyrk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -495,10 +734,13 @@ csyrk.atlas : csyrk.$(SUFFIX)
 csyrk.mkl : csyrk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+csyrk.veclib : csyrk.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zsyrk ####################################################
 
 zsyrk.goto : zsyrk.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zsyrk.acml : zsyrk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -509,10 +751,12 @@ zsyrk.atlas : zsyrk.$(SUFFIX)
 zsyrk.mkl : zsyrk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zsyrk.veclib : zsyrk.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
 ##################################### Ssyr2k ####################################################
 ssyr2k.goto : ssyr2k.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ssyr2k.acml : ssyr2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -523,9 +767,12 @@ ssyr2k.atlas : ssyr2k.$(SUFFIX)
 ssyr2k.mkl : ssyr2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ssyr2k.veclib : ssyr2k.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dsyr2k ####################################################
 dsyr2k.goto : dsyr2k.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dsyr2k.acml : dsyr2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -536,10 +783,13 @@ dsyr2k.atlas : dsyr2k.$(SUFFIX)
 dsyr2k.mkl : dsyr2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dsyr2k.veclib : dsyr2k.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Csyr2k ####################################################
 
 csyr2k.goto : csyr2k.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 csyr2k.acml : csyr2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -550,10 +800,13 @@ csyr2k.atlas : csyr2k.$(SUFFIX)
 csyr2k.mkl : csyr2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+csyr2k.veclib : csyr2k.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zsyr2k ####################################################
 
 zsyr2k.goto : zsyr2k.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zsyr2k.acml : zsyr2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -564,10 +817,13 @@ zsyr2k.atlas : zsyr2k.$(SUFFIX)
 zsyr2k.mkl : zsyr2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zsyr2k.veclib : zsyr2k.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Chemm ####################################################
 
 chemm.goto : chemm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 chemm.acml : chemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -578,10 +834,13 @@ chemm.atlas : chemm.$(SUFFIX)
 chemm.mkl : chemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+chemm.veclib : chemm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zhemm ####################################################
 
 zhemm.goto : zhemm.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zhemm.acml : zhemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -592,10 +851,13 @@ zhemm.atlas : zhemm.$(SUFFIX)
 zhemm.mkl : zhemm.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zhemm.veclib : zhemm.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Cherk ####################################################
 
 cherk.goto : cherk.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 cherk.acml : cherk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -606,10 +868,13 @@ cherk.atlas : cherk.$(SUFFIX)
 cherk.mkl : cherk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+cherk.veclib : cherk.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zherk ####################################################
 
 zherk.goto : zherk.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zherk.acml : zherk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -620,10 +885,13 @@ zherk.atlas : zherk.$(SUFFIX)
 zherk.mkl : zherk.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zherk.veclib : zherk.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Cher2k ####################################################
 
 cher2k.goto : cher2k.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 cher2k.acml : cher2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -634,10 +902,13 @@ cher2k.atlas : cher2k.$(SUFFIX)
 cher2k.mkl : cher2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+cher2k.veclib : cher2k.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zher2k ####################################################
 
 zher2k.goto : zher2k.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zher2k.acml : zher2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -648,9 +919,12 @@ zher2k.atlas : zher2k.$(SUFFIX)
 zher2k.mkl : zher2k.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zher2k.veclib : zher2k.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Sgemv ####################################################
 sgemv.goto : sgemv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 sgemv.acml : sgemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -661,9 +935,12 @@ sgemv.atlas : sgemv.$(SUFFIX)
 sgemv.mkl : sgemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+sgemv.veclib : sgemv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dgemv ####################################################
 dgemv.goto : dgemv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dgemv.acml : dgemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -674,10 +951,13 @@ dgemv.atlas : dgemv.$(SUFFIX)
 dgemv.mkl : dgemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dgemv.veclib : dgemv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Cgemv ####################################################
 
 cgemv.goto : cgemv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 cgemv.acml : cgemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -688,10 +968,13 @@ cgemv.atlas : cgemv.$(SUFFIX)
 cgemv.mkl : cgemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+cgemv.veclib : cgemv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zgemv ####################################################
 
 zgemv.goto : zgemv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zgemv.acml : zgemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -702,9 +985,12 @@ zgemv.atlas : zgemv.$(SUFFIX)
 zgemv.mkl : zgemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zgemv.veclib : zgemv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Sger ####################################################
 sger.goto : sger.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 sger.acml : sger.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -715,9 +1001,12 @@ sger.atlas : sger.$(SUFFIX)
 sger.mkl : sger.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+sger.veclib : sger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dger ####################################################
 dger.goto : dger.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dger.acml : dger.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -728,9 +1017,44 @@ dger.atlas : dger.$(SUFFIX)
 dger.mkl : dger.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dger.veclib : dger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Cger ####################################################
+cger.goto : cger.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+cger.acml : cger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cger.atlas : cger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cger.mkl : cger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cger.veclib : cger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Zger ####################################################
+zger.goto : zger.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+zger.acml : zger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zger.atlas : zger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zger.mkl : zger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zger.veclib : zger.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ssymv ####################################################
 ssymv.goto : ssymv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ssymv.acml : ssymv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -741,9 +1065,12 @@ ssymv.atlas : ssymv.$(SUFFIX)
 ssymv.mkl : ssymv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ssymv.veclib : ssymv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dsymv ####################################################
 dsymv.goto : dsymv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dsymv.acml : dsymv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -754,9 +1081,12 @@ dsymv.atlas : dsymv.$(SUFFIX)
 dsymv.mkl : dsymv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dsymv.veclib : dsymv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Csymv ####################################################
 csymv.goto : csymv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 csymv.acml : csymv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -767,9 +1097,12 @@ csymv.atlas : csymv.$(SUFFIX)
 csymv.mkl : csymv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+csymv.veclib : csymv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dsymv ####################################################
 zsymv.goto : zsymv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zsymv.acml : zsymv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -780,9 +1113,12 @@ zsymv.atlas : zsymv.$(SUFFIX)
 zsymv.mkl : zsymv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zsymv.veclib : zsymv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Sgeev ####################################################
 sgeev.goto : sgeev.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 sgeev.acml : sgeev.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -793,9 +1129,12 @@ sgeev.atlas : sgeev.$(SUFFIX)
 sgeev.mkl : sgeev.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+sgeev.veclib : sgeev.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dgeev ####################################################
 dgeev.goto : dgeev.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dgeev.acml : dgeev.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -806,10 +1145,13 @@ dgeev.atlas : dgeev.$(SUFFIX)
 dgeev.mkl : dgeev.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dgeev.veclib : dgeev.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Cgeev ####################################################
 
 cgeev.goto : cgeev.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 cgeev.acml : cgeev.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -820,10 +1162,13 @@ cgeev.atlas : cgeev.$(SUFFIX)
 cgeev.mkl : cgeev.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+cgeev.veclib : cgeev.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zgeev ####################################################
 
 zgeev.goto : zgeev.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zgeev.acml : zgeev.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -834,10 +1179,12 @@ zgeev.atlas : zgeev.$(SUFFIX)
 zgeev.mkl : zgeev.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zgeev.veclib : zgeev.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
 ##################################### Sgetri ####################################################
 sgetri.goto : sgetri.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 sgetri.acml : sgetri.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -848,9 +1195,12 @@ sgetri.atlas : sgetri.$(SUFFIX)
 sgetri.mkl : sgetri.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+sgetri.veclib : sgetri.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dgetri ####################################################
 dgetri.goto : dgetri.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dgetri.acml : dgetri.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -861,10 +1211,13 @@ dgetri.atlas : dgetri.$(SUFFIX)
 dgetri.mkl : dgetri.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dgetri.veclib : dgetri.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Cgetri ####################################################
 
 cgetri.goto : cgetri.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 cgetri.acml : cgetri.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -875,10 +1228,13 @@ cgetri.atlas : cgetri.$(SUFFIX)
 cgetri.mkl : cgetri.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+cgetri.veclib : cgetri.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zgetri ####################################################
 
 zgetri.goto : zgetri.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zgetri.acml : zgetri.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -889,10 +1245,12 @@ zgetri.atlas : zgetri.$(SUFFIX)
 zgetri.mkl : zgetri.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zgetri.veclib : zgetri.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
 ##################################### Spotrf ####################################################
 spotrf.goto : spotrf.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 spotrf.acml : spotrf.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -903,9 +1261,12 @@ spotrf.atlas : spotrf.$(SUFFIX)
 spotrf.mkl : spotrf.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+spotrf.veclib : spotrf.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Dpotrf ####################################################
 dpotrf.goto : dpotrf.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 dpotrf.acml : dpotrf.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -916,10 +1277,13 @@ dpotrf.atlas : dpotrf.$(SUFFIX)
 dpotrf.mkl : dpotrf.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+dpotrf.veclib : dpotrf.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Cpotrf ####################################################
 
 cpotrf.goto : cpotrf.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 cpotrf.acml : cpotrf.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -930,10 +1294,13 @@ cpotrf.atlas : cpotrf.$(SUFFIX)
 cpotrf.mkl : cpotrf.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+cpotrf.veclib : cpotrf.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zpotrf ####################################################
 
 zpotrf.goto : zpotrf.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zpotrf.acml : zpotrf.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -944,10 +1311,13 @@ zpotrf.atlas : zpotrf.$(SUFFIX)
 zpotrf.mkl : zpotrf.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zpotrf.veclib : zpotrf.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Chemv ####################################################
 
 chemv.goto : chemv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 chemv.acml : chemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -958,10 +1328,13 @@ chemv.atlas : chemv.$(SUFFIX)
 chemv.mkl : chemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+chemv.veclib : chemv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zhemv ####################################################
 
 zhemv.goto : zhemv.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zhemv.acml : zhemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -972,9 +1345,12 @@ zhemv.atlas : zhemv.$(SUFFIX)
 zhemv.mkl : zhemv.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zhemv.veclib : zhemv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Sdot ####################################################
 sdot.goto : sdot.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 sdot.acml : sdot.$(SUFFIX)
 	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -985,9 +1361,12 @@ sdot.atlas : sdot.$(SUFFIX)
 sdot.mkl : sdot.$(SUFFIX)
 	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+sdot.veclib : sdot.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Ddot ####################################################
 ddot.goto : ddot.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 ddot.acml : ddot.$(SUFFIX)
 	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -998,9 +1377,44 @@ ddot.atlas : ddot.$(SUFFIX)
 ddot.mkl : ddot.$(SUFFIX)
 	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+ddot.veclib : ddot.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Cdot ####################################################
+cdot.goto : cdot.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+cdot.acml : cdot.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cdot.atlas : cdot.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cdot.mkl : cdot-intel.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cdot.veclib : cdot-intel.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Zdot ####################################################
+zdot.goto : zdot.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+zdot.acml : zdot.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zdot.atlas : zdot.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zdot.mkl : zdot-intel.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zdot.veclib : zdot-intel.$(SUFFIX)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Saxpy ####################################################
 saxpy.goto : saxpy.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 saxpy.acml : saxpy.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -1011,9 +1425,12 @@ saxpy.atlas : saxpy.$(SUFFIX)
 saxpy.mkl : saxpy.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+saxpy.veclib : saxpy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Daxpy ####################################################
 daxpy.goto : daxpy.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 daxpy.acml : daxpy.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -1024,10 +1441,13 @@ daxpy.atlas : daxpy.$(SUFFIX)
 daxpy.mkl : daxpy.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+daxpy.veclib : daxpy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Caxpy ####################################################
 
 caxpy.goto : caxpy.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 caxpy.acml : caxpy.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -1038,10 +1458,13 @@ caxpy.atlas : caxpy.$(SUFFIX)
 caxpy.mkl : caxpy.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+caxpy.veclib : caxpy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zaxpy ####################################################
 
 zaxpy.goto : zaxpy.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zaxpy.acml : zaxpy.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
@@ -1052,23 +1475,363 @@ zaxpy.atlas : zaxpy.$(SUFFIX)
 zaxpy.mkl : zaxpy.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zaxpy.veclib : zaxpy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+
+##################################### Scopy ####################################################
+scopy.goto : scopy.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+scopy.acml : scopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+scopy.atlas : scopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+scopy.mkl : scopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+scopy.veclib : scopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Dcopy ####################################################
+dcopy.goto : dcopy.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+dcopy.acml : dcopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dcopy.atlas : dcopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dcopy.mkl : dcopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dcopy.veclib : dcopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Ccopy ####################################################
+
+ccopy.goto : ccopy.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+ccopy.acml : ccopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+ccopy.atlas : ccopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+ccopy.mkl : ccopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+ccopy.veclib : ccopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Zcopy ####################################################
+
+zcopy.goto : zcopy.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+zcopy.acml : zcopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zcopy.atlas : zcopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zcopy.mkl : zcopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zcopy.veclib : zcopy.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Sscal ####################################################
+sscal.goto : sscal.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+sscal.acml : sscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sscal.atlas : sscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sscal.mkl : sscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sscal.veclib : sscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Dscal ####################################################
+dscal.goto : dscal.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+dscal.acml : dscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dscal.atlas : dscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dscal.mkl : dscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dscal.veclib : dscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Cscal ####################################################
+
+cscal.goto : cscal.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+cscal.acml : cscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cscal.atlas : cscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cscal.mkl : cscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cscal.veclib : cscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Zscal ####################################################
+
+zscal.goto : zscal.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+zscal.acml : zscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zscal.atlas : zscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zscal.mkl : zscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zscal.veclib : zscal.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Sasum ####################################################
+sasum.goto : sasum.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+sasum.acml : sasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sasum.atlas : sasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sasum.mkl : sasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sasum.veclib : sasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Dasum ####################################################
+dasum.goto : dasum.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+dasum.acml : dasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dasum.atlas : dasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dasum.mkl : dasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dasum.veclib : dasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Casum ####################################################
+
+casum.goto : casum.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+casum.acml : casum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+casum.atlas : casum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+casum.mkl : casum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+casum.veclib : casum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Zasum ####################################################
+
+zasum.goto : zasum.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+zasum.acml : zasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zasum.atlas : zasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zasum.mkl : zasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zasum.veclib : zasum.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Sswap ####################################################
+sswap.goto : sswap.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+sswap.acml : sswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sswap.atlas : sswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sswap.mkl : sswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sswap.veclib : sswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Dswap ####################################################
+dswap.goto : dswap.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+dswap.acml : dswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dswap.atlas : dswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dswap.mkl : dswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dswap.veclib : dswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Cswap ####################################################
+
+cswap.goto : cswap.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+cswap.acml : cswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cswap.atlas : cswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cswap.mkl : cswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cswap.veclib : cswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Zswap ####################################################
+
+zswap.goto : zswap.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+zswap.acml : zswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zswap.atlas : zswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zswap.mkl : zswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zswap.veclib : zswap.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+
+##################################### Sgesv ####################################################
+sgesv.goto : sgesv.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+sgesv.acml : sgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sgesv.atlas : sgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sgesv.mkl : sgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+sgesv.veclib : sgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Dgesv ####################################################
+dgesv.goto : dgesv.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+dgesv.acml : dgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dgesv.atlas : dgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dgesv.mkl : dgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+dgesv.veclib : dgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Cgesv ####################################################
+
+cgesv.goto : cgesv.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+cgesv.acml : cgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cgesv.atlas : cgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cgesv.mkl : cgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+cgesv.veclib : cgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+##################################### Zgesv ####################################################
+
+zgesv.goto : zgesv.$(SUFFIX) ../$(LIBNAME)
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
+
+zgesv.acml : zgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zgesv.atlas : zgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zgesv.mkl : zgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
+zgesv.veclib : zgesv.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 
 ##################################### Cgemm3m ####################################################
 
 cgemm3m.goto : cgemm3m.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 cgemm3m.mkl : cgemm3m.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+cgemm3m.veclib : cgemm3m.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
+
 ##################################### Zgemm3m ####################################################
 
 zgemm3m.goto : zgemm3m.$(SUFFIX) ../$(LIBNAME)
-	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
+	$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB) -lm
 
 zgemm3m.mkl : zgemm3m.$(SUFFIX)
 	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
+zgemm3m.veclib : zgemm3m.$(SUFFIX)
+	-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBVECLIB) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
 
 ###################################################################################################
 
@@ -1204,6 +1967,13 @@ sger.$(SUFFIX) : ger.c
 dger.$(SUFFIX) : ger.c
 	$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
 
+cger.$(SUFFIX) : ger.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
+
+zger.$(SUFFIX) : ger.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
+
+
 ssymv.$(SUFFIX) : symv.c
 	$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^
 
@@ -1264,6 +2034,20 @@ sdot.$(SUFFIX) : dot.c
 ddot.$(SUFFIX) : dot.c
 	$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
 
+cdot.$(SUFFIX) : zdot.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
+
+zdot.$(SUFFIX) : zdot.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
+
+cdot-intel.$(SUFFIX) : zdot-intel.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
+
+zdot-intel.$(SUFFIX) : zdot-intel.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
+
+
+
 saxpy.$(SUFFIX) : axpy.c
 	$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^
 
@@ -1276,6 +2060,72 @@ caxpy.$(SUFFIX) : axpy.c
 zaxpy.$(SUFFIX) : axpy.c
 	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
 
+scopy.$(SUFFIX) : copy.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^
+
+dcopy.$(SUFFIX) : copy.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
+
+ccopy.$(SUFFIX) : copy.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
+
+zcopy.$(SUFFIX) : copy.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
+
+sswap.$(SUFFIX) : swap.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^
+
+dswap.$(SUFFIX) : swap.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
+
+cswap.$(SUFFIX) : swap.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
+
+zswap.$(SUFFIX) : swap.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
+
+
+
+sscal.$(SUFFIX) : scal.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^
+
+dscal.$(SUFFIX) : scal.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
+
+cscal.$(SUFFIX) : scal.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
+
+zscal.$(SUFFIX) : scal.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
+
+sasum.$(SUFFIX) : asum.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^
+
+dasum.$(SUFFIX) : asum.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
+
+casum.$(SUFFIX) : asum.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
+
+zasum.$(SUFFIX) : asum.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
+
+
+sgesv.$(SUFFIX) : gesv.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^
+
+dgesv.$(SUFFIX) : gesv.c
+	$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
+
+cgesv.$(SUFFIX) : gesv.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
+
+zgesv.$(SUFFIX) : gesv.c
+	$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
+
+
+
+
 cgemm3m.$(SUFFIX) : gemm3m.c
 	$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
 
@@ -1284,7 +2134,7 @@ zgemm3m.$(SUFFIX) : gemm3m.c
 
 
 clean ::
-	@rm -f *.goto *.mkl *.acml *.atlas
+	@rm -f *.goto *.mkl *.acml *.atlas *.veclib
 
 include $(TOPDIR)/Makefile.tail
 
diff --git a/benchmark/asum.c b/benchmark/asum.c
new file mode 100644
index 000000000..beb6402f4
--- /dev/null
+++ b/benchmark/asum.c
@@ -0,0 +1,196 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include <stdio.h>
+#include <stdlib.h>
+#ifdef __CYGWIN32__
+#include <sys/time.h>
+#endif
+#include "common.h"
+
+
+#undef ASUM
+
+#ifdef COMPLEX
+#ifdef DOUBLE
+#define ASUM   BLASFUNC(dzasum)
+#else
+#define ASUM   BLASFUNC(scasum)
+#endif
+#else
+#ifdef DOUBLE
+#define ASUM   BLASFUNC(dasum)
+#else
+#define ASUM   BLASFUNC(sasum)
+#endif
+#endif
+
+#if defined(__WIN32__) || defined(__WIN64__)
+
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
+int gettimeofday(struct timeval *tv, void *tz){
+
+  FILETIME ft;
+  unsigned __int64 tmpres = 0;
+  static int tzflag;
+
+  if (NULL != tv)
+    {
+      GetSystemTimeAsFileTime(&ft);
+
+      tmpres |= ft.dwHighDateTime;
+      tmpres <<= 32;
+      tmpres |= ft.dwLowDateTime;
+
+      /*converting file time to unix epoch*/
+      tmpres /= 10;  /*convert into microseconds*/
+      tmpres -= DELTA_EPOCH_IN_MICROSECS;
+      tv->tv_sec = (long)(tmpres / 1000000UL);
+      tv->tv_usec = (long)(tmpres % 1000000UL);
+    }
+
+  return 0;
+}
+
+#endif
+
+#if !defined(__WIN32__) && !defined(__WIN64__) && !defined(__CYGWIN32__) && 0
+
+static void *huge_malloc(BLASLONG size){
+  int shmid;
+  void *address;
+
+#ifndef SHM_HUGETLB
+#define SHM_HUGETLB 04000
+#endif
+
+  if ((shmid =shmget(IPC_PRIVATE,
+		     (size + HUGE_PAGESIZE) & ~(HUGE_PAGESIZE - 1),
+		     SHM_HUGETLB | IPC_CREAT |0600)) < 0) {
+    printf( "Memory allocation failed(shmget).\n");
+    exit(1);
+  }
+
+  address = shmat(shmid, NULL, SHM_RND);
+
+  if ((BLASLONG)address == -1){
+    printf( "Memory allocation failed(shmat).\n");
+    exit(1);
+  }
+
+  shmctl(shmid, IPC_RMID, 0);
+
+  return address;
+}
+
+#define malloc huge_malloc
+
+#endif
+
+int main(int argc, char *argv[]){
+
+  FLOAT *x;
+  FLOAT result;
+  blasint m, i;
+  blasint inc_x=1;
+  int loops = 1;
+  int l;
+  char *p;
+
+  int from =   1;
+  int to   = 200;
+  int step =   1;
+
+  struct timeval start, stop;
+  double time1,timeg;
+
+  argc--;argv++;
+
+  if (argc > 0) { from     = atol(*argv);		argc--; argv++;}
+  if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
+  if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
+
+  if ((p = getenv("OPENBLAS_LOOPS")))  loops = atoi(p);
+  if ((p = getenv("OPENBLAS_INCX")))   inc_x = atoi(p);
+
+  fprintf(stderr, "From : %3d  To : %3d Step = %3d Inc_x = %d Loops = %d\n", from, to, step,inc_x,loops);
+
+  if (( x = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_x) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+
+#ifdef linux
+  srandom(getpid());
+#endif
+
+  fprintf(stderr, "   SIZE       Flops\n");
+
+  for(m = from; m <= to; m += step)
+  {
+
+   timeg=0;
+
+   fprintf(stderr, " %6d : ", (int)m);
+
+
+   for (l=0; l<loops; l++)
+   {
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_x); i++){
+			x[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+
+    	gettimeofday( &start, (struct timezone *)0);
+
+    	result = ASUM (&m, x, &inc_x);
+
+    	gettimeofday( &stop, (struct timezone *)0);
+
+    	time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+
+	timeg += time1;
+
+    }
+
+    timeg /= loops;
+
+#ifdef COMPLEX
+    fprintf(stderr, " %10.2f MFlops\n", 4. * (double)m / timeg * 1.e-6);
+#else
+    fprintf(stderr, " %10.2f MFlops\n", 2. * (double)m / timeg * 1.e-6);
+#endif
+
+  }
+
+  return 0;
+}
+
+// void main(int argc, char *argv[]) __attribute__((weak, alias("MAIN__")));
diff --git a/benchmark/cholesky.c b/benchmark/cholesky.c
index c8b96d80f..8d121efb3 100644
--- a/benchmark/cholesky.c
+++ b/benchmark/cholesky.c
@@ -71,8 +71,14 @@ double fabs(double);
 #endif
 #endif
 
+
+
 #if defined(__WIN32__) || defined(__WIN64__)
 
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
 int gettimeofday(struct timeval *tv, void *tz){
 
   FILETIME ft;
@@ -99,6 +105,7 @@ int gettimeofday(struct timeval *tv, void *tz){
 
 #endif
 
+
 static __inline double getmflops(int ratio, int m, double secs){
 
   double mm = (double)m;
diff --git a/benchmark/copy.c b/benchmark/copy.c
new file mode 100644
index 000000000..15c45201c
--- /dev/null
+++ b/benchmark/copy.c
@@ -0,0 +1,201 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include <stdio.h>
+#include <stdlib.h>
+#ifdef __CYGWIN32__
+#include <sys/time.h>
+#endif
+#include "common.h"
+
+
+#undef COPY
+
+#ifdef COMPLEX
+#ifdef DOUBLE
+#define COPY   BLASFUNC(zcopy)
+#else
+#define COPY   BLASFUNC(ccopy)
+#endif
+#else
+#ifdef DOUBLE
+#define COPY   BLASFUNC(dcopy)
+#else
+#define COPY   BLASFUNC(scopy)
+#endif
+#endif
+
+#if defined(__WIN32__) || defined(__WIN64__)
+
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
+int gettimeofday(struct timeval *tv, void *tz){
+
+  FILETIME ft;
+  unsigned __int64 tmpres = 0;
+  static int tzflag;
+
+  if (NULL != tv)
+    {
+      GetSystemTimeAsFileTime(&ft);
+
+      tmpres |= ft.dwHighDateTime;
+      tmpres <<= 32;
+      tmpres |= ft.dwLowDateTime;
+
+      /*converting file time to unix epoch*/
+      tmpres /= 10;  /*convert into microseconds*/
+      tmpres -= DELTA_EPOCH_IN_MICROSECS;
+      tv->tv_sec = (long)(tmpres / 1000000UL);
+      tv->tv_usec = (long)(tmpres % 1000000UL);
+    }
+
+  return 0;
+}
+
+#endif
+
+#if !defined(__WIN32__) && !defined(__WIN64__) && !defined(__CYGWIN32__) && 0
+
+static void *huge_malloc(BLASLONG size){
+  int shmid;
+  void *address;
+
+#ifndef SHM_HUGETLB
+#define SHM_HUGETLB 04000
+#endif
+
+  if ((shmid =shmget(IPC_PRIVATE,
+		     (size + HUGE_PAGESIZE) & ~(HUGE_PAGESIZE - 1),
+		     SHM_HUGETLB | IPC_CREAT |0600)) < 0) {
+    printf( "Memory allocation failed(shmget).\n");
+    exit(1);
+  }
+
+  address = shmat(shmid, NULL, SHM_RND);
+
+  if ((BLASLONG)address == -1){
+    printf( "Memory allocation failed(shmat).\n");
+    exit(1);
+  }
+
+  shmctl(shmid, IPC_RMID, 0);
+
+  return address;
+}
+
+#define malloc huge_malloc
+
+#endif
+
+int main(int argc, char *argv[]){
+
+  FLOAT *x, *y;
+  FLOAT alpha[2] = { 2.0, 2.0 };
+  blasint m, i;
+  blasint inc_x=1,inc_y=1;
+  int loops = 1;
+  int l;
+  char *p;
+
+  int from =   1;
+  int to   = 200;
+  int step =   1;
+
+  struct timeval start, stop;
+  double time1,timeg;
+
+  argc--;argv++;
+
+  if (argc > 0) { from     = atol(*argv);		argc--; argv++;}
+  if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
+  if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
+
+  if ((p = getenv("OPENBLAS_LOOPS")))  loops = atoi(p);
+  if ((p = getenv("OPENBLAS_INCX")))   inc_x = atoi(p);
+  if ((p = getenv("OPENBLAS_INCY")))   inc_y = atoi(p);
+
+  fprintf(stderr, "From : %3d  To : %3d Step = %3d Inc_x = %d Inc_y = %d Loops = %d\n", from, to, step,inc_x,inc_y,loops);
+
+  if (( x = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_x) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+  if (( y = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_y) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+#ifdef linux
+  srandom(getpid());
+#endif
+
+  fprintf(stderr, "   SIZE       Flops\n");
+
+  for(m = from; m <= to; m += step)
+  {
+
+   timeg=0;
+
+   fprintf(stderr, " %6d : ", (int)m);
+
+
+   for (l=0; l<loops; l++)
+   {
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_x); i++){
+			x[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_y); i++){
+			y[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+    	gettimeofday( &start, (struct timezone *)0);
+
+    	COPY (&m, x, &inc_x, y, &inc_y );
+
+    	gettimeofday( &stop, (struct timezone *)0);
+
+    	time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+
+	timeg += time1;
+
+    }
+
+    timeg /= loops;
+
+    fprintf(stderr,
+	    " %10.2f MBytes\n",
+	    COMPSIZE * sizeof(FLOAT) * 1. * (double)m / timeg * 1.e-6);
+
+  }
+
+  return 0;
+}
+
+// void main(int argc, char *argv[]) __attribute__((weak, alias("MAIN__")));
diff --git a/benchmark/geev.c b/benchmark/geev.c
index a2ca2c315..d3751defb 100644
--- a/benchmark/geev.c
+++ b/benchmark/geev.c
@@ -144,6 +144,7 @@ int main(int argc, char *argv[]){
   FLOAT *a,*vl,*vr,*wi,*wr,*work,*rwork;
   FLOAT wkopt[4];
   char job='V';
+  char jobr='N';
   char *p;
 
   blasint m, i, j, info,lwork;
@@ -202,9 +203,9 @@ int main(int argc, char *argv[]){
     lwork = -1;
     m=to;
 #ifndef COMPLEX
-    GEEV (&job, &job, &m, a, &m, wr, wi, vl, &m, vr, &m, wkopt, &lwork, &info);
+    GEEV (&job, &jobr, &m, a, &m, wr, wi, vl, &m, vr, &m, wkopt, &lwork, &info);
 #else
-    GEEV (&job, &job, &m, a, &m, wr, vl, &m, vr, &m, wkopt, &lwork,rwork, &info);
+    GEEV (&job, &jobr, &m, a, &m, wr, vl, &m, vr, &m, wkopt, &lwork,rwork, &info);
 #endif
 
   lwork = (blasint)wkopt[0];
@@ -226,16 +227,16 @@ int main(int argc, char *argv[]){
 
     lwork = -1;
 #ifndef COMPLEX
-    GEEV (&job, &job, &m, a, &m, wr, wi, vl, &m, vr, &m, wkopt, &lwork, &info);
+    GEEV (&job, &jobr, &m, a, &m, wr, wi, vl, &m, vr, &m, wkopt, &lwork, &info);
 #else
-    GEEV (&job, &job, &m, a, &m, wr, vl, &m, vr, &m, wkopt, &lwork,rwork, &info);
+    GEEV (&job, &jobr, &m, a, &m, wr, vl, &m, vr, &m, wkopt, &lwork,rwork, &info);
 #endif
 
     lwork = (blasint)wkopt[0];
 #ifndef COMPLEX
-    GEEV (&job, &job, &m, a, &m, wr, wi, vl, &m, vr, &m, work, &lwork, &info);
+    GEEV (&job, &jobr, &m, a, &m, wr, wi, vl, &m, vr, &m, work, &lwork, &info);
 #else
-    GEEV (&job, &job, &m, a, &m, wr, vl, &m, vr, &m, work, &lwork,rwork, &info);
+    GEEV (&job, &jobr, &m, a, &m, wr, vl, &m, vr, &m, work, &lwork,rwork, &info);
 #endif
 
     gettimeofday( &stop, (struct timezone *)0);
diff --git a/benchmark/gemm.c b/benchmark/gemm.c
index 5a3587622..9348018dc 100644
--- a/benchmark/gemm.c
+++ b/benchmark/gemm.c
@@ -122,7 +122,7 @@ int main(int argc, char *argv[]){
 
   FLOAT *a, *b, *c;
   FLOAT alpha[] = {1.0, 1.0};
-  FLOAT beta [] = {1.0, 1.0};
+  FLOAT beta [] = {0.0, 0.0};
   char trans='N';
   blasint m, n, i, j;
   int loops = 1;
@@ -168,12 +168,21 @@ int main(int argc, char *argv[]){
 	  has_param_n=1;	  
   }
 
-
 #ifdef linux
   srandom(getpid());
 #endif
+ 
+	for(j = 0; j < m; j++){
+      		for(i = 0; i < to * COMPSIZE; i++){
+			a[i + j * to * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+			b[i + j * to * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+			c[i + j * to * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+      		}
+    	}
 
-  fprintf(stderr, "   SIZE       Flops\n");
+
+
+  fprintf(stderr, "   SIZE          Flops          Time\n");
 
   for(m = from; m <= to; m += step)
   {
@@ -188,34 +197,23 @@ int main(int argc, char *argv[]){
 
 
     fprintf(stderr, " %6dx%d : ", (int)m, (int)n);
+    gettimeofday( &start, (struct timezone *)0);
 
     for (l=0; l<loops; l++)
     {
-  
-    	for(j = 0; j < m; j++){
-      		for(i = 0; i < m * COMPSIZE; i++){
-			a[i + j * m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
-			b[i + j * m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
-			c[i + j * m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
-      		}
-    	}
-
-    	gettimeofday( &start, (struct timezone *)0);
 
     	GEMM (&trans, &trans, &m, &n, &m, alpha, a, &m, b, &m, beta, c, &m );
 
-    	gettimeofday( &stop, (struct timezone *)0);
 
-    	time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
-
-	timeg += time1;
 
     }
+   gettimeofday( &stop, (struct timezone *)0);
+   time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
 
-    timeg /= loops;
+    timeg = time1/loops;
     fprintf(stderr,
-	    " %10.2f MFlops\n",
-	    COMPSIZE * COMPSIZE * 2. * (double)m * (double)m * (double)n / timeg * 1.e-6);
+	    " %10.2f MFlops %10.6f sec\n",
+	    COMPSIZE * COMPSIZE * 2. * (double)m * (double)m * (double)n / timeg * 1.e-6, time1);
 
   }
 
diff --git a/benchmark/ger.c b/benchmark/ger.c
index 354281006..a752a3c3e 100644
--- a/benchmark/ger.c
+++ b/benchmark/ger.c
@@ -35,12 +35,19 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #undef GER
 
-
+#ifdef COMPLEX
+#ifdef DOUBLE
+#define GER   BLASFUNC(zgeru)
+#else
+#define GER   BLASFUNC(cgeru)
+#endif
+#else
 #ifdef DOUBLE
 #define GER   BLASFUNC(dger)
 #else
 #define GER   BLASFUNC(sger)
 #endif
+#endif
 
 
 #if defined(__WIN32__) || defined(__WIN64__)
diff --git a/benchmark/gesv.c b/benchmark/gesv.c
new file mode 100644
index 000000000..26ff8bc1a
--- /dev/null
+++ b/benchmark/gesv.c
@@ -0,0 +1,218 @@
+/*********************************************************************/
+/* Copyright 2009, 2010 The University of Texas at Austin.           */
+/* All rights reserved.                                              */
+/*                                                                   */
+/* Redistribution and use in source and binary forms, with or        */
+/* without modification, are permitted provided that the following   */
+/* conditions are met:                                               */
+/*                                                                   */
+/*   1. Redistributions of source code must retain the above         */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer.                                                  */
+/*                                                                   */
+/*   2. Redistributions in binary form must reproduce the above      */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer in the documentation and/or other materials       */
+/*      provided with the distribution.                              */
+/*                                                                   */
+/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
+/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
+/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
+/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
+/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
+/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
+/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
+/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
+/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
+/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
+/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
+/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
+/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
+/*    POSSIBILITY OF SUCH DAMAGE.                                    */
+/*                                                                   */
+/* The views and conclusions contained in the software and           */
+/* documentation are those of the authors and should not be          */
+/* interpreted as representing official policies, either expressed   */
+/* or implied, of The University of Texas at Austin.                 */
+/*********************************************************************/
+
+#include <stdio.h>
+#include <stdlib.h>
+#ifdef __CYGWIN32__
+#include <sys/time.h>
+#endif
+#include "common.h"
+
+double fabs(double);
+
+#undef GESV
+#undef GETRS
+
+#ifndef COMPLEX
+#ifdef XDOUBLE
+#define GESV   BLASFUNC(qgesv)
+#elif defined(DOUBLE)
+#define GESV   BLASFUNC(dgesv)
+#else
+#define GESV   BLASFUNC(sgesv)
+#endif
+#else
+#ifdef XDOUBLE
+#define GESV   BLASFUNC(xgesv)
+#elif defined(DOUBLE)
+#define GESV   BLASFUNC(zgesv)
+#else
+#define GESV   BLASFUNC(cgesv)
+#endif
+#endif
+
+#if defined(__WIN32__) || defined(__WIN64__)
+
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
+int gettimeofday(struct timeval *tv, void *tz){
+
+  FILETIME ft;
+  unsigned __int64 tmpres = 0;
+  static int tzflag;
+
+  if (NULL != tv)
+    {
+      GetSystemTimeAsFileTime(&ft);
+
+      tmpres |= ft.dwHighDateTime;
+      tmpres <<= 32;
+      tmpres |= ft.dwLowDateTime;
+
+      /*converting file time to unix epoch*/
+      tmpres /= 10;  /*convert into microseconds*/
+      tmpres -= DELTA_EPOCH_IN_MICROSECS;
+      tv->tv_sec = (long)(tmpres / 1000000UL);
+      tv->tv_usec = (long)(tmpres % 1000000UL);
+    }
+
+  return 0;
+}
+
+#endif
+
+#if !defined(__WIN32__) && !defined(__WIN64__) && !defined(__CYGWIN32__) && 0
+
+static void *huge_malloc(BLASLONG size){
+  int shmid;
+  void *address;
+
+#ifndef SHM_HUGETLB
+#define SHM_HUGETLB 04000
+#endif
+
+  if ((shmid =shmget(IPC_PRIVATE,
+		     (size + HUGE_PAGESIZE) & ~(HUGE_PAGESIZE - 1),
+		     SHM_HUGETLB | IPC_CREAT |0600)) < 0) {
+    printf( "Memory allocation failed(shmget).\n");
+    exit(1);
+  }
+
+  address = shmat(shmid, NULL, SHM_RND);
+
+  if ((BLASLONG)address == -1){
+    printf( "Memory allocation failed(shmat).\n");
+    exit(1);
+  }
+
+  shmctl(shmid, IPC_RMID, 0);
+
+  return address;
+}
+
+#define malloc huge_malloc
+
+#endif
+
+int main(int argc, char *argv[]){
+
+  FLOAT *a, *b;
+  blasint *ipiv;
+
+  blasint m, i, j, info;
+
+  int from =   1;
+  int to   = 200;
+  int step =   1;
+
+  struct timeval start, stop;
+  double time1;
+
+  argc--;argv++;
+
+  if (argc > 0) { from     = atol(*argv);		argc--; argv++;}
+  if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
+  if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
+
+  fprintf(stderr, "From : %3d  To : %3d Step = %3d\n", from, to, step);
+
+  if (( a = (FLOAT *)malloc(sizeof(FLOAT) * to * to * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+  if (( b = (FLOAT *)malloc(sizeof(FLOAT) * to * to * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+  if (( ipiv = (blasint *)malloc(sizeof(blasint) * to * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+#ifdef linux
+  srandom(getpid());
+#endif
+
+  fprintf(stderr, "   SIZE       Flops              Time\n");
+
+  for(m = from; m <= to; m += step){
+
+    fprintf(stderr, " %dx%d : ", (int)m, (int)m);
+
+    for(j = 0; j < m; j++){
+      for(i = 0; i < m * COMPSIZE; i++){
+	a[i + j * m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+      }
+    }
+
+    for(j = 0; j < m; j++){
+      for(i = 0; i < m * COMPSIZE; i++){
+	b[i + j * m * COMPSIZE] = 0.0;
+      }
+    }
+
+
+    for (j = 0; j < m; ++j) {
+      for (i = 0; i < m * COMPSIZE; ++i) {
+	b[i] += a[i + j * m * COMPSIZE];
+      }
+    }
+
+    gettimeofday( &start, (struct timezone *)0);
+
+    GESV (&m, &m, a, &m, ipiv, b, &m,  &info);
+
+    gettimeofday( &stop, (struct timezone *)0);
+
+
+    time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+
+
+
+    fprintf(stderr,
+	    "%10.2f MFlops %10.6f s\n",
+	    COMPSIZE * COMPSIZE * (2. / 3. * (double)m * (double)m * (double)m + 2. * (double)m * (double)m * (double)m ) / (time1) * 1.e-6 , time1);
+
+
+  }
+
+  return 0;
+}
+
+// void main(int argc, char *argv[]) __attribute__((weak, alias("MAIN__")));
diff --git a/benchmark/plot-filter.sh b/benchmark/plot-filter.sh
index 420ec9b02..73fbe97b6 100755
--- a/benchmark/plot-filter.sh
+++ b/benchmark/plot-filter.sh
@@ -52,6 +52,11 @@ C)
 	awk '/MFlops/ { print $3,int($9) }'|tail --lines=+2
 	;;
 
+B)
+	# Copy Benchmark	
+	awk '/MBytes/ { print $1,int($3) }'|tail --lines=+2
+	;;
+
 
 *)
 	awk '/MFlops/ { print $1,int($3) }'|tail --lines=+2
diff --git a/benchmark/potrf.c b/benchmark/potrf.c
index 3caf61caa..1d714549b 100644
--- a/benchmark/potrf.c
+++ b/benchmark/potrf.c
@@ -88,6 +88,10 @@ double fabs(double);
 
 #if defined(__WIN32__) || defined(__WIN64__)
 
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
 int gettimeofday(struct timeval *tv, void *tz){
 
   FILETIME ft;
diff --git a/benchmark/scal.c b/benchmark/scal.c
new file mode 100644
index 000000000..4c2da4d30
--- /dev/null
+++ b/benchmark/scal.c
@@ -0,0 +1,202 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include <stdio.h>
+#include <stdlib.h>
+#ifdef __CYGWIN32__
+#include <sys/time.h>
+#endif
+#include "common.h"
+
+
+#undef SCAL
+
+#ifdef COMPLEX
+#ifdef DOUBLE
+#define SCAL   BLASFUNC(zscal)
+#else
+#define SCAL   BLASFUNC(cscal)
+#endif
+#else
+#ifdef DOUBLE
+#define SCAL   BLASFUNC(dscal)
+#else
+#define SCAL   BLASFUNC(sscal)
+#endif
+#endif
+
+#if defined(__WIN32__) || defined(__WIN64__)
+
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
+int gettimeofday(struct timeval *tv, void *tz){
+
+  FILETIME ft;
+  unsigned __int64 tmpres = 0;
+  static int tzflag;
+
+  if (NULL != tv)
+    {
+      GetSystemTimeAsFileTime(&ft);
+
+      tmpres |= ft.dwHighDateTime;
+      tmpres <<= 32;
+      tmpres |= ft.dwLowDateTime;
+
+      /*converting file time to unix epoch*/
+      tmpres /= 10;  /*convert into microseconds*/
+      tmpres -= DELTA_EPOCH_IN_MICROSECS;
+      tv->tv_sec = (long)(tmpres / 1000000UL);
+      tv->tv_usec = (long)(tmpres % 1000000UL);
+    }
+
+  return 0;
+}
+
+#endif
+
+#if !defined(__WIN32__) && !defined(__WIN64__) && !defined(__CYGWIN32__) && 0
+
+static void *huge_malloc(BLASLONG size){
+  int shmid;
+  void *address;
+
+#ifndef SHM_HUGETLB
+#define SHM_HUGETLB 04000
+#endif
+
+  if ((shmid =shmget(IPC_PRIVATE,
+		     (size + HUGE_PAGESIZE) & ~(HUGE_PAGESIZE - 1),
+		     SHM_HUGETLB | IPC_CREAT |0600)) < 0) {
+    printf( "Memory allocation failed(shmget).\n");
+    exit(1);
+  }
+
+  address = shmat(shmid, NULL, SHM_RND);
+
+  if ((BLASLONG)address == -1){
+    printf( "Memory allocation failed(shmat).\n");
+    exit(1);
+  }
+
+  shmctl(shmid, IPC_RMID, 0);
+
+  return address;
+}
+
+#define malloc huge_malloc
+
+#endif
+
+int main(int argc, char *argv[]){
+
+  FLOAT *x, *y;
+  FLOAT alpha[2] = { 2.0, 2.0 };
+  blasint m, i;
+  blasint inc_x=1,inc_y=1;
+  int loops = 1;
+  int l;
+  char *p;
+
+  int from =   1;
+  int to   = 200;
+  int step =   1;
+
+  struct timeval start, stop;
+  double time1,timeg;
+
+  argc--;argv++;
+
+  if (argc > 0) { from     = atol(*argv);		argc--; argv++;}
+  if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
+  if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
+
+  if ((p = getenv("OPENBLAS_LOOPS")))  loops = atoi(p);
+  if ((p = getenv("OPENBLAS_INCX")))   inc_x = atoi(p);
+
+  fprintf(stderr, "From : %3d  To : %3d Step = %3d Inc_x = %d Inc_y = %d Loops = %d\n", from, to, step,inc_x,inc_y,loops);
+
+  if (( x = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_x) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+  if (( y = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_y) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+#ifdef linux
+  srandom(getpid());
+#endif
+
+  fprintf(stderr, "   SIZE       Flops\n");
+
+  for(m = from; m <= to; m += step)
+  {
+
+   timeg=0;
+
+   fprintf(stderr, " %6d : ", (int)m);
+
+
+   for (l=0; l<loops; l++)
+   {
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_x); i++){
+			x[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_y); i++){
+			y[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+    	gettimeofday( &start, (struct timezone *)0);
+
+    	SCAL (&m, alpha, x, &inc_x);
+
+    	gettimeofday( &stop, (struct timezone *)0);
+
+    	time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+
+	timeg += time1;
+
+    }
+
+    timeg /= loops;
+
+#ifdef COMPLEX
+    fprintf(stderr, " %10.2f MFlops\n", 6. * (double)m / timeg * 1.e-6);
+#else
+    fprintf(stderr, " %10.2f MFlops\n", 1. * (double)m / timeg * 1.e-6);
+#endif
+
+  }
+
+  return 0;
+}
+
+// void main(int argc, char *argv[]) __attribute__((weak, alias("MAIN__")));
diff --git a/benchmark/scripts/NUMPY/cgemm.py b/benchmark/scripts/NUMPY/cgemm.py
new file mode 100755
index 000000000..b35d3b896
--- /dev/null
+++ b/benchmark/scripts/NUMPY/cgemm.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_cgemm(N,l):
+
+	A = randn(N,N).astype('float32') + randn(N,N).astype('float32') * 1j;
+	B = randn(N,N).astype('float32') + randn(N,N).astype('float32') * 1j;
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 8*N*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_cgemm(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/cgemv.py b/benchmark/scripts/NUMPY/cgemv.py
new file mode 100755
index 000000000..aa0ac9d64
--- /dev/null
+++ b/benchmark/scripts/NUMPY/cgemv.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_cgemv(N,l):
+
+	A = randn(N,N).astype('float32') + randn(N,N).astype('float32') * 1j;
+	B = randn(N).astype('float32') + randn(N).astype('float32') * 1j;
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 8*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_cgemv(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/daxpy.py b/benchmark/scripts/NUMPY/daxpy.py
new file mode 100755
index 000000000..db2e0e607
--- /dev/null
+++ b/benchmark/scripts/NUMPY/daxpy.py
@@ -0,0 +1,58 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+from scipy.linalg.blas import daxpy
+
+
+def run_daxpy(N,l):
+
+	x = randn(N).astype('float64')
+	y = randn(N).astype('float64')
+
+	start = time.time();
+	for i in range(0,l):
+		y = daxpy(x,y, a=2.0 )
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 2*N ) *l / timediff
+	mflops *= 1e-6
+
+	size = "%d" % (N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_daxpy(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/ddot.py b/benchmark/scripts/NUMPY/ddot.py
new file mode 100755
index 000000000..0f4ced339
--- /dev/null
+++ b/benchmark/scripts/NUMPY/ddot.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_ddot(N,l):
+
+	A = randn(N).astype('float64')
+	B = randn(N).astype('float64')
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 2*N ) *l / timediff
+	mflops *= 1e-6
+
+	size = "%d" % (N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_ddot(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/deig.py b/benchmark/scripts/NUMPY/deig.py
new file mode 100755
index 000000000..aac7abe15
--- /dev/null
+++ b/benchmark/scripts/NUMPY/deig.py
@@ -0,0 +1,55 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_deig(N,l):
+
+	A = randn(N,N).astype('float64')
+
+	start = time.time();
+	for i in range(0,l):
+		la,v = numpy.linalg.eig(A)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 26.33 *N*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_deig(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/dgemm.py b/benchmark/scripts/NUMPY/dgemm.py
new file mode 100755
index 000000000..a31248786
--- /dev/null
+++ b/benchmark/scripts/NUMPY/dgemm.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_dgemm(N,l):
+
+	A = randn(N,N).astype('float64')
+	B = randn(N,N).astype('float64')
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 2*N*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_dgemm(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/dgemv.py b/benchmark/scripts/NUMPY/dgemv.py
new file mode 100755
index 000000000..bbc295e8e
--- /dev/null
+++ b/benchmark/scripts/NUMPY/dgemv.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_dgemv(N,l):
+
+	A = randn(N,N).astype('float64')
+	B = randn(N).astype('float64')
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 2*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_dgemv(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/dgesv.py b/benchmark/scripts/NUMPY/dgesv.py
new file mode 100755
index 000000000..8adabd146
--- /dev/null
+++ b/benchmark/scripts/NUMPY/dgesv.py
@@ -0,0 +1,58 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+from scipy.linalg.lapack import dgesv
+
+def run_dgesv(N,l):
+
+	a = randn(N,N).astype('float64')
+	b = randn(N,N).astype('float64')
+
+	start = time.time();
+	for i in range(0,l):
+		dgesv(a,b,1,1)
+	end = time.time()
+	
+	timediff = (end -start) 
+
+	mflops = ( 2.0/3.0 *N*N*N + 2.0*N*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_dgesv(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/dsolve.py b/benchmark/scripts/NUMPY/dsolve.py
new file mode 100755
index 000000000..1b067a84f
--- /dev/null
+++ b/benchmark/scripts/NUMPY/dsolve.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_dsolve(N,l):
+
+	A = randn(N,N).astype('float64')
+	B = randn(N,N).astype('float64')
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.linalg.solve(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 2.0/3.0 *N*N*N + 2.0*N*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_dsolve(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/sdot.py b/benchmark/scripts/NUMPY/sdot.py
new file mode 100755
index 000000000..4fe6b8cc9
--- /dev/null
+++ b/benchmark/scripts/NUMPY/sdot.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_sdot(N,l):
+
+	A = randn(N).astype('float32')
+	B = randn(N).astype('float32')
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 2*N ) *l / timediff
+	mflops *= 1e-6
+
+	size = "%d" % (N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_sdot(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/sgemm.py b/benchmark/scripts/NUMPY/sgemm.py
new file mode 100755
index 000000000..1680ec28d
--- /dev/null
+++ b/benchmark/scripts/NUMPY/sgemm.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_sgemm(N,l):
+
+	A = randn(N,N).astype('float32')
+	B = randn(N,N).astype('float32')
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 2*N*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_sgemm(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/sgemv.py b/benchmark/scripts/NUMPY/sgemv.py
new file mode 100755
index 000000000..3fe6add55
--- /dev/null
+++ b/benchmark/scripts/NUMPY/sgemv.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_sgemv(N,l):
+
+	A = randn(N,N).astype('float32')
+	B = randn(N).astype('float32')
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 2*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_sgemv(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/zgemm.py b/benchmark/scripts/NUMPY/zgemm.py
new file mode 100755
index 000000000..4556d4fd2
--- /dev/null
+++ b/benchmark/scripts/NUMPY/zgemm.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_zgemm(N,l):
+
+	A = randn(N,N).astype('float64') + randn(N,N).astype('float64') * 1j;
+	B = randn(N,N).astype('float64') + randn(N,N).astype('float64') * 1j;
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 8*N*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_zgemm(i,LOOPS)
+
diff --git a/benchmark/scripts/NUMPY/zgemv.py b/benchmark/scripts/NUMPY/zgemv.py
new file mode 100755
index 000000000..ea69a19bb
--- /dev/null
+++ b/benchmark/scripts/NUMPY/zgemv.py
@@ -0,0 +1,56 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import numpy
+from numpy.random import randn
+
+def run_zgemv(N,l):
+
+	A = randn(N,N).astype('float64') + randn(N,N).astype('float64') * 1j;
+	B = randn(N).astype('float64') + randn(N).astype('float64') * 1j;
+
+	start = time.time();
+	for i in range(0,l):
+		ref = numpy.dot(A,B)
+	end = time.time()
+	
+	timediff = (end -start) 
+	mflops = ( 8*N*N) *l / timediff
+	mflops *= 1e-6
+
+	size = "%dx%d" % (N,N)
+	print("%14s :\t%20f MFlops\t%20f sec" % (size,mflops,timediff))
+
+
+if __name__ == "__main__":
+	N=128
+	NMAX=2048
+	NINC=128
+	LOOPS=1
+
+	z=0
+	for arg in sys.argv:
+		if z == 1:
+			N = int(arg)
+		elif z == 2:
+			NMAX = int(arg)
+		elif z == 3:
+			NINC = int(arg)
+		elif z == 4:
+			LOOPS = int(arg)
+
+		z = z + 1
+
+	if 'OPENBLAS_LOOPS' in os.environ:
+		p = os.environ['OPENBLAS_LOOPS']
+		if p:
+			LOOPS = int(p);
+
+	print("From: %d To: %d Step=%d Loops=%d" % (N, NMAX, NINC, LOOPS))
+	print("\tSIZE\t\t\tFlops\t\t\t\t\tTime")
+
+	for i in range (N,NMAX+NINC,NINC):
+		run_zgemv(i,LOOPS)
+
diff --git a/benchmark/scripts/OCTAVE/cgemm.m b/benchmark/scripts/OCTAVE/cgemm.m
new file mode 100755
index 000000000..0e79e71ad
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/cgemm.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = single(rand(n,n)) + single(rand(n,n)) * 1i;
+	B = single(rand(n,n)) + single(rand(n,n)) * 1i;
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		C = A * B;
+		l = l + 1;
+
+	endwhile
+
+	timeg = etime(clock(), start);
+	mflops = ( 4.0 * 2.0*n*n*n *loops ) / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg);
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/cgemv.m b/benchmark/scripts/OCTAVE/cgemv.m
new file mode 100755
index 000000000..7237983b6
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/cgemv.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = single(rand(n,n)) + single(rand(n,n)) * 1i;
+	B = single(rand(n,1)) + single(rand(n,1)) * 1i;
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		C = A * B;
+		l = l + 1;
+
+	endwhile
+
+	timeg = etime(clock(), start);
+	mflops = ( 4.0 * 2.0*n*n *loops ) / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg);
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/deig.m b/benchmark/scripts/OCTAVE/deig.m
new file mode 100755
index 000000000..15c85af39
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/deig.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = double(rand(n,n));
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		[V,lambda] = eig(A);
+		l = l + 1;
+
+	endwhile
+
+
+	timeg = etime(clock(), start);
+	mflops = ( 26.33 *n*n*n ) *loops / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg );
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/dgemm.m b/benchmark/scripts/OCTAVE/dgemm.m
new file mode 100755
index 000000000..da4f127ec
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/dgemm.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = double(rand(n,n));
+	B = double(rand(n,n));
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		C = A * B;
+		l = l + 1;
+
+	endwhile
+
+	timeg = etime(clock(), start);
+	mflops = ( 2.0*n*n*n *loops ) / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg);
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/dgemv.m b/benchmark/scripts/OCTAVE/dgemv.m
new file mode 100755
index 000000000..139b14159
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/dgemv.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = double(rand(n,n));
+	B = double(rand(n,1));
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		C = A * B;
+		l = l + 1;
+
+	endwhile
+
+	timeg = etime(clock(), start);
+	mflops = ( 2.0*n*n *loops ) / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg);
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/dsolve.m b/benchmark/scripts/OCTAVE/dsolve.m
new file mode 100755
index 000000000..fff4b0847
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/dsolve.m
@@ -0,0 +1,59 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = double(rand(n,n));
+	B = double(rand(n,n));
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		x = linsolve(A,B);
+		#x = A / B;
+		l = l + 1;
+
+	endwhile
+
+
+	timeg = etime(clock(), start);
+	#r = norm(A*x - B)/norm(B)
+	mflops = ( 2.0/3.0 *n*n*n + 2.0*n*n*n ) *loops / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg );
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/sgemm.m b/benchmark/scripts/OCTAVE/sgemm.m
new file mode 100755
index 000000000..b79548b72
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/sgemm.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = single(rand(n,n));
+	B = single(rand(n,n));
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		C = A * B;
+		l = l + 1;
+
+	endwhile
+
+	timeg = etime(clock(), start);
+	mflops = ( 2.0*n*n*n *loops ) / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg);
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/sgemv.m b/benchmark/scripts/OCTAVE/sgemv.m
new file mode 100755
index 000000000..13152886e
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/sgemv.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = single(rand(n,n));
+	B = single(rand(n,1));
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		C = A * B;
+		l = l + 1;
+
+	endwhile
+
+	timeg = etime(clock(), start);
+	mflops = ( 2.0*n*n *loops ) / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg);
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/zgemm.m b/benchmark/scripts/OCTAVE/zgemm.m
new file mode 100755
index 000000000..a748437eb
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/zgemm.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = double(rand(n,n)) + double(rand(n,n)) * 1i;
+	B = double(rand(n,n)) + double(rand(n,n)) * 1i;
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		C = A * B;
+		l = l + 1;
+
+	endwhile
+
+	timeg = etime(clock(), start);
+	mflops = ( 4.0 * 2.0*n*n*n *loops ) / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg);
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/OCTAVE/zgemv.m b/benchmark/scripts/OCTAVE/zgemv.m
new file mode 100755
index 000000000..10a0089ec
--- /dev/null
+++ b/benchmark/scripts/OCTAVE/zgemv.m
@@ -0,0 +1,56 @@
+#!/usr/bin/octave --silent 
+
+nfrom = 128 ;
+nto   = 2048;
+nstep = 128;
+loops = 1;
+
+
+arg_list = argv();
+for i = 1:nargin
+
+	switch(i)
+		case 1
+			nfrom = str2num(arg_list{i});
+		case 2
+			nto   = str2num(arg_list{i});
+		case 3
+			nstep = str2num(arg_list{i});
+		case 4
+			loops = str2num(arg_list{i});
+
+	endswitch
+
+endfor
+
+p = getenv("OPENBLAS_LOOPS");
+if p 
+	loops = str2num(p);
+endif
+
+printf("From %d To %d Step=%d Loops=%d\n",nfrom, nto, nstep, loops);
+printf("        SIZE             FLOPS             TIME\n");
+
+n = nfrom;
+while n <= nto
+
+	A = double(rand(n,n)) + double(rand(n,n)) * 1i;
+	B = double(rand(n,1)) + double(rand(n,1)) * 1i;
+	start = clock();
+
+	l=0;
+	while l < loops
+
+		C = A * B;
+		l = l + 1;
+
+	endwhile
+
+	timeg = etime(clock(), start);
+	mflops = ( 4.0 * 2.0*n*n *loops ) / ( timeg * 1.0e6 );
+
+	st1 = sprintf("%dx%d : ", n,n);
+	printf("%20s %10.2f MFlops %10.6f sec\n", st1, mflops, timeg);
+	n = n + nstep;
+
+endwhile
diff --git a/benchmark/scripts/R/deig.R b/benchmark/scripts/R/deig.R
new file mode 100755
index 000000000..3521c7c5a
--- /dev/null
+++ b/benchmark/scripts/R/deig.R
@@ -0,0 +1,62 @@
+#!/usr/bin/Rscript
+
+argv <- commandArgs(trailingOnly = TRUE)
+
+nfrom = 128
+nto = 2048
+nstep = 128
+loops = 1
+
+if ( length(argv) > 0 ) {
+
+	for ( z in 1:length(argv) ) {
+
+		if ( z == 1 ) {
+			nfrom <- as.numeric(argv[z])
+		} else if ( z==2 ) {
+			nto <- as.numeric(argv[z])
+		} else if ( z==3 ) {
+			nstep <- as.numeric(argv[z])
+		} else if ( z==4 ) {
+			loops <- as.numeric(argv[z])
+		}
+	}
+
+}
+
+p=Sys.getenv("OPENBLAS_LOOPS")
+if ( p != "" ) {
+	loops <- as.numeric(p)
+}	
+
+
+cat(sprintf("From %.0f To %.0f Step=%.0f Loops=%.0f\n",nfrom, nto, nstep, loops))
+cat(sprintf("      SIZE             Flops                   Time\n"))
+
+n = nfrom
+while ( n <= nto ) {
+
+	A <- matrix(runif(n*n), ncol = n, nrow = n, byrow = TRUE)
+	
+	l = 1
+
+	start <- proc.time()[3]
+
+	while ( l <= loops ) {
+
+		ev <- eigen(A)
+		l = l + 1
+	}
+
+	end <- proc.time()[3]
+	timeg = end - start
+	mflops = (26.66 *n*n*n ) * loops / ( timeg * 1.0e6 )
+
+	st = sprintf("%.0fx%.0f :",n , n)
+	cat(sprintf("%20s %10.2f MFlops %10.6f sec\n", st, mflops, timeg))
+
+	n = n + nstep
+
+}
+
+
diff --git a/benchmark/scripts/R/dgemm.R b/benchmark/scripts/R/dgemm.R
new file mode 100755
index 000000000..f1c09c38d
--- /dev/null
+++ b/benchmark/scripts/R/dgemm.R
@@ -0,0 +1,63 @@
+#!/usr/bin/Rscript
+
+argv <- commandArgs(trailingOnly = TRUE)
+
+nfrom = 128
+nto = 2048
+nstep = 128
+loops = 1
+
+if ( length(argv) > 0 ) {
+
+	for ( z in 1:length(argv) ) {
+
+		if ( z == 1 ) {
+			nfrom <- as.numeric(argv[z])
+		} else if ( z==2 ) {
+			nto <- as.numeric(argv[z])
+		} else if ( z==3 ) {
+			nstep <- as.numeric(argv[z])
+		} else if ( z==4 ) {
+			loops <- as.numeric(argv[z])
+		}
+	}
+
+}
+
+p=Sys.getenv("OPENBLAS_LOOPS")
+if ( p != "" ) {
+	loops <- as.numeric(p)
+}	
+
+
+cat(sprintf("From %.0f To %.0f Step=%.0f Loops=%.0f\n",nfrom, nto, nstep, loops))
+cat(sprintf("      SIZE             Flops                   Time\n"))
+
+n = nfrom
+while ( n <= nto ) {
+
+	A <- matrix(runif(n*n), ncol = n, nrow = n, byrow = TRUE)
+	B <- matrix(runif(n*n), ncol = n, nrow = n, byrow = TRUE)
+	
+	l = 1
+
+	start <- proc.time()[3]
+
+	while ( l <= loops ) {
+
+		C <- A %*% B
+		l = l + 1
+	}
+
+	end <- proc.time()[3]
+	timeg = end - start
+	mflops = ( 2.0 *n*n*n ) * loops / ( timeg * 1.0e6 )
+
+	st = sprintf("%.0fx%.0f :",n , n)
+	cat(sprintf("%20s %10.2f MFlops %10.6f sec\n", st, mflops, timeg))
+
+	n = n + nstep
+
+}
+
+
diff --git a/benchmark/scripts/R/dsolve.R b/benchmark/scripts/R/dsolve.R
new file mode 100755
index 000000000..6c6b77f70
--- /dev/null
+++ b/benchmark/scripts/R/dsolve.R
@@ -0,0 +1,63 @@
+#!/usr/bin/Rscript
+
+argv <- commandArgs(trailingOnly = TRUE)
+
+nfrom = 128
+nto = 2048
+nstep = 128
+loops = 1
+
+if ( length(argv) > 0 ) {
+
+	for ( z in 1:length(argv) ) {
+
+		if ( z == 1 ) {
+			nfrom <- as.numeric(argv[z])
+		} else if ( z==2 ) {
+			nto <- as.numeric(argv[z])
+		} else if ( z==3 ) {
+			nstep <- as.numeric(argv[z])
+		} else if ( z==4 ) {
+			loops <- as.numeric(argv[z])
+		}
+	}
+
+}
+
+p=Sys.getenv("OPENBLAS_LOOPS")
+if ( p != "" ) {
+	loops <- as.numeric(p)
+}	
+
+
+cat(sprintf("From %.0f To %.0f Step=%.0f Loops=%.0f\n",nfrom, nto, nstep, loops))
+cat(sprintf("      SIZE             Flops                   Time\n"))
+
+n = nfrom
+while ( n <= nto ) {
+
+	A <- matrix(runif(n*n), ncol = n, nrow = n, byrow = TRUE)
+	B <- matrix(runif(n*n), ncol = n, nrow = n, byrow = TRUE)
+	
+	l = 1
+
+	start <- proc.time()[3]
+
+	while ( l <= loops ) {
+
+		solve(A,B)
+		l = l + 1
+	}
+
+	end <- proc.time()[3]
+	timeg = end - start
+	mflops = (2.0/3.0 *n*n*n + 2.0 *n*n*n ) * loops / ( timeg * 1.0e6 )
+
+	st = sprintf("%.0fx%.0f :",n , n)
+	cat(sprintf("%20s %10.2f MFlops %10.6f sec\n", st, mflops, timeg))
+
+	n = n + nstep
+
+}
+
+
diff --git a/benchmark/swap.c b/benchmark/swap.c
new file mode 100644
index 000000000..9f108ef50
--- /dev/null
+++ b/benchmark/swap.c
@@ -0,0 +1,201 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above swapright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above swapright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE SWAPRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include <stdio.h>
+#include <stdlib.h>
+#ifdef __CYGWIN32__
+#include <sys/time.h>
+#endif
+#include "common.h"
+
+
+#undef SWAP
+
+#ifdef COMPLEX
+#ifdef DOUBLE
+#define SWAP   BLASFUNC(zswap)
+#else
+#define SWAP   BLASFUNC(cswap)
+#endif
+#else
+#ifdef DOUBLE
+#define SWAP   BLASFUNC(dswap)
+#else
+#define SWAP   BLASFUNC(sswap)
+#endif
+#endif
+
+#if defined(__WIN32__) || defined(__WIN64__)
+
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
+int gettimeofday(struct timeval *tv, void *tz){
+
+  FILETIME ft;
+  unsigned __int64 tmpres = 0;
+  static int tzflag;
+
+  if (NULL != tv)
+    {
+      GetSystemTimeAsFileTime(&ft);
+
+      tmpres |= ft.dwHighDateTime;
+      tmpres <<= 32;
+      tmpres |= ft.dwLowDateTime;
+
+      /*converting file time to unix epoch*/
+      tmpres /= 10;  /*convert into microseconds*/
+      tmpres -= DELTA_EPOCH_IN_MICROSECS;
+      tv->tv_sec = (long)(tmpres / 1000000UL);
+      tv->tv_usec = (long)(tmpres % 1000000UL);
+    }
+
+  return 0;
+}
+
+#endif
+
+#if !defined(__WIN32__) && !defined(__WIN64__) && !defined(__CYGWIN32__) && 0
+
+static void *huge_malloc(BLASLONG size){
+  int shmid;
+  void *address;
+
+#ifndef SHM_HUGETLB
+#define SHM_HUGETLB 04000
+#endif
+
+  if ((shmid =shmget(IPC_PRIVATE,
+		     (size + HUGE_PAGESIZE) & ~(HUGE_PAGESIZE - 1),
+		     SHM_HUGETLB | IPC_CREAT |0600)) < 0) {
+    printf( "Memory allocation failed(shmget).\n");
+    exit(1);
+  }
+
+  address = shmat(shmid, NULL, SHM_RND);
+
+  if ((BLASLONG)address == -1){
+    printf( "Memory allocation failed(shmat).\n");
+    exit(1);
+  }
+
+  shmctl(shmid, IPC_RMID, 0);
+
+  return address;
+}
+
+#define malloc huge_malloc
+
+#endif
+
+int main(int argc, char *argv[]){
+
+  FLOAT *x, *y;
+  FLOAT alpha[2] = { 2.0, 2.0 };
+  blasint m, i;
+  blasint inc_x=1,inc_y=1;
+  int loops = 1;
+  int l;
+  char *p;
+
+  int from =   1;
+  int to   = 200;
+  int step =   1;
+
+  struct timeval start, stop;
+  double time1,timeg;
+
+  argc--;argv++;
+
+  if (argc > 0) { from     = atol(*argv);		argc--; argv++;}
+  if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
+  if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
+
+  if ((p = getenv("OPENBLAS_LOOPS")))  loops = atoi(p);
+  if ((p = getenv("OPENBLAS_INCX")))   inc_x = atoi(p);
+  if ((p = getenv("OPENBLAS_INCY")))   inc_y = atoi(p);
+
+  fprintf(stderr, "From : %3d  To : %3d Step = %3d Inc_x = %d Inc_y = %d Loops = %d\n", from, to, step,inc_x,inc_y,loops);
+
+  if (( x = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_x) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+  if (( y = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_y) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+#ifdef linux
+  srandom(getpid());
+#endif
+
+  fprintf(stderr, "   SIZE       Flops\n");
+
+  for(m = from; m <= to; m += step)
+  {
+
+   timeg=0;
+
+   fprintf(stderr, " %6d : ", (int)m);
+
+
+   for (l=0; l<loops; l++)
+   {
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_x); i++){
+			x[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_y); i++){
+			y[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+    	gettimeofday( &start, (struct timezone *)0);
+
+    	SWAP (&m, x, &inc_x, y, &inc_y );
+
+    	gettimeofday( &stop, (struct timezone *)0);
+
+    	time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+
+	timeg += time1;
+
+    }
+
+    timeg /= loops;
+
+    fprintf(stderr,
+	    " %10.2f MBytes\n",
+	    COMPSIZE * sizeof(FLOAT) * 1. * (double)m / timeg * 1.e-6);
+
+  }
+
+  return 0;
+}
+
+// void main(int argc, char *argv[]) __attribute__((weak, alias("MAIN__")));
diff --git a/benchmark/trsm.c b/benchmark/trsm.c
index ed969b707..9eae3380c 100644
--- a/benchmark/trsm.c
+++ b/benchmark/trsm.c
@@ -130,11 +130,21 @@ int main(int argc, char *argv[]){
   char trans='N';
   char diag ='U';
 
+
+  int l;
+  int loops = 1;
+  double timeg;
+
   if ((p = getenv("OPENBLAS_SIDE"))) side=*p; 
   if ((p = getenv("OPENBLAS_UPLO"))) uplo=*p;
   if ((p = getenv("OPENBLAS_TRANS"))) trans=*p;
   if ((p = getenv("OPENBLAS_DIAG"))) diag=*p;
 
+  p = getenv("OPENBLAS_LOOPS");
+  if ( p != NULL )
+        loops = atoi(p);
+
+
   blasint m, i, j;
 
   int from =   1;
@@ -150,7 +160,7 @@ int main(int argc, char *argv[]){
   if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
   if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
 
-  fprintf(stderr, "From : %3d  To : %3d Step = %3d Side = %c Uplo = %c Trans = %c Diag = %c\n", from, to, step,side,uplo,trans,diag);
+  fprintf(stderr, "From : %3d  To : %3d Step = %3d Side = %c Uplo = %c Trans = %c Diag = %c Loops = %d\n", from, to, step,side,uplo,trans,diag,loops);
 
   if (( a = (FLOAT *)malloc(sizeof(FLOAT) * to * to * COMPSIZE)) == NULL){
     fprintf(stderr,"Out of Memory!!\n");exit(1);
@@ -171,28 +181,35 @@ int main(int argc, char *argv[]){
   for(m = from; m <= to; m += step)
   {
 
-    fprintf(stderr, " %6d : ", (int)m);
+	timeg=0.0;
 
-    for(j = 0; j < m; j++){
-      for(i = 0; i < m * COMPSIZE; i++){
-	a[i + j * m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
-	b[i + j * m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
-      }
-    }
+        fprintf(stderr, " %6d : ", (int)m);
 
-    gettimeofday( &start, (struct timezone *)0);
+	for (l=0; l<loops; l++)
+    	{
 
-    TRSM (&side, &uplo, &trans, &diag, &m, &m, alpha, a, &m, b, &m);
 
-    gettimeofday( &stop, (struct timezone *)0);
+   		 for(j = 0; j < m; j++){
+      			for(i = 0; i < m * COMPSIZE; i++){
+				a[i + j * m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+				b[i + j * m * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+      		 	}
+    		 }
 
-    time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+    		gettimeofday( &start, (struct timezone *)0);
 
-    gettimeofday( &start, (struct timezone *)0);
+    		TRSM (&side, &uplo, &trans, &diag, &m, &m, alpha, a, &m, b, &m);
 
-    fprintf(stderr,
-	    " %10.2f MFlops\n",
-	    COMPSIZE * COMPSIZE * 1. * (double)m * (double)m * (double)m / time1 * 1.e-6);
+    		gettimeofday( &stop, (struct timezone *)0);
+
+    		time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+
+		timeg += time1;
+        }
+
+	time1 = timeg/loops;
+
+        fprintf(stderr, " %10.2f MFlops\n", COMPSIZE * COMPSIZE * 1. * (double)m * (double)m * (double)m / time1 * 1.e-6);
 
   }
 
diff --git a/benchmark/zdot-intel.c b/benchmark/zdot-intel.c
new file mode 100644
index 000000000..bb2c40f38
--- /dev/null
+++ b/benchmark/zdot-intel.c
@@ -0,0 +1,196 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include <stdio.h>
+#include <stdlib.h>
+#ifdef __CYGWIN32__
+#include <sys/time.h>
+#endif
+#define RETURN_BY_STACK 1
+#include "common.h"
+
+
+#undef DOT
+
+
+#ifdef DOUBLE
+#define DOT   BLASFUNC(zdotu)
+#else
+#define DOT   BLASFUNC(cdotu)
+#endif
+
+
+#if defined(__WIN32__) || defined(__WIN64__)
+
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
+int gettimeofday(struct timeval *tv, void *tz){
+
+  FILETIME ft;
+  unsigned __int64 tmpres = 0;
+  static int tzflag;
+
+  if (NULL != tv)
+    {
+      GetSystemTimeAsFileTime(&ft);
+
+      tmpres |= ft.dwHighDateTime;
+      tmpres <<= 32;
+      tmpres |= ft.dwLowDateTime;
+
+      /*converting file time to unix epoch*/
+      tmpres /= 10;  /*convert into microseconds*/
+      tmpres -= DELTA_EPOCH_IN_MICROSECS;
+      tv->tv_sec = (long)(tmpres / 1000000UL);
+      tv->tv_usec = (long)(tmpres % 1000000UL);
+    }
+
+  return 0;
+}
+
+#endif
+
+#if !defined(__WIN32__) && !defined(__WIN64__) && !defined(__CYGWIN32__) && 0
+
+static void *huge_malloc(BLASLONG size){
+  int shmid;
+  void *address;
+
+#ifndef SHM_HUGETLB
+#define SHM_HUGETLB 04000
+#endif
+
+  if ((shmid =shmget(IPC_PRIVATE,
+		     (size + HUGE_PAGESIZE) & ~(HUGE_PAGESIZE - 1),
+		     SHM_HUGETLB | IPC_CREAT |0600)) < 0) {
+    printf( "Memory allocation failed(shmget).\n");
+    exit(1);
+  }
+
+  address = shmat(shmid, NULL, SHM_RND);
+
+  if ((BLASLONG)address == -1){
+    printf( "Memory allocation failed(shmat).\n");
+    exit(1);
+  }
+
+  shmctl(shmid, IPC_RMID, 0);
+
+  return address;
+}
+
+#define malloc huge_malloc
+
+#endif
+
+int main(int argc, char *argv[]){
+
+  FLOAT *x, *y;
+  FLOAT _Complex result;
+  blasint m, i;
+  blasint inc_x=1,inc_y=1;
+  int loops = 1;
+  int l;
+  char *p;
+
+  int from =   1;
+  int to   = 200;
+  int step =   1;
+
+  struct timeval start, stop;
+  double time1,timeg;
+
+  argc--;argv++;
+
+  if (argc > 0) { from     = atol(*argv);		argc--; argv++;}
+  if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
+  if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
+
+  if ((p = getenv("OPENBLAS_LOOPS")))  loops = atoi(p);
+  if ((p = getenv("OPENBLAS_INCX")))   inc_x = atoi(p);
+  if ((p = getenv("OPENBLAS_INCY")))   inc_y = atoi(p);
+
+  fprintf(stderr, "From : %3d  To : %3d Step = %3d Inc_x = %d Inc_y = %d Loops = %d\n", from, to, step,inc_x,inc_y,loops);
+
+  if (( x = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_x) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+  if (( y = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_y) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+#ifdef linux
+  srandom(getpid());
+#endif
+
+  fprintf(stderr, "   SIZE       Flops\n");
+
+  for(m = from; m <= to; m += step)
+  {
+
+   timeg=0;
+
+   fprintf(stderr, " %6d : ", (int)m);
+
+
+   for (l=0; l<loops; l++)
+   {
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_x); i++){
+			x[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_y); i++){
+			y[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+    	gettimeofday( &start, (struct timezone *)0);
+
+    	DOT (&result, &m, x, &inc_x, y, &inc_y );
+
+    	gettimeofday( &stop, (struct timezone *)0);
+
+    	time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+
+	timeg += time1;
+
+    }
+
+    timeg /= loops;
+
+    fprintf(stderr,
+	    " %10.2f MFlops\n",
+	    COMPSIZE * COMPSIZE * 2. * (double)m / timeg * 1.e-6);
+
+  }
+
+  return 0;
+}
+
+// void main(int argc, char *argv[]) __attribute__((weak, alias("MAIN__")));
diff --git a/benchmark/zdot.c b/benchmark/zdot.c
new file mode 100644
index 000000000..d5ec99726
--- /dev/null
+++ b/benchmark/zdot.c
@@ -0,0 +1,195 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include <stdio.h>
+#include <stdlib.h>
+#ifdef __CYGWIN32__
+#include <sys/time.h>
+#endif
+#include "common.h"
+
+
+#undef DOT
+
+
+#ifdef DOUBLE
+#define DOT   BLASFUNC(zdotu)
+#else
+#define DOT   BLASFUNC(cdotu)
+#endif
+
+
+#if defined(__WIN32__) || defined(__WIN64__)
+
+#ifndef DELTA_EPOCH_IN_MICROSECS
+#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
+#endif
+
+int gettimeofday(struct timeval *tv, void *tz){
+
+  FILETIME ft;
+  unsigned __int64 tmpres = 0;
+  static int tzflag;
+
+  if (NULL != tv)
+    {
+      GetSystemTimeAsFileTime(&ft);
+
+      tmpres |= ft.dwHighDateTime;
+      tmpres <<= 32;
+      tmpres |= ft.dwLowDateTime;
+
+      /*converting file time to unix epoch*/
+      tmpres /= 10;  /*convert into microseconds*/
+      tmpres -= DELTA_EPOCH_IN_MICROSECS;
+      tv->tv_sec = (long)(tmpres / 1000000UL);
+      tv->tv_usec = (long)(tmpres % 1000000UL);
+    }
+
+  return 0;
+}
+
+#endif
+
+#if !defined(__WIN32__) && !defined(__WIN64__) && !defined(__CYGWIN32__) && 0
+
+static void *huge_malloc(BLASLONG size){
+  int shmid;
+  void *address;
+
+#ifndef SHM_HUGETLB
+#define SHM_HUGETLB 04000
+#endif
+
+  if ((shmid =shmget(IPC_PRIVATE,
+		     (size + HUGE_PAGESIZE) & ~(HUGE_PAGESIZE - 1),
+		     SHM_HUGETLB | IPC_CREAT |0600)) < 0) {
+    printf( "Memory allocation failed(shmget).\n");
+    exit(1);
+  }
+
+  address = shmat(shmid, NULL, SHM_RND);
+
+  if ((BLASLONG)address == -1){
+    printf( "Memory allocation failed(shmat).\n");
+    exit(1);
+  }
+
+  shmctl(shmid, IPC_RMID, 0);
+
+  return address;
+}
+
+#define malloc huge_malloc
+
+#endif
+
+int main(int argc, char *argv[]){
+
+  FLOAT *x, *y;
+  FLOAT _Complex result;
+  blasint m, i;
+  blasint inc_x=1,inc_y=1;
+  int loops = 1;
+  int l;
+  char *p;
+
+  int from =   1;
+  int to   = 200;
+  int step =   1;
+
+  struct timeval start, stop;
+  double time1,timeg;
+
+  argc--;argv++;
+
+  if (argc > 0) { from     = atol(*argv);		argc--; argv++;}
+  if (argc > 0) { to       = MAX(atol(*argv), from);	argc--; argv++;}
+  if (argc > 0) { step     = atol(*argv);		argc--; argv++;}
+
+  if ((p = getenv("OPENBLAS_LOOPS")))  loops = atoi(p);
+  if ((p = getenv("OPENBLAS_INCX")))   inc_x = atoi(p);
+  if ((p = getenv("OPENBLAS_INCY")))   inc_y = atoi(p);
+
+  fprintf(stderr, "From : %3d  To : %3d Step = %3d Inc_x = %d Inc_y = %d Loops = %d\n", from, to, step,inc_x,inc_y,loops);
+
+  if (( x = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_x) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+  if (( y = (FLOAT *)malloc(sizeof(FLOAT) * to * abs(inc_y) * COMPSIZE)) == NULL){
+    fprintf(stderr,"Out of Memory!!\n");exit(1);
+  }
+
+#ifdef linux
+  srandom(getpid());
+#endif
+
+  fprintf(stderr, "   SIZE       Flops\n");
+
+  for(m = from; m <= to; m += step)
+  {
+
+   timeg=0;
+
+   fprintf(stderr, " %6d : ", (int)m);
+
+
+   for (l=0; l<loops; l++)
+   {
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_x); i++){
+			x[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+
+   	for(i = 0; i < m * COMPSIZE * abs(inc_y); i++){
+			y[i] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
+   	}
+    	gettimeofday( &start, (struct timezone *)0);
+
+    	result = DOT (&m, x, &inc_x, y, &inc_y );
+
+    	gettimeofday( &stop, (struct timezone *)0);
+
+    	time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
+
+	timeg += time1;
+
+    }
+
+    timeg /= loops;
+
+    fprintf(stderr,
+	    " %10.2f MFlops\n",
+	    COMPSIZE * COMPSIZE * 2. * (double)m / timeg * 1.e-6);
+
+  }
+
+  return 0;
+}
+
+// void main(int argc, char *argv[]) __attribute__((weak, alias("MAIN__")));
diff --git a/c_check b/c_check
index fbe9c9fab..d694e7411 100644
--- a/c_check
+++ b/c_check
@@ -4,6 +4,8 @@
 $hostos   = `uname -s | sed -e s/\-.*//`;    chop($hostos);
 $hostarch = `uname -m | sed -e s/i.86/x86/`;chop($hostarch);
 $hostarch = "x86_64" if ($hostarch eq "amd64");
+$hostarch = "arm" if ($hostarch =~ /^arm.*/);
+$hostarch = "arm64" if ($hostarch eq "aarch64");
 
 $binary = $ENV{"BINARY"};
 
@@ -28,7 +30,7 @@ if ($ARGV[0] =~ /(.*)(-[.\d]+)/) {
 	$cross_suffix = $1;
     }
 } else {
-    if ($ARGV[0] =~ /(.*-)(.*)/) {
+    if ($ARGV[0] =~ /([^\/]*-)([^\/]*$)/) {
 	$cross_suffix = $1;
     }
 }
@@ -55,6 +57,7 @@ $os = osf             if ($data =~ /OS_OSF/);
 $os = WINNT           if ($data =~ /OS_WINNT/);
 $os = CYGWIN_NT       if ($data =~ /OS_CYGWIN_NT/);
 $os = Interix         if ($data =~ /OS_INTERIX/);
+$os = Android         if ($data =~ /OS_ANDROID/);
 
 $architecture = x86    if ($data =~ /ARCH_X86/);
 $architecture = x86_64 if ($data =~ /ARCH_X86_64/);
diff --git a/cblas_noconst.h b/cblas_noconst.h
deleted file mode 100644
index 4451c304e..000000000
--- a/cblas_noconst.h
+++ /dev/null
@@ -1,350 +0,0 @@
-#ifndef CBLAS_H
-#define CBLAS_H
-
-#include <stddef.h>
-#include "common.h"
-
-#ifdef __cplusplus
-extern "C" {
-	/* Assume C declarations for C++ */
-#endif  /* __cplusplus */
-
-/*Set the number of threads on runtime.*/
-void openblas_set_num_threads(int num_threads);
-void goto_set_num_threads(int num_threads);
-
-/*Get the number of threads on runtime.*/
-int openblas_get_num_threads(void);
-
-/*Get the number of physical processors (cores).*/
-int openblas_get_num_procs(void);
-
-/*Get the build configure on runtime.*/
-char* openblas_get_config(void);
-
-/* Get the parallelization type which is used by OpenBLAS */
-int openblas_get_parallel(void);
-/* OpenBLAS is compiled for sequential use  */
-#define OPENBLAS_SEQUENTIAL  0
-/* OpenBLAS is compiled using normal threading model */
-#define OPENBLAS_THREAD  1
-/* OpenBLAS is compiled using OpenMP threading model */
-#define OPENBLAS_OPENMP 2
-
-
-#define CBLAS_INDEX size_t
-
-typedef enum CBLAS_ORDER     {CblasRowMajor=101, CblasColMajor=102} CBLAS_ORDER;
-typedef enum CBLAS_TRANSPOSE {CblasNoTrans=111, CblasTrans=112, CblasConjTrans=113, CblasConjNoTrans=114} CBLAS_TRANSPOSE;
-typedef enum CBLAS_UPLO      {CblasUpper=121, CblasLower=122} CBLAS_UPLO;
-typedef enum CBLAS_DIAG      {CblasNonUnit=131, CblasUnit=132} CBLAS_DIAG;
-typedef enum CBLAS_SIDE      {CblasLeft=141, CblasRight=142} CBLAS_SIDE;
-
-float  cblas_sdsdot(blasint n, float alpha, float *x, blasint incx, float *y, blasint incy);
-double cblas_dsdot (blasint n, float *x, blasint incx, float *y, blasint incy);
-float  cblas_sdot(blasint n, float  *x, blasint incx, float  *y, blasint incy);
-double cblas_ddot(blasint n, double *x, blasint incx, double *y, blasint incy);
-
-openblas_complex_float  cblas_cdotu(blasint n, float  *x, blasint incx, float  *y, blasint incy);
-openblas_complex_float  cblas_cdotc(blasint n, float  *x, blasint incx, float  *y, blasint incy);
-openblas_complex_double cblas_zdotu(blasint n, double *x, blasint incx, double *y, blasint incy);
-openblas_complex_double cblas_zdotc(blasint n, double *x, blasint incx, double *y, blasint incy);
-
-void  cblas_cdotu_sub(blasint n, float  *x, blasint incx, float  *y, blasint incy, openblas_complex_float  *ret);
-void  cblas_cdotc_sub(blasint n, float  *x, blasint incx, float  *y, blasint incy, openblas_complex_float  *ret);
-void  cblas_zdotu_sub(blasint n, double *x, blasint incx, double *y, blasint incy, openblas_complex_double *ret);
-void  cblas_zdotc_sub(blasint n, double *x, blasint incx, double *y, blasint incy, openblas_complex_double *ret);
-
-float  cblas_sasum (blasint n, float  *x, blasint incx);
-double cblas_dasum (blasint n, double *x, blasint incx);
-float  cblas_scasum(blasint n, float  *x, blasint incx);
-double cblas_dzasum(blasint n, double *x, blasint incx);
-
-float  cblas_snrm2 (blasint N, float  *X, blasint incX);
-double cblas_dnrm2 (blasint N, double *X, blasint incX);
-float  cblas_scnrm2(blasint N, float  *X, blasint incX);
-double cblas_dznrm2(blasint N, double *X, blasint incX);
-
-CBLAS_INDEX cblas_isamax(blasint n, float  *x, blasint incx);
-CBLAS_INDEX cblas_idamax(blasint n, double *x, blasint incx);
-CBLAS_INDEX cblas_icamax(blasint n, float  *x, blasint incx);
-CBLAS_INDEX cblas_izamax(blasint n, double *x, blasint incx);
-
-void cblas_saxpy(blasint n, float alpha, float *x, blasint incx, float *y, blasint incy);
-void cblas_daxpy(blasint n, double alpha, double *x, blasint incx, double *y, blasint incy);
-void cblas_caxpy(blasint n, float *alpha, float *x, blasint incx, float *y, blasint incy);
-void cblas_zaxpy(blasint n, double *alpha, double *x, blasint incx, double *y, blasint incy);
-
-void cblas_scopy(blasint n, float *x, blasint incx, float *y, blasint incy);
-void cblas_dcopy(blasint n, double *x, blasint incx, double *y, blasint incy);
-void cblas_ccopy(blasint n, float *x, blasint incx, float *y, blasint incy);
-void cblas_zcopy(blasint n, double *x, blasint incx, double *y, blasint incy);
-
-void cblas_sswap(blasint n, float *x, blasint incx, float *y, blasint incy);
-void cblas_dswap(blasint n, double *x, blasint incx, double *y, blasint incy);
-void cblas_cswap(blasint n, float *x, blasint incx, float *y, blasint incy);
-void cblas_zswap(blasint n, double *x, blasint incx, double *y, blasint incy);
-
-void cblas_srot(blasint N, float *X, blasint incX, float *Y, blasint incY, float c, float s);
-void cblas_drot(blasint N, double *X, blasint incX, double *Y, blasint incY, double c, double  s);
-
-void cblas_srotg(float *a, float *b, float *c, float *s);
-void cblas_drotg(double *a, double *b, double *c, double *s);
-
-void cblas_srotm(blasint N, float *X, blasint incX, float *Y, blasint incY, float *P);
-void cblas_drotm(blasint N, double *X, blasint incX, double *Y, blasint incY, double *P);
-
-void cblas_srotmg(float *d1, float *d2, float *b1, float b2, float *P);
-void cblas_drotmg(double *d1, double *d2, double *b1, double b2, double *P);
-
-void cblas_sscal(blasint N, float alpha, float *X, blasint incX);
-void cblas_dscal(blasint N, double alpha, double *X, blasint incX);
-void cblas_cscal(blasint N, float *alpha, float *X, blasint incX);
-void cblas_zscal(blasint N, double *alpha, double *X, blasint incX);
-void cblas_csscal(blasint N, float alpha, float *X, blasint incX);
-void cblas_zdscal(blasint N, double alpha, double *X, blasint incX);
-
-void cblas_sgemv(enum CBLAS_ORDER order,  enum CBLAS_TRANSPOSE trans,  blasint m, blasint n,
-		 float alpha, float  *a, blasint lda,  float  *x, blasint incx,  float beta,  float  *y, blasint incy);
-void cblas_dgemv(enum CBLAS_ORDER order,  enum CBLAS_TRANSPOSE trans,  blasint m, blasint n,
-		 double alpha, double  *a, blasint lda,  double  *x, blasint incx,  double beta,  double  *y, blasint incy);
-void cblas_cgemv(enum CBLAS_ORDER order,  enum CBLAS_TRANSPOSE trans,  blasint m, blasint n,
-		 float *alpha, float  *a, blasint lda,  float  *x, blasint incx,  float *beta,  float  *y, blasint incy);
-void cblas_zgemv(enum CBLAS_ORDER order,  enum CBLAS_TRANSPOSE trans,  blasint m, blasint n,
-		 double *alpha, double  *a, blasint lda,  double  *x, blasint incx,  double *beta,  double  *y, blasint incy);
-
-void cblas_sger (enum CBLAS_ORDER order, blasint M, blasint N, float   alpha, float  *X, blasint incX, float  *Y, blasint incY, float  *A, blasint lda);
-void cblas_dger (enum CBLAS_ORDER order, blasint M, blasint N, double  alpha, double *X, blasint incX, double *Y, blasint incY, double *A, blasint lda);
-void cblas_cgeru(enum CBLAS_ORDER order, blasint M, blasint N, float  *alpha, float  *X, blasint incX, float  *Y, blasint incY, float  *A, blasint lda);
-void cblas_cgerc(enum CBLAS_ORDER order, blasint M, blasint N, float  *alpha, float  *X, blasint incX, float  *Y, blasint incY, float  *A, blasint lda);
-void cblas_zgeru(enum CBLAS_ORDER order, blasint M, blasint N, double *alpha, double *X, blasint incX, double *Y, blasint incY, double *A, blasint lda);
-void cblas_zgerc(enum CBLAS_ORDER order, blasint M, blasint N, double *alpha, double *X, blasint incX, double *Y, blasint incY, double *A, blasint lda);
-
-void cblas_strsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, float *A, blasint lda, float *X, blasint incX);
-void cblas_dtrsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, double *A, blasint lda, double *X, blasint incX);
-void cblas_ctrsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, float *A, blasint lda, float *X, blasint incX);
-void cblas_ztrsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, double *A, blasint lda, double *X, blasint incX);
-
-void cblas_strmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, float *A, blasint lda, float *X, blasint incX);
-void cblas_dtrmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, double *A, blasint lda, double *X, blasint incX);
-void cblas_ctrmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, float *A, blasint lda, float *X, blasint incX);
-void cblas_ztrmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag, blasint N, double *A, blasint lda, double *X, blasint incX);
-
-void cblas_ssyr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *A, blasint lda);
-void cblas_dsyr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X, blasint incX, double *A, blasint lda);
-void cblas_cher(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *A, blasint lda);
-void cblas_zher(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X, blasint incX, double *A, blasint lda);
-
-void cblas_ssyr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo,blasint N, float alpha, float *X,
-                blasint incX, float *Y, blasint incY, float *A, blasint lda);
-void cblas_dsyr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X,
-                blasint incX, double *Y, blasint incY, double *A, blasint lda);
-void cblas_cher2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float *alpha, float *X, blasint incX,
-                float *Y, blasint incY, float *A, blasint lda);
-void cblas_zher2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double *alpha, double *X, blasint incX,
-                double *Y, blasint incY, double *A, blasint lda);
-
-void cblas_sgbmv(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, blasint M, blasint N,
-                 blasint KL, blasint KU, float alpha, float *A, blasint lda, float *X, blasint incX, float beta, float *Y, blasint incY);
-void cblas_dgbmv(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, blasint M, blasint N,
-                 blasint KL, blasint KU, double alpha, double *A, blasint lda, double *X, blasint incX, double beta, double *Y, blasint incY);
-void cblas_cgbmv(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, blasint M, blasint N,
-                 blasint KL, blasint KU, float *alpha, float *A, blasint lda, float *X, blasint incX, float *beta, float *Y, blasint incY);
-void cblas_zgbmv(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, blasint M, blasint N,
-                 blasint KL, blasint KU, double *alpha, double *A, blasint lda, double *X, blasint incX, double *beta, double *Y, blasint incY);
-
-void cblas_ssbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, blasint K, float alpha, float *A,
-                 blasint lda, float *X, blasint incX, float beta, float *Y, blasint incY);
-void cblas_dsbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, blasint K, double alpha, double *A,
-                 blasint lda, double *X, blasint incX, double beta, double *Y, blasint incY);
-
-
-void cblas_stbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, float *A, blasint lda, float *X, blasint incX);
-void cblas_dtbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, double *A, blasint lda, double *X, blasint incX);
-void cblas_ctbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, float *A, blasint lda, float *X, blasint incX);
-void cblas_ztbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, double *A, blasint lda, double *X, blasint incX);
-
-void cblas_stbsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, float *A, blasint lda, float *X, blasint incX);
-void cblas_dtbsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, double *A, blasint lda, double *X, blasint incX);
-void cblas_ctbsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, float *A, blasint lda, float *X, blasint incX);
-void cblas_ztbsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, blasint K, double *A, blasint lda, double *X, blasint incX);
-
-void cblas_stpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, float *Ap, float *X, blasint incX);
-void cblas_dtpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, double *Ap, double *X, blasint incX);
-void cblas_ctpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, float *Ap, float *X, blasint incX);
-void cblas_ztpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, double *Ap, double *X, blasint incX);
-
-void cblas_stpsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, float *Ap, float *X, blasint incX);
-void cblas_dtpsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, double *Ap, double *X, blasint incX);
-void cblas_ctpsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, float *Ap, float *X, blasint incX);
-void cblas_ztpsv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA, enum CBLAS_DIAG Diag,
-                 blasint N, double *Ap, double *X, blasint incX);
-
-void cblas_ssymv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *A,
-                 blasint lda, float *X, blasint incX, float beta, float *Y, blasint incY);
-void cblas_dsymv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *A,
-                 blasint lda, double *X, blasint incX, double beta, double *Y, blasint incY);
-void cblas_chemv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float *alpha, float *A,
-                 blasint lda, float *X, blasint incX, float *beta, float *Y, blasint incY);
-void cblas_zhemv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double *alpha, double *A,
-                 blasint lda, double *X, blasint incX, double *beta, double *Y, blasint incY);
-
-
-void cblas_sspmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *Ap,
-                 float *X, blasint incX, float beta, float *Y, blasint incY);
-void cblas_dspmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *Ap,
-                 double *X, blasint incX, double beta, double *Y, blasint incY);
-
-void cblas_sspr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *Ap);
-void cblas_dspr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X, blasint incX, double *Ap);
-
-void cblas_chpr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *A);
-void cblas_zhpr(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X,blasint incX, double *A);
-
-void cblas_sspr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float alpha, float *X, blasint incX, float *Y, blasint incY, float *A);
-void cblas_dspr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double alpha, double *X, blasint incX, double *Y, blasint incY, double *A);
-void cblas_chpr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, float *alpha, float *X, blasint incX, float *Y, blasint incY, float *Ap);
-void cblas_zhpr2(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, double *alpha, double *X, blasint incX, double *Y, blasint incY, double *Ap);
-
-void cblas_chbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, blasint K,
-		 float *alpha, float *A, blasint lda, float *X, blasint incX, float *beta, float *Y, blasint incY);
-void cblas_zhbmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N, blasint K,
-		 double *alpha, double *A, blasint lda, double *X, blasint incX, double *beta, double *Y, blasint incY);
-
-void cblas_chpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N,
-		 float *alpha, float *Ap, float *X, blasint incX, float *beta, float *Y, blasint incY);
-void cblas_zhpmv(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, blasint N,
-		 double *alpha, double *Ap, double *X, blasint incX, double *beta, double *Y, blasint incY);
-
-void cblas_sgemm(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 float alpha, float *A, blasint lda, float *B, blasint ldb, float beta, float *C, blasint ldc);
-void cblas_dgemm(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 double alpha, double *A, blasint lda, double *B, blasint ldb, double beta, double *C, blasint ldc);
-void cblas_cgemm(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_cgemm3m(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_zgemm(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-void cblas_zgemm3m(enum CBLAS_ORDER Order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANSPOSE TransB, blasint M, blasint N, blasint K,
-		 double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-
-void cblas_ssymm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 float alpha, float *A, blasint lda, float *B, blasint ldb, float beta, float *C, blasint ldc);
-void cblas_dsymm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 double alpha, double *A, blasint lda, double *B, blasint ldb, double beta, double *C, blasint ldc);
-void cblas_csymm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_zsymm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-
-void cblas_ssyrk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		 blasint N, blasint K, float alpha, float *A, blasint lda, float beta, float *C, blasint ldc);
-void cblas_dsyrk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		 blasint N, blasint K, double alpha, double *A, blasint lda, double beta, double *C, blasint ldc);
-void cblas_csyrk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		 blasint N, blasint K, float *alpha, float *A, blasint lda, float *beta, float *C, blasint ldc);
-void cblas_zsyrk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		 blasint N, blasint K, double *alpha, double *A, blasint lda, double *beta, double *C, blasint ldc);
-
-void cblas_ssyr2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		  blasint N, blasint K, float alpha, float *A, blasint lda, float *B, blasint ldb, float beta, float *C, blasint ldc);
-void cblas_dsyr2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		  blasint N, blasint K, double alpha, double *A, blasint lda, double *B, blasint ldb, double beta, double *C, blasint ldc);
-void cblas_csyr2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		  blasint N, blasint K, float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_zsyr2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans,
-		  blasint N, blasint K, double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-
-void cblas_strmm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, float alpha, float *A, blasint lda, float *B, blasint ldb);
-void cblas_dtrmm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, double alpha, double *A, blasint lda, double *B, blasint ldb);
-void cblas_ctrmm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, float *alpha, float *A, blasint lda, float *B, blasint ldb);
-void cblas_ztrmm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, double *alpha, double *A, blasint lda, double *B, blasint ldb);
-
-void cblas_strsm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, float alpha, float *A, blasint lda, float *B, blasint ldb);
-void cblas_dtrsm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, double alpha, double *A, blasint lda, double *B, blasint ldb);
-void cblas_ctrsm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, float *alpha, float *A, blasint lda, float *B, blasint ldb);
-void cblas_ztrsm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE TransA,
-                 enum CBLAS_DIAG Diag, blasint M, blasint N, double *alpha, double *A, blasint lda, double *B, blasint ldb);
-
-void cblas_chemm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 float *alpha, float *A, blasint lda, float *B, blasint ldb, float *beta, float *C, blasint ldc);
-void cblas_zhemm(enum CBLAS_ORDER Order, enum CBLAS_SIDE Side, enum CBLAS_UPLO Uplo, blasint M, blasint N,
-                 double *alpha, double *A, blasint lda, double *B, blasint ldb, double *beta, double *C, blasint ldc);
-
-void cblas_cherk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans, blasint N, blasint K,
-                 float alpha, float *A, blasint lda, float beta, float *C, blasint ldc);
-void cblas_zherk(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans, blasint N, blasint K,
-                 double alpha, double *A, blasint lda, double beta, double *C, blasint ldc);
-
-void cblas_cher2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans, blasint N, blasint K,
-                  float *alpha, float *A, blasint lda, float *B, blasint ldb, float beta, float *C, blasint ldc);
-void cblas_zher2k(enum CBLAS_ORDER Order, enum CBLAS_UPLO Uplo, enum CBLAS_TRANSPOSE Trans, blasint N, blasint K,
-                  double *alpha, double *A, blasint lda, double *B, blasint ldb, double beta, double *C, blasint ldc);
-
-void cblas_xerbla(blasint p, char *rout, char *form, ...);
-
-/*** BLAS extensions ***/
-
-void cblas_saxpby(blasint n, float alpha, float *x, blasint incx,float beta, float *y, blasint incy);
-
-void cblas_daxpby(blasint n, double alpha, double *x, blasint incx,double beta, double *y, blasint incy);
-
-void cblas_caxpby(blasint n, float *alpha, float *x, blasint incx,float *beta, float *y, blasint incy);
-
-void cblas_zaxpby(blasint n, double *alpha, double *x, blasint incx,double *beta, double *y, blasint incy);
-
-void cblas_somatcopy( enum CBLAS_ORDER CORDER,  enum CBLAS_TRANSPOSE CTRANS,  blasint crows,  blasint ccols,  float calpha,  float *a, 
-		      blasint clda, float *b,  blasint cldb); 
-void cblas_domatcopy( enum CBLAS_ORDER CORDER,  enum CBLAS_TRANSPOSE CTRANS,  blasint crows,  blasint ccols,  double calpha,  double *a,
-		      blasint clda, double *b,  blasint cldb); 
-void cblas_comatcopy( enum CBLAS_ORDER CORDER,  enum CBLAS_TRANSPOSE CTRANS,  blasint crows,  blasint ccols,  void* calpha,  void* a, 
-		      blasint clda, void *b,  blasint cldb); 
-void cblas_zomatcopy( enum CBLAS_ORDER CORDER,  enum CBLAS_TRANSPOSE CTRANS,  blasint crows,  blasint ccols,  void* calpha,  void* a, 
-		      blasint clda,  void *b,  blasint cldb); 
-
-void cblas_simatcopy( enum CBLAS_ORDER CORDER,  enum CBLAS_TRANSPOSE CTRANS,  blasint crows,  blasint ccols,  float calpha, float *a, 
-		      blasint clda,  blasint cldb); 
-void cblas_dimatcopy( enum CBLAS_ORDER CORDER,  enum CBLAS_TRANSPOSE CTRANS,  blasint crows,  blasint ccols,  double calpha, double *a,
-		      blasint clda,  blasint cldb); 
-void cblas_cimatcopy( enum CBLAS_ORDER CORDER,  enum CBLAS_TRANSPOSE CTRANS,  blasint crows,  blasint ccols,  float* calpha, float* a, 
-		      blasint clda,  blasint cldb); 
-void cblas_zimatcopy( enum CBLAS_ORDER CORDER,  enum CBLAS_TRANSPOSE CTRANS,  blasint crows,  blasint ccols,  double* calpha, double* a, 
-		      blasint clda,  blasint cldb); 
-
-void cblas_sgeadd( enum CBLAS_ORDER CORDER, blasint crows,  blasint ccols,  float calpha, float *a,  blasint clda,  float cbeta, 
-		  float *c,  blasint cldc); 
-void cblas_dgeadd( enum CBLAS_ORDER CORDER, blasint crows,  blasint ccols,  double calpha, double *a,  blasint clda,  double cbeta, 
-		  double *c,  blasint cldc); 
-void cblas_cgeadd( enum CBLAS_ORDER CORDER, blasint crows,  blasint ccols,  float *calpha, float *a,  blasint clda,  float *cbeta, 
-		  float *c,  blasint cldc); 
-void cblas_zgeadd( enum CBLAS_ORDER CORDER, blasint crows,  blasint ccols,  double *calpha, double *a,  blasint clda,  double *cbeta, 
-		  double *c,  blasint cldc); 
-
-#ifdef __cplusplus
-}
-#endif  /* __cplusplus */
-
-#endif
diff --git a/cmake/arch.cmake b/cmake/arch.cmake
new file mode 100644
index 000000000..d6fa3ed5d
--- /dev/null
+++ b/cmake/arch.cmake
@@ -0,0 +1,115 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+## Description: Ported from portion of OpenBLAS/Makefile.system
+##              Sets various variables based on architecture.
+
+if (${ARCH} STREQUAL "x86" OR ${ARCH} STREQUAL "x86_64")
+
+  if (${ARCH} STREQUAL "x86")
+    if (NOT BINARY)
+      set(NO_BINARY_MODE 1)
+    endif ()
+  endif ()
+
+  if (NOT NO_EXPRECISION)
+    if (${F_COMPILER} MATCHES "GFORTRAN")
+      # N.B. I'm not sure if CMake differentiates between GCC and LSB -hpa
+      if (${CMAKE_C_COMPILER} STREQUAL "GNU" OR ${CMAKE_C_COMPILER} STREQUAL "LSB")
+        set(EXPRECISION	1)
+        set(CCOMMON_OPT "${CCOMMON_OPT} -DEXPRECISION -m128bit-long-double")
+        set(FCOMMON_OPT	"${FCOMMON_OPT} -m128bit-long-double")
+      endif ()
+      if (${CMAKE_C_COMPILER} STREQUAL "Clang")
+        set(EXPRECISION	1)
+        set(CCOMMON_OPT "${CCOMMON_OPT} -DEXPRECISION")
+        set(FCOMMON_OPT	"${FCOMMON_OPT} -m128bit-long-double")
+      endif ()
+    endif ()
+  endif ()
+endif ()
+
+if (${CMAKE_C_COMPILER} STREQUAL "Intel")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -wd981")
+endif ()
+
+if (USE_OPENMP)
+
+  if (${CMAKE_C_COMPILER} STREQUAL "GNU" OR ${CMAKE_C_COMPILER} STREQUAL "LSB")
+    set(CCOMMON_OPT "${CCOMMON_OPT} -fopenmp")
+  endif ()
+
+  if (${CMAKE_C_COMPILER} STREQUAL "Clang")
+    message(WARNING "Clang doesn't support OpenMP yet.")
+    set(CCOMMON_OPT "${CCOMMON_OPT} -fopenmp")
+  endif ()
+
+  if (${CMAKE_C_COMPILER} STREQUAL "Intel")
+    set(CCOMMON_OPT "${CCOMMON_OPT} -openmp")
+  endif ()
+
+  if (${CMAKE_C_COMPILER} STREQUAL "PGI")
+    set(CCOMMON_OPT "${CCOMMON_OPT} -mp")
+  endif ()
+
+  if (${CMAKE_C_COMPILER} STREQUAL "OPEN64")
+    set(CCOMMON_OPT "${CCOMMON_OPT} -mp")
+    set(CEXTRALIB "${CEXTRALIB} -lstdc++")
+  endif ()
+
+  if (${CMAKE_C_COMPILER} STREQUAL "PATHSCALE")
+    set(CCOMMON_OPT "${CCOMMON_OPT} -mp")
+  endif ()
+endif ()
+
+
+if (DYNAMIC_ARCH)
+  if (${ARCH} STREQUAL "x86")
+    set(DYNAMIC_CORE "KATMAI COPPERMINE NORTHWOOD PRESCOTT BANIAS CORE2 PENRYN DUNNINGTON NEHALEM ATHLON OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO")
+  endif ()
+
+  if (${ARCH} STREQUAL "x86_64")
+    set(DYNAMIC_CORE "PRESCOTT CORE2 PENRYN DUNNINGTON NEHALEM OPTERON OPTERON_SSE3 BARCELONA BOBCAT ATOM NANO")
+    if (NOT NO_AVX)
+      set(DYNAMIC_CORE "${DYNAMIC_CORE} SANDYBRIDGE BULLDOZER PILEDRIVER STEAMROLLER")
+    endif ()
+    if (NOT NO_AVX2)
+      set(DYNAMIC_CORE "${DYNAMIC_CORE} HASWELL")
+    endif ()
+  endif ()
+
+  if (NOT DYNAMIC_CORE)
+    unset(DYNAMIC_ARCH)
+  endif ()
+endif ()
+
+if (${ARCH} STREQUAL "ia64")
+  set(NO_BINARY_MODE 1)
+  set(BINARY_DEFINED 1)
+
+  if (${F_COMPILER} MATCHES "GFORTRAN")
+    if (${CMAKE_C_COMPILER} STREQUAL "GNU")
+      # EXPRECISION	= 1
+      # CCOMMON_OPT	+= -DEXPRECISION
+    endif ()
+  endif ()
+endif ()
+
+if (${ARCH} STREQUAL "mips64")
+  set(NO_BINARY_MODE 1)
+endif ()
+
+if (${ARCH} STREQUAL "alpha")
+  set(NO_BINARY_MODE 1)
+  set(BINARY_DEFINED 1)
+endif ()
+
+if (${ARCH} STREQUAL "arm")
+  set(NO_BINARY_MODE 1)
+  set(BINARY_DEFINED 1)
+endif ()
+
+if (${ARCH} STREQUAL "arm64")
+  set(NO_BINARY_MODE 1)
+  set(BINARY_DEFINED 1)
+endif ()
+
diff --git a/cmake/c_check.cmake b/cmake/c_check.cmake
new file mode 100644
index 000000000..89ec31446
--- /dev/null
+++ b/cmake/c_check.cmake
@@ -0,0 +1,89 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+## Description: Ported from the OpenBLAS/c_check perl script.
+##              This is triggered by prebuild.cmake and runs before any of the code is built.
+##              Creates config.h and Makefile.conf.
+
+# CMake vars set by this file:
+# OSNAME (use CMAKE_SYSTEM_NAME)
+# ARCH
+# C_COMPILER (use CMAKE_C_COMPILER)
+# BINARY32
+# BINARY64
+# FU
+# CROSS_SUFFIX
+# CROSS
+# CEXTRALIB
+
+# Defines set by this file:
+# OS_
+# ARCH_
+# C_
+# __32BIT__
+# __64BIT__
+# FUNDERSCORE
+# PTHREAD_CREATE_FUNC
+
+# N.B. c_check (and ctest.c) is not cross-platform, so instead try to use CMake variables.
+set(FU "")
+if(APPLE)
+set(FU "_")
+elseif(MSVC)
+set(FU "_")
+elseif(UNIX)
+set(FU "")
+endif()
+
+# Convert CMake vars into the format that OpenBLAS expects
+string(TOUPPER ${CMAKE_SYSTEM_NAME} HOST_OS)
+if (${HOST_OS} STREQUAL "WINDOWS")
+  set(HOST_OS WINNT)
+endif ()
+
+# added by hpa - check size of void ptr to detect 64-bit compile
+if (NOT DEFINED BINARY)
+  set(BINARY 32)
+  if (CMAKE_SIZEOF_VOID_P EQUAL 8)
+    set(BINARY 64)
+  endif ()
+endif ()
+
+if (BINARY EQUAL 64)
+  set(BINARY64 1)
+else ()
+  set(BINARY32 1)
+endif ()
+
+# CMake docs define these:
+# CMAKE_SYSTEM_PROCESSOR - The name of the CPU CMake is building for.
+# CMAKE_HOST_SYSTEM_PROCESSOR - The name of the CPU CMake is running on.
+#
+# TODO: CMAKE_SYSTEM_PROCESSOR doesn't seem to be correct - instead get it from the compiler a la c_check
+set(ARCH ${CMAKE_SYSTEM_PROCESSOR})
+if (${ARCH} STREQUAL "AMD64")
+  set(ARCH "x86_64")
+endif ()
+
+# If you are using a 32-bit compiler on a 64-bit system CMAKE_SYSTEM_PROCESSOR will be wrong
+if (${ARCH} STREQUAL "x86_64" AND BINARY EQUAL 32)
+  set(ARCH x86)
+endif ()
+
+if (${ARCH} STREQUAL "X86")
+  set(ARCH x86)
+endif ()
+
+set(COMPILER_ID ${CMAKE_CXX_COMPILER_ID})
+if (${COMPILER_ID} STREQUAL "GNU")
+  set(COMPILER_ID "GCC")
+endif ()
+
+string(TOUPPER ${ARCH} UC_ARCH)
+
+file(WRITE ${TARGET_CONF}
+  "#define OS_${HOST_OS}\t1\n"
+  "#define ARCH_${UC_ARCH}\t1\n"
+  "#define C_${COMPILER_ID}\t1\n"
+  "#define __${BINARY}BIT__\t1\n"
+  "#define FUNDERSCORE\t${FU}\n")
+
diff --git a/cmake/cc.cmake b/cmake/cc.cmake
new file mode 100644
index 000000000..de196524f
--- /dev/null
+++ b/cmake/cc.cmake
@@ -0,0 +1,103 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+## Description: Ported from portion of OpenBLAS/Makefile.system
+##              Sets C related variables.
+
+if (${CMAKE_C_COMPILER} STREQUAL "GNU" OR ${CMAKE_C_COMPILER} STREQUAL "LSB" OR ${CMAKE_C_COMPILER} STREQUAL "Clang")
+
+  set(CCOMMON_OPT "${CCOMMON_OPT} -Wall")
+  set(COMMON_PROF "${COMMON_PROF} -fno-inline")
+  set(NO_UNINITIALIZED_WARN "-Wno-uninitialized")
+
+  if (QUIET_MAKE)
+    set(CCOMMON_OPT "${CCOMMON_OPT} ${NO_UNINITIALIZED_WARN} -Wno-unused")
+  endif ()
+
+  if (NO_BINARY_MODE)
+
+    if (${ARCH} STREQUAL "mips64")
+      if (BINARY64)
+        set(CCOMMON_OPT "${CCOMMON_OPT} -mabi=64")
+      else ()
+        set(CCOMMON_OPT "${CCOMMON_OPT} -mabi=n32")
+      endif ()
+      set(BINARY_DEFINED 1)
+    endif ()
+
+    if (${CORE} STREQUAL "LOONGSON3A")
+      set(CCOMMON_OPT "${CCOMMON_OPT} -march=mips64")
+      set(FCOMMON_OPT "${FCOMMON_OPT} -march=mips64")
+    endif ()
+
+    if (${CORE} STREQUAL "LOONGSON3B")
+      set(CCOMMON_OPT "${CCOMMON_OPT} -march=mips64")
+      set(FCOMMON_OPT "${FCOMMON_OPT} -march=mips64")
+    endif ()
+
+    if (${OSNAME} STREQUAL "AIX")
+      set(BINARY_DEFINED 1)
+    endif ()
+  endif ()
+
+  if (NOT BINARY_DEFINED)
+    if (BINARY64)
+      set(CCOMMON_OPT "${CCOMMON_OPT} -m64")
+    else ()
+      set(CCOMMON_OPT "${CCOMMON_OPT} -m32")
+    endif ()
+  endif ()
+endif ()
+
+if (${CMAKE_C_COMPILER} STREQUAL "PGI")
+  if (BINARY64)
+    set(CCOMMON_OPT "${CCOMMON_OPT} -tp p7-64")
+  else ()
+    set(CCOMMON_OPT "${CCOMMON_OPT} -tp p7")
+  endif ()
+endif ()
+
+if (${CMAKE_C_COMPILER} STREQUAL "PATHSCALE")
+  if (BINARY64)
+    set(CCOMMON_OPT "${CCOMMON_OPT} -m64")
+  else ()
+    set(CCOMMON_OPT "${CCOMMON_OPT} -m32")
+  endif ()
+endif ()
+
+if (${CMAKE_C_COMPILER} STREQUAL "OPEN64")
+
+  if (${ARCH} STREQUAL "mips64")
+
+    if (NOT BINARY64)
+      set(CCOMMON_OPT "${CCOMMON_OPT} -n32")
+    else ()
+      set(CCOMMON_OPT "${CCOMMON_OPT} -n64")
+    endif ()
+
+    if (${CORE} STREQUAL "LOONGSON3A")
+      set(CCOMMON_OPT "${CCOMMON_OPT} -loongson3 -static")
+    endif ()
+
+    if (${CORE} STREQUAL "LOONGSON3B")
+      set(CCOMMON_OPT "${CCOMMON_OPT} -loongson3 -static")
+    endif ()
+
+  else ()
+
+    if (BINARY64)
+      set(CCOMMON_OPT "${CCOMMON_OPT} -m32")
+    else ()
+      set(CCOMMON_OPT "${CCOMMON_OPT} -m64")
+    endif ()
+  endif ()
+endif ()
+
+if (${CMAKE_C_COMPILER} STREQUAL "SUN")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -w")
+  if (${ARCH} STREQUAL "x86")
+    set(CCOMMON_OPT "${CCOMMON_OPT} -m32")
+  else ()
+    set(FCOMMON_OPT "${FCOMMON_OPT} -m64")
+  endif ()
+endif ()
+
diff --git a/cmake/export.cmake b/cmake/export.cmake
new file mode 100644
index 000000000..adf59101f
--- /dev/null
+++ b/cmake/export.cmake
@@ -0,0 +1,60 @@
+
+#Only generate .def for dll on MSVC
+if(MSVC)
+
+set_source_files_properties(${OpenBLAS_DEF_FILE} PROPERTIES GENERATED 1)
+
+if (NOT DEFINED ARCH)
+  set(ARCH_IN "x86_64")
+else()
+  set(ARCH_IN ${ARCH})
+endif()
+
+if (${CORE} STREQUAL "generic")
+  set(ARCH_IN "GENERIC")
+endif ()
+
+if (NOT DEFINED EXPRECISION)
+  set(EXPRECISION_IN 0)
+else()
+  set(EXPRECISION_IN ${EXPRECISION})
+endif()
+
+if (NOT DEFINED NO_CBLAS)
+  set(NO_CBLAS_IN 0)
+else()
+  set(NO_CBLAS_IN ${NO_CBLAS})
+endif()
+
+if (NOT DEFINED NO_LAPACK)
+  set(NO_LAPACK_IN 0)
+else()
+  set(NO_LAPACK_IN ${NO_LAPACK})
+endif()
+
+if (NOT DEFINED NO_LAPACKE)
+  set(NO_LAPACKE_IN 0)
+else()
+  set(NO_LAPACKE_IN ${NO_LAPACKE})
+endif()
+
+if (NOT DEFINED NEED2UNDERSCORES)
+  set(NEED2UNDERSCORES_IN 0)
+else()
+  set(NEED2UNDERSCORES_IN ${NEED2UNDERSCORES})
+endif()
+
+if (NOT DEFINED ONLY_CBLAS)
+  set(ONLY_CBLAS_IN 0)
+else()
+  set(ONLY_CBLAS_IN ${ONLY_CBLAS})
+endif()
+
+add_custom_command(
+  TARGET ${OpenBLAS_LIBNAME} PRE_LINK 
+  COMMAND perl 
+  ARGS "${CMAKE_SOURCE_DIR}/exports/gensymbol" "win2k" "${ARCH_IN}" "dummy" "${EXPRECISION_IN}" "${NO_CBLAS_IN}" "${NO_LAPACK_IN}" "${NO_LAPACKE_IN}" "${NEED2UNDERSCORES_IN}" "${ONLY_CBLAS_IN}" "${SYMBOLPREFIX}" "${SYMBOLSUFFIX}" > "${PROJECT_BINARY_DIR}/openblas.def"
+  COMMENT "Create openblas.def file"
+  VERBATIM)
+
+endif()
\ No newline at end of file
diff --git a/cmake/f_check.cmake b/cmake/f_check.cmake
new file mode 100644
index 000000000..e8fe4bfa7
--- /dev/null
+++ b/cmake/f_check.cmake
@@ -0,0 +1,66 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+## Copyright: (c) Stat-Ease, Inc.
+## Created: 12/29/14
+## Last Modified: 12/29/14
+## Description: Ported from the OpenBLAS/f_check perl script.
+##              This is triggered by prebuild.cmake and runs before any of the code is built.
+##              Appends Fortran information to config.h and Makefile.conf.
+
+# CMake vars set by this file:
+# F_COMPILER
+# FC
+# BU
+# NOFORTRAN
+# NEED2UNDERSCORES
+# FEXTRALIB
+
+# Defines set by this file:
+# BUNDERSCORE
+# NEEDBUNDERSCORE
+# NEED2UNDERSCORES
+
+if (MSVC)
+  # had to do this for MSVC, else CMake automatically assumes I have ifort... -hpa
+  include(CMakeForceCompiler)
+  CMAKE_FORCE_Fortran_COMPILER(gfortran GNU)
+endif ()
+
+if (NOT NO_LAPACK)
+  enable_language(Fortran)
+else()
+  include(CMakeForceCompiler)
+  CMAKE_FORCE_Fortran_COMPILER(gfortran GNU)
+endif()
+
+if (NOT ONLY_CBLAS)
+  # N.B. f_check is not cross-platform, so instead try to use CMake variables
+  # run f_check (appends to TARGET files)
+#  message(STATUS "Running f_check...")
+#  execute_process(COMMAND perl f_check ${TARGET_MAKE} ${TARGET_CONF} ${CMAKE_Fortran_COMPILER}
+#    WORKING_DIRECTORY ${PROJECT_SOURCE_DIR})
+
+  # TODO: detect whether underscore needed, set #defines and BU appropriately - use try_compile
+  # TODO: set FEXTRALIB flags a la f_check?
+
+  set(BU "_")
+  file(APPEND ${TARGET_CONF}
+    "#define BUNDERSCORE _\n"
+    "#define NEEDBUNDERSCORE 1\n"
+    "#define NEED2UNDERSCORES 0\n")
+
+else ()
+
+  #When we only build CBLAS, we set NOFORTRAN=2
+  set(NOFORTRAN 2)
+  set(NO_FBLAS 1)
+  #set(F_COMPILER GFORTRAN) # CMake handles the fortran compiler
+  set(BU "_")
+  file(APPEND ${TARGET_CONF}
+    "#define BUNDERSCORE _\n"
+    "#define NEEDBUNDERSCORE 1\n")
+endif()
+
+get_filename_component(F_COMPILER ${CMAKE_Fortran_COMPILER} NAME_WE)
+string(TOUPPER ${F_COMPILER} F_COMPILER)
+
diff --git a/cmake/fc.cmake b/cmake/fc.cmake
new file mode 100644
index 000000000..ba156c210
--- /dev/null
+++ b/cmake/fc.cmake
@@ -0,0 +1,200 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+## Description: Ported from portion of OpenBLAS/Makefile.system
+##              Sets Fortran related variables.
+
+if (${F_COMPILER} STREQUAL "G77")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_G77")
+  set(FCOMMON_OPT "${FCOMMON_OPT} -Wall")
+  if (NOT NO_BINARY_MODE)
+    if (BINARY64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m64")
+    else ()
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m32")
+    endif ()
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "G95")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_G95")
+  set(FCOMMON_OPT "${FCOMMON_OPT} -Wall")
+  if (NOT NO_BINARY_MODE)
+    if (BINARY64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m64")
+    else ()
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m32")
+    endif ()
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "GFORTRAN")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_GFORT")
+  set(FCOMMON_OPT "${FCOMMON_OPT} -Wall")
+  #Don't include -lgfortran, when NO_LAPACK=1 or lsbcc
+  if (NOT NO_LAPACK)
+    set(EXTRALIB "{EXTRALIB} -lgfortran")
+  endif ()
+  if (NO_BINARY_MODE)
+    if (${ARCH} STREQUAL "mips64")
+      if (BINARY64)
+        set(FCOMMON_OPT "${FCOMMON_OPT} -mabi=64")
+      else ()
+        set(FCOMMON_OPT "${FCOMMON_OPT} -mabi=n32")
+      endif ()
+    endif ()
+  else ()
+    if (BINARY64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m64")
+      if (INTERFACE64)
+        set(FCOMMON_OPT "${FCOMMON_OPT} -fdefault-integer-8")
+      endif ()
+    else ()
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m32")
+    endif ()
+  endif ()
+
+  if (USE_OPENMP)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -fopenmp")
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "INTEL")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_INTEL")
+  if (INTERFACE64)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -i8")
+  endif ()
+  if (USE_OPENMP)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -openmp")
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "FUJITSU")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_FUJITSU")
+  if (USE_OPENMP)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -openmp")
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "IBM")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_IBM")
+  # FCOMMON_OPT	+= -qarch=440
+  if (BINARY64)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -q64")
+    if (INTERFACE64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -qintsize=8")
+    endif ()
+  else ()
+    set(FCOMMON_OPT "${FCOMMON_OPT} -q32")
+  endif ()
+  if (USE_OPENMP)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -openmp")
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "PGI")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_PGI")
+  set(COMMON_PROF "${COMMON_PROF} -DPGICOMPILER")
+  if (BINARY64)
+    if (INTERFACE64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -i8")
+    endif ()
+    set(FCOMMON_OPT "${FCOMMON_OPT} -tp p7-64")
+  else ()
+    set(FCOMMON_OPT "${FCOMMON_OPT} -tp p7")
+  endif ()
+  if (USE_OPENMP)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -mp")
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "PATHSCALE")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_PATHSCALE")
+  if (BINARY64)
+    if (INTERFACE64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -i8")
+    endif ()
+  endif ()
+
+  if (NOT ${ARCH} STREQUAL "mips64")
+    if (NOT BINARY64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m32")
+    else ()
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m64")
+    endif ()
+  else ()
+    if (BINARY64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -mabi=64")
+    else ()
+    set(FCOMMON_OPT "${FCOMMON_OPT} -mabi=n32")
+    endif ()
+  endif ()
+
+  if (USE_OPENMP)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -mp")
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "OPEN64")
+
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_OPEN64")
+  if (BINARY64)
+    if (INTERFACE64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -i8")
+    endif ()
+  endif ()
+
+  if (${ARCH} STREQUAL "mips64")
+
+    if (NOT BINARY64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -n32")
+    else ()
+      set(FCOMMON_OPT "${FCOMMON_OPT} -n64")
+    endif ()
+
+    if (${CORE} STREQUAL "LOONGSON3A")
+      set(FCOMMON_OPT "${FCOMMON_OPT} -loongson3 -static")
+    endif ()
+
+    if (${CORE} STREQUAL "LOONGSON3B")
+    set(FCOMMON_OPT "${FCOMMON_OPT} -loongson3 -static")
+    endif ()
+  else ()
+    if (NOT BINARY64)
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m32")
+    else ()
+      set(FCOMMON_OPT "${FCOMMON_OPT} -m64")
+    endif ()
+  endif ()
+
+  if (USE_OPENMP)
+    set(FEXTRALIB "${FEXTRALIB} -lstdc++")
+    set(FCOMMON_OPT "${FCOMMON_OPT} -mp")
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "SUN")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_SUN")
+  if (${ARCH} STREQUAL "x86")
+    set(FCOMMON_OPT "${FCOMMON_OPT} -m32")
+  else ()
+    set(FCOMMON_OPT "${FCOMMON_OPT} -m64")
+  endif ()
+  if (USE_OPENMP)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -xopenmp=parallel")
+  endif ()
+endif ()
+
+if (${F_COMPILER} STREQUAL "COMPAQ")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DF_INTERFACE_COMPAQ")
+  if (USE_OPENMP)
+    set(FCOMMON_OPT "${FCOMMON_OPT} -openmp")
+  endif ()
+endif ()
+
+# from the root Makefile - this is for lapack-netlib to compile the correct secnd file.
+if (${F_COMPILER} STREQUAL "GFORTRAN")
+  set(TIMER "INT_ETIME")
+else ()
+  set(TIMER "NONE")
+endif ()
+
diff --git a/cmake/kernel.cmake b/cmake/kernel.cmake
new file mode 100644
index 000000000..fad84de51
--- /dev/null
+++ b/cmake/kernel.cmake
@@ -0,0 +1,165 @@
+# helper functions for the kernel CMakeLists.txt
+
+
+# Set the default filenames for L1 objects. Most of these will be overriden by the appropriate KERNEL file.
+macro(SetDefaultL1)
+  set(SAMAXKERNEL amax.S)
+  set(DAMAXKERNEL amax.S)
+  set(QAMAXKERNEL amax.S)
+  set(CAMAXKERNEL zamax.S)
+  set(ZAMAXKERNEL zamax.S)
+  set(XAMAXKERNEL zamax.S)
+  set(SAMINKERNEL amin.S)
+  set(DAMINKERNEL amin.S)
+  set(QAMINKERNEL amin.S)
+  set(CAMINKERNEL zamin.S)
+  set(ZAMINKERNEL zamin.S)
+  set(XAMINKERNEL zamin.S)
+  set(SMAXKERNEL max.S)
+  set(DMAXKERNEL max.S)
+  set(QMAXKERNEL max.S)
+  set(SMINKERNEL min.S)
+  set(DMINKERNEL min.S)
+  set(QMINKERNEL min.S)
+  set(ISAMAXKERNEL iamax.S)
+  set(IDAMAXKERNEL iamax.S)
+  set(IQAMAXKERNEL iamax.S)
+  set(ICAMAXKERNEL izamax.S)
+  set(IZAMAXKERNEL izamax.S)
+  set(IXAMAXKERNEL izamax.S)
+  set(ISAMINKERNEL iamin.S)
+  set(IDAMINKERNEL iamin.S)
+  set(IQAMINKERNEL iamin.S)
+  set(ICAMINKERNEL izamin.S)
+  set(IZAMINKERNEL izamin.S)
+  set(IXAMINKERNEL izamin.S)
+  set(ISMAXKERNEL iamax.S)
+  set(IDMAXKERNEL iamax.S)
+  set(IQMAXKERNEL iamax.S)
+  set(ISMINKERNEL iamin.S)
+  set(IDMINKERNEL iamin.S)
+  set(IQMINKERNEL iamin.S)
+  set(SASUMKERNEL asum.S)
+  set(DASUMKERNEL asum.S)
+  set(CASUMKERNEL zasum.S)
+  set(ZASUMKERNEL zasum.S)
+  set(QASUMKERNEL asum.S)
+  set(XASUMKERNEL zasum.S)
+  set(SAXPYKERNEL axpy.S)
+  set(DAXPYKERNEL axpy.S)
+  set(CAXPYKERNEL zaxpy.S)
+  set(ZAXPYKERNEL zaxpy.S)
+  set(QAXPYKERNEL axpy.S)
+  set(XAXPYKERNEL zaxpy.S)
+  set(SCOPYKERNEL copy.S)
+  set(DCOPYKERNEL copy.S)
+  set(CCOPYKERNEL zcopy.S)
+  set(ZCOPYKERNEL zcopy.S)
+  set(QCOPYKERNEL copy.S)
+  set(XCOPYKERNEL zcopy.S)
+  set(SDOTKERNEL dot.S)
+  set(DDOTKERNEL dot.S)
+  set(CDOTKERNEL zdot.S)
+  set(ZDOTKERNEL zdot.S)
+  set(QDOTKERNEL dot.S)
+  set(XDOTKERNEL zdot.S)
+  set(SNRM2KERNEL nrm2.S)
+  set(DNRM2KERNEL nrm2.S)
+  set(QNRM2KERNEL nrm2.S)
+  set(CNRM2KERNEL znrm2.S)
+  set(ZNRM2KERNEL znrm2.S)
+  set(XNRM2KERNEL znrm2.S)
+  set(SROTKERNEL rot.S)
+  set(DROTKERNEL rot.S)
+  set(QROTKERNEL rot.S)
+  set(CROTKERNEL zrot.S)
+  set(ZROTKERNEL zrot.S)
+  set(XROTKERNEL zrot.S)
+  set(SSCALKERNEL scal.S)
+  set(DSCALKERNEL scal.S)
+  set(CSCALKERNEL zscal.S)
+  set(ZSCALKERNEL zscal.S)
+  set(QSCALKERNEL scal.S)
+  set(XSCALKERNEL zscal.S)
+  set(SSWAPKERNEL swap.S)
+  set(DSWAPKERNEL swap.S)
+  set(CSWAPKERNEL zswap.S)
+  set(ZSWAPKERNEL zswap.S)
+  set(QSWAPKERNEL swap.S)
+  set(XSWAPKERNEL zswap.S)
+  set(SGEMVNKERNEL gemv_n.S)
+  set(SGEMVTKERNEL gemv_t.S)
+  set(DGEMVNKERNEL gemv_n.S)
+  set(DGEMVTKERNEL gemv_t.S)
+  set(CGEMVNKERNEL zgemv_n.S)
+  set(CGEMVTKERNEL zgemv_t.S)
+  set(ZGEMVNKERNEL zgemv_n.S)
+  set(ZGEMVTKERNEL zgemv_t.S)
+  set(QGEMVNKERNEL gemv_n.S)
+  set(QGEMVTKERNEL gemv_t.S)
+  set(XGEMVNKERNEL zgemv_n.S)
+  set(XGEMVTKERNEL zgemv_t.S)
+  set(SCABS_KERNEL ../generic/cabs.c)
+  set(DCABS_KERNEL ../generic/cabs.c)
+  set(QCABS_KERNEL ../generic/cabs.c)
+  set(LSAME_KERNEL ../generic/lsame.c)
+  set(SAXPBYKERNEL ../arm/axpby.c)
+  set(DAXPBYKERNEL ../arm/axpby.c)
+  set(CAXPBYKERNEL ../arm/zaxpby.c)
+  set(ZAXPBYKERNEL ../arm/zaxpby.c)
+endmacro ()
+
+macro(SetDefaultL2)
+  set(SGEMVNKERNEL gemv_n.S)
+  set(SGEMVTKERNEL gemv_t.S)
+  set(DGEMVNKERNEL gemv_n.S)
+  set(DGEMVTKERNEL gemv_t.S)
+  set(CGEMVNKERNEL zgemv_n.S)
+  set(CGEMVTKERNEL zgemv_t.S)
+  set(ZGEMVNKERNEL zgemv_n.S)
+  set(ZGEMVTKERNEL zgemv_t.S)
+  set(QGEMVNKERNEL gemv_n.S)
+  set(QGEMVTKERNEL gemv_t.S)
+  set(XGEMVNKERNEL zgemv_n.S)
+  set(XGEMVTKERNEL zgemv_t.S)
+  set(SGERKERNEL ../generic/ger.c)
+  set(DGERKERNEL ../generic/ger.c)
+  set(QGERKERNEL ../generic/ger.c)
+  set(CGERUKERNEL ../generic/zger.c)
+  set(CGERCKERNEL ../generic/zger.c)
+  set(ZGERUKERNEL ../generic/zger.c)
+  set(ZGERCKERNEL ../generic/zger.c)
+  set(XGERUKERNEL ../generic/zger.c)
+  set(XGERCKERNEL ../generic/zger.c)
+  set(SSYMV_U_KERNEL ../generic/symv_k.c)
+  set(SSYMV_L_KERNEL ../generic/symv_k.c)
+  set(DSYMV_U_KERNEL ../generic/symv_k.c)
+  set(DSYMV_L_KERNEL ../generic/symv_k.c)
+  set(QSYMV_U_KERNEL ../generic/symv_k.c)
+  set(QSYMV_L_KERNEL ../generic/symv_k.c)
+  set(CSYMV_U_KERNEL ../generic/zsymv_k.c)
+  set(CSYMV_L_KERNEL ../generic/zsymv_k.c)
+  set(ZSYMV_U_KERNEL ../generic/zsymv_k.c)
+  set(ZSYMV_L_KERNEL ../generic/zsymv_k.c)
+  set(XSYMV_U_KERNEL ../generic/zsymv_k.c)
+  set(XSYMV_L_KERNEL ../generic/zsymv_k.c)
+  set(CHEMV_U_KERNEL ../generic/zhemv_k.c)
+  set(CHEMV_L_KERNEL ../generic/zhemv_k.c)
+  set(CHEMV_V_KERNEL ../generic/zhemv_k.c)
+  set(CHEMV_M_KERNEL ../generic/zhemv_k.c)
+  set(ZHEMV_U_KERNEL ../generic/zhemv_k.c)
+  set(ZHEMV_L_KERNEL ../generic/zhemv_k.c)
+  set(ZHEMV_V_KERNEL ../generic/zhemv_k.c)
+  set(ZHEMV_M_KERNEL ../generic/zhemv_k.c)
+  set(XHEMV_U_KERNEL ../generic/zhemv_k.c)
+  set(XHEMV_L_KERNEL ../generic/zhemv_k.c)
+  set(XHEMV_V_KERNEL ../generic/zhemv_k.c)
+  set(XHEMV_M_KERNEL ../generic/zhemv_k.c)
+endmacro ()
+
+macro(SetDefaultL3)
+  set(SGEADD_KERNEL ../generic/geadd.c)
+  set(DGEADD_KERNEL ../generic/geadd.c)
+  set(CGEADD_KERNEL ../generic/zgeadd.c)
+  set(ZGEADD_KERNEL ../generic/zgeadd.c)
+endmacro ()
\ No newline at end of file
diff --git a/cmake/lapack.cmake b/cmake/lapack.cmake
new file mode 100644
index 000000000..3e81611ab
--- /dev/null
+++ b/cmake/lapack.cmake
@@ -0,0 +1,347 @@
+# Sources for compiling lapack-netlib. Can't use CMakeLists.txt because lapack-netlib already has its own cmake files.
+
+set(ALLAUX
+  ilaenv.f ieeeck.f lsamen.f xerbla_array.f iparmq.f
+  ilaprec.f ilatrans.f ilauplo.f iladiag.f chla_transtype.f
+  ../INSTALL/ilaver.f ../INSTALL/slamch.f
+)
+
+set(SCLAUX
+  sbdsdc.f
+  sbdsqr.f sdisna.f slabad.f slacpy.f sladiv.f slae2.f  slaebz.f
+  slaed0.f slaed1.f slaed2.f slaed3.f slaed4.f slaed5.f slaed6.f
+  slaed7.f slaed8.f slaed9.f slaeda.f slaev2.f slagtf.f
+  slagts.f slamrg.f slanst.f
+  slapy2.f slapy3.f slarnv.f
+  slarra.f slarrb.f slarrc.f slarrd.f slarre.f slarrf.f slarrj.f
+  slarrk.f slarrr.f slaneg.f
+  slartg.f slaruv.f slas2.f  slascl.f
+  slasd0.f slasd1.f slasd2.f slasd3.f slasd4.f slasd5.f slasd6.f
+  slasd7.f slasd8.f slasda.f slasdq.f slasdt.f
+  slaset.f slasq1.f slasq2.f slasq3.f slasq4.f slasq5.f slasq6.f
+  slasr.f  slasrt.f slassq.f slasv2.f spttrf.f sstebz.f sstedc.f
+  ssteqr.f ssterf.f slaisnan.f sisnan.f
+  slartgp.f slartgs.f
+  ../INSTALL/second_${TIMER}.f
+)
+
+set(DZLAUX
+  dbdsdc.f
+  dbdsqr.f ddisna.f dlabad.f dlacpy.f dladiv.f dlae2.f  dlaebz.f
+  dlaed0.f dlaed1.f dlaed2.f dlaed3.f dlaed4.f dlaed5.f dlaed6.f
+  dlaed7.f dlaed8.f dlaed9.f dlaeda.f dlaev2.f dlagtf.f
+  dlagts.f dlamrg.f dlanst.f
+  dlapy2.f dlapy3.f dlarnv.f
+  dlarra.f dlarrb.f dlarrc.f dlarrd.f dlarre.f dlarrf.f dlarrj.f
+  dlarrk.f dlarrr.f dlaneg.f
+  dlartg.f dlaruv.f dlas2.f  dlascl.f
+  dlasd0.f dlasd1.f dlasd2.f dlasd3.f dlasd4.f dlasd5.f dlasd6.f
+  dlasd7.f dlasd8.f dlasda.f dlasdq.f dlasdt.f
+  dlaset.f dlasq1.f dlasq2.f dlasq3.f dlasq4.f dlasq5.f dlasq6.f
+  dlasr.f  dlasrt.f dlassq.f dlasv2.f dpttrf.f dstebz.f dstedc.f
+  dsteqr.f dsterf.f dlaisnan.f disnan.f
+  dlartgp.f dlartgs.f
+  ../INSTALL/dlamch.f ../INSTALL/dsecnd_${TIMER}.f
+)
+
+set(SLASRC
+  sgbbrd.f sgbcon.f sgbequ.f sgbrfs.f sgbsv.f
+  sgbsvx.f sgbtf2.f sgbtrf.f sgbtrs.f sgebak.f sgebal.f sgebd2.f
+  sgebrd.f sgecon.f sgeequ.f sgees.f  sgeesx.f sgeev.f  sgeevx.f
+  sgegs.f  sgegv.f  sgehd2.f sgehrd.f sgelq2.f sgelqf.f
+  sgels.f  sgelsd.f sgelss.f sgelsx.f sgelsy.f sgeql2.f sgeqlf.f
+  sgeqp3.f sgeqpf.f sgeqr2.f sgeqr2p.f sgeqrf.f sgeqrfp.f sgerfs.f
+  sgerq2.f sgerqf.f sgesc2.f sgesdd.f  sgesvd.f sgesvx.f
+  sgetc2.f sgetri.f
+  sggbak.f sggbal.f sgges.f  sggesx.f sggev.f  sggevx.f
+  sggglm.f sgghrd.f sgglse.f sggqrf.f
+  sggrqf.f sggsvd.f sggsvp.f sgtcon.f sgtrfs.f sgtsv.f
+  sgtsvx.f sgttrf.f sgttrs.f sgtts2.f shgeqz.f
+  shsein.f shseqr.f slabrd.f slacon.f slacn2.f
+  slaein.f slaexc.f slag2.f  slags2.f slagtm.f slagv2.f slahqr.f
+  slahrd.f slahr2.f slaic1.f slaln2.f slals0.f slalsa.f slalsd.f
+  slangb.f slange.f slangt.f slanhs.f slansb.f slansp.f
+  slansy.f slantb.f slantp.f slantr.f slanv2.f
+  slapll.f slapmt.f
+  slaqgb.f slaqge.f slaqp2.f slaqps.f slaqsb.f slaqsp.f slaqsy.f
+  slaqr0.f slaqr1.f slaqr2.f slaqr3.f slaqr4.f slaqr5.f
+  slaqtr.f slar1v.f slar2v.f ilaslr.f ilaslc.f
+  slarf.f  slarfb.f slarfg.f slarfgp.f slarft.f slarfx.f slargv.f
+  slarrv.f slartv.f
+  slarz.f  slarzb.f slarzt.f slasy2.f slasyf.f slasyf_rook.f
+  slatbs.f slatdf.f slatps.f slatrd.f slatrs.f slatrz.f slatzm.f
+  sopgtr.f sopmtr.f sorg2l.f sorg2r.f
+  sorgbr.f sorghr.f sorgl2.f sorglq.f sorgql.f sorgqr.f sorgr2.f
+  sorgrq.f sorgtr.f sorm2l.f sorm2r.f
+  sormbr.f sormhr.f sorml2.f sormlq.f sormql.f sormqr.f sormr2.f
+  sormr3.f sormrq.f sormrz.f sormtr.f spbcon.f spbequ.f spbrfs.f
+  spbstf.f spbsv.f  spbsvx.f
+  spbtf2.f spbtrf.f spbtrs.f spocon.f spoequ.f sporfs.f sposv.f
+  sposvx.f spstrf.f spstf2.f
+  sppcon.f sppequ.f
+  spprfs.f sppsv.f  sppsvx.f spptrf.f spptri.f spptrs.f sptcon.f
+  spteqr.f sptrfs.f sptsv.f  sptsvx.f spttrs.f sptts2.f srscl.f
+  ssbev.f  ssbevd.f ssbevx.f ssbgst.f ssbgv.f  ssbgvd.f ssbgvx.f
+  ssbtrd.f sspcon.f sspev.f  sspevd.f sspevx.f sspgst.f
+  sspgv.f  sspgvd.f sspgvx.f ssprfs.f sspsv.f  sspsvx.f ssptrd.f
+  ssptrf.f ssptri.f ssptrs.f sstegr.f sstein.f sstev.f  sstevd.f sstevr.f
+  sstevx.f
+  ssycon.f ssyev.f  ssyevd.f ssyevr.f ssyevx.f ssygs2.f
+  ssygst.f ssygv.f  ssygvd.f ssygvx.f ssyrfs.f ssysv.f  ssysvx.f
+  ssytd2.f ssytf2.f ssytrd.f ssytrf.f ssytri.f ssytri2.f ssytri2x.f
+  ssyswapr.f ssytrs.f ssytrs2.f ssyconv.f
+  ssytf2_rook.f ssytrf_rook.f ssytrs_rook.f
+  ssytri_rook.f ssycon_rook.f ssysv_rook.f
+  stbcon.f
+  stbrfs.f stbtrs.f stgevc.f stgex2.f stgexc.f stgsen.f
+  stgsja.f stgsna.f stgsy2.f stgsyl.f stpcon.f stprfs.f stptri.f
+  stptrs.f
+  strcon.f strevc.f strexc.f strrfs.f strsen.f strsna.f strsyl.f
+  strtrs.f stzrqf.f stzrzf.f sstemr.f
+  slansf.f spftrf.f spftri.f spftrs.f ssfrk.f stfsm.f stftri.f stfttp.f
+  stfttr.f stpttf.f stpttr.f strttf.f strttp.f
+  sgejsv.f  sgesvj.f  sgsvj0.f  sgsvj1.f
+  sgeequb.f ssyequb.f spoequb.f sgbequb.f
+  sbbcsd.f slapmr.f sorbdb.f sorbdb1.f sorbdb2.f sorbdb3.f sorbdb4.f
+  sorbdb5.f sorbdb6.f sorcsd.f sorcsd2by1.f
+  sgeqrt.f sgeqrt2.f sgeqrt3.f sgemqrt.f
+  stpqrt.f stpqrt2.f stpmqrt.f stprfb.f spotri.f
+)
+
+set(DSLASRC spotrs.f)
+
+set(CLASRC
+  cbdsqr.f cgbbrd.f cgbcon.f cgbequ.f cgbrfs.f cgbsv.f  cgbsvx.f
+  cgbtf2.f cgbtrf.f cgbtrs.f cgebak.f cgebal.f cgebd2.f cgebrd.f
+  cgecon.f cgeequ.f cgees.f  cgeesx.f cgeev.f  cgeevx.f
+  cgegs.f  cgegv.f  cgehd2.f cgehrd.f cgelq2.f cgelqf.f
+  cgels.f  cgelsd.f cgelss.f cgelsx.f cgelsy.f cgeql2.f cgeqlf.f cgeqp3.f
+  cgeqpf.f cgeqr2.f cgeqr2p.f cgeqrf.f cgeqrfp.f cgerfs.f
+  cgerq2.f cgerqf.f cgesc2.f cgesdd.f  cgesvd.f
+  cgesvx.f cgetc2.f cgetri.f
+  cggbak.f cggbal.f cgges.f  cggesx.f cggev.f  cggevx.f cggglm.f
+  cgghrd.f cgglse.f cggqrf.f cggrqf.f
+  cggsvd.f cggsvp.f
+  cgtcon.f cgtrfs.f cgtsv.f  cgtsvx.f cgttrf.f cgttrs.f cgtts2.f chbev.f
+  chbevd.f chbevx.f chbgst.f chbgv.f  chbgvd.f chbgvx.f chbtrd.f
+  checon.f cheev.f  cheevd.f cheevr.f cheevx.f chegs2.f chegst.f
+  chegv.f  chegvd.f chegvx.f cherfs.f chesv.f  chesvx.f chetd2.f
+  chetf2.f chetrd.f
+  chetrf.f chetri.f chetri2.f chetri2x.f cheswapr.f
+  chetrs.f chetrs2.f
+  chetf2_rook.f chetrf_rook.f chetri_rook.f chetrs_rook.f checon_rook.f chesv_rook.f
+  chgeqz.f chpcon.f chpev.f  chpevd.f
+  chpevx.f chpgst.f chpgv.f  chpgvd.f chpgvx.f chprfs.f chpsv.f
+  chpsvx.f
+  chptrd.f chptrf.f chptri.f chptrs.f chsein.f chseqr.f clabrd.f
+  clacgv.f clacon.f clacn2.f clacp2.f clacpy.f clacrm.f clacrt.f cladiv.f
+  claed0.f claed7.f claed8.f
+  claein.f claesy.f claev2.f clags2.f clagtm.f
+  clahef.f clahef_rook.f clahqr.f
+  clahrd.f clahr2.f claic1.f clals0.f clalsa.f clalsd.f clangb.f clange.f clangt.f
+  clanhb.f clanhe.f
+  clanhp.f clanhs.f clanht.f clansb.f clansp.f clansy.f clantb.f
+  clantp.f clantr.f clapll.f clapmt.f clarcm.f claqgb.f claqge.f
+  claqhb.f claqhe.f claqhp.f claqp2.f claqps.f claqsb.f
+  claqr0.f claqr1.f claqr2.f claqr3.f claqr4.f claqr5.f
+  claqsp.f claqsy.f clar1v.f clar2v.f ilaclr.f ilaclc.f
+  clarf.f  clarfb.f clarfg.f clarft.f clarfgp.f
+  clarfx.f clargv.f clarnv.f clarrv.f clartg.f clartv.f
+  clarz.f  clarzb.f clarzt.f clascl.f claset.f clasr.f  classq.f
+  clasyf.f clasyf_rook.f clatbs.f clatdf.f clatps.f clatrd.f clatrs.f clatrz.f
+  clatzm.f cpbcon.f cpbequ.f cpbrfs.f cpbstf.f cpbsv.f
+  cpbsvx.f cpbtf2.f cpbtrf.f cpbtrs.f cpocon.f cpoequ.f cporfs.f
+  cposv.f  cposvx.f cpstrf.f cpstf2.f
+  cppcon.f cppequ.f cpprfs.f cppsv.f  cppsvx.f cpptrf.f cpptri.f cpptrs.f
+  cptcon.f cpteqr.f cptrfs.f cptsv.f  cptsvx.f cpttrf.f cpttrs.f cptts2.f
+  crot.f   cspcon.f csprfs.f cspsv.f
+  cspsvx.f csptrf.f csptri.f csptrs.f csrscl.f cstedc.f
+  cstegr.f cstein.f csteqr.f
+  csycon.f
+  csyrfs.f csysv.f  csysvx.f csytf2.f csytrf.f csytri.f csytri2.f csytri2x.f
+  csyswapr.f csytrs.f csytrs2.f csyconv.f
+  csytf2_rook.f csytrf_rook.f csytrs_rook.f
+  csytri_rook.f csycon_rook.f csysv_rook.f
+  ctbcon.f ctbrfs.f ctbtrs.f ctgevc.f ctgex2.f
+  ctgexc.f ctgsen.f ctgsja.f ctgsna.f ctgsy2.f ctgsyl.f ctpcon.f
+  ctprfs.f ctptri.f
+  ctptrs.f ctrcon.f ctrevc.f ctrexc.f ctrrfs.f ctrsen.f ctrsna.f
+  ctrsyl.f ctrtrs.f ctzrqf.f ctzrzf.f cung2l.f cung2r.f
+  cungbr.f cunghr.f cungl2.f cunglq.f cungql.f cungqr.f cungr2.f
+  cungrq.f cungtr.f cunm2l.f cunm2r.f cunmbr.f cunmhr.f cunml2.f
+  cunmlq.f cunmql.f cunmqr.f cunmr2.f cunmr3.f cunmrq.f cunmrz.f
+  cunmtr.f cupgtr.f cupmtr.f icmax1.f scsum1.f cstemr.f
+  chfrk.f ctfttp.f clanhf.f cpftrf.f cpftri.f cpftrs.f ctfsm.f ctftri.f
+  ctfttr.f ctpttf.f ctpttr.f ctrttf.f ctrttp.f
+  cgeequb.f cgbequb.f csyequb.f cpoequb.f cheequb.f
+  cbbcsd.f clapmr.f cunbdb.f cunbdb1.f cunbdb2.f cunbdb3.f cunbdb4.f
+  cunbdb5.f cunbdb6.f cuncsd.f cuncsd2by1.f
+  cgeqrt.f cgeqrt2.f cgeqrt3.f cgemqrt.f
+  ctpqrt.f ctpqrt2.f ctpmqrt.f ctprfb.f cpotri.f
+)
+
+set(ZCLASRC cpotrs.f)
+
+set(DLASRC
+  dgbbrd.f dgbcon.f dgbequ.f dgbrfs.f dgbsv.f
+  dgbsvx.f dgbtf2.f dgbtrf.f dgbtrs.f dgebak.f dgebal.f dgebd2.f
+  dgebrd.f dgecon.f dgeequ.f dgees.f  dgeesx.f dgeev.f  dgeevx.f
+  dgegs.f  dgegv.f  dgehd2.f dgehrd.f dgelq2.f dgelqf.f
+  dgels.f  dgelsd.f dgelss.f dgelsx.f dgelsy.f dgeql2.f dgeqlf.f
+  dgeqp3.f dgeqpf.f dgeqr2.f dgeqr2p.f dgeqrf.f dgeqrfp.f dgerfs.f
+  dgerq2.f dgerqf.f dgesc2.f dgesdd.f  dgesvd.f dgesvx.f
+  dgetc2.f dgetri.f
+  dggbak.f dggbal.f dgges.f  dggesx.f dggev.f  dggevx.f
+  dggglm.f dgghrd.f dgglse.f dggqrf.f
+  dggrqf.f dggsvd.f dggsvp.f dgtcon.f dgtrfs.f dgtsv.f
+  dgtsvx.f dgttrf.f dgttrs.f dgtts2.f dhgeqz.f
+  dhsein.f dhseqr.f dlabrd.f dlacon.f dlacn2.f
+  dlaein.f dlaexc.f dlag2.f  dlags2.f dlagtm.f dlagv2.f dlahqr.f
+  dlahrd.f dlahr2.f dlaic1.f dlaln2.f dlals0.f dlalsa.f dlalsd.f
+  dlangb.f dlange.f dlangt.f dlanhs.f dlansb.f dlansp.f
+  dlansy.f dlantb.f dlantp.f dlantr.f dlanv2.f
+  dlapll.f dlapmt.f
+  dlaqgb.f dlaqge.f dlaqp2.f dlaqps.f dlaqsb.f dlaqsp.f dlaqsy.f
+  dlaqr0.f dlaqr1.f dlaqr2.f dlaqr3.f dlaqr4.f dlaqr5.f
+  dlaqtr.f dlar1v.f dlar2v.f iladlr.f iladlc.f
+  dlarf.f  dlarfb.f dlarfg.f dlarfgp.f dlarft.f dlarfx.f
+  dlargv.f dlarrv.f dlartv.f
+  dlarz.f  dlarzb.f dlarzt.f dlasy2.f dlasyf.f dlasyf_rook.f
+  dlatbs.f dlatdf.f dlatps.f dlatrd.f dlatrs.f dlatrz.f dlatzm.f
+  dopgtr.f dopmtr.f dorg2l.f dorg2r.f
+  dorgbr.f dorghr.f dorgl2.f dorglq.f dorgql.f dorgqr.f dorgr2.f
+  dorgrq.f dorgtr.f dorm2l.f dorm2r.f
+  dormbr.f dormhr.f dorml2.f dormlq.f dormql.f dormqr.f dormr2.f
+  dormr3.f dormrq.f dormrz.f dormtr.f dpbcon.f dpbequ.f dpbrfs.f
+  dpbstf.f dpbsv.f  dpbsvx.f
+  dpbtf2.f dpbtrf.f dpbtrs.f dpocon.f dpoequ.f dporfs.f dposv.f
+  dposvx.f dpotrs.f dpstrf.f dpstf2.f
+  dppcon.f dppequ.f
+  dpprfs.f dppsv.f  dppsvx.f dpptrf.f dpptri.f dpptrs.f dptcon.f
+  dpteqr.f dptrfs.f dptsv.f  dptsvx.f dpttrs.f dptts2.f drscl.f
+  dsbev.f  dsbevd.f dsbevx.f dsbgst.f dsbgv.f  dsbgvd.f dsbgvx.f
+  dsbtrd.f  dspcon.f dspev.f  dspevd.f dspevx.f dspgst.f
+  dspgv.f  dspgvd.f dspgvx.f dsprfs.f dspsv.f  dspsvx.f dsptrd.f
+  dsptrf.f dsptri.f dsptrs.f dstegr.f dstein.f dstev.f  dstevd.f dstevr.f
+  dstevx.f
+  dsycon.f dsyev.f  dsyevd.f dsyevr.f
+  dsyevx.f dsygs2.f dsygst.f dsygv.f  dsygvd.f dsygvx.f dsyrfs.f
+  dsysv.f  dsysvx.f
+  dsytd2.f dsytf2.f dsytrd.f dsytrf.f dsytri.f dsytri2.f dsytri2x.f
+  dsyswapr.f dsytrs.f dsytrs2.f dsyconv.f
+  dsytf2_rook.f dsytrf_rook.f dsytrs_rook.f
+  dsytri_rook.f dsycon_rook.f dsysv_rook.f
+  dtbcon.f dtbrfs.f dtbtrs.f dtgevc.f dtgex2.f dtgexc.f dtgsen.f
+  dtgsja.f dtgsna.f dtgsy2.f dtgsyl.f dtpcon.f dtprfs.f dtptri.f
+  dtptrs.f
+  dtrcon.f dtrevc.f dtrexc.f dtrrfs.f dtrsen.f dtrsna.f dtrsyl.f
+  dtrtrs.f dtzrqf.f dtzrzf.f dstemr.f
+  dsgesv.f dsposv.f dlag2s.f slag2d.f dlat2s.f
+  dlansf.f dpftrf.f dpftri.f dpftrs.f dsfrk.f dtfsm.f dtftri.f dtfttp.f
+  dtfttr.f dtpttf.f dtpttr.f dtrttf.f dtrttp.f
+  dgejsv.f  dgesvj.f  dgsvj0.f  dgsvj1.f
+  dgeequb.f dsyequb.f dpoequb.f dgbequb.f
+  dbbcsd.f dlapmr.f dorbdb.f dorbdb1.f dorbdb2.f dorbdb3.f dorbdb4.f
+  dorbdb5.f dorbdb6.f dorcsd.f dorcsd2by1.f
+  dgeqrt.f dgeqrt2.f dgeqrt3.f dgemqrt.f
+  dtpqrt.f dtpqrt2.f dtpmqrt.f dtprfb.f dpotri.f
+)
+
+set(ZLASRC
+  zbdsqr.f zgbbrd.f zgbcon.f zgbequ.f zgbrfs.f zgbsv.f  zgbsvx.f
+  zgbtf2.f zgbtrf.f zgbtrs.f zgebak.f zgebal.f zgebd2.f zgebrd.f
+  zgecon.f zgeequ.f zgees.f  zgeesx.f zgeev.f  zgeevx.f
+  zgegs.f  zgegv.f  zgehd2.f zgehrd.f zgelq2.f zgelqf.f
+  zgels.f  zgelsd.f zgelss.f zgelsx.f zgelsy.f zgeql2.f zgeqlf.f zgeqp3.f
+  zgeqpf.f zgeqr2.f zgeqr2p.f zgeqrf.f zgeqrfp.f zgerfs.f zgerq2.f zgerqf.f
+  zgesc2.f zgesdd.f zgesvd.f zgesvx.f zgetc2.f
+  zgetri.f
+  zggbak.f zggbal.f zgges.f  zggesx.f zggev.f  zggevx.f zggglm.f
+  zgghrd.f zgglse.f zggqrf.f zggrqf.f
+  zggsvd.f zggsvp.f
+  zgtcon.f zgtrfs.f zgtsv.f  zgtsvx.f zgttrf.f zgttrs.f zgtts2.f zhbev.f
+  zhbevd.f zhbevx.f zhbgst.f zhbgv.f  zhbgvd.f zhbgvx.f zhbtrd.f
+  zhecon.f zheev.f  zheevd.f zheevr.f zheevx.f zhegs2.f zhegst.f
+  zhegv.f  zhegvd.f zhegvx.f zherfs.f zhesv.f  zhesvx.f zhetd2.f
+  zhetf2.f zhetrd.f
+  zhetrf.f zhetri.f zhetri2.f zhetri2x.f zheswapr.f
+  zhetrs.f zhetrs2.f
+  zhetf2_rook.f zhetrf_rook.f zhetri_rook.f zhetrs_rook.f zhecon_rook.f zhesv_rook.f
+  zhgeqz.f zhpcon.f zhpev.f  zhpevd.f
+  zhpevx.f zhpgst.f zhpgv.f  zhpgvd.f zhpgvx.f zhprfs.f zhpsv.f
+  zhpsvx.f
+  zhptrd.f zhptrf.f zhptri.f zhptrs.f zhsein.f zhseqr.f zlabrd.f
+  zlacgv.f zlacon.f zlacn2.f zlacp2.f zlacpy.f zlacrm.f zlacrt.f zladiv.f
+  zlaed0.f zlaed7.f zlaed8.f
+  zlaein.f zlaesy.f zlaev2.f zlags2.f zlagtm.f
+  zlahef.f zlahef_rook.f zlahqr.f
+  zlahrd.f zlahr2.f zlaic1.f zlals0.f zlalsa.f zlalsd.f zlangb.f zlange.f
+  zlangt.f zlanhb.f
+  zlanhe.f
+  zlanhp.f zlanhs.f zlanht.f zlansb.f zlansp.f zlansy.f zlantb.f
+  zlantp.f zlantr.f zlapll.f zlapmt.f zlaqgb.f zlaqge.f
+  zlaqhb.f zlaqhe.f zlaqhp.f zlaqp2.f zlaqps.f zlaqsb.f
+  zlaqr0.f zlaqr1.f zlaqr2.f zlaqr3.f zlaqr4.f zlaqr5.f
+  zlaqsp.f zlaqsy.f zlar1v.f zlar2v.f ilazlr.f ilazlc.f
+  zlarcm.f zlarf.f  zlarfb.f
+  zlarfg.f zlarft.f zlarfgp.f
+  zlarfx.f zlargv.f zlarnv.f zlarrv.f zlartg.f zlartv.f
+  zlarz.f  zlarzb.f zlarzt.f zlascl.f zlaset.f zlasr.f
+  zlassq.f zlasyf.f zlasyf_rook.f
+  zlatbs.f zlatdf.f zlatps.f zlatrd.f zlatrs.f zlatrz.f zlatzm.f
+  zpbcon.f zpbequ.f zpbrfs.f zpbstf.f zpbsv.f
+  zpbsvx.f zpbtf2.f zpbtrf.f zpbtrs.f zpocon.f zpoequ.f zporfs.f
+  zposv.f  zposvx.f zpotrs.f zpstrf.f zpstf2.f
+  zppcon.f zppequ.f zpprfs.f zppsv.f  zppsvx.f zpptrf.f zpptri.f zpptrs.f
+  zptcon.f zpteqr.f zptrfs.f zptsv.f  zptsvx.f zpttrf.f zpttrs.f zptts2.f
+  zrot.f   zspcon.f zsprfs.f zspsv.f
+  zspsvx.f zsptrf.f zsptri.f zsptrs.f zdrscl.f zstedc.f
+  zstegr.f zstein.f zsteqr.f
+  zsycon.f
+  zsyrfs.f zsysv.f  zsysvx.f zsytf2.f zsytrf.f zsytri.f zsytri2.f zsytri2x.f
+  zsyswapr.f zsytrs.f zsytrs2.f zsyconv.f
+  zsytf2_rook.f zsytrf_rook.f zsytrs_rook.f
+  zsytri_rook.f zsycon_rook.f zsysv_rook.f
+  ztbcon.f ztbrfs.f ztbtrs.f ztgevc.f ztgex2.f
+  ztgexc.f ztgsen.f ztgsja.f ztgsna.f ztgsy2.f ztgsyl.f ztpcon.f
+  ztprfs.f ztptri.f
+  ztptrs.f ztrcon.f ztrevc.f ztrexc.f ztrrfs.f ztrsen.f ztrsna.f
+  ztrsyl.f ztrtrs.f ztzrqf.f ztzrzf.f zung2l.f
+  zung2r.f zungbr.f zunghr.f zungl2.f zunglq.f zungql.f zungqr.f zungr2.f
+  zungrq.f zungtr.f zunm2l.f zunm2r.f zunmbr.f zunmhr.f zunml2.f
+  zunmlq.f zunmql.f zunmqr.f zunmr2.f zunmr3.f zunmrq.f zunmrz.f
+  zunmtr.f zupgtr.f
+  zupmtr.f izmax1.f dzsum1.f zstemr.f
+  zcgesv.f zcposv.f zlag2c.f clag2z.f zlat2c.f
+  zhfrk.f ztfttp.f zlanhf.f zpftrf.f zpftri.f zpftrs.f ztfsm.f ztftri.f
+  ztfttr.f ztpttf.f ztpttr.f ztrttf.f ztrttp.f
+  zgeequb.f zgbequb.f zsyequb.f zpoequb.f zheequb.f
+  zbbcsd.f zlapmr.f zunbdb.f zunbdb1.f zunbdb2.f zunbdb3.f zunbdb4.f
+  zunbdb5.f zunbdb6.f zuncsd.f zuncsd2by1.f
+  zgeqrt.f zgeqrt2.f zgeqrt3.f zgemqrt.f
+  ztpqrt.f ztpqrt2.f ztpmqrt.f ztprfb.f zpotri.f
+)
+
+set(LA_REL_SRC ${ALLAUX})
+if (BUILD_SINGLE)
+  list(APPEND LA_REL_SRC ${SLASRC} ${DSLASRC} ${SCLAUX})
+endif ()
+
+if (BUILD_DOUBLE)
+  list(APPEND LA_REL_SRC ${DLASRC} ${DSLASRC} ${DZLAUX})
+endif ()
+
+if (BUILD_COMPLEX)
+  list(APPEND LA_REL_SRC ${CLASRC} ${ZCLASRC} ${SCLAUX})
+endif ()
+
+if (BUILD_COMPLEX16)
+  list(APPEND LA_REL_SRC ${ZLASRC} ${ZCLASRC} ${DZLAUX})
+endif ()
+
+# add lapack-netlib folder to the sources
+set(LA_SOURCES "")
+foreach (LA_FILE ${LA_REL_SRC})
+  list(APPEND LA_SOURCES "${NETLIB_LAPACK_DIR}/SRC/${LA_FILE}")
+endforeach ()
+set_source_files_properties(${LA_SOURCES} PROPERTIES COMPILE_FLAGS "${LAPACK_FFLAGS}")
diff --git a/cmake/lapacke.cmake b/cmake/lapacke.cmake
new file mode 100644
index 000000000..39ade0577
--- /dev/null
+++ b/cmake/lapacke.cmake
@@ -0,0 +1,2067 @@
+
+set(C_SRC
+  lapacke_cbbcsd.c 
+  lapacke_cbbcsd_work.c 
+  lapacke_cbdsqr.c 
+  lapacke_cbdsqr_work.c 
+  lapacke_cgbbrd.c 
+  lapacke_cgbbrd_work.c 
+  lapacke_cgbcon.c 
+  lapacke_cgbcon_work.c 
+  lapacke_cgbequ.c 
+  lapacke_cgbequ_work.c 
+  lapacke_cgbequb.c 
+  lapacke_cgbequb_work.c 
+  lapacke_cgbrfs.c 
+  lapacke_cgbrfs_work.c 
+  lapacke_cgbsv.c 
+  lapacke_cgbsv_work.c 
+  lapacke_cgbsvx.c 
+  lapacke_cgbsvx_work.c 
+  lapacke_cgbtrf.c 
+  lapacke_cgbtrf_work.c 
+  lapacke_cgbtrs.c 
+  lapacke_cgbtrs_work.c 
+  lapacke_cgebak.c 
+  lapacke_cgebak_work.c 
+  lapacke_cgebal.c 
+  lapacke_cgebal_work.c 
+  lapacke_cgebrd.c 
+  lapacke_cgebrd_work.c 
+  lapacke_cgecon.c 
+  lapacke_cgecon_work.c 
+  lapacke_cgeequ.c 
+  lapacke_cgeequ_work.c 
+  lapacke_cgeequb.c 
+  lapacke_cgeequb_work.c 
+  lapacke_cgees.c 
+  lapacke_cgees_work.c 
+  lapacke_cgeesx.c 
+  lapacke_cgeesx_work.c 
+  lapacke_cgeev.c 
+  lapacke_cgeev_work.c 
+  lapacke_cgeevx.c 
+  lapacke_cgeevx_work.c 
+  lapacke_cgehrd.c 
+  lapacke_cgehrd_work.c 
+  lapacke_cgelq2.c 
+  lapacke_cgelq2_work.c 
+  lapacke_cgelqf.c 
+  lapacke_cgelqf_work.c 
+  lapacke_cgels.c 
+  lapacke_cgels_work.c 
+  lapacke_cgelsd.c 
+  lapacke_cgelsd_work.c 
+  lapacke_cgelss.c 
+  lapacke_cgelss_work.c 
+  lapacke_cgelsy.c 
+  lapacke_cgelsy_work.c 
+  lapacke_cgemqrt.c 
+  lapacke_cgemqrt_work.c 
+  lapacke_cgeqlf.c 
+  lapacke_cgeqlf_work.c 
+  lapacke_cgeqp3.c 
+  lapacke_cgeqp3_work.c 
+  lapacke_cgeqpf.c 
+  lapacke_cgeqpf_work.c 
+  lapacke_cgeqr2.c 
+  lapacke_cgeqr2_work.c 
+  lapacke_cgeqrf.c 
+  lapacke_cgeqrf_work.c 
+  lapacke_cgeqrfp.c 
+  lapacke_cgeqrfp_work.c 
+  lapacke_cgeqrt.c 
+  lapacke_cgeqrt2.c 
+  lapacke_cgeqrt2_work.c 
+  lapacke_cgeqrt3.c 
+  lapacke_cgeqrt3_work.c 
+  lapacke_cgeqrt_work.c 
+  lapacke_cgerfs.c 
+  lapacke_cgerfs_work.c 
+  lapacke_cgerqf.c 
+  lapacke_cgerqf_work.c 
+  lapacke_cgesdd.c 
+  lapacke_cgesdd_work.c 
+  lapacke_cgesv.c 
+  lapacke_cgesv_work.c 
+  lapacke_cgesvd.c 
+  lapacke_cgesvd_work.c 
+  lapacke_cgesvx.c 
+  lapacke_cgesvx_work.c 
+  lapacke_cgetf2.c 
+  lapacke_cgetf2_work.c 
+  lapacke_cgetrf.c 
+  lapacke_cgetrf_work.c 
+  lapacke_cgetri.c 
+  lapacke_cgetri_work.c 
+  lapacke_cgetrs.c 
+  lapacke_cgetrs_work.c 
+  lapacke_cggbak.c 
+  lapacke_cggbak_work.c 
+  lapacke_cggbal.c 
+  lapacke_cggbal_work.c 
+  lapacke_cgges.c 
+  lapacke_cgges_work.c 
+  lapacke_cggesx.c 
+  lapacke_cggesx_work.c 
+  lapacke_cggev.c 
+  lapacke_cggev_work.c 
+  lapacke_cggevx.c 
+  lapacke_cggevx_work.c 
+  lapacke_cggglm.c 
+  lapacke_cggglm_work.c 
+  lapacke_cgghrd.c 
+  lapacke_cgghrd_work.c 
+  lapacke_cgglse.c 
+  lapacke_cgglse_work.c 
+  lapacke_cggqrf.c 
+  lapacke_cggqrf_work.c 
+  lapacke_cggrqf.c 
+  lapacke_cggrqf_work.c 
+  lapacke_cggsvd.c 
+  lapacke_cggsvd_work.c 
+  lapacke_cggsvp.c 
+  lapacke_cggsvp_work.c 
+  lapacke_cgtcon.c 
+  lapacke_cgtcon_work.c 
+  lapacke_cgtrfs.c 
+  lapacke_cgtrfs_work.c 
+  lapacke_cgtsv.c 
+  lapacke_cgtsv_work.c 
+  lapacke_cgtsvx.c 
+  lapacke_cgtsvx_work.c 
+  lapacke_cgttrf.c 
+  lapacke_cgttrf_work.c 
+  lapacke_cgttrs.c 
+  lapacke_cgttrs_work.c 
+  lapacke_chbev.c 
+  lapacke_chbev_work.c 
+  lapacke_chbevd.c 
+  lapacke_chbevd_work.c 
+  lapacke_chbevx.c 
+  lapacke_chbevx_work.c 
+  lapacke_chbgst.c 
+  lapacke_chbgst_work.c 
+  lapacke_chbgv.c 
+  lapacke_chbgv_work.c 
+  lapacke_chbgvd.c 
+  lapacke_chbgvd_work.c 
+  lapacke_chbgvx.c 
+  lapacke_chbgvx_work.c 
+  lapacke_chbtrd.c 
+  lapacke_chbtrd_work.c 
+  lapacke_checon.c 
+  lapacke_checon_work.c 
+  lapacke_cheequb.c 
+  lapacke_cheequb_work.c 
+  lapacke_cheev.c 
+  lapacke_cheev_work.c 
+  lapacke_cheevd.c 
+  lapacke_cheevd_work.c 
+  lapacke_cheevr.c 
+  lapacke_cheevr_work.c 
+  lapacke_cheevx.c 
+  lapacke_cheevx_work.c 
+  lapacke_chegst.c 
+  lapacke_chegst_work.c 
+  lapacke_chegv.c 
+  lapacke_chegv_work.c 
+  lapacke_chegvd.c 
+  lapacke_chegvd_work.c 
+  lapacke_chegvx.c 
+  lapacke_chegvx_work.c 
+  lapacke_cherfs.c 
+  lapacke_cherfs_work.c 
+  lapacke_chesv.c 
+  lapacke_chesv_work.c 
+  lapacke_chesvx.c 
+  lapacke_chesvx_work.c 
+  lapacke_cheswapr.c 
+  lapacke_cheswapr_work.c 
+  lapacke_chetrd.c 
+  lapacke_chetrd_work.c 
+  lapacke_chetrf.c 
+  lapacke_chetrf_work.c 
+  lapacke_chetri.c 
+  lapacke_chetri2.c 
+  lapacke_chetri2_work.c 
+  lapacke_chetri2x.c 
+  lapacke_chetri2x_work.c 
+  lapacke_chetri_work.c 
+  lapacke_chetrs.c 
+  lapacke_chetrs2.c 
+  lapacke_chetrs2_work.c 
+  lapacke_chetrs_work.c 
+  lapacke_chfrk.c 
+  lapacke_chfrk_work.c 
+  lapacke_chgeqz.c 
+  lapacke_chgeqz_work.c 
+  lapacke_chpcon.c 
+  lapacke_chpcon_work.c 
+  lapacke_chpev.c 
+  lapacke_chpev_work.c 
+  lapacke_chpevd.c 
+  lapacke_chpevd_work.c 
+  lapacke_chpevx.c 
+  lapacke_chpevx_work.c 
+  lapacke_chpgst.c 
+  lapacke_chpgst_work.c 
+  lapacke_chpgv.c 
+  lapacke_chpgv_work.c 
+  lapacke_chpgvd.c 
+  lapacke_chpgvd_work.c 
+  lapacke_chpgvx.c 
+  lapacke_chpgvx_work.c 
+  lapacke_chprfs.c 
+  lapacke_chprfs_work.c 
+  lapacke_chpsv.c 
+  lapacke_chpsv_work.c 
+  lapacke_chpsvx.c 
+  lapacke_chpsvx_work.c 
+  lapacke_chptrd.c 
+  lapacke_chptrd_work.c 
+  lapacke_chptrf.c 
+  lapacke_chptrf_work.c 
+  lapacke_chptri.c 
+  lapacke_chptri_work.c 
+  lapacke_chptrs.c 
+  lapacke_chptrs_work.c 
+  lapacke_chsein.c 
+  lapacke_chsein_work.c 
+  lapacke_chseqr.c 
+  lapacke_chseqr_work.c 
+  lapacke_clacgv.c 
+  lapacke_clacgv_work.c 
+  lapacke_clacn2.c 
+  lapacke_clacn2_work.c 
+  lapacke_clacp2.c 
+  lapacke_clacp2_work.c 
+  lapacke_clacpy.c 
+  lapacke_clacpy_work.c 
+  lapacke_clag2z.c 
+  lapacke_clag2z_work.c 
+  lapacke_clange.c 
+  lapacke_clange_work.c 
+  lapacke_clanhe.c 
+  lapacke_clanhe_work.c 
+  lapacke_clansy.c 
+  lapacke_clansy_work.c 
+  lapacke_clantr.c 
+  lapacke_clantr_work.c 
+  lapacke_clapmr.c 
+  lapacke_clapmr_work.c 
+  lapacke_clarfb.c 
+  lapacke_clarfb_work.c 
+  lapacke_clarfg.c 
+  lapacke_clarfg_work.c 
+  lapacke_clarft.c 
+  lapacke_clarft_work.c 
+  lapacke_clarfx.c 
+  lapacke_clarfx_work.c 
+  lapacke_clarnv.c 
+  lapacke_clarnv_work.c 
+  lapacke_claset.c 
+  lapacke_claset_work.c 
+  lapacke_claswp.c 
+  lapacke_claswp_work.c 
+  lapacke_clauum.c 
+  lapacke_clauum_work.c 
+  lapacke_cpbcon.c 
+  lapacke_cpbcon_work.c 
+  lapacke_cpbequ.c 
+  lapacke_cpbequ_work.c 
+  lapacke_cpbrfs.c 
+  lapacke_cpbrfs_work.c 
+  lapacke_cpbstf.c 
+  lapacke_cpbstf_work.c 
+  lapacke_cpbsv.c 
+  lapacke_cpbsv_work.c 
+  lapacke_cpbsvx.c 
+  lapacke_cpbsvx_work.c 
+  lapacke_cpbtrf.c 
+  lapacke_cpbtrf_work.c 
+  lapacke_cpbtrs.c 
+  lapacke_cpbtrs_work.c 
+  lapacke_cpftrf.c 
+  lapacke_cpftrf_work.c 
+  lapacke_cpftri.c 
+  lapacke_cpftri_work.c 
+  lapacke_cpftrs.c 
+  lapacke_cpftrs_work.c 
+  lapacke_cpocon.c 
+  lapacke_cpocon_work.c 
+  lapacke_cpoequ.c 
+  lapacke_cpoequ_work.c 
+  lapacke_cpoequb.c 
+  lapacke_cpoequb_work.c 
+  lapacke_cporfs.c 
+  lapacke_cporfs_work.c 
+  lapacke_cposv.c 
+  lapacke_cposv_work.c 
+  lapacke_cposvx.c 
+  lapacke_cposvx_work.c 
+  lapacke_cpotrf.c 
+  lapacke_cpotrf_work.c 
+  lapacke_cpotri.c 
+  lapacke_cpotri_work.c 
+  lapacke_cpotrs.c 
+  lapacke_cpotrs_work.c 
+  lapacke_cppcon.c 
+  lapacke_cppcon_work.c 
+  lapacke_cppequ.c 
+  lapacke_cppequ_work.c 
+  lapacke_cpprfs.c 
+  lapacke_cpprfs_work.c 
+  lapacke_cppsv.c 
+  lapacke_cppsv_work.c 
+  lapacke_cppsvx.c 
+  lapacke_cppsvx_work.c 
+  lapacke_cpptrf.c 
+  lapacke_cpptrf_work.c 
+  lapacke_cpptri.c 
+  lapacke_cpptri_work.c 
+  lapacke_cpptrs.c 
+  lapacke_cpptrs_work.c 
+  lapacke_cpstrf.c 
+  lapacke_cpstrf_work.c 
+  lapacke_cptcon.c 
+  lapacke_cptcon_work.c 
+  lapacke_cpteqr.c 
+  lapacke_cpteqr_work.c 
+  lapacke_cptrfs.c 
+  lapacke_cptrfs_work.c 
+  lapacke_cptsv.c 
+  lapacke_cptsv_work.c 
+  lapacke_cptsvx.c 
+  lapacke_cptsvx_work.c 
+  lapacke_cpttrf.c 
+  lapacke_cpttrf_work.c 
+  lapacke_cpttrs.c 
+  lapacke_cpttrs_work.c 
+  lapacke_cspcon.c 
+  lapacke_cspcon_work.c 
+  lapacke_csprfs.c 
+  lapacke_csprfs_work.c 
+  lapacke_cspsv.c 
+  lapacke_cspsv_work.c 
+  lapacke_cspsvx.c 
+  lapacke_cspsvx_work.c 
+  lapacke_csptrf.c 
+  lapacke_csptrf_work.c 
+  lapacke_csptri.c 
+  lapacke_csptri_work.c 
+  lapacke_csptrs.c 
+  lapacke_csptrs_work.c 
+  lapacke_cstedc.c 
+  lapacke_cstedc_work.c 
+  lapacke_cstegr.c 
+  lapacke_cstegr_work.c 
+  lapacke_cstein.c 
+  lapacke_cstein_work.c 
+  lapacke_cstemr.c 
+  lapacke_cstemr_work.c 
+  lapacke_csteqr.c 
+  lapacke_csteqr_work.c 
+  lapacke_csycon.c 
+  lapacke_csycon_work.c 
+  lapacke_csyconv.c 
+  lapacke_csyconv_work.c 
+  lapacke_csyequb.c 
+  lapacke_csyequb_work.c 
+  lapacke_csyrfs.c 
+  lapacke_csyrfs_work.c 
+  lapacke_csysv.c 
+  lapacke_csysv_rook.c 
+  lapacke_csysv_rook_work.c 
+  lapacke_csysv_work.c 
+  lapacke_csysvx.c 
+  lapacke_csysvx_work.c 
+  lapacke_csyswapr.c 
+  lapacke_csyswapr_work.c 
+  lapacke_csytrf.c 
+  lapacke_csytrf_work.c 
+  lapacke_csytri.c 
+  lapacke_csytri2.c 
+  lapacke_csytri2_work.c 
+  lapacke_csytri2x.c 
+  lapacke_csytri2x_work.c 
+  lapacke_csytri_work.c 
+  lapacke_csytrs.c 
+  lapacke_csytrs2.c 
+  lapacke_csytrs2_work.c 
+  lapacke_csytrs_work.c 
+  lapacke_ctbcon.c 
+  lapacke_ctbcon_work.c 
+  lapacke_ctbrfs.c 
+  lapacke_ctbrfs_work.c 
+  lapacke_ctbtrs.c 
+  lapacke_ctbtrs_work.c 
+  lapacke_ctfsm.c 
+  lapacke_ctfsm_work.c 
+  lapacke_ctftri.c 
+  lapacke_ctftri_work.c 
+  lapacke_ctfttp.c 
+  lapacke_ctfttp_work.c 
+  lapacke_ctfttr.c 
+  lapacke_ctfttr_work.c 
+  lapacke_ctgevc.c 
+  lapacke_ctgevc_work.c 
+  lapacke_ctgexc.c 
+  lapacke_ctgexc_work.c 
+  lapacke_ctgsen.c 
+  lapacke_ctgsen_work.c 
+  lapacke_ctgsja.c 
+  lapacke_ctgsja_work.c 
+  lapacke_ctgsna.c 
+  lapacke_ctgsna_work.c 
+  lapacke_ctgsyl.c 
+  lapacke_ctgsyl_work.c 
+  lapacke_ctpcon.c 
+  lapacke_ctpcon_work.c 
+  lapacke_ctpmqrt.c 
+  lapacke_ctpmqrt_work.c 
+  lapacke_ctpqrt.c 
+  lapacke_ctpqrt2.c 
+  lapacke_ctpqrt2_work.c 
+  lapacke_ctpqrt_work.c 
+  lapacke_ctprfb.c 
+  lapacke_ctprfb_work.c 
+  lapacke_ctprfs.c 
+  lapacke_ctprfs_work.c 
+  lapacke_ctptri.c 
+  lapacke_ctptri_work.c 
+  lapacke_ctptrs.c 
+  lapacke_ctptrs_work.c 
+  lapacke_ctpttf.c 
+  lapacke_ctpttf_work.c 
+  lapacke_ctpttr.c 
+  lapacke_ctpttr_work.c 
+  lapacke_ctrcon.c 
+  lapacke_ctrcon_work.c 
+  lapacke_ctrevc.c 
+  lapacke_ctrevc_work.c 
+  lapacke_ctrexc.c 
+  lapacke_ctrexc_work.c 
+  lapacke_ctrrfs.c 
+  lapacke_ctrrfs_work.c 
+  lapacke_ctrsen.c 
+  lapacke_ctrsen_work.c 
+  lapacke_ctrsna.c 
+  lapacke_ctrsna_work.c 
+  lapacke_ctrsyl.c 
+  lapacke_ctrsyl_work.c 
+  lapacke_ctrtri.c 
+  lapacke_ctrtri_work.c 
+  lapacke_ctrtrs.c 
+  lapacke_ctrtrs_work.c 
+  lapacke_ctrttf.c 
+  lapacke_ctrttf_work.c 
+  lapacke_ctrttp.c 
+  lapacke_ctrttp_work.c 
+  lapacke_ctzrzf.c 
+  lapacke_ctzrzf_work.c 
+  lapacke_cunbdb.c 
+  lapacke_cunbdb_work.c 
+  lapacke_cuncsd.c 
+  lapacke_cuncsd_work.c 
+  lapacke_cungbr.c 
+  lapacke_cungbr_work.c 
+  lapacke_cunghr.c 
+  lapacke_cunghr_work.c 
+  lapacke_cunglq.c 
+  lapacke_cunglq_work.c 
+  lapacke_cungql.c 
+  lapacke_cungql_work.c 
+  lapacke_cungqr.c 
+  lapacke_cungqr_work.c 
+  lapacke_cungrq.c 
+  lapacke_cungrq_work.c 
+  lapacke_cungtr.c 
+  lapacke_cungtr_work.c 
+  lapacke_cunmbr.c 
+  lapacke_cunmbr_work.c 
+  lapacke_cunmhr.c 
+  lapacke_cunmhr_work.c 
+  lapacke_cunmlq.c 
+  lapacke_cunmlq_work.c 
+  lapacke_cunmql.c 
+  lapacke_cunmql_work.c 
+  lapacke_cunmqr.c 
+  lapacke_cunmqr_work.c 
+  lapacke_cunmrq.c 
+  lapacke_cunmrq_work.c 
+  lapacke_cunmrz.c 
+  lapacke_cunmrz_work.c 
+  lapacke_cunmtr.c 
+  lapacke_cunmtr_work.c 
+  lapacke_cupgtr.c 
+  lapacke_cupgtr_work.c 
+  lapacke_cupmtr.c 
+  lapacke_cupmtr_work.c
+)
+
+set(DSRC
+  lapacke_dbbcsd.c 
+  lapacke_dbbcsd_work.c 
+  lapacke_dbdsdc.c 
+  lapacke_dbdsdc_work.c 
+  lapacke_dbdsqr.c 
+  lapacke_dbdsqr_work.c 
+  lapacke_ddisna.c 
+  lapacke_ddisna_work.c 
+  lapacke_dgbbrd.c 
+  lapacke_dgbbrd_work.c 
+  lapacke_dgbcon.c 
+  lapacke_dgbcon_work.c 
+  lapacke_dgbequ.c 
+  lapacke_dgbequ_work.c 
+  lapacke_dgbequb.c 
+  lapacke_dgbequb_work.c 
+  lapacke_dgbrfs.c 
+  lapacke_dgbrfs_work.c 
+  lapacke_dgbsv.c 
+  lapacke_dgbsv_work.c 
+  lapacke_dgbsvx.c 
+  lapacke_dgbsvx_work.c 
+  lapacke_dgbtrf.c 
+  lapacke_dgbtrf_work.c 
+  lapacke_dgbtrs.c 
+  lapacke_dgbtrs_work.c 
+  lapacke_dgebak.c 
+  lapacke_dgebak_work.c 
+  lapacke_dgebal.c 
+  lapacke_dgebal_work.c 
+  lapacke_dgebrd.c 
+  lapacke_dgebrd_work.c 
+  lapacke_dgecon.c 
+  lapacke_dgecon_work.c 
+  lapacke_dgeequ.c 
+  lapacke_dgeequ_work.c 
+  lapacke_dgeequb.c 
+  lapacke_dgeequb_work.c 
+  lapacke_dgees.c 
+  lapacke_dgees_work.c 
+  lapacke_dgeesx.c 
+  lapacke_dgeesx_work.c 
+  lapacke_dgeev.c 
+  lapacke_dgeev_work.c 
+  lapacke_dgeevx.c 
+  lapacke_dgeevx_work.c 
+  lapacke_dgehrd.c 
+  lapacke_dgehrd_work.c 
+  lapacke_dgejsv.c 
+  lapacke_dgejsv_work.c 
+  lapacke_dgelq2.c 
+  lapacke_dgelq2_work.c 
+  lapacke_dgelqf.c 
+  lapacke_dgelqf_work.c 
+  lapacke_dgels.c 
+  lapacke_dgels_work.c 
+  lapacke_dgelsd.c 
+  lapacke_dgelsd_work.c 
+  lapacke_dgelss.c 
+  lapacke_dgelss_work.c 
+  lapacke_dgelsy.c 
+  lapacke_dgelsy_work.c 
+  lapacke_dgemqrt.c 
+  lapacke_dgemqrt_work.c 
+  lapacke_dgeqlf.c 
+  lapacke_dgeqlf_work.c 
+  lapacke_dgeqp3.c 
+  lapacke_dgeqp3_work.c 
+  lapacke_dgeqpf.c 
+  lapacke_dgeqpf_work.c 
+  lapacke_dgeqr2.c 
+  lapacke_dgeqr2_work.c 
+  lapacke_dgeqrf.c 
+  lapacke_dgeqrf_work.c 
+  lapacke_dgeqrfp.c 
+  lapacke_dgeqrfp_work.c 
+  lapacke_dgeqrt.c 
+  lapacke_dgeqrt2.c 
+  lapacke_dgeqrt2_work.c 
+  lapacke_dgeqrt3.c 
+  lapacke_dgeqrt3_work.c 
+  lapacke_dgeqrt_work.c 
+  lapacke_dgerfs.c 
+  lapacke_dgerfs_work.c 
+  lapacke_dgerqf.c 
+  lapacke_dgerqf_work.c 
+  lapacke_dgesdd.c 
+  lapacke_dgesdd_work.c 
+  lapacke_dgesv.c 
+  lapacke_dgesv_work.c 
+  lapacke_dgesvd.c 
+  lapacke_dgesvd_work.c 
+  lapacke_dgesvj.c 
+  lapacke_dgesvj_work.c 
+  lapacke_dgesvx.c 
+  lapacke_dgesvx_work.c 
+  lapacke_dgetf2.c 
+  lapacke_dgetf2_work.c 
+  lapacke_dgetrf.c 
+  lapacke_dgetrf_work.c 
+  lapacke_dgetri.c 
+  lapacke_dgetri_work.c 
+  lapacke_dgetrs.c 
+  lapacke_dgetrs_work.c 
+  lapacke_dggbak.c 
+  lapacke_dggbak_work.c 
+  lapacke_dggbal.c 
+  lapacke_dggbal_work.c 
+  lapacke_dgges.c 
+  lapacke_dgges_work.c 
+  lapacke_dggesx.c 
+  lapacke_dggesx_work.c 
+  lapacke_dggev.c 
+  lapacke_dggev_work.c 
+  lapacke_dggevx.c 
+  lapacke_dggevx_work.c 
+  lapacke_dggglm.c 
+  lapacke_dggglm_work.c 
+  lapacke_dgghrd.c 
+  lapacke_dgghrd_work.c 
+  lapacke_dgglse.c 
+  lapacke_dgglse_work.c 
+  lapacke_dggqrf.c 
+  lapacke_dggqrf_work.c 
+  lapacke_dggrqf.c 
+  lapacke_dggrqf_work.c 
+  lapacke_dggsvd.c 
+  lapacke_dggsvd_work.c 
+  lapacke_dggsvp.c 
+  lapacke_dggsvp_work.c 
+  lapacke_dgtcon.c 
+  lapacke_dgtcon_work.c 
+  lapacke_dgtrfs.c 
+  lapacke_dgtrfs_work.c 
+  lapacke_dgtsv.c 
+  lapacke_dgtsv_work.c 
+  lapacke_dgtsvx.c 
+  lapacke_dgtsvx_work.c 
+  lapacke_dgttrf.c 
+  lapacke_dgttrf_work.c 
+  lapacke_dgttrs.c 
+  lapacke_dgttrs_work.c 
+  lapacke_dhgeqz.c 
+  lapacke_dhgeqz_work.c 
+  lapacke_dhsein.c 
+  lapacke_dhsein_work.c 
+  lapacke_dhseqr.c 
+  lapacke_dhseqr_work.c 
+  lapacke_dlacn2.c 
+  lapacke_dlacn2_work.c 
+  lapacke_dlacpy.c 
+  lapacke_dlacpy_work.c 
+  lapacke_dlag2s.c 
+  lapacke_dlag2s_work.c 
+  lapacke_dlamch.c 
+  lapacke_dlamch_work.c 
+  lapacke_dlange.c 
+  lapacke_dlange_work.c 
+  lapacke_dlansy.c 
+  lapacke_dlansy_work.c 
+  lapacke_dlantr.c 
+  lapacke_dlantr_work.c 
+  lapacke_dlapmr.c 
+  lapacke_dlapmr_work.c 
+  lapacke_dlapy2.c 
+  lapacke_dlapy2_work.c 
+  lapacke_dlapy3.c 
+  lapacke_dlapy3_work.c 
+  lapacke_dlarfb.c 
+  lapacke_dlarfb_work.c 
+  lapacke_dlarfg.c 
+  lapacke_dlarfg_work.c 
+  lapacke_dlarft.c 
+  lapacke_dlarft_work.c 
+  lapacke_dlarfx.c 
+  lapacke_dlarfx_work.c 
+  lapacke_dlarnv.c 
+  lapacke_dlarnv_work.c 
+  lapacke_dlartgp.c 
+  lapacke_dlartgp_work.c 
+  lapacke_dlartgs.c 
+  lapacke_dlartgs_work.c 
+  lapacke_dlaset.c 
+  lapacke_dlaset_work.c 
+  lapacke_dlasrt.c 
+  lapacke_dlasrt_work.c 
+  lapacke_dlaswp.c 
+  lapacke_dlaswp_work.c 
+  lapacke_dlauum.c 
+  lapacke_dlauum_work.c 
+  lapacke_dopgtr.c 
+  lapacke_dopgtr_work.c 
+  lapacke_dopmtr.c 
+  lapacke_dopmtr_work.c 
+  lapacke_dorbdb.c 
+  lapacke_dorbdb_work.c 
+  lapacke_dorcsd.c 
+  lapacke_dorcsd_work.c 
+  lapacke_dorgbr.c 
+  lapacke_dorgbr_work.c 
+  lapacke_dorghr.c 
+  lapacke_dorghr_work.c 
+  lapacke_dorglq.c 
+  lapacke_dorglq_work.c 
+  lapacke_dorgql.c 
+  lapacke_dorgql_work.c 
+  lapacke_dorgqr.c 
+  lapacke_dorgqr_work.c 
+  lapacke_dorgrq.c 
+  lapacke_dorgrq_work.c 
+  lapacke_dorgtr.c 
+  lapacke_dorgtr_work.c 
+  lapacke_dormbr.c 
+  lapacke_dormbr_work.c 
+  lapacke_dormhr.c 
+  lapacke_dormhr_work.c 
+  lapacke_dormlq.c 
+  lapacke_dormlq_work.c 
+  lapacke_dormql.c 
+  lapacke_dormql_work.c 
+  lapacke_dormqr.c 
+  lapacke_dormqr_work.c 
+  lapacke_dormrq.c 
+  lapacke_dormrq_work.c 
+  lapacke_dormrz.c 
+  lapacke_dormrz_work.c 
+  lapacke_dormtr.c 
+  lapacke_dormtr_work.c 
+  lapacke_dpbcon.c 
+  lapacke_dpbcon_work.c 
+  lapacke_dpbequ.c 
+  lapacke_dpbequ_work.c 
+  lapacke_dpbrfs.c 
+  lapacke_dpbrfs_work.c 
+  lapacke_dpbstf.c 
+  lapacke_dpbstf_work.c 
+  lapacke_dpbsv.c 
+  lapacke_dpbsv_work.c 
+  lapacke_dpbsvx.c 
+  lapacke_dpbsvx_work.c 
+  lapacke_dpbtrf.c 
+  lapacke_dpbtrf_work.c 
+  lapacke_dpbtrs.c 
+  lapacke_dpbtrs_work.c 
+  lapacke_dpftrf.c 
+  lapacke_dpftrf_work.c 
+  lapacke_dpftri.c 
+  lapacke_dpftri_work.c 
+  lapacke_dpftrs.c 
+  lapacke_dpftrs_work.c 
+  lapacke_dpocon.c 
+  lapacke_dpocon_work.c 
+  lapacke_dpoequ.c 
+  lapacke_dpoequ_work.c 
+  lapacke_dpoequb.c 
+  lapacke_dpoequb_work.c 
+  lapacke_dporfs.c 
+  lapacke_dporfs_work.c 
+  lapacke_dposv.c 
+  lapacke_dposv_work.c 
+  lapacke_dposvx.c 
+  lapacke_dposvx_work.c 
+  lapacke_dpotrf.c 
+  lapacke_dpotrf_work.c 
+  lapacke_dpotri.c 
+  lapacke_dpotri_work.c 
+  lapacke_dpotrs.c 
+  lapacke_dpotrs_work.c 
+  lapacke_dppcon.c 
+  lapacke_dppcon_work.c 
+  lapacke_dppequ.c 
+  lapacke_dppequ_work.c 
+  lapacke_dpprfs.c 
+  lapacke_dpprfs_work.c 
+  lapacke_dppsv.c 
+  lapacke_dppsv_work.c 
+  lapacke_dppsvx.c 
+  lapacke_dppsvx_work.c 
+  lapacke_dpptrf.c 
+  lapacke_dpptrf_work.c 
+  lapacke_dpptri.c 
+  lapacke_dpptri_work.c 
+  lapacke_dpptrs.c 
+  lapacke_dpptrs_work.c 
+  lapacke_dpstrf.c 
+  lapacke_dpstrf_work.c 
+  lapacke_dptcon.c 
+  lapacke_dptcon_work.c 
+  lapacke_dpteqr.c 
+  lapacke_dpteqr_work.c 
+  lapacke_dptrfs.c 
+  lapacke_dptrfs_work.c 
+  lapacke_dptsv.c 
+  lapacke_dptsv_work.c 
+  lapacke_dptsvx.c 
+  lapacke_dptsvx_work.c 
+  lapacke_dpttrf.c 
+  lapacke_dpttrf_work.c 
+  lapacke_dpttrs.c 
+  lapacke_dpttrs_work.c 
+  lapacke_dsbev.c 
+  lapacke_dsbev_work.c 
+  lapacke_dsbevd.c 
+  lapacke_dsbevd_work.c 
+  lapacke_dsbevx.c 
+  lapacke_dsbevx_work.c 
+  lapacke_dsbgst.c 
+  lapacke_dsbgst_work.c 
+  lapacke_dsbgv.c 
+  lapacke_dsbgv_work.c 
+  lapacke_dsbgvd.c 
+  lapacke_dsbgvd_work.c 
+  lapacke_dsbgvx.c 
+  lapacke_dsbgvx_work.c 
+  lapacke_dsbtrd.c 
+  lapacke_dsbtrd_work.c 
+  lapacke_dsfrk.c 
+  lapacke_dsfrk_work.c 
+  lapacke_dsgesv.c 
+  lapacke_dsgesv_work.c 
+  lapacke_dspcon.c 
+  lapacke_dspcon_work.c 
+  lapacke_dspev.c 
+  lapacke_dspev_work.c 
+  lapacke_dspevd.c 
+  lapacke_dspevd_work.c 
+  lapacke_dspevx.c 
+  lapacke_dspevx_work.c 
+  lapacke_dspgst.c 
+  lapacke_dspgst_work.c 
+  lapacke_dspgv.c 
+  lapacke_dspgv_work.c 
+  lapacke_dspgvd.c 
+  lapacke_dspgvd_work.c 
+  lapacke_dspgvx.c 
+  lapacke_dspgvx_work.c 
+  lapacke_dsposv.c 
+  lapacke_dsposv_work.c 
+  lapacke_dsprfs.c 
+  lapacke_dsprfs_work.c 
+  lapacke_dspsv.c 
+  lapacke_dspsv_work.c 
+  lapacke_dspsvx.c 
+  lapacke_dspsvx_work.c 
+  lapacke_dsptrd.c 
+  lapacke_dsptrd_work.c 
+  lapacke_dsptrf.c 
+  lapacke_dsptrf_work.c 
+  lapacke_dsptri.c 
+  lapacke_dsptri_work.c 
+  lapacke_dsptrs.c 
+  lapacke_dsptrs_work.c 
+  lapacke_dstebz.c 
+  lapacke_dstebz_work.c 
+  lapacke_dstedc.c 
+  lapacke_dstedc_work.c 
+  lapacke_dstegr.c 
+  lapacke_dstegr_work.c 
+  lapacke_dstein.c 
+  lapacke_dstein_work.c 
+  lapacke_dstemr.c 
+  lapacke_dstemr_work.c 
+  lapacke_dsteqr.c 
+  lapacke_dsteqr_work.c 
+  lapacke_dsterf.c 
+  lapacke_dsterf_work.c 
+  lapacke_dstev.c 
+  lapacke_dstev_work.c 
+  lapacke_dstevd.c 
+  lapacke_dstevd_work.c 
+  lapacke_dstevr.c 
+  lapacke_dstevr_work.c 
+  lapacke_dstevx.c 
+  lapacke_dstevx_work.c 
+  lapacke_dsycon.c 
+  lapacke_dsycon_work.c 
+  lapacke_dsyconv.c 
+  lapacke_dsyconv_work.c 
+  lapacke_dsyequb.c 
+  lapacke_dsyequb_work.c 
+  lapacke_dsyev.c 
+  lapacke_dsyev_work.c 
+  lapacke_dsyevd.c 
+  lapacke_dsyevd_work.c 
+  lapacke_dsyevr.c 
+  lapacke_dsyevr_work.c 
+  lapacke_dsyevx.c 
+  lapacke_dsyevx_work.c 
+  lapacke_dsygst.c 
+  lapacke_dsygst_work.c 
+  lapacke_dsygv.c 
+  lapacke_dsygv_work.c 
+  lapacke_dsygvd.c 
+  lapacke_dsygvd_work.c 
+  lapacke_dsygvx.c 
+  lapacke_dsygvx_work.c 
+  lapacke_dsyrfs.c 
+  lapacke_dsyrfs_work.c 
+  lapacke_dsysv.c 
+  lapacke_dsysv_rook.c 
+  lapacke_dsysv_rook_work.c 
+  lapacke_dsysv_work.c 
+  lapacke_dsysvx.c 
+  lapacke_dsysvx_work.c 
+  lapacke_dsyswapr.c 
+  lapacke_dsyswapr_work.c 
+  lapacke_dsytrd.c 
+  lapacke_dsytrd_work.c 
+  lapacke_dsytrf.c 
+  lapacke_dsytrf_work.c 
+  lapacke_dsytri.c 
+  lapacke_dsytri2.c 
+  lapacke_dsytri2_work.c 
+  lapacke_dsytri2x.c 
+  lapacke_dsytri2x_work.c 
+  lapacke_dsytri_work.c 
+  lapacke_dsytrs.c 
+  lapacke_dsytrs2.c 
+  lapacke_dsytrs2_work.c 
+  lapacke_dsytrs_work.c 
+  lapacke_dtbcon.c 
+  lapacke_dtbcon_work.c 
+  lapacke_dtbrfs.c 
+  lapacke_dtbrfs_work.c 
+  lapacke_dtbtrs.c 
+  lapacke_dtbtrs_work.c 
+  lapacke_dtfsm.c 
+  lapacke_dtfsm_work.c 
+  lapacke_dtftri.c 
+  lapacke_dtftri_work.c 
+  lapacke_dtfttp.c 
+  lapacke_dtfttp_work.c 
+  lapacke_dtfttr.c 
+  lapacke_dtfttr_work.c 
+  lapacke_dtgevc.c 
+  lapacke_dtgevc_work.c 
+  lapacke_dtgexc.c 
+  lapacke_dtgexc_work.c 
+  lapacke_dtgsen.c 
+  lapacke_dtgsen_work.c 
+  lapacke_dtgsja.c 
+  lapacke_dtgsja_work.c 
+  lapacke_dtgsna.c 
+  lapacke_dtgsna_work.c 
+  lapacke_dtgsyl.c 
+  lapacke_dtgsyl_work.c 
+  lapacke_dtpcon.c 
+  lapacke_dtpcon_work.c 
+  lapacke_dtpmqrt.c 
+  lapacke_dtpmqrt_work.c 
+  lapacke_dtpqrt.c 
+  lapacke_dtpqrt2.c 
+  lapacke_dtpqrt2_work.c 
+  lapacke_dtpqrt_work.c 
+  lapacke_dtprfb.c 
+  lapacke_dtprfb_work.c 
+  lapacke_dtprfs.c 
+  lapacke_dtprfs_work.c 
+  lapacke_dtptri.c 
+  lapacke_dtptri_work.c 
+  lapacke_dtptrs.c 
+  lapacke_dtptrs_work.c 
+  lapacke_dtpttf.c 
+  lapacke_dtpttf_work.c 
+  lapacke_dtpttr.c 
+  lapacke_dtpttr_work.c 
+  lapacke_dtrcon.c 
+  lapacke_dtrcon_work.c 
+  lapacke_dtrevc.c 
+  lapacke_dtrevc_work.c 
+  lapacke_dtrexc.c 
+  lapacke_dtrexc_work.c 
+  lapacke_dtrrfs.c 
+  lapacke_dtrrfs_work.c 
+  lapacke_dtrsen.c 
+  lapacke_dtrsen_work.c 
+  lapacke_dtrsna.c 
+  lapacke_dtrsna_work.c 
+  lapacke_dtrsyl.c 
+  lapacke_dtrsyl_work.c 
+  lapacke_dtrtri.c 
+  lapacke_dtrtri_work.c 
+  lapacke_dtrtrs.c 
+  lapacke_dtrtrs_work.c 
+  lapacke_dtrttf.c 
+  lapacke_dtrttf_work.c 
+  lapacke_dtrttp.c 
+  lapacke_dtrttp_work.c 
+  lapacke_dtzrzf.c 
+  lapacke_dtzrzf_work.c
+)
+
+set(SSRC
+  lapacke_sbbcsd.c 
+  lapacke_sbbcsd_work.c 
+  lapacke_sbdsdc.c 
+  lapacke_sbdsdc_work.c 
+  lapacke_sbdsqr.c 
+  lapacke_sbdsqr_work.c 
+  lapacke_sdisna.c 
+  lapacke_sdisna_work.c 
+  lapacke_sgbbrd.c 
+  lapacke_sgbbrd_work.c 
+  lapacke_sgbcon.c 
+  lapacke_sgbcon_work.c 
+  lapacke_sgbequ.c 
+  lapacke_sgbequ_work.c 
+  lapacke_sgbequb.c 
+  lapacke_sgbequb_work.c 
+  lapacke_sgbrfs.c 
+  lapacke_sgbrfs_work.c 
+  lapacke_sgbsv.c 
+  lapacke_sgbsv_work.c 
+  lapacke_sgbsvx.c 
+  lapacke_sgbsvx_work.c 
+  lapacke_sgbtrf.c 
+  lapacke_sgbtrf_work.c 
+  lapacke_sgbtrs.c 
+  lapacke_sgbtrs_work.c 
+  lapacke_sgebak.c 
+  lapacke_sgebak_work.c 
+  lapacke_sgebal.c 
+  lapacke_sgebal_work.c 
+  lapacke_sgebrd.c 
+  lapacke_sgebrd_work.c 
+  lapacke_sgecon.c 
+  lapacke_sgecon_work.c 
+  lapacke_sgeequ.c 
+  lapacke_sgeequ_work.c 
+  lapacke_sgeequb.c 
+  lapacke_sgeequb_work.c 
+  lapacke_sgees.c 
+  lapacke_sgees_work.c 
+  lapacke_sgeesx.c 
+  lapacke_sgeesx_work.c 
+  lapacke_sgeev.c 
+  lapacke_sgeev_work.c 
+  lapacke_sgeevx.c 
+  lapacke_sgeevx_work.c 
+  lapacke_sgehrd.c 
+  lapacke_sgehrd_work.c 
+  lapacke_sgejsv.c 
+  lapacke_sgejsv_work.c 
+  lapacke_sgelq2.c 
+  lapacke_sgelq2_work.c 
+  lapacke_sgelqf.c 
+  lapacke_sgelqf_work.c 
+  lapacke_sgels.c 
+  lapacke_sgels_work.c 
+  lapacke_sgelsd.c 
+  lapacke_sgelsd_work.c 
+  lapacke_sgelss.c 
+  lapacke_sgelss_work.c 
+  lapacke_sgelsy.c 
+  lapacke_sgelsy_work.c 
+  lapacke_sgemqrt.c 
+  lapacke_sgemqrt_work.c 
+  lapacke_sgeqlf.c 
+  lapacke_sgeqlf_work.c 
+  lapacke_sgeqp3.c 
+  lapacke_sgeqp3_work.c 
+  lapacke_sgeqpf.c 
+  lapacke_sgeqpf_work.c 
+  lapacke_sgeqr2.c 
+  lapacke_sgeqr2_work.c 
+  lapacke_sgeqrf.c 
+  lapacke_sgeqrf_work.c 
+  lapacke_sgeqrfp.c 
+  lapacke_sgeqrfp_work.c 
+  lapacke_sgeqrt.c 
+  lapacke_sgeqrt2.c 
+  lapacke_sgeqrt2_work.c 
+  lapacke_sgeqrt3.c 
+  lapacke_sgeqrt3_work.c 
+  lapacke_sgeqrt_work.c 
+  lapacke_sgerfs.c 
+  lapacke_sgerfs_work.c 
+  lapacke_sgerqf.c 
+  lapacke_sgerqf_work.c 
+  lapacke_sgesdd.c 
+  lapacke_sgesdd_work.c 
+  lapacke_sgesv.c 
+  lapacke_sgesv_work.c 
+  lapacke_sgesvd.c 
+  lapacke_sgesvd_work.c 
+  lapacke_sgesvj.c 
+  lapacke_sgesvj_work.c 
+  lapacke_sgesvx.c 
+  lapacke_sgesvx_work.c 
+  lapacke_sgetf2.c 
+  lapacke_sgetf2_work.c 
+  lapacke_sgetrf.c 
+  lapacke_sgetrf_work.c 
+  lapacke_sgetri.c 
+  lapacke_sgetri_work.c 
+  lapacke_sgetrs.c 
+  lapacke_sgetrs_work.c 
+  lapacke_sggbak.c 
+  lapacke_sggbak_work.c 
+  lapacke_sggbal.c 
+  lapacke_sggbal_work.c 
+  lapacke_sgges.c 
+  lapacke_sgges_work.c 
+  lapacke_sggesx.c 
+  lapacke_sggesx_work.c 
+  lapacke_sggev.c 
+  lapacke_sggev_work.c 
+  lapacke_sggevx.c 
+  lapacke_sggevx_work.c 
+  lapacke_sggglm.c 
+  lapacke_sggglm_work.c 
+  lapacke_sgghrd.c 
+  lapacke_sgghrd_work.c 
+  lapacke_sgglse.c 
+  lapacke_sgglse_work.c 
+  lapacke_sggqrf.c 
+  lapacke_sggqrf_work.c 
+  lapacke_sggrqf.c 
+  lapacke_sggrqf_work.c 
+  lapacke_sggsvd.c 
+  lapacke_sggsvd_work.c 
+  lapacke_sggsvp.c 
+  lapacke_sggsvp_work.c 
+  lapacke_sgtcon.c 
+  lapacke_sgtcon_work.c 
+  lapacke_sgtrfs.c 
+  lapacke_sgtrfs_work.c 
+  lapacke_sgtsv.c 
+  lapacke_sgtsv_work.c 
+  lapacke_sgtsvx.c 
+  lapacke_sgtsvx_work.c 
+  lapacke_sgttrf.c 
+  lapacke_sgttrf_work.c 
+  lapacke_sgttrs.c 
+  lapacke_sgttrs_work.c 
+  lapacke_shgeqz.c 
+  lapacke_shgeqz_work.c 
+  lapacke_shsein.c 
+  lapacke_shsein_work.c 
+  lapacke_shseqr.c 
+  lapacke_shseqr_work.c 
+  lapacke_slacn2.c 
+  lapacke_slacn2_work.c 
+  lapacke_slacpy.c 
+  lapacke_slacpy_work.c 
+  lapacke_slag2d.c 
+  lapacke_slag2d_work.c 
+  lapacke_slamch.c 
+  lapacke_slamch_work.c 
+  lapacke_slange.c 
+  lapacke_slange_work.c 
+  lapacke_slansy.c 
+  lapacke_slansy_work.c 
+  lapacke_slantr.c 
+  lapacke_slantr_work.c 
+  lapacke_slapmr.c 
+  lapacke_slapmr_work.c 
+  lapacke_slapy2.c 
+  lapacke_slapy2_work.c 
+  lapacke_slapy3.c 
+  lapacke_slapy3_work.c 
+  lapacke_slarfb.c 
+  lapacke_slarfb_work.c 
+  lapacke_slarfg.c 
+  lapacke_slarfg_work.c 
+  lapacke_slarft.c 
+  lapacke_slarft_work.c 
+  lapacke_slarfx.c 
+  lapacke_slarfx_work.c 
+  lapacke_slarnv.c 
+  lapacke_slarnv_work.c 
+  lapacke_slartgp.c 
+  lapacke_slartgp_work.c 
+  lapacke_slartgs.c 
+  lapacke_slartgs_work.c 
+  lapacke_slaset.c 
+  lapacke_slaset_work.c 
+  lapacke_slasrt.c 
+  lapacke_slasrt_work.c 
+  lapacke_slaswp.c 
+  lapacke_slaswp_work.c 
+  lapacke_slauum.c 
+  lapacke_slauum_work.c 
+  lapacke_sopgtr.c 
+  lapacke_sopgtr_work.c 
+  lapacke_sopmtr.c 
+  lapacke_sopmtr_work.c 
+  lapacke_sorbdb.c 
+  lapacke_sorbdb_work.c 
+  lapacke_sorcsd.c 
+  lapacke_sorcsd_work.c 
+  lapacke_sorgbr.c 
+  lapacke_sorgbr_work.c 
+  lapacke_sorghr.c 
+  lapacke_sorghr_work.c 
+  lapacke_sorglq.c 
+  lapacke_sorglq_work.c 
+  lapacke_sorgql.c 
+  lapacke_sorgql_work.c 
+  lapacke_sorgqr.c 
+  lapacke_sorgqr_work.c 
+  lapacke_sorgrq.c 
+  lapacke_sorgrq_work.c 
+  lapacke_sorgtr.c 
+  lapacke_sorgtr_work.c 
+  lapacke_sormbr.c 
+  lapacke_sormbr_work.c 
+  lapacke_sormhr.c 
+  lapacke_sormhr_work.c 
+  lapacke_sormlq.c 
+  lapacke_sormlq_work.c 
+  lapacke_sormql.c 
+  lapacke_sormql_work.c 
+  lapacke_sormqr.c 
+  lapacke_sormqr_work.c 
+  lapacke_sormrq.c 
+  lapacke_sormrq_work.c 
+  lapacke_sormrz.c 
+  lapacke_sormrz_work.c 
+  lapacke_sormtr.c 
+  lapacke_sormtr_work.c 
+  lapacke_spbcon.c 
+  lapacke_spbcon_work.c 
+  lapacke_spbequ.c 
+  lapacke_spbequ_work.c 
+  lapacke_spbrfs.c 
+  lapacke_spbrfs_work.c 
+  lapacke_spbstf.c 
+  lapacke_spbstf_work.c 
+  lapacke_spbsv.c 
+  lapacke_spbsv_work.c 
+  lapacke_spbsvx.c 
+  lapacke_spbsvx_work.c 
+  lapacke_spbtrf.c 
+  lapacke_spbtrf_work.c 
+  lapacke_spbtrs.c 
+  lapacke_spbtrs_work.c 
+  lapacke_spftrf.c 
+  lapacke_spftrf_work.c 
+  lapacke_spftri.c 
+  lapacke_spftri_work.c 
+  lapacke_spftrs.c 
+  lapacke_spftrs_work.c 
+  lapacke_spocon.c 
+  lapacke_spocon_work.c 
+  lapacke_spoequ.c 
+  lapacke_spoequ_work.c 
+  lapacke_spoequb.c 
+  lapacke_spoequb_work.c 
+  lapacke_sporfs.c 
+  lapacke_sporfs_work.c 
+  lapacke_sposv.c 
+  lapacke_sposv_work.c 
+  lapacke_sposvx.c 
+  lapacke_sposvx_work.c 
+  lapacke_spotrf.c 
+  lapacke_spotrf_work.c 
+  lapacke_spotri.c 
+  lapacke_spotri_work.c 
+  lapacke_spotrs.c 
+  lapacke_spotrs_work.c 
+  lapacke_sppcon.c 
+  lapacke_sppcon_work.c 
+  lapacke_sppequ.c 
+  lapacke_sppequ_work.c 
+  lapacke_spprfs.c 
+  lapacke_spprfs_work.c 
+  lapacke_sppsv.c 
+  lapacke_sppsv_work.c 
+  lapacke_sppsvx.c 
+  lapacke_sppsvx_work.c 
+  lapacke_spptrf.c 
+  lapacke_spptrf_work.c 
+  lapacke_spptri.c 
+  lapacke_spptri_work.c 
+  lapacke_spptrs.c 
+  lapacke_spptrs_work.c 
+  lapacke_spstrf.c 
+  lapacke_spstrf_work.c 
+  lapacke_sptcon.c 
+  lapacke_sptcon_work.c 
+  lapacke_spteqr.c 
+  lapacke_spteqr_work.c 
+  lapacke_sptrfs.c 
+  lapacke_sptrfs_work.c 
+  lapacke_sptsv.c 
+  lapacke_sptsv_work.c 
+  lapacke_sptsvx.c 
+  lapacke_sptsvx_work.c 
+  lapacke_spttrf.c 
+  lapacke_spttrf_work.c 
+  lapacke_spttrs.c 
+  lapacke_spttrs_work.c 
+  lapacke_ssbev.c 
+  lapacke_ssbev_work.c 
+  lapacke_ssbevd.c 
+  lapacke_ssbevd_work.c 
+  lapacke_ssbevx.c 
+  lapacke_ssbevx_work.c 
+  lapacke_ssbgst.c 
+  lapacke_ssbgst_work.c 
+  lapacke_ssbgv.c 
+  lapacke_ssbgv_work.c 
+  lapacke_ssbgvd.c 
+  lapacke_ssbgvd_work.c 
+  lapacke_ssbgvx.c 
+  lapacke_ssbgvx_work.c 
+  lapacke_ssbtrd.c 
+  lapacke_ssbtrd_work.c 
+  lapacke_ssfrk.c 
+  lapacke_ssfrk_work.c 
+  lapacke_sspcon.c 
+  lapacke_sspcon_work.c 
+  lapacke_sspev.c 
+  lapacke_sspev_work.c 
+  lapacke_sspevd.c 
+  lapacke_sspevd_work.c 
+  lapacke_sspevx.c 
+  lapacke_sspevx_work.c 
+  lapacke_sspgst.c 
+  lapacke_sspgst_work.c 
+  lapacke_sspgv.c 
+  lapacke_sspgv_work.c 
+  lapacke_sspgvd.c 
+  lapacke_sspgvd_work.c 
+  lapacke_sspgvx.c 
+  lapacke_sspgvx_work.c 
+  lapacke_ssprfs.c 
+  lapacke_ssprfs_work.c 
+  lapacke_sspsv.c 
+  lapacke_sspsv_work.c 
+  lapacke_sspsvx.c 
+  lapacke_sspsvx_work.c 
+  lapacke_ssptrd.c 
+  lapacke_ssptrd_work.c 
+  lapacke_ssptrf.c 
+  lapacke_ssptrf_work.c 
+  lapacke_ssptri.c 
+  lapacke_ssptri_work.c 
+  lapacke_ssptrs.c 
+  lapacke_ssptrs_work.c 
+  lapacke_sstebz.c 
+  lapacke_sstebz_work.c 
+  lapacke_sstedc.c 
+  lapacke_sstedc_work.c 
+  lapacke_sstegr.c 
+  lapacke_sstegr_work.c 
+  lapacke_sstein.c 
+  lapacke_sstein_work.c 
+  lapacke_sstemr.c 
+  lapacke_sstemr_work.c 
+  lapacke_ssteqr.c 
+  lapacke_ssteqr_work.c 
+  lapacke_ssterf.c 
+  lapacke_ssterf_work.c 
+  lapacke_sstev.c 
+  lapacke_sstev_work.c 
+  lapacke_sstevd.c 
+  lapacke_sstevd_work.c 
+  lapacke_sstevr.c 
+  lapacke_sstevr_work.c 
+  lapacke_sstevx.c 
+  lapacke_sstevx_work.c 
+  lapacke_ssycon.c 
+  lapacke_ssycon_work.c 
+  lapacke_ssyconv.c 
+  lapacke_ssyconv_work.c 
+  lapacke_ssyequb.c 
+  lapacke_ssyequb_work.c 
+  lapacke_ssyev.c 
+  lapacke_ssyev_work.c 
+  lapacke_ssyevd.c 
+  lapacke_ssyevd_work.c 
+  lapacke_ssyevr.c 
+  lapacke_ssyevr_work.c 
+  lapacke_ssyevx.c 
+  lapacke_ssyevx_work.c 
+  lapacke_ssygst.c 
+  lapacke_ssygst_work.c 
+  lapacke_ssygv.c 
+  lapacke_ssygv_work.c 
+  lapacke_ssygvd.c 
+  lapacke_ssygvd_work.c 
+  lapacke_ssygvx.c 
+  lapacke_ssygvx_work.c 
+  lapacke_ssyrfs.c 
+  lapacke_ssyrfs_work.c 
+  lapacke_ssysv.c 
+  lapacke_ssysv_rook.c 
+  lapacke_ssysv_rook_work.c 
+  lapacke_ssysv_work.c 
+  lapacke_ssysvx.c 
+  lapacke_ssysvx_work.c 
+  lapacke_ssyswapr.c 
+  lapacke_ssyswapr_work.c 
+  lapacke_ssytrd.c 
+  lapacke_ssytrd_work.c 
+  lapacke_ssytrf.c 
+  lapacke_ssytrf_work.c 
+  lapacke_ssytri.c 
+  lapacke_ssytri2.c 
+  lapacke_ssytri2_work.c 
+  lapacke_ssytri2x.c 
+  lapacke_ssytri2x_work.c 
+  lapacke_ssytri_work.c 
+  lapacke_ssytrs.c 
+  lapacke_ssytrs2.c 
+  lapacke_ssytrs2_work.c 
+  lapacke_ssytrs_work.c 
+  lapacke_stbcon.c 
+  lapacke_stbcon_work.c 
+  lapacke_stbrfs.c 
+  lapacke_stbrfs_work.c 
+  lapacke_stbtrs.c 
+  lapacke_stbtrs_work.c 
+  lapacke_stfsm.c 
+  lapacke_stfsm_work.c 
+  lapacke_stftri.c 
+  lapacke_stftri_work.c 
+  lapacke_stfttp.c 
+  lapacke_stfttp_work.c 
+  lapacke_stfttr.c 
+  lapacke_stfttr_work.c 
+  lapacke_stgevc.c 
+  lapacke_stgevc_work.c 
+  lapacke_stgexc.c 
+  lapacke_stgexc_work.c 
+  lapacke_stgsen.c 
+  lapacke_stgsen_work.c 
+  lapacke_stgsja.c 
+  lapacke_stgsja_work.c 
+  lapacke_stgsna.c 
+  lapacke_stgsna_work.c 
+  lapacke_stgsyl.c 
+  lapacke_stgsyl_work.c 
+  lapacke_stpcon.c 
+  lapacke_stpcon_work.c 
+  lapacke_stpmqrt.c 
+  lapacke_stpmqrt_work.c 
+  lapacke_stpqrt2.c 
+  lapacke_stpqrt2_work.c 
+  lapacke_stprfb.c 
+  lapacke_stprfb_work.c 
+  lapacke_stprfs.c 
+  lapacke_stprfs_work.c 
+  lapacke_stptri.c 
+  lapacke_stptri_work.c 
+  lapacke_stptrs.c 
+  lapacke_stptrs_work.c 
+  lapacke_stpttf.c 
+  lapacke_stpttf_work.c 
+  lapacke_stpttr.c 
+  lapacke_stpttr_work.c 
+  lapacke_strcon.c 
+  lapacke_strcon_work.c 
+  lapacke_strevc.c 
+  lapacke_strevc_work.c 
+  lapacke_strexc.c 
+  lapacke_strexc_work.c 
+  lapacke_strrfs.c 
+  lapacke_strrfs_work.c 
+  lapacke_strsen.c 
+  lapacke_strsen_work.c 
+  lapacke_strsna.c 
+  lapacke_strsna_work.c 
+  lapacke_strsyl.c 
+  lapacke_strsyl_work.c 
+  lapacke_strtri.c 
+  lapacke_strtri_work.c 
+  lapacke_strtrs.c 
+  lapacke_strtrs_work.c 
+  lapacke_strttf.c 
+  lapacke_strttf_work.c 
+  lapacke_strttp.c 
+  lapacke_strttp_work.c 
+  lapacke_stzrzf.c 
+  lapacke_stzrzf_work.c
+)
+
+set(ZSRC
+  lapacke_zbbcsd.c 
+  lapacke_zbbcsd_work.c 
+  lapacke_zbdsqr.c 
+  lapacke_zbdsqr_work.c 
+  lapacke_zcgesv.c 
+  lapacke_zcgesv_work.c 
+  lapacke_zcposv.c 
+  lapacke_zcposv_work.c 
+  lapacke_zgbbrd.c 
+  lapacke_zgbbrd_work.c 
+  lapacke_zgbcon.c 
+  lapacke_zgbcon_work.c 
+  lapacke_zgbequ.c 
+  lapacke_zgbequ_work.c 
+  lapacke_zgbequb.c 
+  lapacke_zgbequb_work.c 
+  lapacke_zgbrfs.c 
+  lapacke_zgbrfs_work.c 
+  lapacke_zgbsv.c 
+  lapacke_zgbsv_work.c 
+  lapacke_zgbsvx.c 
+  lapacke_zgbsvx_work.c 
+  lapacke_zgbtrf.c 
+  lapacke_zgbtrf_work.c 
+  lapacke_zgbtrs.c 
+  lapacke_zgbtrs_work.c 
+  lapacke_zgebak.c 
+  lapacke_zgebak_work.c 
+  lapacke_zgebal.c 
+  lapacke_zgebal_work.c 
+  lapacke_zgebrd.c 
+  lapacke_zgebrd_work.c 
+  lapacke_zgecon.c 
+  lapacke_zgecon_work.c 
+  lapacke_zgeequ.c 
+  lapacke_zgeequ_work.c 
+  lapacke_zgeequb.c 
+  lapacke_zgeequb_work.c 
+  lapacke_zgees.c 
+  lapacke_zgees_work.c 
+  lapacke_zgeesx.c 
+  lapacke_zgeesx_work.c 
+  lapacke_zgeev.c 
+  lapacke_zgeev_work.c 
+  lapacke_zgeevx.c 
+  lapacke_zgeevx_work.c 
+  lapacke_zgehrd.c 
+  lapacke_zgehrd_work.c 
+  lapacke_zgelq2.c 
+  lapacke_zgelq2_work.c 
+  lapacke_zgelqf.c 
+  lapacke_zgelqf_work.c 
+  lapacke_zgels.c 
+  lapacke_zgels_work.c 
+  lapacke_zgelsd.c 
+  lapacke_zgelsd_work.c 
+  lapacke_zgelss.c 
+  lapacke_zgelss_work.c 
+  lapacke_zgelsy.c 
+  lapacke_zgelsy_work.c 
+  lapacke_zgemqrt.c 
+  lapacke_zgemqrt_work.c 
+  lapacke_zgeqlf.c 
+  lapacke_zgeqlf_work.c 
+  lapacke_zgeqp3.c 
+  lapacke_zgeqp3_work.c 
+  lapacke_zgeqpf.c 
+  lapacke_zgeqpf_work.c 
+  lapacke_zgeqr2.c 
+  lapacke_zgeqr2_work.c 
+  lapacke_zgeqrf.c 
+  lapacke_zgeqrf_work.c 
+  lapacke_zgeqrfp.c 
+  lapacke_zgeqrfp_work.c 
+  lapacke_zgeqrt.c 
+  lapacke_zgeqrt2.c 
+  lapacke_zgeqrt2_work.c 
+  lapacke_zgeqrt3.c 
+  lapacke_zgeqrt3_work.c 
+  lapacke_zgeqrt_work.c 
+  lapacke_zgerfs.c 
+  lapacke_zgerfs_work.c 
+  lapacke_zgerqf.c 
+  lapacke_zgerqf_work.c 
+  lapacke_zgesdd.c 
+  lapacke_zgesdd_work.c 
+  lapacke_zgesv.c 
+  lapacke_zgesv_work.c 
+  lapacke_zgesvd.c 
+  lapacke_zgesvd_work.c 
+  lapacke_zgesvx.c 
+  lapacke_zgesvx_work.c 
+  lapacke_zgetf2.c 
+  lapacke_zgetf2_work.c 
+  lapacke_zgetrf.c 
+  lapacke_zgetrf_work.c 
+  lapacke_zgetri.c 
+  lapacke_zgetri_work.c 
+  lapacke_zgetrs.c 
+  lapacke_zgetrs_work.c 
+  lapacke_zggbak.c 
+  lapacke_zggbak_work.c 
+  lapacke_zggbal.c 
+  lapacke_zggbal_work.c 
+  lapacke_zgges.c 
+  lapacke_zgges_work.c 
+  lapacke_zggesx.c 
+  lapacke_zggesx_work.c 
+  lapacke_zggev.c 
+  lapacke_zggev_work.c 
+  lapacke_zggevx.c 
+  lapacke_zggevx_work.c 
+  lapacke_zggglm.c 
+  lapacke_zggglm_work.c 
+  lapacke_zgghrd.c 
+  lapacke_zgghrd_work.c 
+  lapacke_zgglse.c 
+  lapacke_zgglse_work.c 
+  lapacke_zggqrf.c 
+  lapacke_zggqrf_work.c 
+  lapacke_zggrqf.c 
+  lapacke_zggrqf_work.c 
+  lapacke_zggsvd.c 
+  lapacke_zggsvd_work.c 
+  lapacke_zggsvp.c 
+  lapacke_zggsvp_work.c 
+  lapacke_zgtcon.c 
+  lapacke_zgtcon_work.c 
+  lapacke_zgtrfs.c 
+  lapacke_zgtrfs_work.c 
+  lapacke_zgtsv.c 
+  lapacke_zgtsv_work.c 
+  lapacke_zgtsvx.c 
+  lapacke_zgtsvx_work.c 
+  lapacke_zgttrf.c 
+  lapacke_zgttrf_work.c 
+  lapacke_zgttrs.c 
+  lapacke_zgttrs_work.c 
+  lapacke_zhbev.c 
+  lapacke_zhbev_work.c 
+  lapacke_zhbevd.c 
+  lapacke_zhbevd_work.c 
+  lapacke_zhbevx.c 
+  lapacke_zhbevx_work.c 
+  lapacke_zhbgst.c 
+  lapacke_zhbgst_work.c 
+  lapacke_zhbgv.c 
+  lapacke_zhbgv_work.c 
+  lapacke_zhbgvd.c 
+  lapacke_zhbgvd_work.c 
+  lapacke_zhbgvx.c 
+  lapacke_zhbgvx_work.c 
+  lapacke_zhbtrd.c 
+  lapacke_zhbtrd_work.c 
+  lapacke_zhecon.c 
+  lapacke_zhecon_work.c 
+  lapacke_zheequb.c 
+  lapacke_zheequb_work.c 
+  lapacke_zheev.c 
+  lapacke_zheev_work.c 
+  lapacke_zheevd.c 
+  lapacke_zheevd_work.c 
+  lapacke_zheevr.c 
+  lapacke_zheevr_work.c 
+  lapacke_zheevx.c 
+  lapacke_zheevx_work.c 
+  lapacke_zhegst.c 
+  lapacke_zhegst_work.c 
+  lapacke_zhegv.c 
+  lapacke_zhegv_work.c 
+  lapacke_zhegvd.c 
+  lapacke_zhegvd_work.c 
+  lapacke_zhegvx.c 
+  lapacke_zhegvx_work.c 
+  lapacke_zherfs.c 
+  lapacke_zherfs_work.c 
+  lapacke_zhesv.c 
+  lapacke_zhesv_work.c 
+  lapacke_zhesvx.c 
+  lapacke_zhesvx_work.c 
+  lapacke_zheswapr.c 
+  lapacke_zheswapr_work.c 
+  lapacke_zhetrd.c 
+  lapacke_zhetrd_work.c 
+  lapacke_zhetrf.c 
+  lapacke_zhetrf_work.c 
+  lapacke_zhetri.c 
+  lapacke_zhetri2.c 
+  lapacke_zhetri2_work.c 
+  lapacke_zhetri2x.c 
+  lapacke_zhetri2x_work.c 
+  lapacke_zhetri_work.c 
+  lapacke_zhetrs.c 
+  lapacke_zhetrs2.c 
+  lapacke_zhetrs2_work.c 
+  lapacke_zhetrs_work.c 
+  lapacke_zhfrk.c 
+  lapacke_zhfrk_work.c 
+  lapacke_zhgeqz.c 
+  lapacke_zhgeqz_work.c 
+  lapacke_zhpcon.c 
+  lapacke_zhpcon_work.c 
+  lapacke_zhpev.c 
+  lapacke_zhpev_work.c 
+  lapacke_zhpevd.c 
+  lapacke_zhpevd_work.c 
+  lapacke_zhpevx.c 
+  lapacke_zhpevx_work.c 
+  lapacke_zhpgst.c 
+  lapacke_zhpgst_work.c 
+  lapacke_zhpgv.c 
+  lapacke_zhpgv_work.c 
+  lapacke_zhpgvd.c 
+  lapacke_zhpgvd_work.c 
+  lapacke_zhpgvx.c 
+  lapacke_zhpgvx_work.c 
+  lapacke_zhprfs.c 
+  lapacke_zhprfs_work.c 
+  lapacke_zhpsv.c 
+  lapacke_zhpsv_work.c 
+  lapacke_zhpsvx.c 
+  lapacke_zhpsvx_work.c 
+  lapacke_zhptrd.c 
+  lapacke_zhptrd_work.c 
+  lapacke_zhptrf.c 
+  lapacke_zhptrf_work.c 
+  lapacke_zhptri.c 
+  lapacke_zhptri_work.c 
+  lapacke_zhptrs.c 
+  lapacke_zhptrs_work.c 
+  lapacke_zhsein.c 
+  lapacke_zhsein_work.c 
+  lapacke_zhseqr.c 
+  lapacke_zhseqr_work.c 
+  lapacke_zlacgv.c 
+  lapacke_zlacgv_work.c 
+  lapacke_zlacn2.c 
+  lapacke_zlacn2_work.c 
+  lapacke_zlacp2.c 
+  lapacke_zlacp2_work.c 
+  lapacke_zlacpy.c 
+  lapacke_zlacpy_work.c 
+  lapacke_zlag2c.c 
+  lapacke_zlag2c_work.c 
+  lapacke_zlange.c 
+  lapacke_zlange_work.c 
+  lapacke_zlanhe.c 
+  lapacke_zlanhe_work.c 
+  lapacke_zlansy.c 
+  lapacke_zlansy_work.c 
+  lapacke_zlantr.c 
+  lapacke_zlantr_work.c 
+  lapacke_zlapmr.c 
+  lapacke_zlapmr_work.c 
+  lapacke_zlarfb.c 
+  lapacke_zlarfb_work.c 
+  lapacke_zlarfg.c 
+  lapacke_zlarfg_work.c 
+  lapacke_zlarft.c 
+  lapacke_zlarft_work.c 
+  lapacke_zlarfx.c 
+  lapacke_zlarfx_work.c 
+  lapacke_zlarnv.c 
+  lapacke_zlarnv_work.c 
+  lapacke_zlaset.c 
+  lapacke_zlaset_work.c 
+  lapacke_zlaswp.c 
+  lapacke_zlaswp_work.c 
+  lapacke_zlauum.c 
+  lapacke_zlauum_work.c 
+  lapacke_zpbcon.c 
+  lapacke_zpbcon_work.c 
+  lapacke_zpbequ.c 
+  lapacke_zpbequ_work.c 
+  lapacke_zpbrfs.c 
+  lapacke_zpbrfs_work.c 
+  lapacke_zpbstf.c 
+  lapacke_zpbstf_work.c 
+  lapacke_zpbsv.c 
+  lapacke_zpbsv_work.c 
+  lapacke_zpbsvx.c 
+  lapacke_zpbsvx_work.c 
+  lapacke_zpbtrf.c 
+  lapacke_zpbtrf_work.c 
+  lapacke_zpbtrs.c 
+  lapacke_zpbtrs_work.c 
+  lapacke_zpftrf.c 
+  lapacke_zpftrf_work.c 
+  lapacke_zpftri.c 
+  lapacke_zpftri_work.c 
+  lapacke_zpftrs.c 
+  lapacke_zpftrs_work.c 
+  lapacke_zpocon.c 
+  lapacke_zpocon_work.c 
+  lapacke_zpoequ.c 
+  lapacke_zpoequ_work.c 
+  lapacke_zpoequb.c 
+  lapacke_zpoequb_work.c 
+  lapacke_zporfs.c 
+  lapacke_zporfs_work.c 
+  lapacke_zposv.c 
+  lapacke_zposv_work.c 
+  lapacke_zposvx.c 
+  lapacke_zposvx_work.c 
+  lapacke_zpotrf.c 
+  lapacke_zpotrf_work.c 
+  lapacke_zpotri.c 
+  lapacke_zpotri_work.c 
+  lapacke_zpotrs.c 
+  lapacke_zpotrs_work.c 
+  lapacke_zppcon.c 
+  lapacke_zppcon_work.c 
+  lapacke_zppequ.c 
+  lapacke_zppequ_work.c 
+  lapacke_zpprfs.c 
+  lapacke_zpprfs_work.c 
+  lapacke_zppsv.c 
+  lapacke_zppsv_work.c 
+  lapacke_zppsvx.c 
+  lapacke_zppsvx_work.c 
+  lapacke_zpptrf.c 
+  lapacke_zpptrf_work.c 
+  lapacke_zpptri.c 
+  lapacke_zpptri_work.c 
+  lapacke_zpptrs.c 
+  lapacke_zpptrs_work.c 
+  lapacke_zpstrf.c 
+  lapacke_zpstrf_work.c 
+  lapacke_zptcon.c 
+  lapacke_zptcon_work.c 
+  lapacke_zpteqr.c 
+  lapacke_zpteqr_work.c 
+  lapacke_zptrfs.c 
+  lapacke_zptrfs_work.c 
+  lapacke_zptsv.c 
+  lapacke_zptsv_work.c 
+  lapacke_zptsvx.c 
+  lapacke_zptsvx_work.c 
+  lapacke_zpttrf.c 
+  lapacke_zpttrf_work.c 
+  lapacke_zpttrs.c 
+  lapacke_zpttrs_work.c 
+  lapacke_zspcon.c 
+  lapacke_zspcon_work.c 
+  lapacke_zsprfs.c 
+  lapacke_zsprfs_work.c 
+  lapacke_zspsv.c 
+  lapacke_zspsv_work.c 
+  lapacke_zspsvx.c 
+  lapacke_zspsvx_work.c 
+  lapacke_zsptrf.c 
+  lapacke_zsptrf_work.c 
+  lapacke_zsptri.c 
+  lapacke_zsptri_work.c 
+  lapacke_zsptrs.c 
+  lapacke_zsptrs_work.c 
+  lapacke_zstedc.c 
+  lapacke_zstedc_work.c 
+  lapacke_zstegr.c 
+  lapacke_zstegr_work.c 
+  lapacke_zstein.c 
+  lapacke_zstein_work.c 
+  lapacke_zstemr.c 
+  lapacke_zstemr_work.c 
+  lapacke_zsteqr.c 
+  lapacke_zsteqr_work.c 
+  lapacke_zsycon.c 
+  lapacke_zsycon_work.c 
+  lapacke_zsyconv.c 
+  lapacke_zsyconv_work.c 
+  lapacke_zsyequb.c 
+  lapacke_zsyequb_work.c 
+  lapacke_zsyrfs.c 
+  lapacke_zsyrfs_work.c 
+  lapacke_zsysv.c 
+  lapacke_zsysv_rook.c 
+  lapacke_zsysv_rook_work.c 
+  lapacke_zsysv_work.c 
+  lapacke_zsysvx.c 
+  lapacke_zsysvx_work.c 
+  lapacke_zsyswapr.c 
+  lapacke_zsyswapr_work.c 
+  lapacke_zsytrf.c 
+  lapacke_zsytrf_work.c 
+  lapacke_zsytri.c 
+  lapacke_zsytri2.c 
+  lapacke_zsytri2_work.c 
+  lapacke_zsytri2x.c 
+  lapacke_zsytri2x_work.c 
+  lapacke_zsytri_work.c 
+  lapacke_zsytrs.c 
+  lapacke_zsytrs2.c 
+  lapacke_zsytrs2_work.c 
+  lapacke_zsytrs_work.c 
+  lapacke_ztbcon.c 
+  lapacke_ztbcon_work.c 
+  lapacke_ztbrfs.c 
+  lapacke_ztbrfs_work.c 
+  lapacke_ztbtrs.c 
+  lapacke_ztbtrs_work.c 
+  lapacke_ztfsm.c 
+  lapacke_ztfsm_work.c 
+  lapacke_ztftri.c 
+  lapacke_ztftri_work.c 
+  lapacke_ztfttp.c 
+  lapacke_ztfttp_work.c 
+  lapacke_ztfttr.c 
+  lapacke_ztfttr_work.c 
+  lapacke_ztgevc.c 
+  lapacke_ztgevc_work.c 
+  lapacke_ztgexc.c 
+  lapacke_ztgexc_work.c 
+  lapacke_ztgsen.c 
+  lapacke_ztgsen_work.c 
+  lapacke_ztgsja.c 
+  lapacke_ztgsja_work.c 
+  lapacke_ztgsna.c 
+  lapacke_ztgsna_work.c 
+  lapacke_ztgsyl.c 
+  lapacke_ztgsyl_work.c 
+  lapacke_ztpcon.c 
+  lapacke_ztpcon_work.c 
+  lapacke_ztpmqrt.c 
+  lapacke_ztpmqrt_work.c 
+  lapacke_ztpqrt.c 
+  lapacke_ztpqrt2.c 
+  lapacke_ztpqrt2_work.c 
+  lapacke_ztpqrt_work.c 
+  lapacke_ztprfb.c 
+  lapacke_ztprfb_work.c 
+  lapacke_ztprfs.c 
+  lapacke_ztprfs_work.c 
+  lapacke_ztptri.c 
+  lapacke_ztptri_work.c 
+  lapacke_ztptrs.c 
+  lapacke_ztptrs_work.c 
+  lapacke_ztpttf.c 
+  lapacke_ztpttf_work.c 
+  lapacke_ztpttr.c 
+  lapacke_ztpttr_work.c 
+  lapacke_ztrcon.c 
+  lapacke_ztrcon_work.c 
+  lapacke_ztrevc.c 
+  lapacke_ztrevc_work.c 
+  lapacke_ztrexc.c 
+  lapacke_ztrexc_work.c 
+  lapacke_ztrrfs.c 
+  lapacke_ztrrfs_work.c 
+  lapacke_ztrsen.c 
+  lapacke_ztrsen_work.c 
+  lapacke_ztrsna.c 
+  lapacke_ztrsna_work.c 
+  lapacke_ztrsyl.c 
+  lapacke_ztrsyl_work.c 
+  lapacke_ztrtri.c 
+  lapacke_ztrtri_work.c 
+  lapacke_ztrtrs.c 
+  lapacke_ztrtrs_work.c 
+  lapacke_ztrttf.c 
+  lapacke_ztrttf_work.c 
+  lapacke_ztrttp.c 
+  lapacke_ztrttp_work.c 
+  lapacke_ztzrzf.c 
+  lapacke_ztzrzf_work.c 
+  lapacke_zunbdb.c 
+  lapacke_zunbdb_work.c 
+  lapacke_zuncsd.c 
+  lapacke_zuncsd_work.c 
+  lapacke_zungbr.c 
+  lapacke_zungbr_work.c 
+  lapacke_zunghr.c 
+  lapacke_zunghr_work.c 
+  lapacke_zunglq.c 
+  lapacke_zunglq_work.c 
+  lapacke_zungql.c 
+  lapacke_zungql_work.c 
+  lapacke_zungqr.c 
+  lapacke_zungqr_work.c 
+  lapacke_zungrq.c 
+  lapacke_zungrq_work.c 
+  lapacke_zungtr.c 
+  lapacke_zungtr_work.c 
+  lapacke_zunmbr.c 
+  lapacke_zunmbr_work.c 
+  lapacke_zunmhr.c 
+  lapacke_zunmhr_work.c 
+  lapacke_zunmlq.c 
+  lapacke_zunmlq_work.c 
+  lapacke_zunmql.c 
+  lapacke_zunmql_work.c 
+  lapacke_zunmqr.c 
+  lapacke_zunmqr_work.c 
+  lapacke_zunmrq.c 
+  lapacke_zunmrq_work.c 
+  lapacke_zunmrz.c 
+  lapacke_zunmrz_work.c 
+  lapacke_zunmtr.c 
+  lapacke_zunmtr_work.c 
+  lapacke_zupgtr.c 
+  lapacke_zupgtr_work.c 
+  lapacke_zupmtr.c 
+  lapacke_zupmtr_work.c 
+  lapacke_zsyr.c 
+  lapacke_csyr.c 
+  lapacke_zsyr_work.c 
+  lapacke_csyr_work.c 
+  lapacke_ilaver.c
+)
+
+set(SRCX
+  lapacke_cgbrfsx.c       lapacke_cporfsx.c       lapacke_dgerfsx.c       lapacke_sgbrfsx.c       lapacke_ssyrfsx.c       lapacke_zherfsx.c 
+  lapacke_cgbrfsx_work.c  lapacke_cporfsx_work.c  lapacke_dgerfsx_work.c  lapacke_sgbrfsx_work.c  lapacke_ssyrfsx_work.c  lapacke_zherfsx_work.c 
+  lapacke_cgerfsx.c       lapacke_csyrfsx.c       lapacke_dporfsx.c       lapacke_sgerfsx.c       lapacke_zgbrfsx.c       lapacke_zporfsx.c 
+  lapacke_cgerfsx_work.c  lapacke_csyrfsx_work.c  lapacke_dporfsx_work.c  lapacke_sgerfsx_work.c  lapacke_zgbrfsx_work.c  lapacke_zporfsx_work.c 
+  lapacke_cherfsx.c       lapacke_dgbrfsx.c       lapacke_dsyrfsx.c       lapacke_sporfsx.c       lapacke_zgerfsx.c       lapacke_zsyrfsx.c 
+  lapacke_cherfsx_work.c  lapacke_dgbrfsx_work.c  lapacke_dsyrfsx_work.c  lapacke_sporfsx_work.c  lapacke_zgerfsx_work.c  lapacke_zsyrfsx_work.c 
+  lapacke_cgbsvxx.c       lapacke_cposvxx.c       lapacke_dgesvxx.c       lapacke_sgbsvxx.c       lapacke_ssysvxx.c       lapacke_zhesvxx.c 
+  lapacke_cgbsvxx_work.c  lapacke_cposvxx_work.c  lapacke_dgesvxx_work.c  lapacke_sgbsvxx_work.c  lapacke_ssysvxx_work.c  lapacke_zhesvxx_work.c 
+  lapacke_cgesvxx.c       lapacke_csysvxx.c       lapacke_dposvxx.c       lapacke_sgesvxx.c       lapacke_zgbsvxx.c       lapacke_zposvxx.c 
+  lapacke_cgesvxx_work.c  lapacke_csysvxx_work.c  lapacke_dposvxx_work.c  lapacke_sgesvxx_work.c  lapacke_zgbsvxx_work.c  lapacke_zposvxx_work.c 
+  lapacke_chesvxx.c       lapacke_dgbsvxx.c       lapacke_dsysvxx.c       lapacke_sposvxx.c       lapacke_zgesvxx.c       lapacke_zsysvxx.c 
+  lapacke_chesvxx_work.c  lapacke_dgbsvxx_work.c  lapacke_dsysvxx_work.c  lapacke_sposvxx_work.c  lapacke_zgesvxx_work.c  lapacke_zsysvxx_work.c
+)
+
+
+# FILE PARTS OF TMGLIB 
+set(MATGEN
+  lapacke_clatms.c 
+  lapacke_clatms_work.c 
+  lapacke_dlatms.c 
+  lapacke_dlatms_work.c 
+  lapacke_slatms.c 
+  lapacke_slatms_work.c 
+  lapacke_zlatms.c 
+  lapacke_zlatms_work.c 
+  lapacke_clagge.c 
+  lapacke_clagge_work.c 
+  lapacke_dlagge.c 
+  lapacke_dlagge_work.c 
+  lapacke_slagge.c 
+  lapacke_slagge_work.c 
+  lapacke_zlagge.c 
+  lapacke_zlagge_work.c 
+  lapacke_claghe.c 
+  lapacke_claghe_work.c 
+  lapacke_zlaghe.c 
+  lapacke_zlaghe_work.c 
+  lapacke_clagsy.c 
+  lapacke_clagsy_work.c 
+  lapacke_dlagsy.c 
+  lapacke_dlagsy_work.c 
+  lapacke_slagsy.c 
+  lapacke_slagsy_work.c 
+  lapacke_zlagsy.c 
+  lapacke_zlagsy_work.c
+)
+
+set(LAPACKE_REL_SRC "")
+if (BUILD_SINGLE)
+  list(APPEND LAPACKE_REL_SRC ${SSRC})
+endif ()
+
+if (BUILD_DOUBLE)
+  list(APPEND LAPACKE_REL_SRC ${DSRC})
+endif ()
+
+if (BUILD_COMPLEX)
+  list(APPEND LAPACKE_REL_SRC ${CSRC})
+endif ()
+
+if (BUILD_COMPLEX16)
+  list(APPEND LAPACKE_REL_SRC ${ZSRC})
+endif ()
+
+# add lapack-netlib folder to the sources
+set(LAPACKE_SOURCES "")
+foreach (LAE_FILE ${LAPACKE_REL_SRC})
+  list(APPEND LAPACKE_SOURCES "${NETLIB_LAPACK_DIR}/lapacke/src/${LAE_FILE}")
+endforeach ()
+
+set(lapacke_include_dir "${NETLIB_LAPACK_DIR}/lapacke/include")
+execute_process(COMMAND ${CMAKE_COMMAND} -E copy "${lapacke_include_dir}/lapacke_mangling_with_flags.h" "${lapacke_include_dir}/lapacke_mangling.h")
+include_directories(${lapacke_include_dir})
+set_source_files_properties(${LAPACKE_SOURCES} PROPERTIES COMPILE_FLAGS "${LAPACK_CFLAGS}")
diff --git a/cmake/os.cmake b/cmake/os.cmake
new file mode 100644
index 000000000..f5a75027c
--- /dev/null
+++ b/cmake/os.cmake
@@ -0,0 +1,104 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+## Description: Ported from portion of OpenBLAS/Makefile.system
+##              Detects the OS and sets appropriate variables.
+
+if (${CMAKE_SYSTEM_NAME} STREQUAL "Darwin")
+  set(ENV{MACOSX_DEPLOYMENT_TARGET} "10.2") # TODO: should be exported as an env var
+  set(MD5SUM "md5 -r")
+endif ()
+
+if (${CMAKE_SYSTEM_NAME} STREQUAL "FreeBSD")
+  set(MD5SUM "md5 -r")
+endif ()
+
+if (${CMAKE_SYSTEM_NAME} STREQUAL "NetBSD")
+  set(MD5SUM "md5 -n")
+endif ()
+
+if (${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
+  set(EXTRALIB "${EXTRALIB} -lm")
+  set(NO_EXPRECISION 1)
+endif ()
+
+if (${CMAKE_SYSTEM_NAME} STREQUAL "AIX")
+  set(EXTRALIB "${EXTRALIB} -lm")
+endif ()
+
+# TODO: this is probably meant for mingw, not other windows compilers
+if (${CMAKE_SYSTEM_NAME} STREQUAL "Windows")
+
+  set(NEED_PIC 0)
+  set(NO_EXPRECISION 1)
+
+  set(EXTRALIB "${EXTRALIB} -defaultlib:advapi32")
+
+  # probably not going to use these
+  set(SUFFIX "obj")
+  set(PSUFFIX "pobj")
+  set(LIBSUFFIX "a")
+
+  if (${CMAKE_C_COMPILER_ID} STREQUAL "Clang")
+    set(CCOMMON_OPT	"${CCOMMON_OPT} -DMS_ABI")
+  endif ()
+
+  if (${CMAKE_C_COMPILER_ID} STREQUAL "GNU")
+
+    # Test for supporting MS_ABI
+    # removed string parsing in favor of CMake's version comparison -hpa
+    execute_process(COMMAND ${CMAKE_C_COMPILER} -dumpversion OUTPUT_VARIABLE GCC_VERSION)
+    if (${GCC_VERSION} VERSION_GREATER 4.7 OR ${GCC_VERSION} VERSION_EQUAL 4.7)
+      # GCC Version >=4.7
+      # It is compatible with MSVC ABI.
+      set(CCOMMON_OPT "${CCOMMON_OPT} -DMS_ABI")
+    endif ()
+  endif ()
+
+  # Ensure the correct stack alignment on Win32
+  # http://permalink.gmane.org/gmane.comp.lib.openblas.general/97
+  if (${ARCH} STREQUAL "x86")
+    if (NOT MSVC AND NOT ${CMAKE_C_COMPILER_ID} STREQUAL "Clang")
+      set(CCOMMON_OPT "${CCOMMON_OPT} -mincoming-stack-boundary=2")
+    endif ()
+    set(FCOMMON_OPT "${FCOMMON_OPT} -mincoming-stack-boundary=2")
+  endif ()
+  
+endif ()
+
+if (${CMAKE_SYSTEM_NAME} STREQUAL "Interix")
+  set(NEED_PIC 0)
+  set(NO_EXPRECISION 1)
+  
+  set(INTERIX_TOOL_DIR STREQUAL "/opt/gcc.3.3/i586-pc-interix3/bin")
+endif ()
+
+if (CYGWIN)
+  set(NEED_PIC 0)
+  set(NO_EXPRECISION 1)
+endif ()
+
+if (NOT ${CMAKE_SYSTEM_NAME} STREQUAL "Windows" AND NOT ${CMAKE_SYSTEM_NAME} STREQUAL "Interix")
+  if (SMP)
+    set(EXTRALIB "${EXTRALIB} -lpthread")
+  endif ()
+endif ()
+
+if (QUAD_PRECISION)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DQUAD_PRECISION")
+  set(NO_EXPRECISION 1)
+endif ()
+
+if (${ARCH} STREQUAL "x86")
+  set(NO_EXPRECISION 1)
+endif ()
+
+if (UTEST_CHECK)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DUTEST_CHECK")
+  set(SANITY_CHECK 1)
+endif ()
+
+if (SANITY_CHECK)
+  # TODO: need some way to get $(*F) (target filename)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DSANITY_CHECK -DREFNAME=$(*F)f${BU}")
+endif ()
+
diff --git a/cmake/prebuild.cmake b/cmake/prebuild.cmake
new file mode 100644
index 000000000..c3fa48655
--- /dev/null
+++ b/cmake/prebuild.cmake
@@ -0,0 +1,113 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+## Description: Ported from OpenBLAS/Makefile.prebuild
+##              This is triggered by system.cmake and runs before any of the code is built.
+##              Creates config.h and Makefile.conf by first running the c_check perl script (which creates those files).
+##              Next it runs f_check and appends some fortran information to the files.
+##              Finally it runs getarch and getarch_2nd for even more environment information.
+
+# CMake vars set by this file:
+# CORE
+# LIBCORE
+# NUM_CORES
+# HAVE_MMX
+# HAVE_SSE
+# HAVE_SSE2
+# HAVE_SSE3
+# MAKE
+# SGEMM_UNROLL_M
+# SGEMM_UNROLL_N
+# DGEMM_UNROLL_M
+# DGEMM_UNROLL_M
+# QGEMM_UNROLL_N
+# QGEMM_UNROLL_N
+# CGEMM_UNROLL_M
+# CGEMM_UNROLL_M
+# ZGEMM_UNROLL_N
+# ZGEMM_UNROLL_N
+# XGEMM_UNROLL_M
+# XGEMM_UNROLL_N
+# CGEMM3M_UNROLL_M
+# CGEMM3M_UNROLL_N
+# ZGEMM3M_UNROLL_M
+# ZGEMM3M_UNROLL_M
+# XGEMM3M_UNROLL_N
+# XGEMM3M_UNROLL_N
+
+# CPUIDEMU = ../../cpuid/table.o
+
+if (DEFINED CPUIDEMU)
+  set(EXFLAGS "-DCPUIDEMU -DVENDOR=99")
+endif ()
+
+if (DEFINED TARGET_CORE)
+  # set the C flags for just this file
+  set(GETARCH2_FLAGS "-DBUILD_KERNEL")
+  set(TARGET_MAKE "Makefile_kernel.conf")
+  set(TARGET_CONF "config_kernel.h")
+else()
+  set(TARGET_MAKE "Makefile.conf")
+  set(TARGET_CONF "config.h")
+endif ()
+
+include("${CMAKE_SOURCE_DIR}/cmake/c_check.cmake")
+
+if (NOT NOFORTRAN)
+  include("${CMAKE_SOURCE_DIR}/cmake/f_check.cmake")
+endif ()
+
+# compile getarch
+set(GETARCH_SRC
+  ${CMAKE_SOURCE_DIR}/getarch.c
+  ${CPUIDEMO}
+)
+
+if (NOT MSVC)
+  list(APPEND GETARCH_SRC ${CMAKE_SOURCE_DIR}/cpuid.S)
+endif ()
+
+if (MSVC)
+#Use generic for MSVC now
+set(GETARCH_FLAGS ${GETARCH_FLAGS} -DFORCE_GENERIC)
+endif()
+
+set(GETARCH_DIR "${PROJECT_BINARY_DIR}/getarch_build")
+set(GETARCH_BIN "getarch${CMAKE_EXECUTABLE_SUFFIX}")
+file(MAKE_DIRECTORY ${GETARCH_DIR})
+try_compile(GETARCH_RESULT ${GETARCH_DIR}
+  SOURCES ${GETARCH_SRC}
+  COMPILE_DEFINITIONS ${EXFLAGS} ${GETARCH_FLAGS} -I${CMAKE_SOURCE_DIR}
+  OUTPUT_VARIABLE GETARCH_LOG
+  COPY_FILE ${PROJECT_BINARY_DIR}/${GETARCH_BIN}
+)
+
+message(STATUS "Running getarch")
+
+# use the cmake binary w/ the -E param to run a shell command in a cross-platform way
+execute_process(COMMAND ${PROJECT_BINARY_DIR}/${GETARCH_BIN} 0 OUTPUT_VARIABLE GETARCH_MAKE_OUT)
+execute_process(COMMAND ${PROJECT_BINARY_DIR}/${GETARCH_BIN} 1 OUTPUT_VARIABLE GETARCH_CONF_OUT)
+
+message(STATUS "GETARCH results:\n${GETARCH_MAKE_OUT}")
+
+# append config data from getarch to the TARGET file and read in CMake vars
+file(APPEND ${TARGET_CONF} ${GETARCH_CONF_OUT})
+ParseGetArchVars(${GETARCH_MAKE_OUT})
+
+set(GETARCH2_DIR "${PROJECT_BINARY_DIR}/getarch2_build")
+set(GETARCH2_BIN "getarch_2nd${CMAKE_EXECUTABLE_SUFFIX}")
+file(MAKE_DIRECTORY ${GETARCH2_DIR})
+try_compile(GETARCH2_RESULT ${GETARCH2_DIR}
+  SOURCES ${CMAKE_SOURCE_DIR}/getarch_2nd.c
+  COMPILE_DEFINITIONS ${EXFLAGS} ${GETARCH_FLAGS} ${GETARCH2_FLAGS} -I${CMAKE_SOURCE_DIR}
+  OUTPUT_VARIABLE GETARCH2_LOG
+  COPY_FILE ${PROJECT_BINARY_DIR}/${GETARCH2_BIN}
+)
+
+# use the cmake binary w/ the -E param to run a shell command in a cross-platform way
+execute_process(COMMAND ${PROJECT_BINARY_DIR}/${GETARCH2_BIN} 0 OUTPUT_VARIABLE GETARCH2_MAKE_OUT)
+execute_process(COMMAND ${PROJECT_BINARY_DIR}/${GETARCH2_BIN} 1 OUTPUT_VARIABLE GETARCH2_CONF_OUT)
+
+# append config data from getarch_2nd to the TARGET file and read in CMake vars
+file(APPEND ${TARGET_CONF} ${GETARCH2_CONF_OUT})
+ParseGetArchVars(${GETARCH2_MAKE_OUT})
+
diff --git a/cmake/system.cmake b/cmake/system.cmake
new file mode 100644
index 000000000..134e9c12d
--- /dev/null
+++ b/cmake/system.cmake
@@ -0,0 +1,552 @@
+##
+## Author: Hank Anderson <hank@statease.com>
+## Description: Ported from OpenBLAS/Makefile.system
+##
+
+set(NETLIB_LAPACK_DIR "${CMAKE_SOURCE_DIR}/lapack-netlib")
+
+# TODO: Makefile.system detects Darwin (mac) and switches to clang here -hpa
+# http://stackoverflow.com/questions/714100/os-detecting-makefile
+
+# TODO: Makefile.system sets HOSTCC = $(CC) here if not already set -hpa
+
+# TARGET_CORE will override TARGET which is used in DYNAMIC_ARCH=1.
+if (DEFINED TARGET_CORE)
+  set(TARGET ${TARGET_CORE})
+endif ()
+
+# Force fallbacks for 32bit
+if (DEFINED BINARY AND DEFINED TARGET AND BINARY EQUAL 32)
+  message(STATUS "Compiling a ${BINARY}-bit binary.")
+  set(NO_AVX 1)
+  if (${TARGET} STREQUAL "HASWELL" OR ${TARGET} STREQUAL "SANDYBRIDGE")
+    set(TARGET "NEHALEM")
+  endif ()
+  if (${TARGET} STREQUAL "BULLDOZER" OR ${TARGET} STREQUAL "PILEDRIVER")
+    set(TARGET "BARCELONA")
+  endif ()
+endif ()
+
+if (DEFINED TARGET)
+  message(STATUS "Targetting the ${TARGET} architecture.")
+  set(GETARCH_FLAGS "-DFORCE_${TARGET}")
+endif ()
+
+if (INTERFACE64)
+  message(STATUS "Using 64-bit integers.")
+  set(GETARCH_FLAGS	"${GETARCH_FLAGS} -DUSE64BITINT")
+endif ()
+
+if (NOT DEFINED GEMM_MULTITHREAD_THRESHOLD)
+  set(GEMM_MULTITHREAD_THRESHOLD 4)
+endif ()
+message(STATUS "GEMM multithread threshold set to ${GEMM_MULTITHREAD_THRESHOLD}.")
+set(GETARCH_FLAGS	"${GETARCH_FLAGS} -DGEMM_MULTITHREAD_THRESHOLD=${GEMM_MULTITHREAD_THRESHOLD}")
+
+if (NO_AVX)
+  message(STATUS "Disabling Advanced Vector Extensions (AVX).")
+  set(GETARCH_FLAGS "${GETARCH_FLAGS} -DNO_AVX")
+endif ()
+
+if (NO_AVX2)
+  message(STATUS "Disabling Advanced Vector Extensions 2 (AVX2).")
+  set(GETARCH_FLAGS "${GETARCH_FLAGS} -DNO_AVX2")
+endif ()
+
+if (CMAKE_BUILD_TYPE STREQUAL Debug)
+  set(GETARCH_FLAGS "${GETARCH_FLAGS} -g")
+endif ()
+
+# TODO: let CMake handle this? -hpa
+#if (${QUIET_MAKE})
+#  set(MAKE "${MAKE} -s")
+#endif()
+
+if (NOT DEFINED NO_PARALLEL_MAKE)
+  set(NO_PARALLEL_MAKE 0)
+endif ()
+set(GETARCH_FLAGS	"${GETARCH_FLAGS} -DNO_PARALLEL_MAKE=${NO_PARALLEL_MAKE}")
+
+if (CMAKE_CXX_COMPILER STREQUAL loongcc)
+  set(GETARCH_FLAGS	"${GETARCH_FLAGS} -static")
+endif ()
+
+#if don't use Fortran, it will only compile CBLAS.
+if (ONLY_CBLAS)
+  set(NO_LAPACK 1)
+else ()
+  set(ONLY_CBLAS 0)
+endif ()
+
+include("${CMAKE_SOURCE_DIR}/cmake/prebuild.cmake")
+
+if (NOT DEFINED NUM_THREADS)
+  set(NUM_THREADS ${NUM_CORES})
+endif ()
+
+if (${NUM_THREADS} EQUAL 1)
+  set(USE_THREAD 0)
+endif ()
+
+if (DEFINED USE_THREAD)
+  if (NOT ${USE_THREAD})
+    unset(SMP)
+  else ()
+    set(SMP 1)
+  endif ()
+else ()
+  # N.B. this is NUM_THREAD in Makefile.system which is probably a bug -hpa
+  if (${NUM_THREADS} EQUAL 1)
+    unset(SMP)
+  else ()
+    set(SMP 1)
+  endif ()
+endif ()
+
+if (${SMP})
+  message(STATUS "SMP enabled.")
+endif ()
+
+if (NOT DEFINED NEED_PIC)
+  set(NEED_PIC 1)
+endif ()
+
+# TODO: I think CMake should be handling all this stuff -hpa
+unset(ARFLAGS)
+set(CPP "${COMPILER} -E")
+set(AR "${CROSS_SUFFIX}ar")
+set(AS "${CROSS_SUFFIX}as")
+set(LD "${CROSS_SUFFIX}ld")
+set(RANLIB "${CROSS_SUFFIX}ranlib")
+set(NM "${CROSS_SUFFIX}nm")
+set(DLLWRAP "${CROSS_SUFFIX}dllwrap")
+set(OBJCOPY "${CROSS_SUFFIX}objcopy")
+set(OBJCONV "${CROSS_SUFFIX}objconv")
+
+# OS dependent settings
+include("${CMAKE_SOURCE_DIR}/cmake/os.cmake")
+
+# Architecture dependent settings
+include("${CMAKE_SOURCE_DIR}/cmake/arch.cmake")
+
+# C Compiler dependent settings
+include("${CMAKE_SOURCE_DIR}/cmake/cc.cmake")
+
+if (NOT NOFORTRAN)
+  # Fortran Compiler dependent settings
+  include("${CMAKE_SOURCE_DIR}/cmake/fc.cmake")
+endif ()
+
+if (BINARY64)
+  if (INTERFACE64)
+    # CCOMMON_OPT += -DUSE64BITINT
+  endif ()
+endif ()
+
+if (NEED_PIC)
+  if (${CMAKE_C_COMPILER} STREQUAL "IBM")
+    set(CCOMMON_OPT "${CCOMMON_OPT} -qpic=large")
+  else ()
+    set(CCOMMON_OPT "${CCOMMON_OPT} -fPIC")
+  endif ()
+
+  if (${F_COMPILER} STREQUAL "SUN")
+    set(FCOMMON_OPT "${FCOMMON_OPT} -pic")
+  else ()
+    set(FCOMMON_OPT "${FCOMMON_OPT} -fPIC")
+  endif ()
+endif ()
+
+if (DYNAMIC_ARCH)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_ARCH")
+endif ()
+
+if (NO_LAPACK)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_LAPACK")
+  #Disable LAPACK C interface
+  set(NO_LAPACKE 1)
+endif ()
+
+if (NO_LAPACKE)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_LAPACKE")
+endif ()
+
+if (NO_AVX)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_AVX")
+endif ()
+
+if (${ARCH} STREQUAL "x86")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_AVX")
+endif ()
+
+if (NO_AVX2)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_AVX2")
+endif ()
+
+if (SMP)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DSMP_SERVER")
+
+  if (${ARCH} STREQUAL "mips64")
+    if (NOT ${CORE} STREQUAL "LOONGSON3B")
+      set(USE_SIMPLE_THREADED_LEVEL3 1)
+    endif ()
+  endif ()
+
+  if (USE_OPENMP)
+    # USE_SIMPLE_THREADED_LEVEL3 = 1
+    # NO_AFFINITY = 1
+    set(CCOMMON_OPT "${CCOMMON_OPT} -DUSE_OPENMP")
+  endif ()
+
+  if (BIGNUMA)
+    set(CCOMMON_OPT "${CCOMMON_OPT} -DBIGNUMA")
+  endif ()
+
+endif ()
+
+if (NO_WARMUP)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_WARMUP")
+endif ()
+
+if (CONSISTENT_FPCSR)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DCONSISTENT_FPCSR")
+endif ()
+
+# Only for development
+# set(CCOMMON_OPT "${CCOMMON_OPT} -DPARAMTEST")
+# set(CCOMMON_OPT "${CCOMMON_OPT} -DPREFETCHTEST")
+# set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_SWITCHING")
+# set(USE_PAPI 1)
+
+if (USE_PAPI)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DUSE_PAPI")
+  set(EXTRALIB "${EXTRALIB} -lpapi -lperfctr")
+endif ()
+
+if (DYNAMIC_THREADS)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_THREADS")
+endif ()
+
+set(CCOMMON_OPT "${CCOMMON_OPT} -DMAX_CPU_NUMBER=${NUM_THREADS}")
+
+if (USE_SIMPLE_THREADED_LEVEL3)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DUSE_SIMPLE_THREADED_LEVEL3")
+endif ()
+
+if (DEFINED LIBNAMESUFFIX)
+  set(LIBPREFIX "libopenblas_${LIBNAMESUFFIX}")
+else ()
+  set(LIBPREFIX "libopenblas")
+endif ()
+
+if (NOT DEFINED SYMBOLPREFIX)
+  set(SYMBOLPREFIX "")
+endif ()
+
+if (NOT DEFINED SYMBOLSUFFIX)
+  set(SYMBOLSUFFIX "")
+endif ()
+
+set(KERNELDIR	"${CMAKE_SOURCE_DIR}/kernel/${ARCH}")
+
+# TODO: nead to convert these Makefiles
+# include ${CMAKE_SOURCE_DIR}/cmake/${ARCH}.cmake
+
+if (${CORE} STREQUAL "PPC440")
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DALLOC_QALLOC")
+endif ()
+
+if (${CORE} STREQUAL "PPC440FP2")
+  set(STATIC_ALLOCATION 1)
+endif ()
+
+if (NOT ${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
+  set(NO_AFFINITY 1)
+endif ()
+
+if (NOT ${ARCH} STREQUAL "x86_64" AND NOT ${ARCH} STREQUAL "x86" AND NOT ${CORE} STREQUAL "LOONGSON3B")
+  set(NO_AFFINITY 1)
+endif ()
+
+if (NO_AFFINITY)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DNO_AFFINITY")
+endif ()
+
+if (FUNCTION_PROFILE)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DFUNCTION_PROFILE")
+endif ()
+
+if (HUGETLB_ALLOCATION)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DALLOC_HUGETLB")
+endif ()
+
+if (DEFINED HUGETLBFILE_ALLOCATION)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DALLOC_HUGETLBFILE -DHUGETLB_FILE_NAME=${HUGETLBFILE_ALLOCATION})")
+endif ()
+
+if (STATIC_ALLOCATION)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DALLOC_STATIC")
+endif ()
+
+if (DEVICEDRIVER_ALLOCATION)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DALLOC_DEVICEDRIVER -DDEVICEDRIVER_NAME=\"/dev/mapper\"")
+endif ()
+
+if (MIXED_MEMORY_ALLOCATION)
+  set(CCOMMON_OPT "${CCOMMON_OPT} -DMIXED_MEMORY_ALLOCATION")
+endif ()
+
+if (${CMAKE_SYSTEM_NAME} STREQUAL "SunOS")
+  set(TAR	gtar)
+  set(PATCH	gpatch)
+  set(GREP ggrep)
+else ()
+  set(TAR tar)
+  set(PATCH patch)
+  set(GREP grep)
+endif ()
+
+if (NOT DEFINED MD5SUM)
+  set(MD5SUM md5sum)
+endif ()
+
+set(AWK awk)
+
+set(REVISION "-r${OpenBLAS_VERSION}")
+set(MAJOR_VERSION ${OpenBLAS_MAJOR_VERSION})
+
+if (DEBUG)
+  set(COMMON_OPT "${COMMON_OPT} -g")
+endif ()
+
+if (NOT DEFINED COMMON_OPT)
+  set(COMMON_OPT "-O2")
+endif ()
+
+#For x86 32-bit
+if (DEFINED BINARY AND BINARY EQUAL 32)
+if (NOT MSVC)
+  set(COMMON_OPT "${COMMON_OPT} -m32")
+endif()
+endif()
+
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${COMMON_OPT} ${CCOMMON_OPT}")
+if(NOT MSVC)
+set(CMAKE_ASM_FLAGS "${CMAKE_ASM_FLAGS} ${COMMON_OPT} ${CCOMMON_OPT}")
+endif()
+# TODO: not sure what PFLAGS is -hpa
+set(PFLAGS "${PFLAGS} ${COMMON_OPT} ${CCOMMON_OPT} -I${TOPDIR} -DPROFILE ${COMMON_PROF}")
+
+set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} ${COMMON_OPT} ${FCOMMON_OPT}")
+# TODO: not sure what FPFLAGS is -hpa
+set(FPFLAGS "${FPFLAGS} ${COMMON_OPT} ${FCOMMON_OPT} ${COMMON_PROF}")
+
+#For LAPACK Fortran codes.
+set(LAPACK_FFLAGS "${LAPACK_FFLAGS} ${CMAKE_Fortran_FLAGS}")
+set(LAPACK_FPFLAGS "${LAPACK_FPFLAGS} ${FPFLAGS}")
+
+#Disable -fopenmp for LAPACK Fortran codes on Windows.
+if (${CMAKE_SYSTEM_NAME} STREQUAL "Windows")
+  set(FILTER_FLAGS "-fopenmp;-mp;-openmp;-xopenmp=parralel")
+  foreach (FILTER_FLAG ${FILTER_FLAGS})
+    string(REPLACE ${FILTER_FLAG} "" LAPACK_FFLAGS ${LAPACK_FFLAGS})
+    string(REPLACE ${FILTER_FLAG} "" LAPACK_FPFLAGS ${LAPACK_FPFLAGS})
+  endforeach ()
+endif ()
+
+if ("${F_COMPILER}" STREQUAL "GFORTRAN")
+  # lapack-netlib is rife with uninitialized warnings -hpa
+  set(LAPACK_FFLAGS "${LAPACK_FFLAGS} -Wno-maybe-uninitialized")
+endif ()
+
+set(LAPACK_CFLAGS "${CMAKE_C_CFLAGS} -DHAVE_LAPACK_CONFIG_H")
+if (INTERFACE64)
+  set(LAPACK_CFLAGS "${LAPACK_CFLAGS} -DLAPACK_ILP64")
+endif ()
+
+if (${CMAKE_SYSTEM_NAME} STREQUAL "Windows")
+  set(LAPACK_CFLAGS "${LAPACK_CFLAGS} -DOPENBLAS_OS_WINDOWS")
+endif ()
+
+if (${CMAKE_C_COMPILER} STREQUAL "LSB" OR ${CMAKE_SYSTEM_NAME} STREQUAL "Windows")
+  set(LAPACK_CFLAGS "${LAPACK_CFLAGS} -DLAPACK_COMPLEX_STRUCTURE")
+endif ()
+
+if (NOT DEFINED SUFFIX)
+  set(SUFFIX o)
+endif ()
+
+if (NOT DEFINED PSUFFIX)
+  set(PSUFFIX po)
+endif ()
+
+if (NOT DEFINED LIBSUFFIX)
+  set(LIBSUFFIX a)
+endif ()
+
+if (DYNAMIC_ARCH)
+  if (DEFINED SMP)
+    set(LIBNAME "${LIBPREFIX}p${REVISION}.${LIBSUFFIX}")
+    set(LIBNAME_P	"${LIBPREFIX}p${REVISION}_p.${LIBSUFFIX}")
+  else ()
+    set(LIBNAME "${LIBPREFIX}${REVISION}.${LIBSUFFIX}")
+    set(LIBNAME_P	"${LIBPREFIX}${REVISION}_p.${LIBSUFFIX}")
+  endif ()
+else ()
+  if (DEFINED SMP)
+    set(LIBNAME "${LIBPREFIX}_${LIBCORE}p${REVISION}.${LIBSUFFIX}")
+    set(LIBNAME_P	"${LIBPREFIX}_${LIBCORE}p${REVISION}_p.${LIBSUFFIX}")
+  else ()
+    set(LIBNAME	"${LIBPREFIX}_${LIBCORE}${REVISION}.${LIBSUFFIX}")
+    set(LIBNAME_P	"${LIBPREFIX}_${LIBCORE}${REVISION}_p.${LIBSUFFIX}")
+  endif ()
+endif ()
+
+
+set(LIBDLLNAME "${LIBPREFIX}.dll")
+set(LIBSONAME "${LIBNAME}.${LIBSUFFIX}.so")
+set(LIBDYNNAME "${LIBNAME}.${LIBSUFFIX}.dylib")
+set(LIBDEFNAME "${LIBNAME}.${LIBSUFFIX}.def")
+set(LIBEXPNAME "${LIBNAME}.${LIBSUFFIX}.exp")
+set(LIBZIPNAME "${LIBNAME}.${LIBSUFFIX}.zip")
+
+set(LIBS "${CMAKE_SOURCE_DIR}/${LIBNAME}")
+set(LIBS_P "${CMAKE_SOURCE_DIR}/${LIBNAME_P}")
+
+
+set(LIB_COMPONENTS BLAS)
+if (NOT NO_CBLAS)
+  set(LIB_COMPONENTS "${LIB_COMPONENTS} CBLAS")
+endif ()
+
+if (NOT NO_LAPACK)
+  set(LIB_COMPONENTS "${LIB_COMPONENTS} LAPACK")
+  if (NOT NO_LAPACKE)
+    set(LIB_COMPONENTS "${LIB_COMPONENTS} LAPACKE")
+  endif ()
+endif ()
+
+if (ONLY_CBLAS)
+  set(LIB_COMPONENTS CBLAS)
+endif ()
+
+
+# For GEMM3M
+set(USE_GEMM3M 0)
+
+if (DEFINED ARCH)
+  if (${ARCH} STREQUAL "x86" OR ${ARCH} STREQUAL "x86_64" OR ${ARCH} STREQUAL "ia64" OR ${ARCH} STREQUAL "MIPS")
+    set(USE_GEMM3M 1)
+  endif ()
+
+  if (${CORE} STREQUAL "generic")
+    set(USE_GEMM3M 0)
+  endif ()
+endif ()
+
+
+#export OSNAME
+#export ARCH
+#export CORE
+#export LIBCORE
+#export PGCPATH
+#export CONFIG
+#export CC
+#export FC
+#export BU
+#export FU
+#export NEED2UNDERSCORES
+#export USE_THREAD
+#export NUM_THREADS
+#export NUM_CORES
+#export SMP
+#export MAKEFILE_RULE
+#export NEED_PIC
+#export BINARY
+#export BINARY32
+#export BINARY64
+#export F_COMPILER
+#export C_COMPILER
+#export USE_OPENMP
+#export CROSS
+#export CROSS_SUFFIX
+#export NOFORTRAN
+#export NO_FBLAS
+#export EXTRALIB
+#export CEXTRALIB
+#export FEXTRALIB
+#export HAVE_SSE
+#export HAVE_SSE2
+#export HAVE_SSE3
+#export HAVE_SSSE3
+#export HAVE_SSE4_1
+#export HAVE_SSE4_2
+#export HAVE_SSE4A
+#export HAVE_SSE5
+#export HAVE_AVX
+#export HAVE_VFP
+#export HAVE_VFPV3
+#export HAVE_VFPV4
+#export HAVE_NEON
+#export KERNELDIR
+#export FUNCTION_PROFILE
+#export TARGET_CORE
+#
+#export SGEMM_UNROLL_M
+#export SGEMM_UNROLL_N
+#export DGEMM_UNROLL_M
+#export DGEMM_UNROLL_N
+#export QGEMM_UNROLL_M
+#export QGEMM_UNROLL_N
+#export CGEMM_UNROLL_M
+#export CGEMM_UNROLL_N
+#export ZGEMM_UNROLL_M
+#export ZGEMM_UNROLL_N
+#export XGEMM_UNROLL_M
+#export XGEMM_UNROLL_N
+#export CGEMM3M_UNROLL_M
+#export CGEMM3M_UNROLL_N
+#export ZGEMM3M_UNROLL_M
+#export ZGEMM3M_UNROLL_N
+#export XGEMM3M_UNROLL_M
+#export XGEMM3M_UNROLL_N
+
+
+#if (USE_CUDA)
+#  export CUDADIR
+#  export CUCC
+#  export CUFLAGS
+#  export CULIB
+#endif
+
+#.SUFFIXES: .$(PSUFFIX) .$(SUFFIX) .f
+#
+#.f.$(SUFFIX):
+#	$(FC) $(FFLAGS) -c $<  -o $(@F)
+#
+#.f.$(PSUFFIX):
+#	$(FC) $(FPFLAGS) -pg -c $<  -o $(@F)
+
+# these are not cross-platform
+#ifdef BINARY64
+#PATHSCALEPATH	= /opt/pathscale/lib/3.1
+#PGIPATH		= /opt/pgi/linux86-64/7.1-5/lib
+#else
+#PATHSCALEPATH	= /opt/pathscale/lib/3.1/32
+#PGIPATH		= /opt/pgi/linux86/7.1-5/lib
+#endif
+
+#ACMLPATH	= /opt/acml/4.3.0
+#ifneq ($(OSNAME), Darwin)
+#MKLPATH         = /opt/intel/mkl/10.2.2.025/lib
+#else
+#MKLPATH         = /Library/Frameworks/Intel_MKL.framework/Versions/10.0.1.014/lib
+#endif
+#ATLASPATH	= /opt/atlas/3.9.17/opteron
+#FLAMEPATH	= $(HOME)/flame/lib
+#ifneq ($(OSNAME), SunOS)
+#SUNPATH		= /opt/sunstudio12.1
+#else
+#SUNPATH		= /opt/SUNWspro
+#endif
+
diff --git a/cmake/utils.cmake b/cmake/utils.cmake
new file mode 100644
index 000000000..6e2a98069
--- /dev/null
+++ b/cmake/utils.cmake
@@ -0,0 +1,346 @@
+# Functions to help with the OpenBLAS build
+
+# Reads string from getarch into CMake vars. Format of getarch vars is VARNAME=VALUE
+function(ParseGetArchVars GETARCH_IN)
+  string(REGEX MATCHALL "[0-9_a-zA-Z]+=[0-9_a-zA-Z]+" GETARCH_RESULT_LIST "${GETARCH_IN}")
+  foreach (GETARCH_LINE ${GETARCH_RESULT_LIST})
+    # split the line into var and value, then assign the value to a CMake var
+    string(REGEX MATCHALL "[0-9_a-zA-Z]+" SPLIT_VAR "${GETARCH_LINE}")
+    list(GET SPLIT_VAR 0 VAR_NAME)
+    list(GET SPLIT_VAR 1 VAR_VALUE)
+    set(${VAR_NAME} ${VAR_VALUE} PARENT_SCOPE)
+  endforeach ()
+endfunction ()
+
+# Reads a Makefile into CMake vars.
+macro(ParseMakefileVars MAKEFILE_IN)
+  message(STATUS "Reading vars from ${MAKEFILE_IN}...")
+  file(STRINGS ${MAKEFILE_IN} makefile_contents)
+  foreach (makefile_line ${makefile_contents})
+    string(REGEX MATCH "([0-9_a-zA-Z]+)[ \t]*=[ \t]*(.+)$" line_match "${makefile_line}")
+    if (NOT "${line_match}" STREQUAL "")
+      set(var_name ${CMAKE_MATCH_1})
+      set(var_value ${CMAKE_MATCH_2})
+      # check for Makefile variables in the string, e.g. $(TSUFFIX)
+      string(REGEX MATCHALL "\\$\\(([0-9_a-zA-Z]+)\\)" make_var_matches ${var_value})
+      foreach (make_var ${make_var_matches})
+        # strip out Makefile $() markup
+        string(REGEX REPLACE "\\$\\(([0-9_a-zA-Z]+)\\)" "\\1" make_var ${make_var})
+        # now replace the instance of the Makefile variable with the value of the CMake variable (note the double quote)
+        string(REPLACE "$(${make_var})" "${${make_var}}" var_value ${var_value})
+      endforeach ()
+      set(${var_name} ${var_value})
+    else ()
+      string(REGEX MATCH "include \\$\\(KERNELDIR\\)/(.+)$" line_match "${makefile_line}")
+      if (NOT "${line_match}" STREQUAL "")
+        ParseMakefileVars(${KERNELDIR}/${CMAKE_MATCH_1})
+      endif ()
+    endif ()
+  endforeach ()
+endmacro ()
+
+# Returns all combinations of the input list, as a list with colon-separated combinations
+# E.g. input of A B C returns A B C A:B A:C B:C
+# N.B. The input is meant to be a list, and to past a list to a function in CMake you must quote it (e.g. AllCombinations("${LIST_VAR}")).
+# #param absent_codes codes to use when an element is absent from a combination. For example, if you have TRANS;UNIT;UPPER you may want the code to be NNL when nothing is present.
+# @returns LIST_OUT a list of combinations
+#          CODES_OUT a list of codes corresponding to each combination, with N meaning the item is not present, and the first letter of the list item meaning it is presen
+function(AllCombinations list_in absent_codes_in)
+  list(LENGTH list_in list_count)
+  set(num_combos 1)
+  # subtract 1 since we will iterate from 0 to num_combos
+  math(EXPR num_combos "(${num_combos} << ${list_count}) - 1")
+  set(LIST_OUT "")
+  set(CODES_OUT "")
+  foreach (c RANGE 0 ${num_combos})
+
+    set(current_combo "")
+    set(current_code "")
+
+    # this is a little ridiculous just to iterate through a list w/ indices
+    math(EXPR last_list_index "${list_count} - 1")
+    foreach (list_index RANGE 0 ${last_list_index})
+      math(EXPR bit "1 << ${list_index}")
+      math(EXPR combo_has_bit "${c} & ${bit}")
+      list(GET list_in ${list_index} list_elem)
+      if (combo_has_bit)
+        if (current_combo)
+          set(current_combo "${current_combo}:${list_elem}")
+        else ()
+          set(current_combo ${list_elem})
+        endif ()
+        string(SUBSTRING ${list_elem} 0 1 code_char)
+      else ()
+        list(GET absent_codes_in ${list_index} code_char)
+      endif ()
+      set(current_code "${current_code}${code_char}")
+    endforeach ()
+
+    if (current_combo STREQUAL "")
+      list(APPEND LIST_OUT " ") # Empty set is a valid combination, but CMake isn't appending the empty string for some reason, use a space
+    else ()
+      list(APPEND LIST_OUT ${current_combo})
+    endif ()
+    list(APPEND CODES_OUT ${current_code})
+
+  endforeach ()
+
+  set(LIST_OUT ${LIST_OUT} PARENT_SCOPE)
+  set(CODES_OUT ${CODES_OUT} PARENT_SCOPE)
+endfunction ()
+
+# generates object files for each of the sources, using the BLAS naming scheme to pass the funciton name as a preprocessor definition
+# @param sources_in the source files to build from
+# @param defines_in (optional) preprocessor definitions that will be applied to all objects
+# @param name_in (optional) if this is set this name will be used instead of the filename. Use a * to indicate where the float character should go, if no star the character will be prepended.
+#                           e.g. with DOUBLE set, "i*max" will generate the name "idmax", and "max" will be "dmax"
+# @param replace_last_with replaces the last character in the filename with this string (e.g. symm_k should be symm_TU)
+# @param append_with appends the filename with this string (e.g. trmm_R should be trmm_RTUU or some other combination of characters)
+# @param no_float_type turns off the float type define for this build (e.g. SINGLE/DOUBLE/etc)
+# @param complex_filename_scheme some routines have separate source files for complex and non-complex float types.
+#                               0 - compiles for all types
+#                               1 - compiles the sources for non-complex types only (SINGLE/DOUBLE)
+#                               2 - compiles for complex types only (COMPLEX/DOUBLE COMPLEX)
+#                               3 - compiles for all types, but changes source names for complex by prepending z (e.g. axpy.c becomes zaxpy.c)
+#                               4 - compiles for complex types only, but changes source names for complex by prepending z (e.g. hemv.c becomes zhemv.c)
+#                               STRING - compiles only the given type (e.g. DOUBLE)
+function(GenerateNamedObjects sources_in)
+
+  if (DEFINED ARGV1)
+    set(defines_in ${ARGV1})
+  endif ()
+
+  if (DEFINED ARGV2 AND NOT "${ARGV2}" STREQUAL "")
+    set(name_in ${ARGV2})
+    # strip off extension for kernel files that pass in the object name.
+    get_filename_component(name_in ${name_in} NAME_WE)
+  endif ()
+
+  if (DEFINED ARGV3)
+    set(use_cblas ${ARGV3})
+  else ()
+    set(use_cblas false)
+  endif ()
+
+  if (DEFINED ARGV4)
+    set(replace_last_with ${ARGV4})
+  endif ()
+
+  if (DEFINED ARGV5)
+    set(append_with ${ARGV5})
+  endif ()
+
+  if (DEFINED ARGV6)
+    set(no_float_type ${ARGV6})
+  else ()
+    set(no_float_type false)
+  endif ()
+
+  if (no_float_type)
+    set(float_list "DUMMY") # still need to loop once
+  else ()
+    set(float_list "${FLOAT_TYPES}")
+  endif ()
+
+  set(real_only false)
+  set(complex_only false)
+  set(mangle_complex_sources false)
+  if (DEFINED ARGV7 AND NOT "${ARGV7}" STREQUAL "")
+    if (${ARGV7} EQUAL 1)
+      set(real_only true)
+    elseif (${ARGV7} EQUAL 2)
+      set(complex_only true)
+    elseif (${ARGV7} EQUAL 3)
+      set(mangle_complex_sources true)
+    elseif (${ARGV7} EQUAL 4)
+      set(mangle_complex_sources true)
+      set(complex_only true)
+    elseif (NOT ${ARGV7} EQUAL 0)
+      set(float_list ${ARGV7})
+    endif ()
+  endif ()
+
+  if (complex_only)
+    list(REMOVE_ITEM float_list "SINGLE")
+    list(REMOVE_ITEM float_list "DOUBLE")
+  elseif (real_only)
+    list(REMOVE_ITEM float_list "COMPLEX")
+    list(REMOVE_ITEM float_list "ZCOMPLEX")
+  endif ()
+
+  set(float_char "")
+  set(OBJ_LIST_OUT "")
+  foreach (float_type ${float_list})
+    foreach (source_file ${sources_in})
+
+      if (NOT no_float_type)
+        string(SUBSTRING ${float_type} 0 1 float_char)
+        string(TOLOWER ${float_char} float_char)
+      endif ()
+
+      if (NOT name_in)
+        get_filename_component(source_name ${source_file} NAME_WE)
+        set(obj_name "${float_char}${source_name}")
+      else ()
+        # replace * with float_char
+        if (${name_in} MATCHES "\\*")
+          string(REPLACE "*" ${float_char} obj_name ${name_in})
+        else ()
+          set(obj_name "${float_char}${name_in}")
+        endif ()
+      endif ()
+
+      if (replace_last_with)
+        string(REGEX REPLACE ".$" ${replace_last_with} obj_name ${obj_name})
+      else ()
+        set(obj_name "${obj_name}${append_with}")
+      endif ()
+
+      # now add the object and set the defines
+      set(obj_defines ${defines_in})
+
+      if (use_cblas)
+        set(obj_name "cblas_${obj_name}")
+        list(APPEND obj_defines "CBLAS")
+      endif ()
+
+      list(APPEND obj_defines "ASMNAME=${FU}${obj_name};ASMFNAME=${FU}${obj_name}${BU};NAME=${obj_name}${BU};CNAME=${obj_name};CHAR_NAME=\"${obj_name}${BU}\";CHAR_CNAME=\"${obj_name}\"")
+      if (${float_type} STREQUAL "DOUBLE" OR ${float_type} STREQUAL "ZCOMPLEX")
+        list(APPEND obj_defines "DOUBLE")
+      endif ()
+      if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+        list(APPEND obj_defines "COMPLEX")
+        if (mangle_complex_sources)
+          # add a z to the filename
+          get_filename_component(source_name ${source_file} NAME)
+          get_filename_component(source_dir ${source_file} DIRECTORY)
+          string(REPLACE ${source_name} "z${source_name}" source_file ${source_file})
+        endif ()
+      endif ()
+
+      if (VERBOSE_GEN)
+        message(STATUS "${obj_name}:${source_file}")
+        message(STATUS "${obj_defines}")
+      endif ()
+
+      # create a copy of the source to avoid duplicate obj filename problem with ar.exe
+      get_filename_component(source_extension ${source_file} EXT)
+      set(new_source_file "${CMAKE_CURRENT_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/${obj_name}${source_extension}")
+      if (IS_ABSOLUTE ${source_file})
+        set(old_source_file ${source_file})
+      else ()
+        set(old_source_file "${CMAKE_CURRENT_LIST_DIR}/${source_file}")
+      endif ()
+
+      string(REPLACE ";" "\n#define " define_source "${obj_defines}")
+      string(REPLACE "=" " " define_source "${define_source}")
+      file(WRITE ${new_source_file} "#define ${define_source}\n#include \"${old_source_file}\"")
+      list(APPEND SRC_LIST_OUT ${new_source_file})
+
+    endforeach ()
+  endforeach ()
+
+  list(APPEND OPENBLAS_SRC ${SRC_LIST_OUT})
+  set(OPENBLAS_SRC ${OPENBLAS_SRC} PARENT_SCOPE)
+endfunction ()
+
+# generates object files for each of the sources for each of the combinations of the preprocessor definitions passed in
+# @param sources_in the source files to build from
+# @param defines_in the preprocessor definitions that will be combined to create the object files
+# @param all_defines_in (optional) preprocessor definitions that will be applied to all objects
+# @param replace_scheme If 1, replace the "k" in the filename with the define combo letters. E.g. symm_k.c with TRANS and UNIT defined will be symm_TU.
+#                  If 0, it will simply append the code, e.g. symm_L.c with TRANS and UNIT will be symm_LTU.
+#                  If 2, it will append the code with an underscore, e.g. symm.c with TRANS and UNIT will be symm_TU.
+#                  If 3, it will insert the code *around* the last character with an underscore, e.g. symm_L.c with TRANS and UNIT will be symm_TLU (required by BLAS level2 objects).
+#                  If 4, it will insert the code before the last underscore. E.g. trtri_U_parallel with TRANS will be trtri_UT_parallel
+# @param alternate_name replaces the source name as the object name (define codes are still appended)
+# @param no_float_type turns off the float type define for this build (e.g. SINGLE/DOUBLE/etc)
+# @param complex_filename_scheme see GenerateNamedObjects
+function(GenerateCombinationObjects sources_in defines_in absent_codes_in all_defines_in replace_scheme)
+
+  set(alternate_name_in "")
+  if (DEFINED ARGV5)
+    set(alternate_name_in ${ARGV5})
+  endif ()
+
+  set(no_float_type false)
+  if (DEFINED ARGV6)
+    set(no_float_type ${ARGV6})
+  endif ()
+
+  set(complex_filename_scheme "")
+  if (DEFINED ARGV7)
+    set(complex_filename_scheme ${ARGV7})
+  endif ()
+
+  AllCombinations("${defines_in}" "${absent_codes_in}")
+  set(define_combos ${LIST_OUT})
+  set(define_codes ${CODES_OUT})
+
+  list(LENGTH define_combos num_combos)
+  math(EXPR num_combos "${num_combos} - 1")
+
+  foreach (c RANGE 0 ${num_combos})
+
+    list(GET define_combos ${c} define_combo)
+    list(GET define_codes ${c} define_code)
+
+    foreach (source_file ${sources_in})
+
+      set(alternate_name ${alternate_name_in})
+
+      # replace colon separated list with semicolons, this turns it into a CMake list that we can use foreach with
+      string(REPLACE ":" ";" define_combo ${define_combo})
+
+      # now add the object and set the defines
+      set(cur_defines ${define_combo})
+      if ("${cur_defines}" STREQUAL " ")
+        set(cur_defines ${all_defines_in})
+      else ()
+        list(APPEND cur_defines ${all_defines_in})
+      endif ()
+
+      set(replace_code "")
+      set(append_code "")
+      if (replace_scheme EQUAL 1)
+        set(replace_code ${define_code})
+      else ()
+        if (replace_scheme EQUAL 2)
+          set(append_code "_${define_code}")
+        elseif (replace_scheme EQUAL 3)
+          if ("${alternate_name}" STREQUAL "")
+            string(REGEX MATCH "[a-zA-Z]\\." last_letter ${source_file})
+          else ()
+            string(REGEX MATCH "[a-zA-Z]$" last_letter ${alternate_name})
+          endif ()
+          # first extract the last letter
+          string(SUBSTRING ${last_letter} 0 1 last_letter) # remove period from match
+          # break the code up into the first letter and the remaining (should only be 2 anyway)
+          string(SUBSTRING ${define_code} 0 1 define_code_first)
+          string(SUBSTRING ${define_code} 1 -1 define_code_second)
+          set(replace_code "${define_code_first}${last_letter}${define_code_second}")
+        elseif (replace_scheme EQUAL 4)
+          # insert code before the last underscore and pass that in as the alternate_name
+          if ("${alternate_name}" STREQUAL "")
+            get_filename_component(alternate_name ${source_file} NAME_WE)
+          endif ()
+          set(extra_underscore "")
+          # check if filename has two underscores, insert another if not (e.g. getrs_parallel needs to become getrs_U_parallel not getrsU_parallel)
+          string(REGEX MATCH "_[a-zA-Z]+_" underscores ${alternate_name})
+          string(LENGTH "${underscores}" underscores)
+          if (underscores EQUAL 0)
+            set(extra_underscore "_")
+          endif ()
+          string(REGEX REPLACE "(.+)(_[^_]+)$" "\\1${extra_underscore}${define_code}\\2" alternate_name ${alternate_name})
+        else()
+          set(append_code ${define_code}) # replace_scheme should be 0
+        endif ()
+      endif ()
+
+      GenerateNamedObjects("${source_file}" "${cur_defines}" "${alternate_name}" false "${replace_code}" "${append_code}" "${no_float_type}" "${complex_filename_scheme}")
+    endforeach ()
+  endforeach ()
+
+  set(OPENBLAS_SRC ${OPENBLAS_SRC} PARENT_SCOPE)
+endfunction ()
+
diff --git a/common.h b/common.h
index fe2083469..7b81c6fb6 100644
--- a/common.h
+++ b/common.h
@@ -82,7 +82,10 @@ extern "C" {
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+
+#if !defined(_MSC_VER)
 #include <unistd.h>
+#endif
 
 #ifdef OS_LINUX
 #include <malloc.h>
@@ -93,6 +96,14 @@ extern "C" {
 #include <sched.h>
 #endif
 
+#ifdef OS_ANDROID
+#define NO_SYSV_IPC
+//Android NDK only supports complex.h since Android 5.0
+#if __ANDROID_API__ < 21
+#define FORCE_OPENBLAS_COMPLEX_STRUCT
+#endif
+#endif
+
 #ifdef OS_WINDOWS
 #ifdef  ATOM
 #define GOTO_ATOM ATOM
@@ -106,8 +117,11 @@ extern "C" {
 #endif
 #else
 #include <sys/mman.h>
+#ifndef NO_SYSV_IPC
 #include <sys/shm.h>
+#endif
 #include <sys/time.h>
+#include <time.h>
 #include <unistd.h>
 #include <math.h>
 #ifdef SMP
@@ -287,13 +301,6 @@ typedef int blasint;
 #define COMPSIZE  2
 #endif
 
-#if defined(C_PGI) || defined(C_SUN)
-#define CREAL(X)	(*((FLOAT *)&X + 0))
-#define CIMAG(X)	(*((FLOAT *)&X + 1))
-#else
-#define CREAL	__real__
-#define CIMAG	__imag__
-#endif
 
 #define Address_H(x) (((x)+(1<<15))>>16)
 #define Address_L(x) ((x)-((Address_H(x))<<16))
@@ -307,8 +314,12 @@ typedef int blasint;
 #endif
 
 #if defined(OS_WINDOWS)
+#if defined(_MSC_VER) && !defined(__clang__)
+#define YIELDING    YieldProcessor()
+#else
 #define YIELDING	SwitchToThread()
 #endif
+#endif
 
 #if defined(ARMV7) || defined(ARMV6) || defined(ARMV8) || defined(ARMV5)
 #define YIELDING        asm volatile ("nop;nop;nop;nop;nop;nop;nop;nop; \n");
@@ -404,7 +415,51 @@ typedef char env_var_t[MAX_PATH];
 typedef char* env_var_t;
 #define readenv(p, n) ((p)=getenv(n))
 #endif
+
+#if !defined(RPCC_DEFINED) && !defined(OS_WINDOWS)
+#ifdef _POSIX_MONOTONIC_CLOCK
+#if defined(__GLIBC_PREREQ) // cut the if condition if two lines, otherwise will fail at __GLIBC_PREREQ(2, 17)
+#if __GLIBC_PREREQ(2, 17) // don't require -lrt
+#define USE_MONOTONIC
 #endif
+#elif defined(OS_ANDROID)
+#define USE_MONOTONIC
+#endif
+#endif
+/* use similar scale as x86 rdtsc for timeouts to work correctly */
+static inline unsigned long long rpcc(void){
+#ifdef USE_MONOTONIC
+  struct timespec ts;
+  clock_gettime(CLOCK_MONOTONIC, &ts);
+  return (unsigned long long)ts.tv_sec * 1000000000ull + ts.tv_nsec;
+#else
+  struct timeval tv;
+  gettimeofday(&tv,NULL);
+  return (unsigned long long)tv.tv_sec * 1000000000ull + tv.tv_usec * 1000;
+#endif
+}
+#define RPCC_DEFINED
+#define RPCC64BIT
+#endif // !RPCC_DEFINED
+
+#if !defined(BLAS_LOCK_DEFINED) && defined(__GNUC__)
+static void __inline blas_lock(volatile BLASULONG *address){
+
+  do {
+    while (*address) {YIELDING;};
+
+  } while (!__sync_bool_compare_and_swap(address, 0, 1));
+}
+#define BLAS_LOCK_DEFINED
+#endif
+
+#ifndef RPCC_DEFINED
+#error "rpcc() implementation is missing for your platform"
+#endif
+#ifndef BLAS_LOCK_DEFINED
+#error "blas_lock() implementation is missing for your platform"
+#endif
+#endif // !ASSEMBLER
 
 #ifdef OS_LINUX
 #include "common_linux.h"
@@ -450,18 +505,52 @@ typedef char* env_var_t;
 /* C99 supports complex floating numbers natively, which GCC also offers as an
    extension since version 3.0.  If neither are available, use a compatible
    structure as fallback (see Clause 6.2.5.13 of the C99 standard). */
-#if (defined(__STDC_IEC_559_COMPLEX__) || __STDC_VERSION__ >= 199901L || \
-     (__GNUC__ >= 3 && !defined(__cplusplus)))
+#if ((defined(__STDC_IEC_559_COMPLEX__) || __STDC_VERSION__ >= 199901L || \
+      (__GNUC__ >= 3 && !defined(__cplusplus))) && !(defined(FORCE_OPENBLAS_COMPLEX_STRUCT)))
   #define OPENBLAS_COMPLEX_C99
+  #ifndef __cplusplus
+    #include <complex.h>
+  #endif
   typedef float _Complex openblas_complex_float;
   typedef double _Complex openblas_complex_double;
   typedef xdouble _Complex openblas_complex_xdouble;
+  #define openblas_make_complex_float(real, imag)    ((real) + ((imag) * _Complex_I))
+  #define openblas_make_complex_double(real, imag)   ((real) + ((imag) * _Complex_I))
+  #define openblas_make_complex_xdouble(real, imag)  ((real) + ((imag) * _Complex_I))
 #else
   #define OPENBLAS_COMPLEX_STRUCT
   typedef struct { float real, imag; } openblas_complex_float;
   typedef struct { double real, imag; } openblas_complex_double;
   typedef struct { xdouble real, imag; } openblas_complex_xdouble;
+  #define openblas_make_complex_float(real, imag)    {(real), (imag)}
+  #define openblas_make_complex_double(real, imag)   {(real), (imag)}
+  #define openblas_make_complex_xdouble(real, imag)  {(real), (imag)}
 #endif
+
+#ifdef XDOUBLE
+#define OPENBLAS_COMPLEX_FLOAT openblas_complex_xdouble
+#define OPENBLAS_MAKE_COMPLEX_FLOAT(r,i) openblas_make_complex_xdouble(r,i)
+#elif defined(DOUBLE)
+#define OPENBLAS_COMPLEX_FLOAT openblas_complex_double
+#define OPENBLAS_MAKE_COMPLEX_FLOAT(r,i) openblas_make_complex_double(r,i)
+#else
+#define OPENBLAS_COMPLEX_FLOAT openblas_complex_float
+#define OPENBLAS_MAKE_COMPLEX_FLOAT(r,i) openblas_make_complex_float(r,i)
+#endif
+
+#if defined(C_PGI) || defined(C_SUN)
+#define CREAL(X)	(*((FLOAT *)&X + 0))
+#define CIMAG(X)	(*((FLOAT *)&X + 1))
+#else
+#ifdef OPENBLAS_COMPLEX_STRUCT
+#define CREAL(Z)	((Z).real)
+#define CIMAG(Z)	((Z).imag)
+#else
+#define CREAL	__real__
+#define CIMAG	__imag__
+#endif
+#endif
+
 #endif  // ASSEMBLER
 
 #ifndef IFLUSH
@@ -478,6 +567,10 @@ typedef char* env_var_t;
 #endif
 #endif
 
+#if defined(C_MSVC)
+#define inline __inline
+#endif
+
 #ifndef ASSEMBLER
 
 #ifndef MIN
@@ -499,6 +592,8 @@ void  blas_set_parameter(void);
 int   blas_get_cpu_number(void);
 void *blas_memory_alloc  (int);
 void  blas_memory_free   (void *);
+void *blas_memory_alloc_nolock  (int); //use malloc without blas_lock
+void  blas_memory_free_nolock   (void *);
 
 int  get_num_procs (void);
 
diff --git a/common_alpha.h b/common_alpha.h
index 845fb316a..9739c941d 100644
--- a/common_alpha.h
+++ b/common_alpha.h
@@ -76,6 +76,7 @@ static void __inline blas_lock(unsigned long *address){
     "30:", address);
 #endif
 }
+#define BLAS_LOCK_DEFINED
 
 static __inline unsigned int rpcc(void){
 
@@ -89,6 +90,7 @@ static __inline unsigned int rpcc(void){
 
   return r0;
 }
+#define RPCC_DEFINED
 
 
 #define HALT 	ldq	$0, 0($0)
diff --git a/common_arm.h b/common_arm.h
index 3cf15848a..6bf836835 100644
--- a/common_arm.h
+++ b/common_arm.h
@@ -1,5 +1,5 @@
 /*****************************************************************************
-Copyright (c) 2011-2014, The OpenBLAS Project
+Copyright (c) 2011-2015, The OpenBLAS Project
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -30,56 +30,29 @@ OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
 USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 **********************************************************************************/
 
-/*********************************************************************/
-/* Copyright 2009, 2010 The University of Texas at Austin.           */
-/* All rights reserved.                                              */
-/*                                                                   */
-/* Redistribution and use in source and binary forms, with or        */
-/* without modification, are permitted provided that the following   */
-/* conditions are met:                                               */
-/*                                                                   */
-/*   1. Redistributions of source code must retain the above         */
-/*      copyright notice, this list of conditions and the following  */
-/*      disclaimer.                                                  */
-/*                                                                   */
-/*   2. Redistributions in binary form must reproduce the above      */
-/*      copyright notice, this list of conditions and the following  */
-/*      disclaimer in the documentation and/or other materials       */
-/*      provided with the distribution.                              */
-/*                                                                   */
-/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
-/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
-/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
-/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
-/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
-/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
-/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
-/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
-/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
-/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
-/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
-/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
-/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
-/*    POSSIBILITY OF SUCH DAMAGE.                                    */
-/*                                                                   */
-/* The views and conclusions contained in the software and           */
-/* documentation are those of the authors and should not be          */
-/* interpreted as representing official policies, either expressed   */
-/* or implied, of The University of Texas at Austin.                 */
-/*********************************************************************/
-
 #ifndef COMMON_ARM
 #define COMMON_ARM
 
+#if defined(ARMV5) || defined(ARMV6)
+
 #define MB
 #define WMB
 
+#else
+
+#define MB   __asm__ __volatile__ ("dmb  ish" : : : "memory")
+#define WMB  __asm__ __volatile__ ("dmb  ishst" : : : "memory")
+
+#endif
+
 #define INLINE inline
 
 #define RETURN_BY_COMPLEX
 
 #ifndef ASSEMBLER
 
+#if defined(ARMV6) || defined(ARMV7) || defined(ARMV8)
+
 static void __inline blas_lock(volatile BLASULONG *address){
 
   int register ret;
@@ -88,37 +61,29 @@ static void __inline blas_lock(volatile BLASULONG *address){
     while (*address) {YIELDING;};
 
     __asm__ __volatile__(
-                         "ldrex r2, [%1]                                                \n\t"
-                         "mov   r2, #0                                                  \n\t"
-                         "strex r3, r2, [%1]                                            \n\t"
-			 "mov	%0 , r3							\n\t"
-                         : "=r"(ret), "=r"(address)
-                         : "1"(address)
-                         : "memory", "r2" , "r3"
-
-
+                         "ldrex r2, [%1]      \n\t"
+                         "strex %0, %2, [%1]  \n\t"
+                         "orr   %0, r2        \n\t"
+                         : "=&r"(ret)
+                         : "r"(address), "r"(1)
+                         : "memory", "r2"
     );
 
   } while (ret);
-
+  MB;
 }
 
-
-static inline unsigned long long rpcc(void){
-  unsigned long long ret=0;
-  double v;
-  struct timeval tv;
-  gettimeofday(&tv,NULL);
-  v=(double) tv.tv_sec + (double) tv.tv_usec * 1e-6;
-  ret = (unsigned long long) ( v * 1000.0d );
-  return ret;
-}
+#define BLAS_LOCK_DEFINED
+#endif
 
 static inline int blas_quickdivide(blasint x, blasint y){
   return x / y;
 }
 
-#if defined(DOUBLE)
+#if !defined(HAVE_VFP)
+/* no FPU, soft float */
+#define GET_IMAGE(res)
+#elif defined(DOUBLE)
 #define GET_IMAGE(res)  __asm__ __volatile__("vstr.f64 d1, %0" : "=m"(res) : : "memory")
 #else
 #define GET_IMAGE(res)  __asm__ __volatile__("vstr.f32 s1, %0" : "=m"(res) : : "memory")
@@ -166,4 +131,8 @@ REALNAME:
 #define MAP_ANONYMOUS MAP_ANON
 #endif
 
+#if !defined(ARMV5) && !defined(ARMV6) && !defined(ARMV7) && !defined(ARMV8)
+#error "you must define ARMV5, ARMV6, ARMV7 or ARMV8"
+#endif
+
 #endif
diff --git a/common_arm64.h b/common_arm64.h
index ae79c5309..15987c677 100644
--- a/common_arm64.h
+++ b/common_arm64.h
@@ -1,5 +1,5 @@
 /*****************************************************************************
-Copyright (c) 2011-2014, The OpenBLAS Project
+Copyright (c) 2011-2015, The OpenBLAS Project
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
@@ -30,49 +30,12 @@ OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
 USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 **********************************************************************************/
 
-/*********************************************************************/
-/* Copyright 2009, 2010 The University of Texas at Austin.           */
-/* All rights reserved.                                              */
-/*                                                                   */
-/* Redistribution and use in source and binary forms, with or        */
-/* without modification, are permitted provided that the following   */
-/* conditions are met:                                               */
-/*                                                                   */
-/*   1. Redistributions of source code must retain the above         */
-/*      copyright notice, this list of conditions and the following  */
-/*      disclaimer.                                                  */
-/*                                                                   */
-/*   2. Redistributions in binary form must reproduce the above      */
-/*      copyright notice, this list of conditions and the following  */
-/*      disclaimer in the documentation and/or other materials       */
-/*      provided with the distribution.                              */
-/*                                                                   */
-/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
-/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
-/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
-/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
-/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
-/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
-/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
-/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
-/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
-/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
-/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
-/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
-/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
-/*    POSSIBILITY OF SUCH DAMAGE.                                    */
-/*                                                                   */
-/* The views and conclusions contained in the software and           */
-/* documentation are those of the authors and should not be          */
-/* interpreted as representing official policies, either expressed   */
-/* or implied, of The University of Texas at Austin.                 */
-/*********************************************************************/
-
 #ifndef COMMON_ARM64
 #define COMMON_ARM64
 
-#define MB
-#define WMB
+#define MB   __asm__ __volatile__ ("dmb  ish" : : : "memory")
+#define WMB  __asm__ __volatile__ ("dmb  ishst" : : : "memory")
+
 
 #define INLINE inline
 
@@ -81,39 +44,27 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifndef ASSEMBLER
 
 static void __inline blas_lock(volatile BLASULONG *address){
-/*
-  int register ret;
+
+  long register ret;
 
   do {
     while (*address) {YIELDING;};
 
     __asm__ __volatile__(
-                         "ldrex r2, [%1]                                                \n\t"
-                         "mov   r2, #0                                                  \n\t"
-                         "strex r3, r2, [%1]                                            \n\t"
-			 "mov	%0 , r3							\n\t"
-                         : "=r"(ret), "=r"(address)
-                         : "1"(address)
-                         : "memory", "r2" , "r3"
-
-
+                         "ldaxr %0, [%1]      \n\t"
+                         "stlxr w2, %2, [%1]  \n\t"
+                         "orr   %0, %0, x2    \n\t"
+                         : "=r"(ret)
+                         : "r"(address), "r"(1l)
+                         : "memory", "x2"
     );
 
   } while (ret);
-*/
+  MB;
 }
+#define BLAS_LOCK_DEFINED
 
 
-static inline unsigned long long rpcc(void){
-  unsigned long long ret=0;
-  double v;
-  struct timeval tv;
-  gettimeofday(&tv,NULL);
-  v=(double) tv.tv_sec + (double) tv.tv_usec * 1e-6;
-  ret = (unsigned long long) ( v * 1000.0d );
-  return ret;
-}
-
 static inline int blas_quickdivide(blasint x, blasint y){
   return x / y;
 }
@@ -166,3 +117,4 @@ REALNAME:
 #endif
 
 #endif
+
diff --git a/common_c.h b/common_c.h
index 741d7d087..ce0f2a5bd 100644
--- a/common_c.h
+++ b/common_c.h
@@ -220,6 +220,15 @@
 #define COMATCOPY_K_CTC         comatcopy_k_ctc
 #define COMATCOPY_K_RTC         comatcopy_k_rtc
 
+#define CIMATCOPY_K_CN          cimatcopy_k_cn
+#define CIMATCOPY_K_RN          cimatcopy_k_rn
+#define CIMATCOPY_K_CT          cimatcopy_k_ct
+#define CIMATCOPY_K_RT          cimatcopy_k_rt
+#define CIMATCOPY_K_CNC         cimatcopy_k_cnc
+#define CIMATCOPY_K_RNC         cimatcopy_k_rnc
+#define CIMATCOPY_K_CTC         cimatcopy_k_ctc
+#define CIMATCOPY_K_RTC         cimatcopy_k_rtc
+
 #define CGEADD_K                cgeadd_k 
 
 #else
@@ -403,6 +412,16 @@
 #define COMATCOPY_K_RNC         gotoblas -> comatcopy_k_rnc
 #define COMATCOPY_K_CTC         gotoblas -> comatcopy_k_ctc
 #define COMATCOPY_K_RTC         gotoblas -> comatcopy_k_rtc
+
+#define CIMATCOPY_K_CN          gotoblas -> cimatcopy_k_cn
+#define CIMATCOPY_K_RN          gotoblas -> cimatcopy_k_rn
+#define CIMATCOPY_K_CT          gotoblas -> cimatcopy_k_ct
+#define CIMATCOPY_K_RT          gotoblas -> cimatcopy_k_rt
+#define CIMATCOPY_K_CNC         gotoblas -> cimatcopy_k_cnc
+#define CIMATCOPY_K_RNC         gotoblas -> cimatcopy_k_rnc
+#define CIMATCOPY_K_CTC         gotoblas -> cimatcopy_k_ctc
+#define CIMATCOPY_K_RTC         gotoblas -> cimatcopy_k_rtc
+
 #define CGEADD_K                gotoblas -> cgeadd_k 
 
 #endif
diff --git a/common_d.h b/common_d.h
index d6dfd7f04..ad9945186 100644
--- a/common_d.h
+++ b/common_d.h
@@ -149,6 +149,11 @@
 #define DOMATCOPY_K_RN		domatcopy_k_rn
 #define DOMATCOPY_K_CT		domatcopy_k_ct
 #define DOMATCOPY_K_RT		domatcopy_k_rt
+
+#define DIMATCOPY_K_CN		dimatcopy_k_cn
+#define DIMATCOPY_K_RN		dimatcopy_k_rn
+#define DIMATCOPY_K_CT      dimatcopy_k_ct
+#define DIMATCOPY_K_RT      dimatcopy_k_rt
 #define DGEADD_K                dgeadd_k 
 
 #else
@@ -267,6 +272,10 @@
 #define DOMATCOPY_K_RN		gotoblas -> domatcopy_k_rn
 #define DOMATCOPY_K_CT		gotoblas -> domatcopy_k_ct
 #define DOMATCOPY_K_RT		gotoblas -> domatcopy_k_rt
+#define DIMATCOPY_K_CN		gotoblas -> dimatcopy_k_cn
+#define DIMATCOPY_K_RN		gotoblas -> dimatcopy_k_rn
+#define DIMATCOPY_K_CT		gotoblas -> dimatcopy_k_ct
+#define DIMATCOPY_K_RT		gotoblas -> dimatcopy_k_rt
 
 #define DGEADD_K                gotoblas -> dgeadd_k 
 
diff --git a/common_ia64.h b/common_ia64.h
index 8e92b5992..72b75fc4e 100644
--- a/common_ia64.h
+++ b/common_ia64.h
@@ -68,6 +68,7 @@ static __inline void blas_lock(volatile unsigned long *address){
 			  : "ar.ccv", "memory");
   } while (ret);
 }
+#define BLAS_LOCK_DEFINED
 
 static __inline unsigned long rpcc(void) {
   unsigned long clocks;
@@ -75,6 +76,7 @@ static __inline unsigned long rpcc(void) {
   __asm__ __volatile__ ("mov %0=ar.itc" : "=r"(clocks));
   return clocks;
 }
+#define RPCC_DEFINED
 
 
 static __inline unsigned long stmxcsr(void){
@@ -99,10 +101,12 @@ static __inline void blas_lock(volatile unsigned long *address){
   while (*address || _InterlockedCompareExchange((volatile int *) address,1,0))
     ;
 }
+#define BLAS_LOCK_DEFINED
 
 static __inline unsigned int rpcc(void) {
   return __getReg(_IA64_REG_AR_ITC);
 }
+#define RPCC_DEFINED
 
 static __inline unsigned int stmxcsr(void) {
   return __getReg(_IA64_REG_AR_FPSR);
diff --git a/common_level1.h b/common_level1.h
index 2a1b4f1cf..32ffd6f18 100644
--- a/common_level1.h
+++ b/common_level1.h
@@ -47,12 +47,12 @@ double dsdot_k(BLASLONG, float   *, BLASLONG, float *, BLASLONG);
 double  ddot_k(BLASLONG, double  *, BLASLONG, double  *, BLASLONG);
 xdouble qdot_k(BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
 
-float   _Complex cdotc_k (BLASLONG, float  *, BLASLONG, float  *, BLASLONG);
-float   _Complex cdotu_k (BLASLONG, float  *, BLASLONG, float  *, BLASLONG);
-double  _Complex zdotc_k (BLASLONG, double *, BLASLONG, double *, BLASLONG);
-double  _Complex zdotu_k (BLASLONG, double *, BLASLONG, double *, BLASLONG);
-xdouble _Complex xdotc_k (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
-xdouble _Complex xdotu_k (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
+openblas_complex_float cdotc_k (BLASLONG, float  *, BLASLONG, float  *, BLASLONG);
+openblas_complex_float cdotu_k (BLASLONG, float  *, BLASLONG, float  *, BLASLONG);
+openblas_complex_double zdotc_k (BLASLONG, double *, BLASLONG, double *, BLASLONG);
+openblas_complex_double zdotu_k (BLASLONG, double *, BLASLONG, double *, BLASLONG);
+openblas_complex_xdouble xdotc_k (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
+openblas_complex_xdouble xdotu_k (BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
 
 int    saxpy_k (BLASLONG, BLASLONG, BLASLONG, float,
 	       float  *, BLASLONG, float  *, BLASLONG, float  *, BLASLONG);
diff --git a/common_level3.h b/common_level3.h
index e0ecbc4e2..1f5490baa 100644
--- a/common_level3.h
+++ b/common_level3.h
@@ -1736,31 +1736,55 @@ int somatcopy_k_cn(BLASLONG, BLASLONG,  float, float *, BLASLONG, float  *, BLAS
 int somatcopy_k_rn(BLASLONG, BLASLONG,  float, float *, BLASLONG, float  *, BLASLONG);
 int somatcopy_k_ct(BLASLONG, BLASLONG,  float, float *, BLASLONG, float  *, BLASLONG);
 int somatcopy_k_rt(BLASLONG, BLASLONG,  float, float *, BLASLONG, float  *, BLASLONG);
+int simatcopy_k_cn(BLASLONG, BLASLONG,  float, float *, BLASLONG);
+int simatcopy_k_rn(BLASLONG, BLASLONG,  float, float *, BLASLONG);
+int simatcopy_k_ct(BLASLONG, BLASLONG,  float, float *, BLASLONG);
+int simatcopy_k_rt(BLASLONG, BLASLONG,  float, float *, BLASLONG);
 
 int domatcopy_k_cn(BLASLONG, BLASLONG,  double, double *, BLASLONG, double  *, BLASLONG);
 int domatcopy_k_rn(BLASLONG, BLASLONG,  double, double *, BLASLONG, double  *, BLASLONG);
 int domatcopy_k_ct(BLASLONG, BLASLONG,  double, double *, BLASLONG, double  *, BLASLONG);
 int domatcopy_k_rt(BLASLONG, BLASLONG,  double, double *, BLASLONG, double  *, BLASLONG);
+int dimatcopy_k_cn(BLASLONG, BLASLONG,  double, double *, BLASLONG);
+int dimatcopy_k_rn(BLASLONG, BLASLONG,  double, double *, BLASLONG);
+int dimatcopy_k_ct(BLASLONG, BLASLONG,  double, double *, BLASLONG);
+int dimatcopy_k_rt(BLASLONG, BLASLONG,  double, double *, BLASLONG);
 
 int comatcopy_k_cn(BLASLONG, BLASLONG,  float, float, float *, BLASLONG, float  *, BLASLONG);
 int comatcopy_k_rn(BLASLONG, BLASLONG,  float, float, float *, BLASLONG, float  *, BLASLONG);
 int comatcopy_k_ct(BLASLONG, BLASLONG,  float, float, float *, BLASLONG, float  *, BLASLONG);
 int comatcopy_k_rt(BLASLONG, BLASLONG,  float, float, float *, BLASLONG, float  *, BLASLONG);
+int cimatcopy_k_cn(BLASLONG, BLASLONG,  float, float, float *, BLASLONG);
+int cimatcopy_k_rn(BLASLONG, BLASLONG,  float, float, float *, BLASLONG);
+int cimatcopy_k_ct(BLASLONG, BLASLONG,  float, float, float *, BLASLONG);
+int cimatcopy_k_rt(BLASLONG, BLASLONG,  float, float, float *, BLASLONG);
 
 int comatcopy_k_cnc(BLASLONG, BLASLONG,  float, float, float *, BLASLONG, float  *, BLASLONG);
 int comatcopy_k_rnc(BLASLONG, BLASLONG,  float, float, float *, BLASLONG, float  *, BLASLONG);
 int comatcopy_k_ctc(BLASLONG, BLASLONG,  float, float, float *, BLASLONG, float  *, BLASLONG);
 int comatcopy_k_rtc(BLASLONG, BLASLONG,  float, float, float *, BLASLONG, float  *, BLASLONG);
+int cimatcopy_k_cnc(BLASLONG, BLASLONG,  float, float, float *, BLASLONG);
+int cimatcopy_k_rnc(BLASLONG, BLASLONG,  float, float, float *, BLASLONG);
+int cimatcopy_k_ctc(BLASLONG, BLASLONG,  float, float, float *, BLASLONG);
+int cimatcopy_k_rtc(BLASLONG, BLASLONG,  float, float, float *, BLASLONG);
 
 int zomatcopy_k_cn(BLASLONG, BLASLONG,  double, double, double *, BLASLONG, double  *, BLASLONG);
 int zomatcopy_k_rn(BLASLONG, BLASLONG,  double, double, double *, BLASLONG, double  *, BLASLONG);
 int zomatcopy_k_ct(BLASLONG, BLASLONG,  double, double, double *, BLASLONG, double  *, BLASLONG);
 int zomatcopy_k_rt(BLASLONG, BLASLONG,  double, double, double *, BLASLONG, double  *, BLASLONG);
+int zimatcopy_k_cn(BLASLONG, BLASLONG,  double, double, double *, BLASLONG);
+int zimatcopy_k_rn(BLASLONG, BLASLONG,  double, double, double *, BLASLONG);
+int zimatcopy_k_ct(BLASLONG, BLASLONG,  double, double, double *, BLASLONG);
+int zimatcopy_k_rt(BLASLONG, BLASLONG,  double, double, double *, BLASLONG);
 
 int zomatcopy_k_cnc(BLASLONG, BLASLONG,  double, double, double *, BLASLONG, double  *, BLASLONG);
 int zomatcopy_k_rnc(BLASLONG, BLASLONG,  double, double, double *, BLASLONG, double  *, BLASLONG);
 int zomatcopy_k_ctc(BLASLONG, BLASLONG,  double, double, double *, BLASLONG, double  *, BLASLONG);
 int zomatcopy_k_rtc(BLASLONG, BLASLONG,  double, double, double *, BLASLONG, double  *, BLASLONG);
+int zimatcopy_k_cnc(BLASLONG, BLASLONG,  double, double, double *, BLASLONG);
+int zimatcopy_k_rnc(BLASLONG, BLASLONG,  double, double, double *, BLASLONG);
+int zimatcopy_k_ctc(BLASLONG, BLASLONG,  double, double, double *, BLASLONG);
+int zimatcopy_k_rtc(BLASLONG, BLASLONG,  double, double, double *, BLASLONG);
 
 int sgeadd_k(BLASLONG, BLASLONG, float, float*, BLASLONG, float, float *, BLASLONG); 
 int dgeadd_k(BLASLONG, BLASLONG, double, double*, BLASLONG, double, double *, BLASLONG); 
diff --git a/common_macro.h b/common_macro.h
index 8555baa67..4976e766f 100644
--- a/common_macro.h
+++ b/common_macro.h
@@ -634,6 +634,11 @@
 #define OMATCOPY_K_RN		DOMATCOPY_K_RN
 #define OMATCOPY_K_CT		DOMATCOPY_K_CT
 #define OMATCOPY_K_RT		DOMATCOPY_K_RT
+#define IMATCOPY_K_CN		DIMATCOPY_K_CN
+#define IMATCOPY_K_RN		DIMATCOPY_K_RN
+#define IMATCOPY_K_CT		DIMATCOPY_K_CT
+#define IMATCOPY_K_RT		DIMATCOPY_K_RT
+
 #define GEADD_K                 DGEADD_K 
 #else
 
@@ -931,6 +936,10 @@
 #define OMATCOPY_K_RN		SOMATCOPY_K_RN
 #define OMATCOPY_K_CT		SOMATCOPY_K_CT
 #define OMATCOPY_K_RT		SOMATCOPY_K_RT
+#define IMATCOPY_K_CN		SIMATCOPY_K_CN
+#define IMATCOPY_K_RN		SIMATCOPY_K_RN
+#define IMATCOPY_K_CT		SIMATCOPY_K_CT
+#define IMATCOPY_K_RT		SIMATCOPY_K_RT
 
 #define GEADD_K 		SGEADD_K 
 #endif
@@ -1747,6 +1756,15 @@
 #define OMATCOPY_K_RNC		ZOMATCOPY_K_RNC
 #define OMATCOPY_K_CTC		ZOMATCOPY_K_CTC
 #define OMATCOPY_K_RTC		ZOMATCOPY_K_RTC
+#define IMATCOPY_K_CN		ZIMATCOPY_K_CN
+#define IMATCOPY_K_RN		ZIMATCOPY_K_RN
+#define IMATCOPY_K_CT		ZIMATCOPY_K_CT
+#define IMATCOPY_K_RT		ZIMATCOPY_K_RT
+#define IMATCOPY_K_CNC		ZIMATCOPY_K_CNC
+#define IMATCOPY_K_RNC		ZIMATCOPY_K_RNC
+#define IMATCOPY_K_CTC		ZIMATCOPY_K_CTC
+#define IMATCOPY_K_RTC		ZIMATCOPY_K_RTC
+
 #define GEADD_K                 ZGEADD_K 
 
 #else
@@ -2160,6 +2178,14 @@
 #define OMATCOPY_K_RNC		COMATCOPY_K_RNC
 #define OMATCOPY_K_CTC		COMATCOPY_K_CTC
 #define OMATCOPY_K_RTC		COMATCOPY_K_RTC
+#define IMATCOPY_K_CN		CIMATCOPY_K_CN
+#define IMATCOPY_K_RN		CIMATCOPY_K_RN
+#define IMATCOPY_K_CT		CIMATCOPY_K_CT
+#define IMATCOPY_K_RT		CIMATCOPY_K_RT
+#define IMATCOPY_K_CNC		CIMATCOPY_K_CNC
+#define IMATCOPY_K_RNC		CIMATCOPY_K_RNC
+#define IMATCOPY_K_CTC		CIMATCOPY_K_CTC
+#define IMATCOPY_K_RTC		CIMATCOPY_K_RTC
 
 #define GEADD_K                 CGEADD_K 
 
diff --git a/common_mips64.h b/common_mips64.h
index 7cd86b375..f5c0ec7cf 100644
--- a/common_mips64.h
+++ b/common_mips64.h
@@ -98,6 +98,7 @@ static void INLINE blas_lock(volatile unsigned long *address){
 
   } while (ret);
 }
+#define BLAS_LOCK_DEFINED
 
 static inline unsigned int rpcc(void){
   unsigned long ret;
@@ -118,6 +119,7 @@ static inline unsigned int rpcc(void){
 #endif
   return ret;
 }
+#define RPCC_DEFINED
 
 #if defined(LOONGSON3A) || defined(LOONGSON3B)
 #ifndef NO_AFFINITY
diff --git a/common_param.h b/common_param.h
index 1b56e85f0..36d6149ea 100644
--- a/common_param.h
+++ b/common_param.h
@@ -855,6 +855,36 @@ BLASLONG (*ixamin_k)(BLASLONG, xdouble *, BLASLONG);
   int    (*zomatcopy_k_rnc)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG, double*, BLASLONG);
   int    (*zomatcopy_k_rtc)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG, double*, BLASLONG);
 
+  int    (*simatcopy_k_cn)	(BLASLONG, BLASLONG, float, float*, BLASLONG);
+  int    (*simatcopy_k_ct)	(BLASLONG, BLASLONG, float, float*, BLASLONG);
+  int    (*simatcopy_k_rn)	(BLASLONG, BLASLONG, float, float*, BLASLONG);
+  int    (*simatcopy_k_rt)	(BLASLONG, BLASLONG, float, float*, BLASLONG);
+
+  int    (*dimatcopy_k_cn)	(BLASLONG, BLASLONG, double, double*, BLASLONG);
+  int    (*dimatcopy_k_ct)	(BLASLONG, BLASLONG, double, double*, BLASLONG);
+  int    (*dimatcopy_k_rn)	(BLASLONG, BLASLONG, double, double*, BLASLONG);
+  int    (*dimatcopy_k_rt)	(BLASLONG, BLASLONG, double, double*, BLASLONG);
+
+  int    (*cimatcopy_k_cn)	(BLASLONG, BLASLONG, float, float, float*, BLASLONG);
+  int    (*cimatcopy_k_ct)	(BLASLONG, BLASLONG, float, float, float*, BLASLONG);
+  int    (*cimatcopy_k_rn)	(BLASLONG, BLASLONG, float, float, float*, BLASLONG);
+  int    (*cimatcopy_k_rt)	(BLASLONG, BLASLONG, float, float, float*, BLASLONG);
+
+  int    (*cimatcopy_k_cnc)	(BLASLONG, BLASLONG, float, float, float*, BLASLONG);
+  int    (*cimatcopy_k_ctc)	(BLASLONG, BLASLONG, float, float, float*, BLASLONG);
+  int    (*cimatcopy_k_rnc)	(BLASLONG, BLASLONG, float, float, float*, BLASLONG);
+  int    (*cimatcopy_k_rtc)	(BLASLONG, BLASLONG, float, float, float*, BLASLONG);
+
+  int    (*zimatcopy_k_cn)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG);
+  int    (*zimatcopy_k_ct)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG);
+  int    (*zimatcopy_k_rn)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG);
+  int    (*zimatcopy_k_rt)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG);
+
+  int    (*zimatcopy_k_cnc)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG);
+  int    (*zimatcopy_k_ctc)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG);
+  int    (*zimatcopy_k_rnc)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG);
+  int    (*zimatcopy_k_rtc)	(BLASLONG, BLASLONG, double, double, double*, BLASLONG);
+
   int    (*sgeadd_k) (BLASLONG, BLASLONG, float, float *, BLASLONG, float, float *, BLASLONG); 
   int    (*dgeadd_k) (BLASLONG, BLASLONG, double, double *, BLASLONG, double, double *, BLASLONG); 
   int    (*cgeadd_k) (BLASLONG, BLASLONG, float, float,  float *,  BLASLONG, float, float, float *, BLASLONG); 
diff --git a/common_power.h b/common_power.h
index f88f527bd..ab331b04a 100644
--- a/common_power.h
+++ b/common_power.h
@@ -87,6 +87,7 @@ static void INLINE blas_lock(volatile unsigned long *address){
 #endif
   } while (ret);
 }
+#define BLAS_LOCK_DEFINED
 
 static inline unsigned long rpcc(void){
   unsigned long ret;
@@ -103,6 +104,7 @@ static inline unsigned long rpcc(void){
 #endif
 
 }
+#define RPCC_DEFINED
 
 #ifdef __64BIT__
 #define RPCC64BIT
@@ -495,6 +497,15 @@ static inline int blas_quickdivide(blasint x, blasint y){
 REALNAME:
 #define EPILOGUE	.size	REALNAME, .-REALNAME
 #else
+#if _CALL_ELF == 2
+#define PROLOGUE \
+	.section .text;\
+	.align 6;\
+	.globl	REALNAME;\
+	.type	REALNAME, @function;\
+REALNAME:
+#define EPILOGUE	.size	REALNAME, .-REALNAME
+#else
 #define PROLOGUE \
 	.section .text;\
 	.align 5;\
@@ -514,6 +525,7 @@ REALNAME:;\
 	.size	.REALNAME, .-.REALNAME; \
 	.section	.note.GNU-stack,"",@progbits
 #endif
+#endif
 
 #ifdef PROFILE
 #ifndef __64BIT__
@@ -792,4 +804,25 @@ Lmcount$lazy_ptr:
 #ifndef MAP_ANONYMOUS
 #define MAP_ANONYMOUS MAP_ANON
 #endif
+
+#ifdef OS_LINUX
+#ifndef __64BIT__
+#define FRAMESLOT(X) (((X) * 4) + 8)
+#else
+#if _CALL_ELF == 2
+#define FRAMESLOT(X) (((X) * 8) + 96)
+#else
+#define FRAMESLOT(X) (((X) * 8) + 112)
+#endif
+#endif
+#endif
+
+#if defined(OS_AIX) || defined(OS_DARWIN)
+#ifndef __64BIT__
+#define FRAMESLOT(X) (((X) * 4) + 56)
+#else
+#define FRAMESLOT(X) (((X) * 8) + 112)
+#endif
+#endif
+
 #endif
diff --git a/common_s.h b/common_s.h
index a4d8679b7..3c1600859 100644
--- a/common_s.h
+++ b/common_s.h
@@ -152,6 +152,10 @@
 #define SOMATCOPY_K_RN          somatcopy_k_rn
 #define SOMATCOPY_K_CT          somatcopy_k_ct
 #define SOMATCOPY_K_RT          somatcopy_k_rt
+#define SIMATCOPY_K_CN          simatcopy_k_cn
+#define SIMATCOPY_K_RN          simatcopy_k_rn
+#define SIMATCOPY_K_CT          simatcopy_k_ct
+#define SIMATCOPY_K_RT          simatcopy_k_rt
 
 #define SGEADD_K                sgeadd_k 
 
@@ -274,6 +278,10 @@
 #define SOMATCOPY_K_RN          gotoblas -> somatcopy_k_rn
 #define SOMATCOPY_K_CT          gotoblas -> somatcopy_k_ct
 #define SOMATCOPY_K_RT          gotoblas -> somatcopy_k_rt
+#define SIMATCOPY_K_CN          gotoblas -> simatcopy_k_cn
+#define SIMATCOPY_K_RN          gotoblas -> simatcopy_k_rn
+#define SIMATCOPY_K_CT          gotoblas -> simatcopy_k_ct
+#define SIMATCOPY_K_RT          gotoblas -> simatcopy_k_rt
 
 #define SGEADD_K                gotoblas -> sgeadd_k 
 
diff --git a/common_sparc.h b/common_sparc.h
index 87ef75276..f99972db9 100644
--- a/common_sparc.h
+++ b/common_sparc.h
@@ -58,6 +58,7 @@ static void __inline blas_lock(volatile unsigned long *address){
 			 : "memory");
   } while (ret);
 }
+#define BLAS_LOCK_DEFINED
 
 static __inline unsigned long rpcc(void){
   unsigned long clocks;
@@ -66,6 +67,7 @@ static __inline unsigned long rpcc(void){
 
   return clocks;
 };
+#define RPCC_DEFINED
 
 #ifdef __64BIT__
 #define RPCC64BIT
diff --git a/common_x86.h b/common_x86.h
index 9d82090cc..1ace84cad 100644
--- a/common_x86.h
+++ b/common_x86.h
@@ -56,41 +56,67 @@ static void __inline blas_lock(volatile BLASULONG *address){
   do {
     while (*address) {YIELDING;};
 
+#if defined(_MSC_VER) && !defined(__clang__)
+	// use intrinsic instead of inline assembly
+	ret = _InterlockedExchange(address, 1);
+	// inline assembly
+	/*__asm {
+		mov eax, address
+		mov ebx, 1
+		xchg [eax], ebx
+		mov ret, ebx
+	}*/
+#else
     __asm__ __volatile__(
 			 "xchgl %0, %1\n"
 			 : "=r"(ret), "=m"(*address)
 			 : "0"(1), "m"(*address)
 			 : "memory");
+#endif
 
   } while (ret);
 
 }
+#define BLAS_LOCK_DEFINED
 
 static __inline unsigned long long rpcc(void){
+#if defined(_MSC_VER) && !defined(__clang__)
+  return __rdtsc(); // use MSVC intrinsic
+#else
   unsigned int a, d;
 
   __asm__ __volatile__ ("rdtsc" : "=a" (a), "=d" (d));
 
   return ((unsigned long long)a + ((unsigned long long)d << 32));
+#endif
 };
+#define RPCC_DEFINED
 
 static __inline unsigned long getstackaddr(void){
+#if defined(_MSC_VER) && !defined(__clang__)
+  return (unsigned long)_ReturnAddress(); // use MSVC intrinsic
+#else
   unsigned long addr;
 
   __asm__ __volatile__ ("mov %%esp, %0"
 			 : "=r"(addr) : : "memory");
 
   return addr;
+#endif
 };
 
 
 static __inline long double sqrt_long(long double val) {
+#if defined(_MSC_VER) && !defined(__clang__)
+  return sqrt(val); // not sure if this will use fsqrt
+#else
   long double result;
 
   __asm__ __volatile__ ("fldt %1\n"
 		    "fsqrt\n"
 		    "fstpt %0\n" : "=m" (result) : "m"(val));
   return result;
+#endif
 }
 
 #define SQRT(a)  sqrt_long(a)
@@ -100,7 +126,7 @@ void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx);
 
 #define WHEREAMI
 
-static inline int WhereAmI(void){
+static __inline int WhereAmI(void){
   int eax, ebx, ecx, edx;
   int apicid;
 
@@ -146,9 +172,14 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 
   y = blas_quick_divide_table[y];
 
+#if defined(_MSC_VER) && !defined(__clang__)
+  (void*)result;
+  return x*y;
+#else
   __asm__ __volatile__  ("mull %0" :"=d" (result) :"a"(x), "0" (y));
 
   return result;
+#endif
 }
 #endif
 
@@ -171,7 +202,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 #define MMXSTORE	movd
 #endif
 
-#if defined(PILEDRIVER) || defined(BULLDOZER) || defined(STEAMROLLER)
+#if defined(PILEDRIVER) || defined(BULLDOZER) || defined(STEAMROLLER) || defined(EXCAVATOR)
 //Enable some optimazation for barcelona.
 #define BARCELONA_OPTIMIZATION
 #endif
@@ -284,8 +315,12 @@ REALNAME:
 
 #define PROFCODE
 
+#ifdef __clang__
+#define EPILOGUE .end
+#else
 #define EPILOGUE .end	 REALNAME
 #endif
+#endif
 
 #if defined(OS_LINUX) || defined(OS_FREEBSD) || defined(OS_NETBSD) || defined(__ELF__)
 #define PROLOGUE \
diff --git a/common_x86_64.h b/common_x86_64.h
index e0a6c4c42..da9afc0e4 100644
--- a/common_x86_64.h
+++ b/common_x86_64.h
@@ -41,6 +41,10 @@
 
 #ifndef ASSEMBLER
 
+#ifdef C_MSVC
+#include <intrin.h>
+#endif
+
 #ifdef C_SUN
 #define	__asm__ __asm
 #define	__volatile__
@@ -61,30 +65,45 @@
 
 static void __inline blas_lock(volatile BLASULONG *address){
 
+#ifndef C_MSVC
   int ret;
+#else
+  BLASULONG ret;
+#endif
 
   do {
     while (*address) {YIELDING;};
 
+#ifndef C_MSVC
     __asm__ __volatile__(
 			 "xchgl %0, %1\n"
 			 : "=r"(ret), "=m"(*address)
 			 : "0"(1), "m"(*address)
 			 : "memory");
-
+#else
+    ret=InterlockedExchange64((volatile LONG64 *)(address), 1);
+#endif
   } while (ret);
+
 }
+#define BLAS_LOCK_DEFINED
 
 static __inline BLASULONG rpcc(void){
+#ifdef C_MSVC
+  return __rdtsc();
+#else
   BLASULONG a, d;
 
   __asm__ __volatile__ ("rdtsc" : "=a" (a), "=d" (d));
 
   return ((BLASULONG)a + ((BLASULONG)d << 32));
+#endif
 }
+#define RPCC_DEFINED
 
 #define RPCC64BIT
 
+#ifndef C_MSVC
 static __inline BLASULONG getstackaddr(void){
   BLASULONG addr;
 
@@ -93,22 +112,32 @@ static __inline BLASULONG getstackaddr(void){
 
   return addr;
 }
+#endif
 
 static __inline void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx){
 
+#ifdef C_MSVC
+  int cpuinfo[4];
+  __cpuid(cpuinfo, op);
+  *eax=cpuinfo[0];
+  *ebx=cpuinfo[1];
+  *ecx=cpuinfo[2];
+  *edx=cpuinfo[3];
+#else
         __asm__ __volatile__("cpuid"
 			     : "=a" (*eax),
 			     "=b" (*ebx),
 			     "=c" (*ecx),
 			     "=d" (*edx)
 			     : "0" (op));
+#endif
 }
 
 /*
 #define WHEREAMI
 */
 
-static inline int WhereAmI(void){
+static __inline int WhereAmI(void){
   int eax, ebx, ecx, edx;
   int apicid;
 
@@ -150,10 +179,14 @@ static inline int WhereAmI(void){
 #define GET_IMAGE_CANCEL
 
 #ifdef SMP
-#ifdef USE64BITINT
+#if defined(USE64BITINT)
 static __inline blasint blas_quickdivide(blasint x, blasint y){
   return x / y;
 }
+#elif defined (C_MSVC)
+static __inline BLASLONG blas_quickdivide(BLASLONG x, BLASLONG y){
+  return x / y;
+}
 #else
 extern unsigned int blas_quick_divide_table[];
 
@@ -226,7 +259,7 @@ static __inline int blas_quickdivide(unsigned int x, unsigned int y){
 
 #ifdef ASSEMBLER
 
-#if defined(PILEDRIVER) || defined(BULLDOZER) || defined(STEAMROLLER)
+#if defined(PILEDRIVER) || defined(BULLDOZER) || defined(STEAMROLLER) || defined(EXCAVATOR)
 //Enable some optimazation for barcelona.
 #define BARCELONA_OPTIMIZATION
 #endif
diff --git a/common_z.h b/common_z.h
index b17122776..b4f58bb0c 100644
--- a/common_z.h
+++ b/common_z.h
@@ -220,6 +220,15 @@
 #define ZOMATCOPY_K_CTC         zomatcopy_k_ctc
 #define ZOMATCOPY_K_RTC         zomatcopy_k_rtc
 
+#define ZIMATCOPY_K_CN          zimatcopy_k_cn
+#define ZIMATCOPY_K_RN          zimatcopy_k_rn
+#define ZIMATCOPY_K_CT          zimatcopy_k_ct
+#define ZIMATCOPY_K_RT          zimatcopy_k_rt
+#define ZIMATCOPY_K_CNC         zimatcopy_k_cnc
+#define ZIMATCOPY_K_RNC         zimatcopy_k_rnc
+#define ZIMATCOPY_K_CTC         zimatcopy_k_ctc
+#define ZIMATCOPY_K_RTC         zimatcopy_k_rtc
+
 #define ZGEADD_K                zgeadd_k 
 
 #else
@@ -404,6 +413,15 @@
 #define ZOMATCOPY_K_CTC         gotoblas -> zomatcopy_k_ctc
 #define ZOMATCOPY_K_RTC         gotoblas -> zomatcopy_k_rtc
 
+#define ZIMATCOPY_K_CN          gotoblas -> zimatcopy_k_cn
+#define ZIMATCOPY_K_RN          gotoblas -> zimatcopy_k_rn
+#define ZIMATCOPY_K_CT          gotoblas -> zimatcopy_k_ct
+#define ZIMATCOPY_K_RT          gotoblas -> zimatcopy_k_rt
+#define ZIMATCOPY_K_CNC         gotoblas -> zimatcopy_k_cnc
+#define ZIMATCOPY_K_RNC         gotoblas -> zimatcopy_k_rnc
+#define ZIMATCOPY_K_CTC         gotoblas -> zimatcopy_k_ctc
+#define ZIMATCOPY_K_RTC         gotoblas -> zimatcopy_k_rtc
+
 #define ZGEADD_K                gotoblas -> zgeadd_k
 
 #endif
diff --git a/cpuid.h b/cpuid.h
index ab6a3fb32..e9bd2d016 100644
--- a/cpuid.h
+++ b/cpuid.h
@@ -39,6 +39,10 @@
 #ifndef CPUID_H
 #define CPUID_H
 
+#if defined(__i386__) || defined(__x86_64__) || defined(_M_IX86) || defined(_M_X64)
+#define INTEL_AMD
+#endif
+
 #define VENDOR_INTEL      1
 #define VENDOR_UMC        2
 #define VENDOR_AMD        3
@@ -59,7 +63,7 @@
 #define FAMILY_PM     7
 #define FAMILY_IA64   8
 
-#if defined(__i386__) || defined(__x86_64__)
+#ifdef INTEL_AMD
 #define GET_EXFAMILY  1
 #define GET_EXMODEL   2
 #define GET_TYPE      3
@@ -109,6 +113,7 @@
 #define CORE_PILEDRIVER  23
 #define CORE_HASWELL     24
 #define CORE_STEAMROLLER 25
+#define CORE_EXCAVATOR   26
 
 #define HAVE_SSE      (1 <<  0)
 #define HAVE_SSE2     (1 <<  1)
@@ -203,5 +208,6 @@ typedef struct {
 #define CPUTYPE_PILEDRIVER              47
 #define CPUTYPE_HASWELL 		48
 #define CPUTYPE_STEAMROLLER 		49
+#define CPUTYPE_EXCAVATOR 		50
 
 #endif
diff --git a/cpuid_arm.c b/cpuid_arm.c
index 51ba72d70..6485003f3 100644
--- a/cpuid_arm.c
+++ b/cpuid_arm.c
@@ -192,6 +192,7 @@ void get_cpuconfig(void)
 	{
 	       case CPU_CORTEXA9:
     			printf("#define CORTEXA9\n");
+    			printf("#define ARMV7\n");
     			printf("#define HAVE_VFP\n");
     			printf("#define HAVE_VFPV3\n");
 			if ( get_feature("neon"))	printf("#define HAVE_NEON\n");
@@ -207,6 +208,7 @@ void get_cpuconfig(void)
 
 	       case CPU_CORTEXA15:
     			printf("#define CORTEXA15\n");
+    			printf("#define ARMV7\n");
     			printf("#define HAVE_VFP\n");
     			printf("#define HAVE_VFPV3\n");
 			if ( get_feature("neon"))	printf("#define HAVE_NEON\n");
diff --git a/cpuid_power.c b/cpuid_power.c
index 2fc333dd2..366c6ed08 100644
--- a/cpuid_power.c
+++ b/cpuid_power.c
@@ -115,6 +115,7 @@ int detect(void){
   if (!strncasecmp(p, "POWER5", 6)) return CPUTYPE_POWER5;
   if (!strncasecmp(p, "POWER6", 6)) return CPUTYPE_POWER6;
   if (!strncasecmp(p, "POWER7", 6)) return CPUTYPE_POWER6;
+  if (!strncasecmp(p, "POWER8", 6)) return CPUTYPE_POWER6;
   if (!strncasecmp(p, "Cell",   4)) return CPUTYPE_CELL;
   if (!strncasecmp(p, "7447",   4)) return CPUTYPE_PPCG4;
 
diff --git a/cpuid_x86.c b/cpuid_x86.c
index ef90b26d8..a65991041 100644
--- a/cpuid_x86.c
+++ b/cpuid_x86.c
@@ -40,6 +40,12 @@
 #include <string.h>
 #include "cpuid.h"
 
+#if defined(_MSC_VER) && !defined(__clang__)
+#define C_INLINE __inline
+#else
+#define C_INLINE inline
+#endif
+
 /*
 #ifdef NO_AVX
 #define CPUTYPE_HASWELL CPUTYPE_NEHALEM
@@ -53,12 +59,26 @@
 #endif
 */
 
+#if defined(_MSC_VER) && !defined(__clang__)
+
+void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx)
+{
+  int cpuInfo[4] = {-1};
+  __cpuid(cpuInfo, op);
+  *eax = cpuInfo[0];
+  *ebx = cpuInfo[1];
+  *ecx = cpuInfo[2];
+  *edx = cpuInfo[3];
+}
+
+#else
+
 #ifndef CPUIDEMU
 
 #if defined(__APPLE__) && defined(__i386__)
 void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx);
 #else
-static inline void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx){
+static C_INLINE void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx){
 #if defined(__i386__) && defined(__PIC__)
   __asm__ __volatile__
     ("mov %%ebx, %%edi;"
@@ -115,14 +135,16 @@ void cpuid(unsigned int op, unsigned int *eax, unsigned int *ebx, unsigned int *
 
 #endif
 
-static inline int have_cpuid(void){
+#endif // _MSC_VER
+
+static C_INLINE int have_cpuid(void){
   int eax, ebx, ecx, edx;
 
   cpuid(0, &eax, &ebx, &ecx, &edx);
   return eax;
 }
 
-static inline int have_excpuid(void){
+static C_INLINE int have_excpuid(void){
   int eax, ebx, ecx, edx;
 
   cpuid(0x80000000, &eax, &ebx, &ecx, &edx);
@@ -130,10 +152,14 @@ static inline int have_excpuid(void){
 }
 
 #ifndef NO_AVX
-static inline void xgetbv(int op, int * eax, int * edx){
+static C_INLINE void xgetbv(int op, int * eax, int * edx){
   //Use binary code for xgetbv
+#if defined(_MSC_VER) && !defined(__clang__)
+  *eax = __xgetbv(op);
+#else
   __asm__ __volatile__
     (".byte 0x0f, 0x01, 0xd0": "=a" (*eax), "=d" (*edx) : "c" (op) : "cc");
+#endif
 }
 #endif
 
@@ -1098,6 +1124,16 @@ int get_cpuname(void){
             return CPUTYPE_HASWELL;
 #else
 	    return CPUTYPE_SANDYBRIDGE;
+#endif
+          else
+	    return CPUTYPE_NEHALEM;
+	case 13:
+	  //Broadwell
+          if(support_avx())
+#ifndef NO_AVX2
+            return CPUTYPE_HASWELL;
+#else
+	    return CPUTYPE_SANDYBRIDGE;
 #endif
           else
 	    return CPUTYPE_NEHALEM;
@@ -1112,11 +1148,57 @@ int get_cpuname(void){
             return CPUTYPE_HASWELL;
 #else
 	    return CPUTYPE_SANDYBRIDGE;
+#endif
+          else
+	    return CPUTYPE_NEHALEM;
+	case 7:
+	case 15:
+	  //Broadwell
+          if(support_avx())
+#ifndef NO_AVX2
+            return CPUTYPE_HASWELL;
+#else
+	    return CPUTYPE_SANDYBRIDGE;
+#endif
+          else
+	    return CPUTYPE_NEHALEM;
+	case 14:
+	  //Skylake
+          if(support_avx())
+#ifndef NO_AVX2
+            return CPUTYPE_HASWELL;
+#else
+	    return CPUTYPE_SANDYBRIDGE;
 #endif
           else
 	    return CPUTYPE_NEHALEM;
         }
         break;
+      case 5:
+        switch (model) {
+	case 6:
+	  //Broadwell
+          if(support_avx())
+#ifndef NO_AVX2
+            return CPUTYPE_HASWELL;
+#else
+	    return CPUTYPE_SANDYBRIDGE;
+#endif
+          else
+	    return CPUTYPE_NEHALEM;
+	case 5:
+        case 14:
+	  // Skylake
+          if(support_avx())
+#ifndef NO_AVX2
+            return CPUTYPE_HASWELL;
+#else
+	    return CPUTYPE_SANDYBRIDGE;
+#endif
+          else
+	    return CPUTYPE_NEHALEM;
+	}
+	break;
       }
       break;
     case 0x7:
@@ -1163,11 +1245,20 @@ int get_cpuname(void){
 	  else
 	    return CPUTYPE_BARCELONA; //OS don't support AVX.
 	case 0:
-	  if(support_avx())
-	    return CPUTYPE_STEAMROLLER;
-	  else
-	    return CPUTYPE_BARCELONA; //OS don't support AVX.
+	  switch(exmodel){
+	  case 3:
+	    if(support_avx())
+	      return CPUTYPE_STEAMROLLER;
+	    else
+	      return CPUTYPE_BARCELONA; //OS don't support AVX.
 
+	  case 6:
+	    if(support_avx())
+	      return CPUTYPE_EXCAVATOR;
+	    else
+	      return CPUTYPE_BARCELONA; //OS don't support AVX.
+	  }
+	  break;
 	}
 	break;
       case  5:
@@ -1297,6 +1388,7 @@ static char *cpuname[] = {
   "PILEDRIVER",
   "HASWELL",
   "STEAMROLLER",
+  "EXCAVATOR",
 };
 
 static char *lowercpuname[] = {
@@ -1349,6 +1441,7 @@ static char *lowercpuname[] = {
   "piledriver",
   "haswell",
   "steamroller",
+  "excavator",
 };
 
 static char *corename[] = {
@@ -1378,6 +1471,7 @@ static char *corename[] = {
   "PILEDRIVER",
   "HASWELL",
   "STEAMROLLER",
+  "EXCAVATOR",
 };
 
 static char *corename_lower[] = {
@@ -1407,6 +1501,7 @@ static char *corename_lower[] = {
   "piledriver",
   "haswell",
   "steamroller",
+  "excavator",
 };
 
 
@@ -1525,6 +1620,16 @@ int get_coretype(void){
             return CORE_HASWELL;
 #else
 	    return CORE_SANDYBRIDGE;
+#endif
+          else
+	    return CORE_NEHALEM;
+	case 13:
+	  //broadwell
+          if(support_avx())
+#ifndef NO_AVX2
+            return CORE_HASWELL;
+#else
+	    return CORE_SANDYBRIDGE;
 #endif
           else
 	    return CORE_NEHALEM;
@@ -1539,11 +1644,57 @@ int get_coretype(void){
             return CORE_HASWELL;
 #else
 	    return CORE_SANDYBRIDGE;
+#endif
+          else
+	    return CORE_NEHALEM;
+	case 7:
+	case 15:
+	  //broadwell
+          if(support_avx())
+#ifndef NO_AVX2
+            return CORE_HASWELL;
+#else
+	    return CORE_SANDYBRIDGE;
+#endif
+          else
+	    return CORE_NEHALEM;
+	case 14:
+	  //Skylake
+          if(support_avx())
+#ifndef NO_AVX2
+            return CORE_HASWELL;
+#else
+	    return CORE_SANDYBRIDGE;
 #endif
           else
 	    return CORE_NEHALEM;
         }
         break;
+      case 5:
+        switch (model) {
+	case 6:
+	  //broadwell
+          if(support_avx())
+#ifndef NO_AVX2
+            return CORE_HASWELL;
+#else
+	    return CORE_SANDYBRIDGE;
+#endif
+          else
+	    return CORE_NEHALEM;
+	case 5:
+	case 14:
+	  // Skylake
+          if(support_avx())
+#ifndef NO_AVX2
+            return CORE_HASWELL;
+#else
+	    return CORE_SANDYBRIDGE;
+#endif
+          else
+	    return CORE_NEHALEM;
+	}
+	break;
       }
       break;
 
@@ -1574,10 +1725,20 @@ int get_coretype(void){
 	    return CORE_BARCELONA; //OS don't support AVX.
 	
 	case 0:
-	  if(support_avx())
-	    return CORE_STEAMROLLER;
-	  else
-	    return CORE_BARCELONA; //OS don't support AVX.
+	  switch(exmodel){
+	  case 3:
+	    if(support_avx())
+	      return CORE_STEAMROLLER;
+	    else
+	      return CORE_BARCELONA; //OS don't support AVX.
+
+	  case 6:
+	    if(support_avx())
+	      return CORE_EXCAVATOR;
+	    else
+	      return CORE_BARCELONA; //OS don't support AVX.
+	  }
+	  break;
 	}
 
 
diff --git a/ctest.c b/ctest.c
index d5c224726..b5c74f137 100644
--- a/ctest.c
+++ b/ctest.c
@@ -44,6 +44,10 @@ COMPILER_DEC
 COMPILER_GNU
 #endif
 
+#if defined(__ANDROID__)
+OS_ANDROID
+#endif
+
 #if defined(__linux__)
 OS_LINUX
 #endif
diff --git a/ctest/CMakeLists.txt b/ctest/CMakeLists.txt
new file mode 100644
index 000000000..dbe785bcb
--- /dev/null
+++ b/ctest/CMakeLists.txt
@@ -0,0 +1,46 @@
+include_directories(${CMAKE_SOURCE_DIR})
+
+enable_language(Fortran)
+
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -DADD${BU} -DCBLAS")
+
+FILE(WRITE ${CMAKE_CURRENT_BINARY_DIR}/test_cblas_helper.sh
+"$1 < $2\n"
+)
+
+foreach(float_type ${FLOAT_TYPES})
+  string(SUBSTRING ${float_type} 0 1 float_char_upper)
+  string(TOLOWER ${float_char_upper} float_char)
+  #level1
+  add_executable(x${float_char}cblat1
+    c_${float_char}blat1.f
+    c_${float_char}blas1.c)
+  target_link_libraries(x${float_char}cblat1 ${OpenBLAS_LIBNAME}_static)
+  add_test(NAME "x${float_char}cblat1"
+    COMMAND "${CMAKE_CURRENT_BINARY_DIR}/x${float_char}cblat1")
+
+  #level2
+  add_executable(x${float_char}cblat2
+    c_${float_char}blat2.f
+    c_${float_char}blas2.c
+    c_${float_char}2chke.c
+    auxiliary.c
+    c_xerbla.c
+    constant.c)
+  target_link_libraries(x${float_char}cblat2 ${OpenBLAS_LIBNAME}_static)
+  add_test(NAME "x${float_char}cblat2"
+    COMMAND sh "${CMAKE_CURRENT_BINARY_DIR}/test_cblas_helper.sh" "${CMAKE_CURRENT_BINARY_DIR}/x${float_char}cblat2" "${PROJECT_SOURCE_DIR}/ctest/${float_char}in2")
+
+  #level3
+  add_executable(x${float_char}cblat3
+    c_${float_char}blat3.f
+    c_${float_char}blas3.c
+    c_${float_char}3chke.c
+    auxiliary.c
+    c_xerbla.c
+    constant.c)
+  target_link_libraries(x${float_char}cblat3 ${OpenBLAS_LIBNAME}_static)
+  add_test(NAME "x${float_char}cblat3"
+    COMMAND sh "${CMAKE_CURRENT_BINARY_DIR}/test_cblas_helper.sh" "${CMAKE_CURRENT_BINARY_DIR}/x${float_char}cblat3" "${PROJECT_SOURCE_DIR}/ctest/${float_char}in3")
+
+endforeach()
diff --git a/ctest/Makefile b/ctest/Makefile
index 1d9567150..7a5d236aa 100644
--- a/ctest/Makefile
+++ b/ctest/Makefile
@@ -27,12 +27,18 @@ ctestl2o = c_cblas2.o c_c2chke.o auxiliary.o c_xerbla.o constant.o
 
 ctestl3o = c_cblas3.o c_c3chke.o auxiliary.o c_xerbla.o constant.o
 
+ctestl3o_3m = c_cblas3_3m.o c_c3chke_3m.o auxiliary.o c_xerbla.o constant.o
+
 ztestl1o = c_zblas1.o
 
 ztestl2o = c_zblas2.o c_z2chke.o auxiliary.o c_xerbla.o constant.o
 
 ztestl3o = c_zblas3.o c_z3chke.o auxiliary.o c_xerbla.o constant.o
 
+ztestl3o_3m = c_zblas3_3m.o c_z3chke_3m.o auxiliary.o c_xerbla.o constant.o
+
+
+
 all :: all1 all2 all3
 
 all1: xscblat1 xdcblat1 xccblat1 xzcblat1
@@ -115,8 +121,8 @@ xccblat2: $(ctestl2o) c_cblat2.o $(TOPDIR)/$(LIBNAME)
 xccblat3: $(ctestl3o) c_cblat3.o $(TOPDIR)/$(LIBNAME)
 	$(FC) $(FLDFLAGS) -o xccblat3 c_cblat3.o $(ctestl3o) $(LIB) $(EXTRALIB) $(CEXTRALIB)
 
-xccblat3_3m: $(ctestl3o) c_cblat3_3m.o $(TOPDIR)/$(LIBNAME)
-	$(FC) $(FLDFLAGS) -o xccblat3_3m c_cblat3_3m.o $(ctestl3o) $(LIB) $(EXTRALIB) $(CEXTRALIB)
+xccblat3_3m: $(ctestl3o_3m) c_cblat3_3m.o $(TOPDIR)/$(LIBNAME)
+	$(FC) $(FLDFLAGS) -o xccblat3_3m c_cblat3_3m.o $(ctestl3o_3m) $(LIB) $(EXTRALIB) $(CEXTRALIB)
 
 # Double complex
 xzcblat1: $(ztestl1o) c_zblat1.o $(TOPDIR)/$(LIBNAME)
@@ -127,8 +133,8 @@ xzcblat3: $(ztestl3o) c_zblat3.o $(TOPDIR)/$(LIBNAME)
 	$(FC) $(FLDFLAGS) -o xzcblat3 c_zblat3.o $(ztestl3o) $(LIB) $(EXTRALIB) $(CEXTRALIB)
 
 
-xzcblat3_3m: $(ztestl3o) c_zblat3_3m.o $(TOPDIR)/$(LIBNAME)
-	$(FC) $(FLDFLAGS) -o xzcblat3_3m c_zblat3_3m.o $(ztestl3o) $(LIB) $(EXTRALIB) $(CEXTRALIB)
+xzcblat3_3m: $(ztestl3o_3m) c_zblat3_3m.o $(TOPDIR)/$(LIBNAME)
+	$(FC) $(FLDFLAGS) -o xzcblat3_3m c_zblat3_3m.o $(ztestl3o_3m) $(LIB) $(EXTRALIB) $(CEXTRALIB)
 
 
 include $(TOPDIR)/Makefile.tail
diff --git a/ctest/c_c3chke.c b/ctest/c_c3chke.c
index 4d5de5150..3b4764c4a 100644
--- a/ctest/c_c3chke.c
+++ b/ctest/c_c3chke.c
@@ -46,235 +46,7 @@ void  F77_c3chke(char *  rout) {
    }
 
 
-   if (strncmp( sf,"cblas_cgemm3m"   ,13)==0) {
-      cblas_rout = "cblas_cgemm3"   ;
-
-      cblas_info = 1;
-      cblas_cgemm3m( INVALID,  CblasNoTrans, CblasNoTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 1;
-      cblas_cgemm3m( INVALID,  CblasNoTrans, CblasTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 1;
-      cblas_cgemm3m( INVALID,  CblasTrans, CblasNoTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 1;
-      cblas_cgemm3m( INVALID,  CblasTrans, CblasTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 2; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  INVALID, CblasNoTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 2; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  INVALID, CblasTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 3; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, INVALID, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 3; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, INVALID, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 2 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 2 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 2, BETA, C, 1 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = FALSE;
-      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 9;  RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 1, BETA, C, 2 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 2, BETA, C, 2 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 2, BETA, C, 1 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 2, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 2, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = TRUE;
-      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-
-   } else if (strncmp( sf,"cblas_cgemm"   ,11)==0) {
+   if (strncmp( sf,"cblas_cgemm"   ,11)==0) {
             cblas_rout = "cblas_cgemm"   ;
 
 
diff --git a/ctest/c_c3chke_3m.c b/ctest/c_c3chke_3m.c
new file mode 100644
index 000000000..4d5de5150
--- /dev/null
+++ b/ctest/c_c3chke_3m.c
@@ -0,0 +1,1936 @@
+#include <stdio.h>
+#include <string.h>
+#include "common.h"
+#include "cblas_test.h"
+
+int cblas_ok, cblas_lerr, cblas_info;
+int link_xerbla=TRUE;
+char *cblas_rout;
+
+#ifdef F77_Char
+void F77_xerbla(F77_Char F77_srname, void *vinfo);
+#else
+void F77_xerbla(char *srname, void *vinfo);
+#endif
+
+void chkxer(void) {
+   extern int cblas_ok, cblas_lerr, cblas_info;
+   extern int link_xerbla;
+   extern char *cblas_rout;
+   if (cblas_lerr == 1 ) {
+      printf("***** ILLEGAL VALUE OF PARAMETER NUMBER %d NOT DETECTED BY %s *****\n", cblas_info, cblas_rout);
+      cblas_ok = 0 ;
+   }
+   cblas_lerr = 1 ;
+}
+
+void  F77_c3chke(char *  rout) {
+   char *sf = ( rout ) ;
+   float   A[4]     = {0.0,0.0,0.0,0.0},
+           B[4]     = {0.0,0.0,0.0,0.0},
+           C[4]     = {0.0,0.0,0.0,0.0},
+           ALPHA[2] = {0.0,0.0},
+           BETA[2]  = {0.0,0.0},
+           RALPHA   = 0.0, RBETA = 0.0;
+   extern int cblas_info, cblas_lerr, cblas_ok;
+   extern int RowMajorStrg;
+   extern char *cblas_rout;
+
+   cblas_ok = TRUE ;
+   cblas_lerr = PASSED ;
+
+   if (link_xerbla) /* call these first to link */
+   {
+      cblas_xerbla(cblas_info,cblas_rout,"");
+      F77_xerbla(cblas_rout,&cblas_info);
+   }
+
+
+   if (strncmp( sf,"cblas_cgemm3m"   ,13)==0) {
+      cblas_rout = "cblas_cgemm3"   ;
+
+      cblas_info = 1;
+      cblas_cgemm3m( INVALID,  CblasNoTrans, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_cgemm3m( INVALID,  CblasNoTrans, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_cgemm3m( INVALID,  CblasTrans, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_cgemm3m( INVALID,  CblasTrans, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  INVALID, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  INVALID, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, INVALID, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, INVALID, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_cgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9;  RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_cgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_cgemm"   ,11)==0) {
+            cblas_rout = "cblas_cgemm"   ;
+
+
+      cblas_info = 1;
+      cblas_cgemm( INVALID,  CblasNoTrans, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_cgemm( INVALID,  CblasNoTrans, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_cgemm( INVALID,  CblasTrans, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_cgemm( INVALID,  CblasTrans, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  INVALID, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  INVALID, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, INVALID, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, INVALID, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_cgemm( CblasColMajor,  CblasTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9;  RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_cgemm( CblasRowMajor,  CblasTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_chemm"   ,11)==0) {
+            cblas_rout = "cblas_chemm"   ;
+
+      cblas_info = 1;
+      cblas_chemm( INVALID,  CblasRight, CblasLower, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  INVALID, CblasUpper, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_chemm( CblasColMajor,  CblasRight, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasLeft, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_chemm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_csymm"   ,11)==0) {
+            cblas_rout = "cblas_csymm"   ;
+
+      cblas_info = 1;
+      cblas_csymm( INVALID,  CblasRight, CblasLower, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  INVALID, CblasUpper, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_csymm( CblasColMajor,  CblasRight, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasLeft, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_csymm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_ctrmm"   ,11)==0) {
+            cblas_rout = "cblas_ctrmm"   ;
+
+      cblas_info = 1;
+      cblas_ctrmm( INVALID,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  INVALID, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, INVALID, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, INVALID,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   INVALID, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrmm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrmm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_ctrsm"   ,11)==0) {
+            cblas_rout = "cblas_ctrsm"   ;
+
+      cblas_info = 1;
+      cblas_ctrsm( INVALID,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  INVALID, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, INVALID, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, INVALID,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   INVALID, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ctrsm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ctrsm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_cherk"   ,11)==0) {
+            cblas_rout = "cblas_cherk"   ;
+
+      cblas_info = 1;
+      cblas_cherk(INVALID,  CblasUpper, CblasNoTrans, 0, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  INVALID, CblasNoTrans, 0, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasTrans, 0, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasNoTrans, INVALID, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasConjTrans, INVALID, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasLower, CblasNoTrans, INVALID, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasLower, CblasConjTrans, INVALID, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasNoTrans, 0, INVALID,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasConjTrans, 0, INVALID,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasLower, CblasNoTrans, 0, INVALID,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasLower, CblasConjTrans, 0, INVALID,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_cherk(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                  RALPHA, A, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_cherk(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_cherk(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                  RALPHA, A, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_cherk(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 2,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasLower, CblasConjTrans, 0, 2,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cherk(CblasRowMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cherk(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                  RALPHA, A, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cherk(CblasRowMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_cherk(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                  RALPHA, A, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  RALPHA, A, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  RALPHA, A, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_cherk(CblasColMajor,  CblasLower, CblasConjTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_csyrk"   ,11)==0) {
+            cblas_rout = "cblas_csyrk"   ;
+
+      cblas_info = 1;
+      cblas_csyrk(INVALID,  CblasUpper, CblasNoTrans, 0, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  INVALID, CblasNoTrans, 0, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasNoTrans, INVALID, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasTrans, INVALID, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasLower, CblasNoTrans, INVALID, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasLower, CblasTrans, INVALID, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasNoTrans, 0, INVALID,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasTrans, 0, INVALID,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasLower, CblasNoTrans, 0, INVALID,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasLower, CblasTrans, 0, INVALID,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csyrk(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                  ALPHA, A, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csyrk(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csyrk(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                  ALPHA, A, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csyrk(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasTrans, 0, 2,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasLower, CblasTrans, 0, 2,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_csyrk(CblasRowMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_csyrk(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                  ALPHA, A, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_csyrk(CblasRowMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_csyrk(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                  ALPHA, A, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  ALPHA, A, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasUpper, CblasTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  ALPHA, A, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_csyrk(CblasColMajor,  CblasLower, CblasTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_cher2k"   ,12)==0) {
+            cblas_rout = "cblas_cher2k"   ;
+
+      cblas_info = 1;
+      cblas_cher2k(INVALID,  CblasUpper, CblasNoTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  INVALID, CblasNoTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasNoTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasConjTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasNoTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasConjTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasNoTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasConjTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasNoTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasConjTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                   ALPHA, A, 1, B, 2, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                   ALPHA, A, 1, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                   ALPHA, A, 1, B, 2, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                   ALPHA, A, 1, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 2,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasConjTrans, 0, 2,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasConjTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_cher2k(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_cher2k(CblasColMajor,  CblasLower, CblasConjTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_csyr2k"   ,12)==0) {
+            cblas_rout = "cblas_csyr2k"   ;
+
+      cblas_info = 1;
+      cblas_csyr2k(INVALID,  CblasUpper, CblasNoTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  INVALID, CblasNoTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasNoTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasNoTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasTrans, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasTrans, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_csyr2k(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasUpper, CblasTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_csyr2k(CblasColMajor,  CblasLower, CblasTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+
+   }
+
+   if (cblas_ok == 1 )
+       printf(" %-12s PASSED THE TESTS OF ERROR-EXITS\n", cblas_rout);
+   else
+       printf("***** %s FAILED THE TESTS OF ERROR-EXITS *******\n",cblas_rout);
+}
diff --git a/ctest/c_cblas3.c b/ctest/c_cblas3.c
index f1b108c64..9f48c49b1 100644
--- a/ctest/c_cblas3.c
+++ b/ctest/c_cblas3.c
@@ -567,81 +567,3 @@ void F77_ctrsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
 
 
 
-void F77_cgemm3m(int *order, char *transpa, char *transpb, int *m, int *n,
-     int *k, CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
-     CBLAS_TEST_COMPLEX *b, int *ldb, CBLAS_TEST_COMPLEX *beta,
-     CBLAS_TEST_COMPLEX *c, int *ldc ) {
-
-  CBLAS_TEST_COMPLEX *A, *B, *C;
-  int i,j,LDA, LDB, LDC;
-  enum CBLAS_TRANSPOSE transa, transb;
-
-  get_transpose_type(transpa, &transa);
-  get_transpose_type(transpb, &transb);
-
-  if (*order == TEST_ROW_MJR) {
-     if (transa == CblasNoTrans) {
-        LDA = *k+1;
-        A=(CBLAS_TEST_COMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_COMPLEX));
-        for( i=0; i<*m; i++ )
-           for( j=0; j<*k; j++ ) {
-              A[i*LDA+j].real=a[j*(*lda)+i].real;
-              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
-           }
-     }
-     else {
-        LDA = *m+1;
-        A=(CBLAS_TEST_COMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_COMPLEX));
-        for( i=0; i<*k; i++ )
-           for( j=0; j<*m; j++ ) {
-              A[i*LDA+j].real=a[j*(*lda)+i].real;
-              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
-           }
-     }
-
-     if (transb == CblasNoTrans) {
-        LDB = *n+1;
-        B=(CBLAS_TEST_COMPLEX* )malloc((*k)*LDB*sizeof(CBLAS_TEST_COMPLEX) );
-        for( i=0; i<*k; i++ )
-           for( j=0; j<*n; j++ ) {
-              B[i*LDB+j].real=b[j*(*ldb)+i].real;
-              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
-           }
-     }
-     else {
-        LDB = *k+1;
-        B=(CBLAS_TEST_COMPLEX* )malloc(LDB*(*n)*sizeof(CBLAS_TEST_COMPLEX));
-        for( i=0; i<*n; i++ )
-           for( j=0; j<*k; j++ ) {
-              B[i*LDB+j].real=b[j*(*ldb)+i].real;
-              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
-           }
-     }
-
-     LDC = *n+1;
-     C=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_COMPLEX));
-     for( j=0; j<*n; j++ )
-        for( i=0; i<*m; i++ ) {
-           C[i*LDC+j].real=c[j*(*ldc)+i].real;
-           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
-        }
-     cblas_cgemm3m( CblasRowMajor, transa, transb, *m, *n, *k, alpha, A, LDA,
-                  B, LDB, beta, C, LDC );
-     for( j=0; j<*n; j++ )
-        for( i=0; i<*m; i++ ) {
-           c[j*(*ldc)+i].real=C[i*LDC+j].real;
-           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
-        }
-     free(A);
-     free(B);
-     free(C);
-  }
-  else if (*order == TEST_COL_MJR)
-     cblas_cgemm3m( CblasColMajor, transa, transb, *m, *n, *k, alpha, a, *lda,
-                  b, *ldb, beta, c, *ldc );
-  else
-     cblas_cgemm3m( UNDEFINED, transa, transb, *m, *n, *k, alpha, a, *lda,
-                  b, *ldb, beta, c, *ldc );
-}
-
-
diff --git a/ctest/c_cblas3_3m.c b/ctest/c_cblas3_3m.c
new file mode 100644
index 000000000..f1b108c64
--- /dev/null
+++ b/ctest/c_cblas3_3m.c
@@ -0,0 +1,647 @@
+/*
+ *     Written by D.P. Manley, Digital Equipment Corporation.
+ *     Prefixed "C_" to BLAS routines and their declarations.
+ *
+ *     Modified by T. H. Do, 4/15/98, SGI/CRAY Research.
+ */
+#include <stdlib.h>
+#include "common.h"
+#include "cblas_test.h"
+
+#define  TEST_COL_MJR	0
+#define  TEST_ROW_MJR	1
+#define  UNDEFINED     -1
+
+void F77_cgemm(int *order, char *transpa, char *transpb, int *m, int *n,
+     int *k, CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
+     CBLAS_TEST_COMPLEX *b, int *ldb, CBLAS_TEST_COMPLEX *beta,
+     CBLAS_TEST_COMPLEX *c, int *ldc ) {
+
+  CBLAS_TEST_COMPLEX *A, *B, *C;
+  int i,j,LDA, LDB, LDC;
+  enum CBLAS_TRANSPOSE transa, transb;
+
+  get_transpose_type(transpa, &transa);
+  get_transpose_type(transpb, &transb);
+
+  if (*order == TEST_ROW_MJR) {
+     if (transa == CblasNoTrans) {
+        LDA = *k+1;
+        A=(CBLAS_TEST_COMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else {
+        LDA = *m+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+
+     if (transb == CblasNoTrans) {
+        LDB = *n+1;
+        B=(CBLAS_TEST_COMPLEX* )malloc((*k)*LDB*sizeof(CBLAS_TEST_COMPLEX) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ) {
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     else {
+        LDB = *k+1;
+        B=(CBLAS_TEST_COMPLEX* )malloc(LDB*(*n)*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+
+     LDC = *n+1;
+     C=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_COMPLEX));
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_cgemm( CblasRowMajor, transa, transb, *m, *n, *k, alpha, A, LDA,
+                  B, LDB, beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_cgemm( CblasColMajor, transa, transb, *m, *n, *k, alpha, a, *lda,
+                  b, *ldb, beta, c, *ldc );
+  else
+     cblas_cgemm( UNDEFINED, transa, transb, *m, *n, *k, alpha, a, *lda,
+                  b, *ldb, beta, c, *ldc );
+}
+
+void F77_chemm(int *order, char *rtlf, char *uplow, int *m, int *n,
+        CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
+	CBLAS_TEST_COMPLEX *b, int *ldb, CBLAS_TEST_COMPLEX *beta,
+        CBLAS_TEST_COMPLEX *c, int *ldc ) {
+
+  CBLAS_TEST_COMPLEX *A, *B, *C;
+  int i,j,LDA, LDB, LDC;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_SIDE side;
+
+  get_uplo_type(uplow,&uplo);
+  get_side_type(rtlf,&side);
+
+  if (*order == TEST_ROW_MJR) {
+     if (side == CblasLeft) {
+        LDA = *m+1;
+        A= (CBLAS_TEST_COMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDB = *n+1;
+     B=(CBLAS_TEST_COMPLEX* )malloc( (*m)*LDB*sizeof(CBLAS_TEST_COMPLEX ) );
+     for( i=0; i<*m; i++ )
+        for( j=0; j<*n; j++ ) {
+           B[i*LDB+j].real=b[j*(*ldb)+i].real;
+           B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+        }
+     LDC = *n+1;
+     C=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_COMPLEX ) );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_chemm( CblasRowMajor, side, uplo, *m, *n, alpha, A, LDA, B, LDB,
+                  beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_chemm( CblasColMajor, side, uplo, *m, *n, alpha, a, *lda, b, *ldb,
+                  beta, c, *ldc );
+  else
+     cblas_chemm( UNDEFINED, side, uplo, *m, *n, alpha, a, *lda, b, *ldb,
+                  beta, c, *ldc );
+}
+void F77_csymm(int *order, char *rtlf, char *uplow, int *m, int *n,
+          CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
+	  CBLAS_TEST_COMPLEX *b, int *ldb, CBLAS_TEST_COMPLEX *beta,
+          CBLAS_TEST_COMPLEX *c, int *ldc ) {
+
+  CBLAS_TEST_COMPLEX *A, *B, *C;
+  int i,j,LDA, LDB, LDC;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_SIDE side;
+
+  get_uplo_type(uplow,&uplo);
+  get_side_type(rtlf,&side);
+
+  if (*order == TEST_ROW_MJR) {
+     if (side == CblasLeft) {
+        LDA = *m+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*m; j++ )
+              A[i*LDA+j]=a[j*(*lda)+i];
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*n; j++ )
+              A[i*LDA+j]=a[j*(*lda)+i];
+     }
+     LDB = *n+1;
+     B=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_COMPLEX ));
+     for( i=0; i<*m; i++ )
+        for( j=0; j<*n; j++ )
+           B[i*LDB+j]=b[j*(*ldb)+i];
+     LDC = *n+1;
+     C=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_COMPLEX));
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ )
+           C[i*LDC+j]=c[j*(*ldc)+i];
+     cblas_csymm( CblasRowMajor, side, uplo, *m, *n, alpha, A, LDA, B, LDB,
+                  beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ )
+           c[j*(*ldc)+i]=C[i*LDC+j];
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_csymm( CblasColMajor, side, uplo, *m, *n, alpha, a, *lda, b, *ldb,
+                  beta, c, *ldc );
+  else
+     cblas_csymm( UNDEFINED, side, uplo, *m, *n, alpha, a, *lda, b, *ldb,
+                  beta, c, *ldc );
+}
+
+void F77_cherk(int *order, char *uplow, char *transp, int *n, int *k,
+     float *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
+     float *beta, CBLAS_TEST_COMPLEX *c, int *ldc ) {
+
+  int i,j,LDA,LDC;
+  CBLAS_TEST_COMPLEX *A, *C;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+
+  if (*order == TEST_ROW_MJR) {
+     if (trans == CblasNoTrans) {
+        LDA = *k+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*k)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDC = *n+1;
+     C=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDC*sizeof(CBLAS_TEST_COMPLEX ) );
+     for( i=0; i<*n; i++ )
+        for( j=0; j<*n; j++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_cherk(CblasRowMajor, uplo, trans, *n, *k, *alpha, A, LDA, *beta,
+	         C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*n; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_cherk(CblasColMajor, uplo, trans, *n, *k, *alpha, a, *lda, *beta,
+	         c, *ldc );
+  else
+     cblas_cherk(UNDEFINED, uplo, trans, *n, *k, *alpha, a, *lda, *beta,
+	         c, *ldc );
+}
+
+void F77_csyrk(int *order, char *uplow, char *transp, int *n, int *k,
+     CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
+     CBLAS_TEST_COMPLEX *beta, CBLAS_TEST_COMPLEX *c, int *ldc ) {
+
+  int i,j,LDA,LDC;
+  CBLAS_TEST_COMPLEX *A, *C;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+
+  if (*order == TEST_ROW_MJR) {
+     if (trans == CblasNoTrans) {
+        LDA = *k+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*k)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDC = *n+1;
+     C=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDC*sizeof(CBLAS_TEST_COMPLEX ) );
+     for( i=0; i<*n; i++ )
+        for( j=0; j<*n; j++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_csyrk(CblasRowMajor, uplo, trans, *n, *k, alpha, A, LDA, beta,
+	         C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*n; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_csyrk(CblasColMajor, uplo, trans, *n, *k, alpha, a, *lda, beta,
+	         c, *ldc );
+  else
+     cblas_csyrk(UNDEFINED, uplo, trans, *n, *k, alpha, a, *lda, beta,
+	         c, *ldc );
+}
+void F77_cher2k(int *order, char *uplow, char *transp, int *n, int *k,
+        CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
+	CBLAS_TEST_COMPLEX *b, int *ldb, float *beta,
+        CBLAS_TEST_COMPLEX *c, int *ldc ) {
+  int i,j,LDA,LDB,LDC;
+  CBLAS_TEST_COMPLEX *A, *B, *C;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+
+  if (*order == TEST_ROW_MJR) {
+     if (trans == CblasNoTrans) {
+        LDA = *k+1;
+        LDB = *k+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX ));
+        B=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDB*sizeof(CBLAS_TEST_COMPLEX ));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     else {
+        LDA = *n+1;
+        LDB = *n+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc( LDA*(*k)*sizeof(CBLAS_TEST_COMPLEX ) );
+        B=(CBLAS_TEST_COMPLEX* )malloc( LDB*(*k)*sizeof(CBLAS_TEST_COMPLEX ) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ){
+	      A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     LDC = *n+1;
+     C=(CBLAS_TEST_COMPLEX* )malloc( (*n)*LDC*sizeof(CBLAS_TEST_COMPLEX ) );
+     for( i=0; i<*n; i++ )
+        for( j=0; j<*n; j++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_cher2k(CblasRowMajor, uplo, trans, *n, *k, alpha, A, LDA,
+		  B, LDB, *beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*n; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_cher2k(CblasColMajor, uplo, trans, *n, *k, alpha, a, *lda,
+		   b, *ldb, *beta, c, *ldc );
+  else
+     cblas_cher2k(UNDEFINED, uplo, trans, *n, *k, alpha, a, *lda,
+		   b, *ldb, *beta, c, *ldc );
+}
+void F77_csyr2k(int *order, char *uplow, char *transp, int *n, int *k,
+         CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
+	 CBLAS_TEST_COMPLEX *b, int *ldb, CBLAS_TEST_COMPLEX *beta,
+         CBLAS_TEST_COMPLEX *c, int *ldc ) {
+  int i,j,LDA,LDB,LDC;
+  CBLAS_TEST_COMPLEX *A, *B, *C;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+
+  if (*order == TEST_ROW_MJR) {
+     if (trans == CblasNoTrans) {
+        LDA = *k+1;
+        LDB = *k+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        B=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDB*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     else {
+        LDA = *n+1;
+        LDB = *n+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_COMPLEX));
+        B=(CBLAS_TEST_COMPLEX* )malloc(LDB*(*k)*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ){
+	      A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     LDC = *n+1;
+     C=(CBLAS_TEST_COMPLEX* )malloc( (*n)*LDC*sizeof(CBLAS_TEST_COMPLEX));
+     for( i=0; i<*n; i++ )
+        for( j=0; j<*n; j++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_csyr2k(CblasRowMajor, uplo, trans, *n, *k, alpha, A, LDA,
+		  B, LDB, beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*n; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_csyr2k(CblasColMajor, uplo, trans, *n, *k, alpha, a, *lda,
+		   b, *ldb, beta, c, *ldc );
+  else
+     cblas_csyr2k(UNDEFINED, uplo, trans, *n, *k, alpha, a, *lda,
+		   b, *ldb, beta, c, *ldc );
+}
+void F77_ctrmm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
+       int *m, int *n, CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a,
+       int *lda, CBLAS_TEST_COMPLEX *b, int *ldb) {
+  int i,j,LDA,LDB;
+  CBLAS_TEST_COMPLEX *A, *B;
+  enum CBLAS_SIDE side;
+  enum CBLAS_DIAG diag;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+  get_diag_type(diagn,&diag);
+  get_side_type(rtlf,&side);
+
+  if (*order == TEST_ROW_MJR) {
+     if (side == CblasLeft) {
+        LDA = *m+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDB = *n+1;
+     B=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_COMPLEX));
+     for( i=0; i<*m; i++ )
+        for( j=0; j<*n; j++ ) {
+           B[i*LDB+j].real=b[j*(*ldb)+i].real;
+           B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+        }
+     cblas_ctrmm(CblasRowMajor, side, uplo, trans, diag, *m, *n, alpha,
+		 A, LDA, B, LDB );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           b[j*(*ldb)+i].real=B[i*LDB+j].real;
+           b[j*(*ldb)+i].imag=B[i*LDB+j].imag;
+        }
+     free(A);
+     free(B);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_ctrmm(CblasColMajor, side, uplo, trans, diag, *m, *n, alpha,
+		   a, *lda, b, *ldb);
+  else
+     cblas_ctrmm(UNDEFINED, side, uplo, trans, diag, *m, *n, alpha,
+		   a, *lda, b, *ldb);
+}
+
+void F77_ctrsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
+         int *m, int *n, CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a,
+         int *lda, CBLAS_TEST_COMPLEX *b, int *ldb) {
+  int i,j,LDA,LDB;
+  CBLAS_TEST_COMPLEX *A, *B;
+  enum CBLAS_SIDE side;
+  enum CBLAS_DIAG diag;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+  get_diag_type(diagn,&diag);
+  get_side_type(rtlf,&side);
+
+  if (*order == TEST_ROW_MJR) {
+     if (side == CblasLeft) {
+        LDA = *m+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc( (*m)*LDA*sizeof(CBLAS_TEST_COMPLEX ) );
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDB = *n+1;
+     B=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_COMPLEX));
+     for( i=0; i<*m; i++ )
+        for( j=0; j<*n; j++ ) {
+           B[i*LDB+j].real=b[j*(*ldb)+i].real;
+           B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+        }
+     cblas_ctrsm(CblasRowMajor, side, uplo, trans, diag, *m, *n, alpha,
+		 A, LDA, B, LDB );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           b[j*(*ldb)+i].real=B[i*LDB+j].real;
+           b[j*(*ldb)+i].imag=B[i*LDB+j].imag;
+        }
+     free(A);
+     free(B);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_ctrsm(CblasColMajor, side, uplo, trans, diag, *m, *n, alpha,
+		   a, *lda, b, *ldb);
+  else
+     cblas_ctrsm(UNDEFINED, side, uplo, trans, diag, *m, *n, alpha,
+		   a, *lda, b, *ldb);
+}
+
+
+
+void F77_cgemm3m(int *order, char *transpa, char *transpb, int *m, int *n,
+     int *k, CBLAS_TEST_COMPLEX *alpha, CBLAS_TEST_COMPLEX *a, int *lda,
+     CBLAS_TEST_COMPLEX *b, int *ldb, CBLAS_TEST_COMPLEX *beta,
+     CBLAS_TEST_COMPLEX *c, int *ldc ) {
+
+  CBLAS_TEST_COMPLEX *A, *B, *C;
+  int i,j,LDA, LDB, LDC;
+  enum CBLAS_TRANSPOSE transa, transb;
+
+  get_transpose_type(transpa, &transa);
+  get_transpose_type(transpb, &transb);
+
+  if (*order == TEST_ROW_MJR) {
+     if (transa == CblasNoTrans) {
+        LDA = *k+1;
+        A=(CBLAS_TEST_COMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else {
+        LDA = *m+1;
+        A=(CBLAS_TEST_COMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+
+     if (transb == CblasNoTrans) {
+        LDB = *n+1;
+        B=(CBLAS_TEST_COMPLEX* )malloc((*k)*LDB*sizeof(CBLAS_TEST_COMPLEX) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ) {
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     else {
+        LDB = *k+1;
+        B=(CBLAS_TEST_COMPLEX* )malloc(LDB*(*n)*sizeof(CBLAS_TEST_COMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+
+     LDC = *n+1;
+     C=(CBLAS_TEST_COMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_COMPLEX));
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_cgemm3m( CblasRowMajor, transa, transb, *m, *n, *k, alpha, A, LDA,
+                  B, LDB, beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_cgemm3m( CblasColMajor, transa, transb, *m, *n, *k, alpha, a, *lda,
+                  b, *ldb, beta, c, *ldc );
+  else
+     cblas_cgemm3m( UNDEFINED, transa, transb, *m, *n, *k, alpha, a, *lda,
+                  b, *ldb, beta, c, *ldc );
+}
+
+
diff --git a/ctest/c_z3chke.c b/ctest/c_z3chke.c
index 4be4457b4..054e72360 100644
--- a/ctest/c_z3chke.c
+++ b/ctest/c_z3chke.c
@@ -49,237 +49,7 @@ void  F77_z3chke(char *  rout) {
 
 
 
-   if (strncmp( sf,"cblas_zgemm3m"   ,13)==0) {
-      cblas_rout = "cblas_zgemm3"   ;
-
-      cblas_info = 1;
-      cblas_zgemm3m( INVALID,  CblasNoTrans, CblasNoTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 1;
-      cblas_zgemm3m( INVALID,  CblasNoTrans, CblasTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 1;
-      cblas_zgemm3m( INVALID,  CblasTrans, CblasNoTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 1;
-      cblas_zgemm3m( INVALID,  CblasTrans, CblasTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 2; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  INVALID, CblasNoTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 2; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  INVALID, CblasTrans, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 3; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, INVALID, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 3; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, INVALID, 0, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 2 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 2 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 2, BETA, C, 1 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = FALSE;
-      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 4; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 5; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 6; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 9;  RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 1, BETA, C, 2 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 2, BETA, C, 2 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 2, BETA, C, 1 );
-      chkxer();
-      cblas_info = 9; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 2, 0, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
-                   ALPHA, A, 2, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 11; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, 2,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 2, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 2, BETA, C, 1 );
-      chkxer();
-      cblas_info = 14; RowMajorStrg = TRUE;
-      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 2, 0,
-                   ALPHA, A, 1, B, 1, BETA, C, 1 );
-      chkxer();
-
-
-
-   } else if (strncmp( sf,"cblas_zgemm"   ,11)==0) {
+   if (strncmp( sf,"cblas_zgemm"   ,11)==0) {
             cblas_rout = "cblas_zgemm"   ;
 
       cblas_info = 1;
diff --git a/ctest/c_z3chke_3m.c b/ctest/c_z3chke_3m.c
new file mode 100644
index 000000000..4be4457b4
--- /dev/null
+++ b/ctest/c_z3chke_3m.c
@@ -0,0 +1,1940 @@
+#include <stdio.h>
+#include <string.h>
+#include "common.h"
+#include "cblas_test.h"
+
+int cblas_ok, cblas_lerr, cblas_info;
+int link_xerbla=TRUE;
+char *cblas_rout;
+
+#ifdef F77_Char
+void F77_xerbla(F77_Char F77_srname, void *vinfo);
+#else
+void F77_xerbla(char *srname, void *vinfo);
+#endif
+
+void chkxer(void) {
+   extern int cblas_ok, cblas_lerr, cblas_info;
+   extern int link_xerbla;
+   extern char *cblas_rout;
+   if (cblas_lerr == 1 ) {
+      printf("***** ILLEGAL VALUE OF PARAMETER NUMBER %d NOT DETECTED BY %s *****\n", cblas_info, cblas_rout);
+      cblas_ok = 0 ;
+   }
+   cblas_lerr = 1 ;
+}
+
+void  F77_z3chke(char *  rout) {
+   char *sf = ( rout ) ;
+   double  A[4]     = {0.0,0.0,0.0,0.0},
+           B[4]     = {0.0,0.0,0.0,0.0},
+           C[4]     = {0.0,0.0,0.0,0.0},
+           ALPHA[2] = {0.0,0.0},
+           BETA[2]  = {0.0,0.0},
+           RALPHA   = 0.0, RBETA = 0.0;
+   extern int cblas_info, cblas_lerr, cblas_ok;
+   extern int RowMajorStrg;
+   extern char *cblas_rout;
+
+   cblas_ok = TRUE ;
+   cblas_lerr = PASSED ;
+
+   if (link_xerbla) /* call these first to link */
+   {
+      cblas_xerbla(cblas_info,cblas_rout,"");
+      F77_xerbla(cblas_rout,&cblas_info);
+   }
+
+
+
+
+
+   if (strncmp( sf,"cblas_zgemm3m"   ,13)==0) {
+      cblas_rout = "cblas_zgemm3"   ;
+
+      cblas_info = 1;
+      cblas_zgemm3m( INVALID,  CblasNoTrans, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_zgemm3m( INVALID,  CblasNoTrans, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_zgemm3m( INVALID,  CblasTrans, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_zgemm3m( INVALID,  CblasTrans, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  INVALID, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  INVALID, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, INVALID, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, INVALID, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_zgemm3m( CblasColMajor,  CblasTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9;  RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_zgemm3m( CblasRowMajor,  CblasTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+
+
+
+   } else if (strncmp( sf,"cblas_zgemm"   ,11)==0) {
+            cblas_rout = "cblas_zgemm"   ;
+
+      cblas_info = 1;
+      cblas_zgemm( INVALID,  CblasNoTrans, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_zgemm( INVALID,  CblasNoTrans, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_zgemm( INVALID,  CblasTrans, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 1;
+      cblas_zgemm( INVALID,  CblasTrans, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  INVALID, CblasNoTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  INVALID, CblasTrans, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, INVALID, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, INVALID, 0, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasNoTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = FALSE;
+      cblas_zgemm( CblasColMajor,  CblasTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasTrans, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasTrans, 0, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9;  RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 9; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasTrans, 2, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasTrans, 0, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasNoTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasNoTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 14; RowMajorStrg = TRUE;
+      cblas_zgemm( CblasRowMajor,  CblasTrans, CblasTrans, 0, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_zhemm"   ,11)==0) {
+            cblas_rout = "cblas_zhemm"   ;
+
+      cblas_info = 1;
+      cblas_zhemm( INVALID,  CblasRight, CblasLower, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  INVALID, CblasUpper, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zhemm( CblasColMajor,  CblasRight, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasLeft, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zhemm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_zsymm"   ,11)==0) {
+            cblas_rout = "cblas_zsymm"   ;
+
+      cblas_info = 1;
+      cblas_zsymm( INVALID,  CblasRight, CblasLower, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  INVALID, CblasUpper, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, INVALID, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zsymm( CblasColMajor,  CblasRight, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasUpper, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasLower, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasUpper, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasLower, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasUpper, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasLower, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasUpper, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasUpper, 0, 2,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasLeft, CblasLower, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zsymm( CblasRowMajor,  CblasRight, CblasLower, 0, 2,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_ztrmm"   ,11)==0) {
+            cblas_rout = "cblas_ztrmm"   ;
+
+      cblas_info = 1;
+      cblas_ztrmm( INVALID,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  INVALID, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, INVALID, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, INVALID,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   INVALID, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrmm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrmm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_ztrsm"   ,11)==0) {
+            cblas_rout = "cblas_ztrsm"   ;
+
+      cblas_info = 1;
+      cblas_ztrsm( INVALID,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  INVALID, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, INVALID, CblasNoTrans,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, INVALID,
+                   CblasNonUnit, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   INVALID, 0, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = FALSE;
+      cblas_ztrsm( CblasColMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 6; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, INVALID, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 7; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, INVALID, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 2, 0, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 2 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasUpper, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasLeft, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 1, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasLower, CblasNoTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+      cblas_info = 12; RowMajorStrg = TRUE;
+      cblas_ztrsm( CblasRowMajor,  CblasRight, CblasLower, CblasTrans,
+                   CblasNonUnit, 0, 2, ALPHA, A, 2, B, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_zherk"   ,11)==0) {
+            cblas_rout = "cblas_zherk"   ;
+
+      cblas_info = 1;
+      cblas_zherk(INVALID,  CblasUpper, CblasNoTrans, 0, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  INVALID, CblasNoTrans, 0, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasTrans, 0, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasNoTrans, INVALID, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasConjTrans, INVALID, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasLower, CblasNoTrans, INVALID, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasLower, CblasConjTrans, INVALID, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasNoTrans, 0, INVALID,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasConjTrans, 0, INVALID,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasLower, CblasNoTrans, 0, INVALID,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasLower, CblasConjTrans, 0, INVALID,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zherk(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                  RALPHA, A, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zherk(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zherk(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                  RALPHA, A, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zherk(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 2,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasLower, CblasConjTrans, 0, 2,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zherk(CblasRowMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zherk(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                  RALPHA, A, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zherk(CblasRowMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zherk(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                  RALPHA, A, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  RALPHA, A, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  RALPHA, A, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zherk(CblasColMajor,  CblasLower, CblasConjTrans, 2, 0,
+                  RALPHA, A, 1, RBETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_zsyrk"   ,11)==0) {
+            cblas_rout = "cblas_zsyrk"   ;
+
+      cblas_info = 1;
+      cblas_zsyrk(INVALID,  CblasUpper, CblasNoTrans, 0, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  INVALID, CblasNoTrans, 0, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasNoTrans, INVALID, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasTrans, INVALID, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasLower, CblasNoTrans, INVALID, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasLower, CblasTrans, INVALID, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasNoTrans, 0, INVALID,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasTrans, 0, INVALID,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasLower, CblasNoTrans, 0, INVALID,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasLower, CblasTrans, 0, INVALID,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsyrk(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                  ALPHA, A, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsyrk(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsyrk(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                  ALPHA, A, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsyrk(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasTrans, 0, 2,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasLower, CblasTrans, 0, 2,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zsyrk(CblasRowMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zsyrk(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                  ALPHA, A, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zsyrk(CblasRowMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = TRUE;
+      cblas_zsyrk(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                  ALPHA, A, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                  ALPHA, A, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasUpper, CblasTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                  ALPHA, A, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 11; RowMajorStrg = FALSE;
+      cblas_zsyrk(CblasColMajor,  CblasLower, CblasTrans, 2, 0,
+                  ALPHA, A, 1, BETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_zher2k"   ,12)==0) {
+            cblas_rout = "cblas_zher2k"   ;
+
+      cblas_info = 1;
+      cblas_zher2k(INVALID,  CblasUpper, CblasNoTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  INVALID, CblasNoTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasNoTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasConjTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasNoTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasConjTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasNoTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasConjTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasNoTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasConjTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                   ALPHA, A, 1, B, 2, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                   ALPHA, A, 1, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                   ALPHA, A, 1, B, 2, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                   ALPHA, A, 1, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 2,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasConjTrans, 0, 2,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, RBETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasConjTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zher2k(CblasRowMajor,  CblasLower, CblasConjTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasUpper, CblasConjTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, RBETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zher2k(CblasColMajor,  CblasLower, CblasConjTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, RBETA, C, 1 );
+      chkxer();
+
+   } else if (strncmp( sf,"cblas_zsyr2k"   ,12)==0) {
+            cblas_rout = "cblas_zsyr2k"   ;
+
+      cblas_info = 1;
+      cblas_zsyr2k(INVALID,  CblasUpper, CblasNoTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 2; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  INVALID, CblasNoTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 3; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasConjTrans, 0, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasNoTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 4; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasTrans, INVALID, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasNoTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 5; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasTrans, 0, INVALID,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                   ALPHA, A, 1, B, 2, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                   ALPHA, A, 1, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasTrans, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 8; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasTrans, 0, 2,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasLower, CblasNoTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 1, BETA, C, 2 );
+      chkxer();
+      cblas_info = 10; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasTrans, 0, 2,
+                   ALPHA, A, 2, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasUpper, CblasTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = TRUE;
+      cblas_zsyr2k(CblasRowMajor,  CblasLower, CblasTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasUpper, CblasTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasNoTrans, 2, 0,
+                   ALPHA, A, 2, B, 2, BETA, C, 1 );
+      chkxer();
+      cblas_info = 13; RowMajorStrg = FALSE;
+      cblas_zsyr2k(CblasColMajor,  CblasLower, CblasTrans, 2, 0,
+                   ALPHA, A, 1, B, 1, BETA, C, 1 );
+      chkxer();
+
+   }
+
+   if (cblas_ok == 1 )
+       printf(" %-12s PASSED THE TESTS OF ERROR-EXITS\n", cblas_rout);
+   else
+       printf("***** %s FAILED THE TESTS OF ERROR-EXITS *******\n",cblas_rout);
+}
diff --git a/ctest/c_zblas3.c b/ctest/c_zblas3.c
index 46ff467d0..40afa4edf 100644
--- a/ctest/c_zblas3.c
+++ b/ctest/c_zblas3.c
@@ -564,80 +564,3 @@ void F77_ztrsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
 }
 
 
-void F77_zgemm3m(int *order, char *transpa, char *transpb, int *m, int *n,
-     int *k, CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
-     CBLAS_TEST_ZOMPLEX *b, int *ldb, CBLAS_TEST_ZOMPLEX *beta,
-     CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
-
-  CBLAS_TEST_ZOMPLEX *A, *B, *C;
-  int i,j,LDA, LDB, LDC;
-  enum CBLAS_TRANSPOSE transa, transb;
-
-  get_transpose_type(transpa, &transa);
-  get_transpose_type(transpb, &transb);
-
-  if (*order == TEST_ROW_MJR) {
-     if (transa == CblasNoTrans) {
-        LDA = *k+1;
-        A=(CBLAS_TEST_ZOMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
-        for( i=0; i<*m; i++ )
-           for( j=0; j<*k; j++ ) {
-              A[i*LDA+j].real=a[j*(*lda)+i].real;
-              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
-           }
-     }
-     else {
-        LDA = *m+1;
-        A=(CBLAS_TEST_ZOMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
-        for( i=0; i<*k; i++ )
-           for( j=0; j<*m; j++ ) {
-              A[i*LDA+j].real=a[j*(*lda)+i].real;
-              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
-           }
-     }
-
-     if (transb == CblasNoTrans) {
-        LDB = *n+1;
-        B=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*LDB*sizeof(CBLAS_TEST_ZOMPLEX) );
-        for( i=0; i<*k; i++ )
-           for( j=0; j<*n; j++ ) {
-              B[i*LDB+j].real=b[j*(*ldb)+i].real;
-              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
-           }
-     }
-     else {
-        LDB = *k+1;
-        B=(CBLAS_TEST_ZOMPLEX* )malloc(LDB*(*n)*sizeof(CBLAS_TEST_ZOMPLEX));
-        for( i=0; i<*n; i++ )
-           for( j=0; j<*k; j++ ) {
-              B[i*LDB+j].real=b[j*(*ldb)+i].real;
-              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
-           }
-     }
-
-     LDC = *n+1;
-     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_ZOMPLEX));
-     for( j=0; j<*n; j++ )
-        for( i=0; i<*m; i++ ) {
-           C[i*LDC+j].real=c[j*(*ldc)+i].real;
-           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
-        }
-     cblas_zgemm3m( CblasRowMajor, transa, transb, *m, *n, *k, alpha, A, LDA,
-                  B, LDB, beta, C, LDC );
-     for( j=0; j<*n; j++ )
-        for( i=0; i<*m; i++ ) {
-           c[j*(*ldc)+i].real=C[i*LDC+j].real;
-           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
-        }
-     free(A);
-     free(B);
-     free(C);
-  }
-  else if (*order == TEST_COL_MJR)
-     cblas_zgemm3m( CblasColMajor, transa, transb, *m, *n, *k, alpha, a, *lda,
-                  b, *ldb, beta, c, *ldc );
-  else
-     cblas_zgemm3m( UNDEFINED, transa, transb, *m, *n, *k, alpha, a, *lda,
-                  b, *ldb, beta, c, *ldc );
-}
-
diff --git a/ctest/c_zblas3_3m.c b/ctest/c_zblas3_3m.c
new file mode 100644
index 000000000..46ff467d0
--- /dev/null
+++ b/ctest/c_zblas3_3m.c
@@ -0,0 +1,643 @@
+/*
+ *     Written by D.P. Manley, Digital Equipment Corporation.
+ *     Prefixed "C_" to BLAS routines and their declarations.
+ *
+ *     Modified by T. H. Do, 4/15/98, SGI/CRAY Research.
+ */
+#include <stdlib.h>
+#include "common.h"
+#include "cblas_test.h"
+#define  TEST_COL_MJR	0
+#define  TEST_ROW_MJR	1
+#define  UNDEFINED     -1
+
+void F77_zgemm(int *order, char *transpa, char *transpb, int *m, int *n,
+     int *k, CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
+     CBLAS_TEST_ZOMPLEX *b, int *ldb, CBLAS_TEST_ZOMPLEX *beta,
+     CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
+
+  CBLAS_TEST_ZOMPLEX *A, *B, *C;
+  int i,j,LDA, LDB, LDC;
+  enum CBLAS_TRANSPOSE transa, transb;
+
+  get_transpose_type(transpa, &transa);
+  get_transpose_type(transpb, &transb);
+
+  if (*order == TEST_ROW_MJR) {
+     if (transa == CblasNoTrans) {
+        LDA = *k+1;
+        A=(CBLAS_TEST_ZOMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else {
+        LDA = *m+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+
+     if (transb == CblasNoTrans) {
+        LDB = *n+1;
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*LDB*sizeof(CBLAS_TEST_ZOMPLEX) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ) {
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     else {
+        LDB = *k+1;
+        B=(CBLAS_TEST_ZOMPLEX* )malloc(LDB*(*n)*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+
+     LDC = *n+1;
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_ZOMPLEX));
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_zgemm( CblasRowMajor, transa, transb, *m, *n, *k, alpha, A, LDA,
+                  B, LDB, beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_zgemm( CblasColMajor, transa, transb, *m, *n, *k, alpha, a, *lda,
+                  b, *ldb, beta, c, *ldc );
+  else
+     cblas_zgemm( UNDEFINED, transa, transb, *m, *n, *k, alpha, a, *lda,
+                  b, *ldb, beta, c, *ldc );
+}
+void F77_zhemm(int *order, char *rtlf, char *uplow, int *m, int *n,
+        CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
+	CBLAS_TEST_ZOMPLEX *b, int *ldb, CBLAS_TEST_ZOMPLEX *beta,
+        CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
+
+  CBLAS_TEST_ZOMPLEX *A, *B, *C;
+  int i,j,LDA, LDB, LDC;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_SIDE side;
+
+  get_uplo_type(uplow,&uplo);
+  get_side_type(rtlf,&side);
+
+  if (*order == TEST_ROW_MJR) {
+     if (side == CblasLeft) {
+        LDA = *m+1;
+        A= (CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDB = *n+1;
+     B=(CBLAS_TEST_ZOMPLEX* )malloc( (*m)*LDB*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     for( i=0; i<*m; i++ )
+        for( j=0; j<*n; j++ ) {
+           B[i*LDB+j].real=b[j*(*ldb)+i].real;
+           B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+        }
+     LDC = *n+1;
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_zhemm( CblasRowMajor, side, uplo, *m, *n, alpha, A, LDA, B, LDB,
+                  beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_zhemm( CblasColMajor, side, uplo, *m, *n, alpha, a, *lda, b, *ldb,
+                  beta, c, *ldc );
+  else
+     cblas_zhemm( UNDEFINED, side, uplo, *m, *n, alpha, a, *lda, b, *ldb,
+                  beta, c, *ldc );
+}
+void F77_zsymm(int *order, char *rtlf, char *uplow, int *m, int *n,
+          CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
+	  CBLAS_TEST_ZOMPLEX *b, int *ldb, CBLAS_TEST_ZOMPLEX *beta,
+          CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
+
+  CBLAS_TEST_ZOMPLEX *A, *B, *C;
+  int i,j,LDA, LDB, LDC;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_SIDE side;
+
+  get_uplo_type(uplow,&uplo);
+  get_side_type(rtlf,&side);
+
+  if (*order == TEST_ROW_MJR) {
+     if (side == CblasLeft) {
+        LDA = *m+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*m; j++ )
+              A[i*LDA+j]=a[j*(*lda)+i];
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*n; j++ )
+              A[i*LDA+j]=a[j*(*lda)+i];
+     }
+     LDB = *n+1;
+     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_ZOMPLEX ));
+     for( i=0; i<*m; i++ )
+        for( j=0; j<*n; j++ )
+           B[i*LDB+j]=b[j*(*ldb)+i];
+     LDC = *n+1;
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_ZOMPLEX));
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ )
+           C[i*LDC+j]=c[j*(*ldc)+i];
+     cblas_zsymm( CblasRowMajor, side, uplo, *m, *n, alpha, A, LDA, B, LDB,
+                  beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ )
+           c[j*(*ldc)+i]=C[i*LDC+j];
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_zsymm( CblasColMajor, side, uplo, *m, *n, alpha, a, *lda, b, *ldb,
+                  beta, c, *ldc );
+  else
+     cblas_zsymm( UNDEFINED, side, uplo, *m, *n, alpha, a, *lda, b, *ldb,
+                  beta, c, *ldc );
+}
+
+void F77_zherk(int *order, char *uplow, char *transp, int *n, int *k,
+     double *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
+     double *beta, CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
+
+  int i,j,LDA,LDC;
+  CBLAS_TEST_ZOMPLEX *A, *C;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+
+  if (*order == TEST_ROW_MJR) {
+     if (trans == CblasNoTrans) {
+        LDA = *k+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDC = *n+1;
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     for( i=0; i<*n; i++ )
+        for( j=0; j<*n; j++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_zherk(CblasRowMajor, uplo, trans, *n, *k, *alpha, A, LDA, *beta,
+	         C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*n; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_zherk(CblasColMajor, uplo, trans, *n, *k, *alpha, a, *lda, *beta,
+	         c, *ldc );
+  else
+     cblas_zherk(UNDEFINED, uplo, trans, *n, *k, *alpha, a, *lda, *beta,
+	         c, *ldc );
+}
+
+void F77_zsyrk(int *order, char *uplow, char *transp, int *n, int *k,
+     CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
+     CBLAS_TEST_ZOMPLEX *beta, CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
+
+  int i,j,LDA,LDC;
+  CBLAS_TEST_ZOMPLEX *A, *C;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+
+  if (*order == TEST_ROW_MJR) {
+     if (trans == CblasNoTrans) {
+        LDA = *k+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDC = *n+1;
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     for( i=0; i<*n; i++ )
+        for( j=0; j<*n; j++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_zsyrk(CblasRowMajor, uplo, trans, *n, *k, alpha, A, LDA, beta,
+	         C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*n; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_zsyrk(CblasColMajor, uplo, trans, *n, *k, alpha, a, *lda, beta,
+	         c, *ldc );
+  else
+     cblas_zsyrk(UNDEFINED, uplo, trans, *n, *k, alpha, a, *lda, beta,
+	         c, *ldc );
+}
+void F77_zher2k(int *order, char *uplow, char *transp, int *n, int *k,
+        CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
+	CBLAS_TEST_ZOMPLEX *b, int *ldb, double *beta,
+        CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
+  int i,j,LDA,LDB,LDC;
+  CBLAS_TEST_ZOMPLEX *A, *B, *C;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+
+  if (*order == TEST_ROW_MJR) {
+     if (trans == CblasNoTrans) {
+        LDA = *k+1;
+        LDB = *k+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ));
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDB*sizeof(CBLAS_TEST_ZOMPLEX ));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     else {
+        LDA = *n+1;
+        LDB = *n+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc( LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        B=(CBLAS_TEST_ZOMPLEX* )malloc( LDB*(*k)*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ){
+	      A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     LDC = *n+1;
+     C=(CBLAS_TEST_ZOMPLEX* )malloc( (*n)*LDC*sizeof(CBLAS_TEST_ZOMPLEX ) );
+     for( i=0; i<*n; i++ )
+        for( j=0; j<*n; j++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_zher2k(CblasRowMajor, uplo, trans, *n, *k, alpha, A, LDA,
+		  B, LDB, *beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*n; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_zher2k(CblasColMajor, uplo, trans, *n, *k, alpha, a, *lda,
+		   b, *ldb, *beta, c, *ldc );
+  else
+     cblas_zher2k(UNDEFINED, uplo, trans, *n, *k, alpha, a, *lda,
+		   b, *ldb, *beta, c, *ldc );
+}
+void F77_zsyr2k(int *order, char *uplow, char *transp, int *n, int *k,
+         CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
+	 CBLAS_TEST_ZOMPLEX *b, int *ldb, CBLAS_TEST_ZOMPLEX *beta,
+         CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
+  int i,j,LDA,LDB,LDC;
+  CBLAS_TEST_ZOMPLEX *A, *B, *C;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+
+  if (*order == TEST_ROW_MJR) {
+     if (trans == CblasNoTrans) {
+        LDA = *k+1;
+        LDB = *k+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDB*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     else {
+        LDA = *n+1;
+        LDB = *n+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
+        B=(CBLAS_TEST_ZOMPLEX* )malloc(LDB*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ){
+	      A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     LDC = *n+1;
+     C=(CBLAS_TEST_ZOMPLEX* )malloc( (*n)*LDC*sizeof(CBLAS_TEST_ZOMPLEX));
+     for( i=0; i<*n; i++ )
+        for( j=0; j<*n; j++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_zsyr2k(CblasRowMajor, uplo, trans, *n, *k, alpha, A, LDA,
+		  B, LDB, beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*n; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_zsyr2k(CblasColMajor, uplo, trans, *n, *k, alpha, a, *lda,
+		   b, *ldb, beta, c, *ldc );
+  else
+     cblas_zsyr2k(UNDEFINED, uplo, trans, *n, *k, alpha, a, *lda,
+		   b, *ldb, beta, c, *ldc );
+}
+void F77_ztrmm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
+       int *m, int *n, CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a,
+       int *lda, CBLAS_TEST_ZOMPLEX *b, int *ldb) {
+  int i,j,LDA,LDB;
+  CBLAS_TEST_ZOMPLEX *A, *B;
+  enum CBLAS_SIDE side;
+  enum CBLAS_DIAG diag;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+  get_diag_type(diagn,&diag);
+  get_side_type(rtlf,&side);
+
+  if (*order == TEST_ROW_MJR) {
+     if (side == CblasLeft) {
+        LDA = *m+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDB = *n+1;
+     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_ZOMPLEX));
+     for( i=0; i<*m; i++ )
+        for( j=0; j<*n; j++ ) {
+           B[i*LDB+j].real=b[j*(*ldb)+i].real;
+           B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+        }
+     cblas_ztrmm(CblasRowMajor, side, uplo, trans, diag, *m, *n, alpha,
+		 A, LDA, B, LDB );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           b[j*(*ldb)+i].real=B[i*LDB+j].real;
+           b[j*(*ldb)+i].imag=B[i*LDB+j].imag;
+        }
+     free(A);
+     free(B);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_ztrmm(CblasColMajor, side, uplo, trans, diag, *m, *n, alpha,
+		   a, *lda, b, *ldb);
+  else
+     cblas_ztrmm(UNDEFINED, side, uplo, trans, diag, *m, *n, alpha,
+		   a, *lda, b, *ldb);
+}
+
+void F77_ztrsm(int *order, char *rtlf, char *uplow, char *transp, char *diagn,
+         int *m, int *n, CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a,
+         int *lda, CBLAS_TEST_ZOMPLEX *b, int *ldb) {
+  int i,j,LDA,LDB;
+  CBLAS_TEST_ZOMPLEX *A, *B;
+  enum CBLAS_SIDE side;
+  enum CBLAS_DIAG diag;
+  enum CBLAS_UPLO uplo;
+  enum CBLAS_TRANSPOSE trans;
+
+  get_uplo_type(uplow,&uplo);
+  get_transpose_type(transp,&trans);
+  get_diag_type(diagn,&diag);
+  get_side_type(rtlf,&side);
+
+  if (*order == TEST_ROW_MJR) {
+     if (side == CblasLeft) {
+        LDA = *m+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc( (*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else{
+        LDA = *n+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc((*n)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*n; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     LDB = *n+1;
+     B=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDB*sizeof(CBLAS_TEST_ZOMPLEX));
+     for( i=0; i<*m; i++ )
+        for( j=0; j<*n; j++ ) {
+           B[i*LDB+j].real=b[j*(*ldb)+i].real;
+           B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+        }
+     cblas_ztrsm(CblasRowMajor, side, uplo, trans, diag, *m, *n, alpha,
+		 A, LDA, B, LDB );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           b[j*(*ldb)+i].real=B[i*LDB+j].real;
+           b[j*(*ldb)+i].imag=B[i*LDB+j].imag;
+        }
+     free(A);
+     free(B);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_ztrsm(CblasColMajor, side, uplo, trans, diag, *m, *n, alpha,
+		   a, *lda, b, *ldb);
+  else
+     cblas_ztrsm(UNDEFINED, side, uplo, trans, diag, *m, *n, alpha,
+		   a, *lda, b, *ldb);
+}
+
+
+void F77_zgemm3m(int *order, char *transpa, char *transpb, int *m, int *n,
+     int *k, CBLAS_TEST_ZOMPLEX *alpha, CBLAS_TEST_ZOMPLEX *a, int *lda,
+     CBLAS_TEST_ZOMPLEX *b, int *ldb, CBLAS_TEST_ZOMPLEX *beta,
+     CBLAS_TEST_ZOMPLEX *c, int *ldc ) {
+
+  CBLAS_TEST_ZOMPLEX *A, *B, *C;
+  int i,j,LDA, LDB, LDC;
+  enum CBLAS_TRANSPOSE transa, transb;
+
+  get_transpose_type(transpa, &transa);
+  get_transpose_type(transpb, &transb);
+
+  if (*order == TEST_ROW_MJR) {
+     if (transa == CblasNoTrans) {
+        LDA = *k+1;
+        A=(CBLAS_TEST_ZOMPLEX*)malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*m; i++ )
+           for( j=0; j<*k; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+     else {
+        LDA = *m+1;
+        A=(CBLAS_TEST_ZOMPLEX* )malloc(LDA*(*k)*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*m; j++ ) {
+              A[i*LDA+j].real=a[j*(*lda)+i].real;
+              A[i*LDA+j].imag=a[j*(*lda)+i].imag;
+           }
+     }
+
+     if (transb == CblasNoTrans) {
+        LDB = *n+1;
+        B=(CBLAS_TEST_ZOMPLEX* )malloc((*k)*LDB*sizeof(CBLAS_TEST_ZOMPLEX) );
+        for( i=0; i<*k; i++ )
+           for( j=0; j<*n; j++ ) {
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+     else {
+        LDB = *k+1;
+        B=(CBLAS_TEST_ZOMPLEX* )malloc(LDB*(*n)*sizeof(CBLAS_TEST_ZOMPLEX));
+        for( i=0; i<*n; i++ )
+           for( j=0; j<*k; j++ ) {
+              B[i*LDB+j].real=b[j*(*ldb)+i].real;
+              B[i*LDB+j].imag=b[j*(*ldb)+i].imag;
+           }
+     }
+
+     LDC = *n+1;
+     C=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDC*sizeof(CBLAS_TEST_ZOMPLEX));
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           C[i*LDC+j].real=c[j*(*ldc)+i].real;
+           C[i*LDC+j].imag=c[j*(*ldc)+i].imag;
+        }
+     cblas_zgemm3m( CblasRowMajor, transa, transb, *m, *n, *k, alpha, A, LDA,
+                  B, LDB, beta, C, LDC );
+     for( j=0; j<*n; j++ )
+        for( i=0; i<*m; i++ ) {
+           c[j*(*ldc)+i].real=C[i*LDC+j].real;
+           c[j*(*ldc)+i].imag=C[i*LDC+j].imag;
+        }
+     free(A);
+     free(B);
+     free(C);
+  }
+  else if (*order == TEST_COL_MJR)
+     cblas_zgemm3m( CblasColMajor, transa, transb, *m, *n, *k, alpha, a, *lda,
+                  b, *ldb, beta, c, *ldc );
+  else
+     cblas_zgemm3m( UNDEFINED, transa, transb, *m, *n, *k, alpha, a, *lda,
+                  b, *ldb, beta, c, *ldc );
+}
+
diff --git a/driver/level2/CMakeLists.txt b/driver/level2/CMakeLists.txt
new file mode 100644
index 000000000..696767486
--- /dev/null
+++ b/driver/level2/CMakeLists.txt
@@ -0,0 +1,203 @@
+
+include_directories(${CMAKE_SOURCE_DIR})
+
+# sources that need to be compiled twice, once with no flags and once with LOWER
+set(UL_SOURCES
+  sbmv_k.c
+  spmv_k.c
+  spr_k.c
+  spr2_k.c
+  syr_k.c
+  syr2_k.c
+)
+
+# sources that need to be compiled several times, for UNIT, TRANSA
+set(U_SOURCES
+  trmv_U.c
+  tbmv_U.c
+  tbsv_U.c
+  tpmv_U.c
+  tpsv_U.c
+  trsv_U.c
+)
+
+set(L_SOURCES
+  trmv_L.c
+  tbmv_L.c
+  tbsv_L.c
+  tpmv_L.c
+  tpsv_L.c
+  trsv_L.c
+)
+
+set(UL_SMP_SOURCES
+  symv_thread.c
+  syr_thread.c
+  syr2_thread.c
+  spr_thread.c
+  spr2_thread.c
+  spmv_thread.c
+  sbmv_thread.c
+)
+
+set(NU_SMP_SOURCES
+  trmv_thread.c
+  tpmv_thread.c
+  tbmv_thread.c
+)
+
+set(ULVM_COMPLEX_SOURCES
+  hbmv_k.c
+  hpmv_k.c
+  hpr_k.c
+  hpr2_k.c
+  her_k.c
+  her2_k.c
+)
+
+# objects that need LOWER set
+GenerateCombinationObjects("${UL_SOURCES}" "LOWER" "U" "" 1 "" "" 3)
+
+# gbmv uses a lowercase n and t
+GenerateNamedObjects("gbmv_k.c" "" "gbmv_n" false "" "" "" 3)
+GenerateNamedObjects("gbmv_k.c" "TRANS" "gbmv_t" false "" "" "" 3)
+# c/zgbmv 
+GenerateNamedObjects("zgbmv_k.c" "CONJ" "gbmv_r" false "" "" "" 2)
+GenerateNamedObjects("zgbmv_k.c" "TRANS;CONJ" "gbmv_c" false "" "" "" 2)
+GenerateNamedObjects("zgbmv_k.c" "XCONJ" "gbmv_o" false "" "" "" 2)
+GenerateNamedObjects("zgbmv_k.c" "TRANS;XCONJ" "gbmv_u" false "" "" "" 2)
+GenerateNamedObjects("zgbmv_k.c" "CONJ;XCONJ" "gbmv_s" false "" "" "" 2)
+GenerateNamedObjects("zgbmv_k.c" "TRANS;CONJ;XCONJ" "gbmv_d" false "" "" "" 2)
+
+# special defines for complex
+foreach (float_type ${FLOAT_TYPES})
+
+  if (SMP)
+    GenerateNamedObjects("gemv_thread.c" "" "gemv_thread_n" false "" "" false ${float_type})
+    GenerateNamedObjects("gemv_thread.c" "TRANSA" "gemv_thread_t" false "" "" false ${float_type})
+
+    GenerateNamedObjects("gbmv_thread.c" "" "gbmv_thread_n" false "" "" false ${float_type})
+    GenerateNamedObjects("gbmv_thread.c" "TRANSA" "gbmv_thread_t" false "" "" false ${float_type})
+  endif ()
+
+  if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+
+    foreach (u_source ${U_SOURCES})
+      string(REGEX MATCH "[a-z]+" op_name ${u_source})
+      GenerateCombinationObjects("z${u_source}" "UNIT" "N" "TRANSA=1" 0 "${op_name}_NU" false ${float_type})
+      GenerateCombinationObjects("z${u_source}" "UNIT" "N" "TRANSA=2" 0 "${op_name}_TL" false ${float_type})
+      GenerateCombinationObjects("z${u_source}" "UNIT" "N" "TRANSA=3" 0 "${op_name}_RU" false ${float_type})
+      GenerateCombinationObjects("z${u_source}" "UNIT" "N" "TRANSA=4" 0 "${op_name}_CL" false ${float_type})
+    endforeach ()
+
+    foreach (l_source ${L_SOURCES})
+      string(REGEX MATCH "[a-z]+" op_name ${l_source})
+      GenerateCombinationObjects("z${l_source}" "UNIT" "N" "TRANSA=1" 0 "${op_name}_NL" false ${float_type})
+      GenerateCombinationObjects("z${l_source}" "UNIT" "N" "TRANSA=2" 0 "${op_name}_TU" false ${float_type})
+      GenerateCombinationObjects("z${l_source}" "UNIT" "N" "TRANSA=3" 0 "${op_name}_RL" false ${float_type})
+      GenerateCombinationObjects("z${l_source}" "UNIT" "N" "TRANSA=4" 0 "${op_name}_CU" false ${float_type})
+    endforeach ()
+
+    foreach (ulvm_source ${ULVM_COMPLEX_SOURCES})
+      string(REGEX MATCH "[a-z0-9]+" op_name ${ulvm_source})
+      GenerateNamedObjects("z${ulvm_source}" "" "${op_name}_U" false "" "" false ${float_type})
+      GenerateNamedObjects("z${ulvm_source}" "LOWER" "${op_name}_L" false "" "" false ${float_type})
+      GenerateNamedObjects("z${ulvm_source}" "HEMVREV" "${op_name}_V" false "" "" false ${float_type})
+      GenerateNamedObjects("z${ulvm_source}" "LOWER;HEMVREV" "${op_name}_M" false "" "" false ${float_type})
+    endforeach()
+
+    if (SMP)
+
+      GenerateNamedObjects("gemv_thread.c" "CONJ" "gemv_thread_r" false "" "" false ${float_type})
+      GenerateNamedObjects("gemv_thread.c" "CONJ;TRANSA" "gemv_thread_c" false "" "" false ${float_type})
+      GenerateNamedObjects("gemv_thread.c" "XCONJ" "gemv_thread_o" false "" "" false ${float_type})
+      GenerateNamedObjects("gemv_thread.c" "XCONJ;TRANSA" "gemv_thread_u" false "" "" false ${float_type})
+      GenerateNamedObjects("gemv_thread.c" "XCONJ;CONJ" "gemv_thread_s" false "" "" false ${float_type})
+      GenerateNamedObjects("gemv_thread.c" "XCONJ;CONJ;TRANSA" "gemv_thread_d" false "" "" false ${float_type})
+
+      GenerateNamedObjects("gbmv_thread.c" "CONJ" "gbmv_thread_r" false "" "" false ${float_type})
+      GenerateNamedObjects("gbmv_thread.c" "CONJ;TRANSA" "gbmv_thread_c" false "" "" false ${float_type})
+      GenerateNamedObjects("gbmv_thread.c" "XCONJ" "gbmv_thread_o" false "" "" false ${float_type})
+      GenerateNamedObjects("gbmv_thread.c" "XCONJ;TRANSA" "gbmv_thread_u" false "" "" false ${float_type})
+      GenerateNamedObjects("gbmv_thread.c" "XCONJ;CONJ" "gbmv_thread_s" false "" "" false ${float_type})
+      GenerateNamedObjects("gbmv_thread.c" "XCONJ;CONJ;TRANSA" "gbmv_thread_d" false "" "" false ${float_type})
+
+      GenerateNamedObjects("ger_thread.c" "" "ger_thread_U" false "" "" false ${float_type})
+      GenerateNamedObjects("ger_thread.c" "CONJ" "ger_thread_C" false "" "" false ${float_type})
+      GenerateNamedObjects("ger_thread.c" "XCONJ" "ger_thread_V" false "" "" false ${float_type})
+      GenerateNamedObjects("ger_thread.c" "XCONJ;CONJ" "ger_thread_D" false "" "" false ${float_type})
+
+      GenerateNamedObjects("sbmv_thread.c" "HEMV" "hbmv_thread_U" false "" "" false ${float_type})
+      GenerateNamedObjects("sbmv_thread.c" "HEMV;LOWER" "hbmv_thread_L" false "" "" false ${float_type})
+      GenerateNamedObjects("sbmv_thread.c" "HEMVREV" "hbmv_thread_V" false "" "" false ${float_type})
+      GenerateNamedObjects("sbmv_thread.c" "LOWER;HEMVREV" "hbmv_thread_M" false "" "" false ${float_type})
+
+      GenerateNamedObjects("spmv_thread.c" "HEMV" "hpmv_thread_U" false "" "" false ${float_type})
+      GenerateNamedObjects("spmv_thread.c" "HEMV;LOWER" "hpmv_thread_L" false "" "" false ${float_type})
+      GenerateNamedObjects("spmv_thread.c" "HEMVREV" "hpmv_thread_V" false "" "" false ${float_type})
+      GenerateNamedObjects("spmv_thread.c" "LOWER;HEMVREV" "hpmv_thread_M" false "" "" false ${float_type})
+
+      GenerateNamedObjects("spr_thread.c" "HEMV" "hpr_thread_U" false "" "" false ${float_type})
+      GenerateNamedObjects("spr_thread.c" "HEMV;LOWER" "hpr_thread_L" false "" "" false ${float_type})
+      GenerateNamedObjects("spr_thread.c" "HEMVREV" "hpr_thread_V" false "" "" false ${float_type})
+      GenerateNamedObjects("spr_thread.c" "LOWER;HEMVREV" "hpr_thread_M" false "" "" false ${float_type})
+
+      GenerateNamedObjects("spr2_thread.c" "HEMV" "hpr2_thread_U" false "" "" false ${float_type})
+      GenerateNamedObjects("spr2_thread.c" "HEMV;LOWER" "hpr2_thread_L" false "" "" false ${float_type})
+      GenerateNamedObjects("spr2_thread.c" "HEMVREV" "hpr2_thread_V" false "" "" false ${float_type})
+      GenerateNamedObjects("spr2_thread.c" "LOWER;HEMVREV" "hpr2_thread_M" false "" "" false ${float_type})
+
+      GenerateNamedObjects("symv_thread.c" "HEMV" "hemv_thread_U" false "" "" false ${float_type})
+      GenerateNamedObjects("symv_thread.c" "HEMV;LOWER" "hemv_thread_L" false "" "" false ${float_type})
+      GenerateNamedObjects("symv_thread.c" "HEMVREV" "hemv_thread_V" false "" "" false ${float_type})
+      GenerateNamedObjects("symv_thread.c" "LOWER;HEMVREV" "hemv_thread_M" false "" "" false ${float_type})
+
+      GenerateNamedObjects("syr_thread.c" "HER" "her_thread_U" false "" "" false ${float_type})
+      GenerateNamedObjects("syr_thread.c" "HER;LOWER" "her_thread_L" false "" "" false ${float_type})
+      GenerateNamedObjects("syr_thread.c" "HERREV" "her_thread_V" false "" "" false ${float_type})
+      GenerateNamedObjects("syr_thread.c" "LOWER;HERREV" "her_thread_M" false "" "" false ${float_type})
+
+      GenerateNamedObjects("syr2_thread.c" "HER" "her2_thread_U" false "" "" false ${float_type})
+      GenerateNamedObjects("syr2_thread.c" "HER;LOWER" "her2_thread_L" false "" "" false ${float_type})
+      GenerateNamedObjects("syr2_thread.c" "HERREV" "her2_thread_V" false "" "" false ${float_type})
+      GenerateNamedObjects("syr2_thread.c" "LOWER;HERREV" "her2_thread_M" false "" "" false ${float_type})
+
+      foreach (nu_smp_src ${NU_SMP_SOURCES})
+        string(REGEX MATCH "[a-z]+_[a-z]+" op_name ${nu_smp_src})
+        GenerateCombinationObjects("${nu_smp_src}" "LOWER;UNIT" "U;N" "TRANSA=1" 0 "${op_name}_N" false ${float_type})
+        GenerateCombinationObjects("${nu_smp_src}" "LOWER;UNIT" "U;N" "TRANSA=2" 0 "${op_name}_T" false ${float_type})
+        GenerateCombinationObjects("${nu_smp_src}" "LOWER;UNIT" "U;N" "TRANSA=3" 0 "${op_name}_R" false ${float_type})
+        GenerateCombinationObjects("${nu_smp_src}" "LOWER;UNIT" "U;N" "TRANSA=4" 0 "${op_name}_C" false ${float_type})
+      endforeach ()
+    endif ()
+
+  else ()
+    # For real number functions
+    foreach (u_source ${U_SOURCES})
+      string(REGEX MATCH "[a-z]+" op_name ${u_source})
+      GenerateCombinationObjects("${u_source}" "UNIT" "N" "" 0 "${op_name}_NU" false ${float_type})
+      GenerateCombinationObjects("${u_source}" "UNIT" "N" "TRANSA" 0 "${op_name}_TL" false ${float_type})
+    endforeach ()
+
+    foreach (l_source ${L_SOURCES})
+      string(REGEX MATCH "[a-z]+" op_name ${l_source})
+      GenerateCombinationObjects("${l_source}" "UNIT" "N" "" 0 "${op_name}_NL" false ${float_type})
+      GenerateCombinationObjects("${l_source}" "UNIT" "N" "TRANSA" 0 "${op_name}_TU" false ${float_type})
+    endforeach ()
+
+    if (SMP)
+      GenerateNamedObjects("ger_thread.c" "" "" false "" "" false ${float_type})
+      foreach(nu_smp_source ${NU_SMP_SOURCES})
+        string(REGEX MATCH "[a-z]+_[a-z]+" op_name ${nu_smp_source})
+        GenerateCombinationObjects("${nu_smp_source}" "LOWER;UNIT" "U;N" "" 0 "${op_name}_N" false ${float_type})
+        GenerateCombinationObjects("${nu_smp_source}" "LOWER;UNIT" "U;N" "TRANSA" 0 "${op_name}_T" false ${float_type})
+      endforeach()
+    endif ()
+  endif ()
+endforeach ()
+
+if (SMP)
+  GenerateCombinationObjects("${UL_SMP_SOURCES}" "LOWER" "U" "" 2)
+endif ()
+
+add_library(driver_level2 OBJECT ${OPENBLAS_SRC})
diff --git a/driver/level2/gbmv_thread.c b/driver/level2/gbmv_thread.c
index 9efe17092..ef9d58d76 100644
--- a/driver/level2/gbmv_thread.c
+++ b/driver/level2/gbmv_thread.c
@@ -64,7 +64,7 @@ static int gbmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
 #ifndef COMPLEX
   FLOAT result;
 #else
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 #endif
 
diff --git a/driver/level2/gemv_thread.c b/driver/level2/gemv_thread.c
index ddd475367..061454848 100644
--- a/driver/level2/gemv_thread.c
+++ b/driver/level2/gemv_thread.c
@@ -62,6 +62,11 @@
 #endif
 #endif
 
+#ifndef TRANSA
+#define Y_DUMMY_NUM 1024
+static FLOAT y_dummy[Y_DUMMY_NUM];
+#endif
+
 static int gemv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *dummy1, FLOAT *buffer, BLASLONG pos){
 
   FLOAT *a, *x, *y;
@@ -99,10 +104,15 @@ static int gemv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
     a += n_from * lda  * COMPSIZE;
 #ifdef TRANSA
     y += n_from * incy * COMPSIZE;
+#else
+    //for split matrix row (n) direction and vector x of gemv_n
+    x += n_from * incx * COMPSIZE;
+    //store partial result for every thread
+    y += (m_to - m_from) * 1 * COMPSIZE * pos;
 #endif
   }
 
-  //  fprintf(stderr, "M_From = %d  M_To = %d  N_From = %d  N_To = %d\n", m_from, m_to, n_from, n_to);
+  //fprintf(stderr, "M_From = %d  M_To = %d  N_From = %d  N_To = %d POS=%d\n", m_from, m_to, n_from, n_to, pos);
 
   GEMV(m_to - m_from, n_to - n_from, 0,
        *((FLOAT *)args -> alpha + 0),
@@ -126,6 +136,10 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *alpha, FLOAT *a, BLASLONG lda, FLOAT *x
 
   BLASLONG width, i, num_cpu;
 
+#ifndef TRANSA
+  int split_x=0;
+#endif
+
 #ifdef SMP
 #ifndef COMPLEX
 #ifdef XDOUBLE
@@ -198,6 +212,58 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *alpha, FLOAT *a, BLASLONG lda, FLOAT *x
     i -= width;
   }
 
+#ifndef TRANSA
+  //try to split matrix on row direction and x.
+  //Then, reduction.
+  if (num_cpu < nthreads) {
+
+    //too small to split or bigger than the y_dummy buffer.
+    double MN = (double) m * (double) n;
+    if ( MN <= (24.0 * 24.0  * (double) (GEMM_MULTITHREAD_THRESHOLD*GEMM_MULTITHREAD_THRESHOLD))
+	 || m*COMPSIZE*nthreads > Y_DUMMY_NUM)
+      goto Outer;
+
+    num_cpu  = 0;
+    range[0] = 0;
+
+    memset(y_dummy, 0, sizeof(FLOAT) * m * COMPSIZE * nthreads);
+
+    args.ldc = 1;
+    args.c = (void *)y_dummy;
+
+    //split on row (n) and x
+    i=n;
+    split_x=1;
+    while (i > 0){
+
+      width  = blas_quickdivide(i + nthreads - num_cpu - 1, nthreads - num_cpu);
+      if (width < 4) width = 4;
+      if (i < width) width = i;
+
+      range[num_cpu + 1] = range[num_cpu] + width;
+
+      queue[num_cpu].mode    = mode;
+      queue[num_cpu].routine = gemv_kernel;
+      queue[num_cpu].args    = &args;
+
+      queue[num_cpu].position = num_cpu;
+
+      queue[num_cpu].range_m = NULL;
+      queue[num_cpu].range_n = &range[num_cpu];
+
+      queue[num_cpu].sa      = NULL;
+      queue[num_cpu].sb      = NULL;
+      queue[num_cpu].next    = &queue[num_cpu + 1];
+
+      num_cpu ++;
+      i -= width;
+    }
+
+  }
+
+  Outer:
+#endif
+
   if (num_cpu) {
     queue[0].sa = NULL;
     queue[0].sb = buffer;
@@ -206,5 +272,21 @@ int CNAME(BLASLONG m, BLASLONG n, FLOAT *alpha, FLOAT *a, BLASLONG lda, FLOAT *x
     exec_blas(num_cpu, queue);
   }
 
+#ifndef TRANSA
+  if(split_x==1){
+    //reduction
+    for(i=0; i<num_cpu; i++){
+
+      int j;
+      for(j=0; j<m; j++){
+	y[j*incy*COMPSIZE] +=y_dummy[i*m*COMPSIZE + j*COMPSIZE];
+#ifdef COMPLEX
+	y[j*incy*COMPSIZE+1] +=y_dummy[i*m*COMPSIZE + j*COMPSIZE+1];
+#endif
+      }
+    }
+  }
+#endif
+
   return 0;
 }
diff --git a/driver/level2/sbmv_thread.c b/driver/level2/sbmv_thread.c
index 5b7fc7332..a0377d638 100644
--- a/driver/level2/sbmv_thread.c
+++ b/driver/level2/sbmv_thread.c
@@ -60,7 +60,7 @@ static int sbmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
 #ifndef COMPLEX
   FLOAT result;
 #else
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 
   a = (FLOAT *)args -> a;
diff --git a/driver/level2/spmv_thread.c b/driver/level2/spmv_thread.c
index 93a2f44d4..0f47344df 100644
--- a/driver/level2/spmv_thread.c
+++ b/driver/level2/spmv_thread.c
@@ -60,7 +60,7 @@ static int spmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
 #ifndef COMPLEX
   FLOAT result;
 #else
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 
   a = (FLOAT *)args -> a;
diff --git a/driver/level2/tbmv_thread.c b/driver/level2/tbmv_thread.c
index 3c1249448..bbb1c50eb 100644
--- a/driver/level2/tbmv_thread.c
+++ b/driver/level2/tbmv_thread.c
@@ -76,7 +76,7 @@ static int trmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
 #ifndef COMPLEX
   FLOAT          result;
 #else
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 #endif
 
diff --git a/driver/level2/tpmv_thread.c b/driver/level2/tpmv_thread.c
index 3b91cee45..47dc1daf9 100644
--- a/driver/level2/tpmv_thread.c
+++ b/driver/level2/tpmv_thread.c
@@ -81,7 +81,7 @@ static int tpmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
 #ifndef COMPLEX
   FLOAT          result;
 #else
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 #endif
 
diff --git a/driver/level2/trmv_thread.c b/driver/level2/trmv_thread.c
index 29e9799f6..a9dc2dc62 100644
--- a/driver/level2/trmv_thread.c
+++ b/driver/level2/trmv_thread.c
@@ -87,7 +87,7 @@ static int trmv_kernel(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, F
 #ifndef COMPLEX
   FLOAT          result;
 #else
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 #endif
 
diff --git a/driver/level2/zgbmv_k.c b/driver/level2/zgbmv_k.c
index 68d6045bd..d89932e33 100644
--- a/driver/level2/zgbmv_k.c
+++ b/driver/level2/zgbmv_k.c
@@ -77,7 +77,7 @@ void CNAME(BLASLONG m, BLASLONG n, BLASLONG ku, BLASLONG kl, FLOAT alpha_r, FLOA
   FLOAT *bufferY    = gemvbuffer;
   FLOAT *bufferX    = gemvbuffer;
 #ifdef TRANS
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 
   if (incy != 1) {
diff --git a/driver/level2/zhbmv_k.c b/driver/level2/zhbmv_k.c
index 70e92e050..33f70d2c5 100644
--- a/driver/level2/zhbmv_k.c
+++ b/driver/level2/zhbmv_k.c
@@ -56,6 +56,8 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
   FLOAT *bufferX    = sbmvbuffer;
   FLOAT temp[2];
 
+  OPENBLAS_COMPLEX_FLOAT result;
+
   if (incy != 1) {
     Y = bufferY;
     bufferX    = (FLOAT *)(((BLASLONG)bufferY + n * sizeof(FLOAT) * COMPSIZE + 4095) & ~4095);
@@ -93,7 +95,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
     Y[i * 2 + 1] +=  alpha_r * temp[1] + alpha_i * temp[0];
 
     if (length > 0) {
-      FLOAT _Complex result = DOTC_K(length, a + offset * COMPSIZE, 1, X + (i - length) * COMPSIZE, 1);
+      result = DOTC_K(length, a + offset * COMPSIZE, 1, X + (i - length) * COMPSIZE, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
@@ -118,7 +120,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
     Y[i * 2 + 1] +=  alpha_r * temp[1] + alpha_i * temp[0];
 
     if (length > 0) {
-      FLOAT _Complex result = DOTC_K(length, a + COMPSIZE, 1, X + (i + 1) * COMPSIZE, 1);
+      result = DOTC_K(length, a + COMPSIZE, 1, X + (i + 1) * COMPSIZE, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
@@ -143,7 +145,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
     Y[i * 2 + 1] +=  alpha_r * temp[1] + alpha_i * temp[0];
 
     if (length > 0) {
-      FLOAT _Complex result = DOTU_K(length, a + offset * COMPSIZE, 1, X + (i - length) * COMPSIZE, 1);
+      result = DOTU_K(length, a + offset * COMPSIZE, 1, X + (i - length) * COMPSIZE, 1);
 
       Y[i * 2 + 0] +=   alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] +=  alpha_r * CIMAG(result) + alpha_i * CREAL(result);
@@ -168,7 +170,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
     Y[i * 2 + 1] +=  alpha_r * temp[1] + alpha_i * temp[0];
 
     if (length > 0) {
-      FLOAT _Complex result = DOTU_K(length, a + COMPSIZE, 1, X + (i + 1) * COMPSIZE, 1);
+      result = DOTU_K(length, a + COMPSIZE, 1, X + (i + 1) * COMPSIZE, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
diff --git a/driver/level2/zhpmv_k.c b/driver/level2/zhpmv_k.c
index 96bceaaf2..9e7ed7b0e 100644
--- a/driver/level2/zhpmv_k.c
+++ b/driver/level2/zhpmv_k.c
@@ -51,6 +51,8 @@ int CNAME(BLASLONG m, FLOAT alpha_r, FLOAT alpha_i,
   FLOAT *bufferX    = gemvbuffer;
   FLOAT temp[2];
 
+  OPENBLAS_COMPLEX_FLOAT result;
+
   if (incy != 1) {
     Y = bufferY;
     bufferX    = (FLOAT *)(((BLASLONG)bufferY + m * sizeof(FLOAT) * 2 + 4095) & ~4095);
@@ -69,7 +71,7 @@ int CNAME(BLASLONG m, FLOAT alpha_r, FLOAT alpha_i,
 #ifndef HEMVREV
 #ifndef LOWER
     if (i > 0) {
-      FLOAT _Complex result = DOTC_K(i, a, 1, X, 1);
+      result = DOTC_K(i, a, 1, X, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
@@ -93,7 +95,7 @@ int CNAME(BLASLONG m, FLOAT alpha_r, FLOAT alpha_i,
 #else
 
     if (m - i > 1) {
-      FLOAT _Complex result = DOTC_K(m - i - 1, a + (i + 1) * 2, 1, X + (i + 1) * 2, 1);
+      result = DOTC_K(m - i - 1, a + (i + 1) * 2, 1, X + (i + 1) * 2, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
@@ -118,7 +120,7 @@ int CNAME(BLASLONG m, FLOAT alpha_r, FLOAT alpha_i,
 #else
 #ifndef LOWER
     if (i > 0) {
-      FLOAT _Complex result = DOTU_K(i, a, 1, X, 1);
+      result = DOTU_K(i, a, 1, X, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
@@ -142,7 +144,7 @@ int CNAME(BLASLONG m, FLOAT alpha_r, FLOAT alpha_i,
 #else
 
     if (m - i > 1) {
-      FLOAT _Complex result = DOTU_K(m - i - 1, a + (i + 1) * 2, 1, X + (i + 1) * 2, 1);
+      result = DOTU_K(m - i - 1, a + (i + 1) * 2, 1, X + (i + 1) * 2, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
diff --git a/driver/level2/zsbmv_k.c b/driver/level2/zsbmv_k.c
index 30e2f91c3..3ae74ce80 100644
--- a/driver/level2/zsbmv_k.c
+++ b/driver/level2/zsbmv_k.c
@@ -55,6 +55,8 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
   FLOAT *bufferY    = sbmvbuffer;
   FLOAT *bufferX    = sbmvbuffer;
 
+  OPENBLAS_COMPLEX_FLOAT result;
+
   if (incy != 1) {
     Y = bufferY;
     bufferX    = (FLOAT *)(((BLASLONG)bufferY + n * sizeof(FLOAT) * COMPSIZE + 4095) & ~4095);
@@ -83,7 +85,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
 	    a + offset * COMPSIZE, 1, Y + (i - length) * COMPSIZE, 1, NULL, 0);
 
     if (length > 0) {
-      FLOAT _Complex result = DOTU_K(length, a + offset * COMPSIZE, 1, X + (i - length) * COMPSIZE, 1);
+      result = DOTU_K(length, a + offset * COMPSIZE, 1, X + (i - length) * COMPSIZE, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
@@ -100,7 +102,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT alpha_r, FLOAT alpha_i,
 	    a, 1, Y + i * COMPSIZE, 1, NULL, 0);
 
     if (length > 0) {
-      FLOAT _Complex result = DOTU_K(length, a + COMPSIZE, 1, X + (i + 1) * COMPSIZE, 1);
+      result = DOTU_K(length, a + COMPSIZE, 1, X + (i + 1) * COMPSIZE, 1);
 
       Y[i * 2 + 0] += alpha_r * CREAL(result) - alpha_i * CIMAG(result);
       Y[i * 2 + 1] += alpha_r * CIMAG(result) + alpha_i * CREAL(result);
diff --git a/driver/level2/zspmv_k.c b/driver/level2/zspmv_k.c
index 76657eab9..432205e83 100644
--- a/driver/level2/zspmv_k.c
+++ b/driver/level2/zspmv_k.c
@@ -49,7 +49,8 @@ int CNAME(BLASLONG m, FLOAT alpha_r, FLOAT alpha_i,
   FLOAT *gemvbuffer = (FLOAT *)buffer;
   FLOAT *bufferY    = gemvbuffer;
   FLOAT *bufferX    = gemvbuffer;
-  FLOAT _Complex result;
+
+  OPENBLAS_COMPLEX_FLOAT result;
 
   if (incy != 1) {
     Y = bufferY;
diff --git a/driver/level2/ztbmv_L.c b/driver/level2/ztbmv_L.c
index 74ff0bce1..1ac1cdef1 100644
--- a/driver/level2/ztbmv_L.c
+++ b/driver/level2/ztbmv_L.c
@@ -49,7 +49,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG inc
   FLOAT *B = b;
   BLASLONG length;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 #ifndef UNIT
   FLOAT atemp1, atemp2, btemp1, btemp2;
diff --git a/driver/level2/ztbmv_U.c b/driver/level2/ztbmv_U.c
index 933275de3..9aa203396 100644
--- a/driver/level2/ztbmv_U.c
+++ b/driver/level2/ztbmv_U.c
@@ -49,7 +49,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG inc
   FLOAT *B = b;
   BLASLONG length;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 #ifndef UNIT
   FLOAT atemp1, atemp2, btemp1, btemp2;
diff --git a/driver/level2/ztbsv_L.c b/driver/level2/ztbsv_L.c
index 0726bbd16..9aa701841 100644
--- a/driver/level2/ztbsv_L.c
+++ b/driver/level2/ztbsv_L.c
@@ -49,7 +49,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG inc
   FLOAT *B = b;
   BLASLONG length;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 #ifndef UNIT
   FLOAT ar, ai, br, bi, ratio, den;
diff --git a/driver/level2/ztbsv_U.c b/driver/level2/ztbsv_U.c
index d022650bc..3722b1f71 100644
--- a/driver/level2/ztbsv_U.c
+++ b/driver/level2/ztbsv_U.c
@@ -49,7 +49,7 @@ int CNAME(BLASLONG n, BLASLONG k, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG inc
   FLOAT *B = b;
   BLASLONG length;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 #ifndef UNIT
   FLOAT ar, ai, br, bi, ratio, den;
diff --git a/driver/level2/ztpmv_L.c b/driver/level2/ztpmv_L.c
index 12c254c12..47e6df56c 100644
--- a/driver/level2/ztpmv_L.c
+++ b/driver/level2/ztpmv_L.c
@@ -44,7 +44,7 @@ int CNAME(BLASLONG m, FLOAT *a, FLOAT *b, BLASLONG incb, void *buffer){
 
   BLASLONG i;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 #ifndef UNIT
   FLOAT atemp1, atemp2, btemp1, btemp2;
diff --git a/driver/level2/ztpmv_U.c b/driver/level2/ztpmv_U.c
index 59708b8b8..da911fb4e 100644
--- a/driver/level2/ztpmv_U.c
+++ b/driver/level2/ztpmv_U.c
@@ -44,7 +44,7 @@ int CNAME(BLASLONG m, FLOAT *a, FLOAT *b, BLASLONG incb, void *buffer){
 
   BLASLONG i;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 #ifndef UNIT
   FLOAT atemp1, atemp2, btemp1, btemp2;
diff --git a/driver/level2/ztpsv_L.c b/driver/level2/ztpsv_L.c
index 3b8e562ce..a497e42a4 100644
--- a/driver/level2/ztpsv_L.c
+++ b/driver/level2/ztpsv_L.c
@@ -46,7 +46,7 @@ int CNAME(BLASLONG m, FLOAT *a, FLOAT *b, BLASLONG incb, void *buffer){
 
   BLASLONG i;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 #ifndef UNIT
   FLOAT ar, ai, br, bi, ratio, den;
diff --git a/driver/level2/ztpsv_U.c b/driver/level2/ztpsv_U.c
index 601ac2f9d..28b824e3a 100644
--- a/driver/level2/ztpsv_U.c
+++ b/driver/level2/ztpsv_U.c
@@ -44,7 +44,7 @@ int CNAME(BLASLONG m, FLOAT *a, FLOAT *b, BLASLONG incb, void *buffer){
 
   BLASLONG i;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 #ifndef UNIT
   FLOAT ar, ai, br, bi, ratio, den;
diff --git a/driver/level2/ztrmv_L.c b/driver/level2/ztrmv_L.c
index 63522cf81..92c86aec2 100644
--- a/driver/level2/ztrmv_L.c
+++ b/driver/level2/ztrmv_L.c
@@ -46,7 +46,7 @@ int CNAME(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG incb, FLOAT *bu
 
   BLASLONG i, is, min_i;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 #ifndef UNIT
   FLOAT atemp1, atemp2, btemp1, btemp2;
diff --git a/driver/level2/ztrmv_U.c b/driver/level2/ztrmv_U.c
index 8a4494fd7..f9671c9d6 100644
--- a/driver/level2/ztrmv_U.c
+++ b/driver/level2/ztrmv_U.c
@@ -46,7 +46,7 @@ int CNAME(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG incb, FLOAT *bu
 
   BLASLONG i, is, min_i;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex temp;
+  OPENBLAS_COMPLEX_FLOAT temp;
 #endif
 #ifndef UNIT
   FLOAT atemp1, atemp2, btemp1, btemp2;
diff --git a/driver/level2/ztrsv_L.c b/driver/level2/ztrsv_L.c
index 90f1c2c7d..dd3b2786e 100644
--- a/driver/level2/ztrsv_L.c
+++ b/driver/level2/ztrsv_L.c
@@ -46,7 +46,7 @@ int CNAME(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG incb, void *buf
 
   BLASLONG i, is, min_i;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 #ifndef UNIT
   FLOAT ar, ai, br, bi, ratio, den;
diff --git a/driver/level2/ztrsv_U.c b/driver/level2/ztrsv_U.c
index bec8114f3..8803182a8 100644
--- a/driver/level2/ztrsv_U.c
+++ b/driver/level2/ztrsv_U.c
@@ -46,7 +46,7 @@ int CNAME(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b, BLASLONG incb, void *buf
 
   BLASLONG i, is, min_i;
 #if (TRANSA == 2) || (TRANSA == 4)
-  FLOAT _Complex result;
+  OPENBLAS_COMPLEX_FLOAT result;
 #endif
 #ifndef UNIT
   FLOAT ar, ai, br, bi, ratio, den;
diff --git a/driver/level3/CMakeLists.txt b/driver/level3/CMakeLists.txt
new file mode 100644
index 000000000..41d440f7a
--- /dev/null
+++ b/driver/level3/CMakeLists.txt
@@ -0,0 +1,115 @@
+include_directories(${CMAKE_SOURCE_DIR})
+
+# N.B. In the original makefile there was a BLOCKS define used in the compilation of these files but I don't see any evidence of it being set anywhere. -hpa
+
+# loop through gemm.c defines
+set(GEMM_DEFINES NN NT TN TT)
+set(GEMM_COMPLEX_DEFINES RN CN RT CT NR TR RR CR NC TC RC CC)
+foreach (GEMM_DEFINE ${GEMM_DEFINES})
+  string(TOLOWER ${GEMM_DEFINE} GEMM_DEFINE_LC)
+  GenerateNamedObjects("gemm.c" "${GEMM_DEFINE}" "gemm_${GEMM_DEFINE_LC}" 0)
+  if (SMP AND NOT USE_SIMPLE_THREADED_LEVEL3)
+    GenerateNamedObjects("gemm.c" "${GEMM_DEFINE};THREADED_LEVEL3" "gemm_thread_${GEMM_DEFINE_LC}" 0)
+  endif ()
+endforeach ()
+
+
+set(TRMM_TRSM_SOURCES
+  trmm_L.c
+  trmm_R.c
+  trsm_L.c
+  trsm_R.c)
+
+foreach(trmm_trsm_source ${TRMM_TRSM_SOURCES})
+  string(REGEX MATCH "[a-z]+_[A-Z]+" op_name ${trmm_trsm_source})
+  GenerateCombinationObjects("${trmm_trsm_source}" "UPPER;UNIT" "L;N" "" 0 "${op_name}N")
+  GenerateCombinationObjects("${trmm_trsm_source}" "UPPER;UNIT" "L;N" "TRANSA" 0 "${op_name}T")
+endforeach()
+
+GenerateCombinationObjects("symm_k.c" "RSIDE;LOWER" "L;U" "NN" 1)
+GenerateCombinationObjects("syrk_k.c" "LOWER;TRANS" "U;N" "" 1)
+GenerateCombinationObjects("syr2k_k.c" "LOWER;TRANS" "U;N" "" 1)
+GenerateCombinationObjects("syrk_kernel.c" "LOWER" "U" "" 2)
+GenerateCombinationObjects("syr2k_kernel.c" "LOWER" "U" "" 2)
+if (SMP)
+
+  # N.B. these do NOT have a float type (e.g. DOUBLE) defined!
+  GenerateNamedObjects("gemm_thread_m.c;gemm_thread_n.c;gemm_thread_mn.c;gemm_thread_variable.c;syrk_thread.c" "" "" 0 "" "" 1)
+
+  if (NOT USE_SIMPLE_THREADED_LEVEL3)
+    GenerateCombinationObjects("syrk_k.c" "LOWER;TRANS" "U;N" "THREADED_LEVEL3" 2 "syrk_thread")
+    GenerateCombinationObjects("symm_k.c" "RSIDE;LOWER" "L;U" "THREADED_LEVEL3;NN" 2 "symm_thread")
+  endif ()
+endif ()
+
+foreach (float_type ${FLOAT_TYPES})
+  if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+    GenerateCombinationObjects("zherk_kernel.c" "LOWER;CONJ" "U;N" "HERK" 2 "herk_kernel" false ${float_type})
+    # TRANS needs to be set/unset when CONJ is set/unset, so can't use it as a combination
+    GenerateCombinationObjects("zherk_k.c" "LOWER" "U" "HERK" 3 "herk_N" false ${float_type})
+    GenerateCombinationObjects("zherk_k.c" "LOWER" "U" "HERK;TRANS;CONJ" 3 "herk_C" false ${float_type})
+    GenerateCombinationObjects("zherk_k.c" "LOWER" "U" "HERK;THREADED_LEVEL3" 3 "herk_thread_N" false ${float_type})
+    GenerateCombinationObjects("zherk_k.c" "LOWER" "U" "HERK;THREADED_LEVEL3;TRANS;CONJ" 3 "herk_thread_C" false ${float_type})
+    # Need to set CONJ for trmm and trsm
+    GenerateCombinationObjects("trmm_L.c" "UPPER;UNIT" "L;N" "CONJ" 0 "trmm_LR" false ${float_type})
+    GenerateCombinationObjects("trmm_L.c" "UPPER;UNIT" "L;N" "TRANSA;CONJ" 0 "trmm_LC" false ${float_type})
+    GenerateCombinationObjects("trmm_R.c" "UPPER;UNIT" "L;N" "CONJ" 0 "trmm_RR" false ${float_type})
+    GenerateCombinationObjects("trmm_R.c" "UPPER;UNIT" "L;N" "TRANSA;CONJ" 0 "trmm_RC" false ${float_type})
+    GenerateCombinationObjects("trsm_L.c" "UPPER;UNIT" "L;N" "CONJ" 0 "trsm_LR" false ${float_type})
+    GenerateCombinationObjects("trsm_L.c" "UPPER;UNIT" "L;N" "TRANSA;CONJ" 0 "trsm_LC" false ${float_type})
+    GenerateCombinationObjects("trsm_R.c" "UPPER;UNIT" "L;N" "CONJ" 0 "trsm_RR" false ${float_type})
+    GenerateCombinationObjects("trsm_R.c" "UPPER;UNIT" "L;N" "TRANSA;CONJ" 0 "trsm_RC" false ${float_type})
+
+    #hemm
+    GenerateCombinationObjects("zhemm_k.c" "LOWER" "U" "NN" 0 "hemm_L" false ${float_type})
+    GenerateCombinationObjects("zhemm_k.c" "LOWER" "U" "NC;RSIDE" 0 "hemm_R" false ${float_type})
+
+    #her2k
+    GenerateCombinationObjects("zher2k_kernel.c" "LOWER;CONJ" "U;N" "" 2 "her2k_kernel" false ${float_type})
+    GenerateNamedObjects("zher2k_k.c" "HER2K" "her2k_UN" false "" "" false ${float_type})
+    GenerateNamedObjects("zher2k_k.c" "HER2K;TRANS;CONJ" "her2k_UC" false "" "" false ${float_type})
+    GenerateNamedObjects("zher2k_k.c" "HER2K;LOWER" "her2k_LN" false "" "" false ${float_type})
+    GenerateNamedObjects("zher2k_k.c" "HER2K;LOWER;TRANS;CONJ" "her2k_LC" false "" "" false ${float_type})
+
+    if (SMP AND NOT USE_SIMPLE_THREADED_LEVEL3)
+      #hemm
+      GenerateCombinationObjects("zhemm_k.c" "LOWER" "U" "NN;THREADED_LEVEL3" 0 "hemm_thread_L" false ${float_type})
+      GenerateCombinationObjects("zhemm_k.c" "LOWER" "U" "NC;RSIDE;THREADED_LEVEL3" 0 "hemm_thread_R" false ${float_type})
+      #her2k
+      GenerateNamedObjects("zher2k_k.c" "HER2K" "her2k_UN" false "" "" false ${float_type})
+      GenerateNamedObjects("zher2k_k.c" "HER2K;TRANS;CONJ" "her2k_UC" false "" "" false ${float_type})
+      GenerateNamedObjects("zher2k_k.c" "HER2K;LOWER" "her2k_LN" false "" "" false ${float_type})
+      GenerateNamedObjects("zher2k_k.c" "HER2K;LOWER;TRANS;CONJ" "her2k_LC" false "" "" false ${float_type})
+    endif()
+
+    # special gemm defines for complex
+    foreach (gemm_define ${GEMM_COMPLEX_DEFINES})
+      string(TOLOWER ${gemm_define} gemm_define_LC)
+      GenerateNamedObjects("gemm.c" "${gemm_define}" "gemm_${gemm_define_LC}" false "" "" false ${float_type})
+      if(USE_GEMM3M)
+	GenerateNamedObjects("gemm3m.c" "${gemm_define}" "gemm3m_${gemm_define_LC}" false "" "" false ${float_type})
+      endif()
+      if (SMP AND NOT USE_SIMPLE_THREADED_LEVEL3)
+        GenerateNamedObjects("gemm.c" "${gemm_define};THREADED_LEVEL3" "gemm_thread_${gemm_define_LC}" false "" "" false ${float_type})
+	if(USE_GEMM3M)
+	  GenerateNamedObjects("gemm3m.c" "${gemm_define};THREADED_LEVEL3" "gemm3m_thread_${gemm_define_LC}" false "" "" false ${float_type})
+	endif()
+      endif ()
+    endforeach ()
+  endif ()
+endforeach ()
+
+#HPLOBJS =
+#	dgemm_nn.c dgemm_nt.c dgemm_tn.c dgemm_tt.c
+#	dtrsm_LNUU.c dtrsm_LNUN.c dtrsm_LNLU.c dtrsm_LNLN.c
+#	dtrsm_LTUU.c dtrsm_LTUN.c dtrsm_LTLU.c dtrsm_LTLN.c
+#	dtrsm_RNUU.c dtrsm_RNUN.c dtrsm_RNLU.c dtrsm_RNLN.c
+#	dtrsm_RTUU.c dtrsm_RTUN.c dtrsm_RTLU.c dtrsm_RTLN.c
+#
+#if (USE_SIMPLE_THREADED_LEVEL3)
+#  HPLOBJS += dgemm_thread_nn.c dgemm_thread_nt.c
+#	   dgemm_thread_tn.c dgemm_thread_tt.c
+#endif
+#
+
+add_library(driver_level3 OBJECT ${OPENBLAS_SRC})
diff --git a/driver/level3/syr2k_k.c b/driver/level3/syr2k_k.c
index 8df0f122f..09131fbdb 100644
--- a/driver/level3/syr2k_k.c
+++ b/driver/level3/syr2k_k.c
@@ -47,7 +47,7 @@
 #endif
 #endif
 
-static inline int syrk_beta(BLASLONG m_from, BLASLONG m_to, BLASLONG n_from, BLASLONG n_to, FLOAT *alpha, FLOAT *c, BLASLONG ldc) {
+static __inline int syrk_beta(BLASLONG m_from, BLASLONG m_to, BLASLONG n_from, BLASLONG n_to, FLOAT *alpha, FLOAT *c, BLASLONG ldc) {
 
   BLASLONG i;
 
diff --git a/driver/level3/syrk_k.c b/driver/level3/syrk_k.c
index 08751dc8b..8bc817f87 100644
--- a/driver/level3/syrk_k.c
+++ b/driver/level3/syrk_k.c
@@ -49,7 +49,7 @@
 #endif
 #endif
 
-static inline int syrk_beta(BLASLONG m_from, BLASLONG m_to, BLASLONG n_from, BLASLONG n_to, FLOAT *alpha, FLOAT *c, BLASLONG ldc) {
+static __inline int syrk_beta(BLASLONG m_from, BLASLONG m_to, BLASLONG n_from, BLASLONG n_to, FLOAT *alpha, FLOAT *c, BLASLONG ldc) {
 
   BLASLONG i;
 
diff --git a/driver/level3/trmm_R.c b/driver/level3/trmm_R.c
index bdd9370cd..0882aa496 100644
--- a/driver/level3/trmm_R.c
+++ b/driver/level3/trmm_R.c
@@ -70,6 +70,9 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
   BLASLONG ls, is, js;
   BLASLONG min_l, min_i, min_j;
   BLASLONG jjs, min_jj;
+#if !((!defined(UPPER) && !defined(TRANSA)) || (defined(UPPER) && defined(TRANSA)))
+  BLASLONG start_ls;
+#endif
 
   m = args -> m;
   n = args -> n;
@@ -226,8 +229,6 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
   }
 
 #else
-  BLASLONG start_ls;
-
   for(js = n; js > 0; js -= GEMM_R){
     min_j = js;
     if (min_j > GEMM_R) min_j = GEMM_R;
diff --git a/driver/level3/trsm_L.c b/driver/level3/trsm_L.c
index 78da0eb6c..d8130ee7e 100644
--- a/driver/level3/trsm_L.c
+++ b/driver/level3/trsm_L.c
@@ -76,6 +76,9 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
   BLASLONG ls, is, js;
   BLASLONG min_l, min_i, min_j;
   BLASLONG jjs, min_jj;
+#if !((!defined(UPPER) && !defined(TRANSA)) || (defined(UPPER) && defined(TRANSA)))
+  BLASLONG start_is;
+#endif
 
   m = args -> m;
   n = args -> n;
@@ -178,8 +181,6 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
       }
     }
 #else
-    BLASLONG start_is;
-
     for(ls = m; ls > 0; ls -= GEMM_Q){
       min_l = ls;
       if (min_l > GEMM_Q) min_l = GEMM_Q;
diff --git a/driver/level3/trsm_R.c b/driver/level3/trsm_R.c
index 169441d1e..f6a57f93f 100644
--- a/driver/level3/trsm_R.c
+++ b/driver/level3/trsm_R.c
@@ -75,6 +75,9 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
   BLASLONG ls, is, js;
   BLASLONG min_l, min_i, min_j;
   BLASLONG jjs, min_jj;
+#if !((defined(UPPER) && !defined(TRANSA)) || (!defined(UPPER) && defined(TRANSA)))
+  BLASLONG start_ls;
+#endif
 
   m = args -> m;
   n = args -> n;
@@ -226,8 +229,6 @@ int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLO
   }
 
 #else
-  BLASLONG start_ls;
-
   for(js = n; js > 0; js -= GEMM_R){
     min_j = js;
     if (min_j > GEMM_R) min_j = GEMM_R;
diff --git a/driver/others/CMakeLists.txt b/driver/others/CMakeLists.txt
new file mode 100644
index 000000000..b2af55e36
--- /dev/null
+++ b/driver/others/CMakeLists.txt
@@ -0,0 +1,75 @@
+include_directories(${CMAKE_SOURCE_DIR})
+
+if (${CORE} STREQUAL "PPC440")
+  set(MEMORY memory_qalloc.c)
+else ()
+  set(MEMORY memory.c)
+endif ()
+
+if (SMP)
+
+  if (USE_OPENMP)
+    set(BLAS_SERVER blas_server_omp.c)
+  elseif (${CMAKE_SYSTEM_NAME} STREQUAL "Windows")
+    set(BLAS_SERVER blas_server_win32.c)
+  endif ()
+
+  if (NOT DEFINED BLAS_SERVER)
+    set(BLAS_SERVER blas_server.c)
+  endif ()
+
+  set(SMP_SOURCES
+    ${BLAS_SERVER}
+    divtable.c # TODO: Makefile has -UDOUBLE
+    blas_l1_thread.c
+  )
+
+  if (NOT NO_AFFINITY)
+    list(APPEND SMP_SOURCES init.c)
+  endif ()
+endif ()
+
+set(COMMON_SOURCES
+  xerbla.c
+  openblas_set_num_threads.c
+  openblas_error_handle.c
+  openblas_get_num_procs.c
+  openblas_get_num_threads.c
+)
+
+# these need to have NAME/CNAME set, so use GenerateNamedObjects, but don't use standard name mangling
+GenerateNamedObjects("abs.c" "" "c_abs" 0 "" "" 1 )
+GenerateNamedObjects("abs.c" "DOUBLE" "z_abs" 0 "" "" 1)
+GenerateNamedObjects("openblas_get_config.c;openblas_get_parallel.c" "" "" 0 "" "" 1)
+
+if (DYNAMIC_ARCH)
+  list(APPEND COMMON_SOURCES dynamic.c)
+else ()
+  list(APPEND COMMON_SOURCES parameter.c)
+endif ()
+
+#ifdef EXPRECISION
+#COMMONOBJS	+= x_abs.$(SUFFIX) qlamch.$(SUFFIX) qlamc3.$(SUFFIX)
+#endif
+#
+#ifdef QUAD_PRECISION
+#COMMONOBJS	+= addx.$(SUFFIX) mulx.$(SUFFIX)
+#endif
+#
+#ifdef USE_CUDA
+#COMMONOBJS	+= cuda_init.$(SUFFIX)
+#endif
+#
+#ifdef FUNCTION_PROFILE
+#COMMONOBJS	+= profile.$(SUFFIX)
+#endif
+
+#LIBOTHERS = libothers.$(LIBSUFFIX)
+
+#ifeq ($(DYNAMIC_ARCH), 1)
+#HPLOBJS = memory.$(SUFFIX) xerbla.$(SUFFIX) dynamic.$(SUFFIX)
+#else
+#HPLOBJS = memory.$(SUFFIX) xerbla.$(SUFFIX) parameter.$(SUFFIX)
+#endif
+
+add_library(driver_others OBJECT ${OPENBLAS_SRC} ${MEMORY} ${SMP_SOURCES} ${COMMON_SOURCES})
diff --git a/driver/others/blas_server.c b/driver/others/blas_server.c
index b3b1ce7bd..e1c644a80 100644
--- a/driver/others/blas_server.c
+++ b/driver/others/blas_server.c
@@ -70,9 +70,11 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 /*********************************************************************/
 
 #include "common.h"
-#ifdef OS_LINUX
+#if defined(OS_LINUX) || defined(OS_NETBSD) || defined(OS_DARWIN) || defined(OS_ANDROID)
 #include <dlfcn.h>
+#include <signal.h>
 #include <sys/resource.h>
+#include <sys/time.h>
 #endif
 
 #ifndef likely
@@ -265,7 +267,7 @@ int get_node(void);
 
 static int increased_threads = 0;
 
-static int blas_thread_server(void *arg){
+static void* blas_thread_server(void *arg){
 
   /* Thread identifier */
   BLASLONG  cpu = (BLASLONG)arg;
@@ -425,6 +427,10 @@ static int blas_thread_server(void *arg){
       main_status[cpu] = MAIN_FINISH;
 #endif
 
+      // arm: make sure all results are written out _before_
+      // thread is marked as done and other threads use them
+      WMB;
+
       thread_status[cpu].queue = (blas_queue_t * volatile) ((long)thread_status[cpu].queue & 0);  /* Need a trick */
       WMB;
 
@@ -454,7 +460,7 @@ static int blas_thread_server(void *arg){
 
   //pthread_exit(NULL);
 
-  return 0;
+  return NULL;
 }
 
 #ifdef MONITOR
@@ -561,14 +567,23 @@ int blas_thread_init(void){
 
 #ifdef NEED_STACKATTR
       ret=pthread_create(&blas_threads[i], &attr,
-		     (void *)&blas_thread_server, (void *)i);
+		     &blas_thread_server, (void *)i);
 #else
       ret=pthread_create(&blas_threads[i], NULL,
-		     (void *)&blas_thread_server, (void *)i);
+		     &blas_thread_server, (void *)i);
 #endif
       if(ret!=0){
-	fprintf(STDERR,"OpenBLAS: pthread_creat error in blas_thread_init function. Error code:%d\n",ret);
-	exit(1);
+	struct rlimit rlim;
+        const char *msg = strerror(ret);
+        fprintf(STDERR, "OpenBLAS blas_thread_init: pthread_create: %s\n", msg);
+        if(0 == getrlimit(RLIMIT_NPROC, &rlim)) {
+          fprintf(STDERR, "OpenBLAS blas_thread_init: RLIMIT_NPROC "
+                  "%ld current, %ld max\n", (long)(rlim.rlim_cur), (long)(rlim.rlim_max));
+        }
+        if(0 != raise(SIGINT)) {
+          fprintf(STDERR, "OpenBLAS blas_thread_init: calling exit(3)\n");
+          exit(EXIT_FAILURE);
+        }
       }
     }
 
@@ -775,7 +790,12 @@ int exec_blas(BLASLONG num, blas_queue_t *queue){
   stop = rpcc();
 #endif
 
-  if ((num > 1) && queue -> next) exec_blas_async_wait(num - 1, queue -> next);
+  if ((num > 1) && queue -> next) {
+    exec_blas_async_wait(num - 1, queue -> next);
+
+    // arm: make sure results from other threads are visible
+    MB;
+  }
 
 #ifdef TIMING_DEBUG
   fprintf(STDERR, "Thread[0] : %16lu %16lu (%8lu cycles)\n",
@@ -823,10 +843,10 @@ void goto_set_num_threads(int num_threads) {
 
 #ifdef NEED_STACKATTR
       pthread_create(&blas_threads[i], &attr,
-		     (void *)&blas_thread_server, (void *)i);
+		     &blas_thread_server, (void *)i);
 #else
       pthread_create(&blas_threads[i], NULL,
-		     (void *)&blas_thread_server, (void *)i);
+		     &blas_thread_server, (void *)i);
 #endif
     }
 
diff --git a/driver/others/dynamic.c b/driver/others/dynamic.c
index 60b3c72af..c41164559 100644
--- a/driver/others/dynamic.c
+++ b/driver/others/dynamic.c
@@ -67,6 +67,7 @@ extern gotoblas_t  gotoblas_SANDYBRIDGE;
 extern gotoblas_t  gotoblas_BULLDOZER;
 extern gotoblas_t  gotoblas_PILEDRIVER;
 extern gotoblas_t  gotoblas_STEAMROLLER;
+extern gotoblas_t  gotoblas_EXCAVATOR;
 #ifdef NO_AVX2
 #define gotoblas_HASWELL gotoblas_SANDYBRIDGE
 #else
@@ -79,6 +80,7 @@ extern gotoblas_t  gotoblas_HASWELL;
 #define gotoblas_BULLDOZER gotoblas_BARCELONA
 #define gotoblas_PILEDRIVER gotoblas_BARCELONA
 #define gotoblas_STEAMROLLER gotoblas_BARCELONA
+#define gotoblas_EXCAVATOR gotoblas_BARCELONA
 #endif
 
 
@@ -221,6 +223,15 @@ static gotoblas_t *get_coretype(void){
 	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
 	  }
 	}
+	//Intel Broadwell
+	if (model == 13) {
+	  if(support_avx())
+	    return &gotoblas_HASWELL;
+	  else{
+	    openblas_warning(FALLBACK_VERBOSE, NEHALEM_FALLBACK);
+	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
+	  }
+	}
 	return NULL;
       case 4:
 		//Intel Haswell
@@ -232,6 +243,44 @@ static gotoblas_t *get_coretype(void){
 	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
 	  }
 	}
+	//Intel Broadwell
+	if (model == 7 || model == 15) {
+	  if(support_avx())
+	    return &gotoblas_HASWELL;
+	  else{
+	    openblas_warning(FALLBACK_VERBOSE, NEHALEM_FALLBACK);
+	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
+	  }
+	}
+	//Intel Skylake
+	if (model == 14) {
+	  if(support_avx())
+	    return &gotoblas_HASWELL;
+	  else{
+	    openblas_warning(FALLBACK_VERBOSE, NEHALEM_FALLBACK);
+	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
+	  }
+	}
+	return NULL;
+      case 5:
+	//Intel Broadwell
+	if (model == 6) {
+	  if(support_avx())
+	    return &gotoblas_HASWELL;
+	  else{
+	    openblas_warning(FALLBACK_VERBOSE, NEHALEM_FALLBACK);
+	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
+	  }
+	}
+	//Intel Skylake
+	if (model == 14 || model == 5) {
+	  if(support_avx())
+	    return &gotoblas_HASWELL;
+	  else{
+	    openblas_warning(FALLBACK_VERBOSE, NEHALEM_FALLBACK);
+	    return &gotoblas_NEHALEM; //OS doesn't support AVX. Use old kernels.
+	  }
+	}
 	return NULL;
       }
       case 0xf:
@@ -278,12 +327,22 @@ static gotoblas_t *get_coretype(void){
 	    return &gotoblas_BARCELONA; //OS doesn't support AVX. Use old kernels.
 	  }
 	}else if(model == 0){
-	  //AMD STEAMROLLER
-	  if(support_avx())
-	    return &gotoblas_STEAMROLLER;
-	  else{
-	    openblas_warning(FALLBACK_VERBOSE, BARCELONA_FALLBACK);
-	    return &gotoblas_BARCELONA; //OS doesn't support AVX. Use old kernels.
+	  if (exmodel == 3) {
+	    //AMD STEAMROLLER
+	    if(support_avx())
+	      return &gotoblas_STEAMROLLER;
+	    else{
+	      openblas_warning(FALLBACK_VERBOSE, BARCELONA_FALLBACK);
+	      return &gotoblas_BARCELONA; //OS doesn't support AVX. Use old kernels.
+	    }
+	  }else if (exmodel == 6) {
+	    if(support_avx())
+	      return &gotoblas_EXCAVATOR;
+	    else{
+	      openblas_warning(FALLBACK_VERBOSE, BARCELONA_FALLBACK);
+	      return &gotoblas_BARCELONA; //OS doesn't support AVX. Use old kernels.
+	    }
+
 	  }
 	}
 
@@ -328,6 +387,7 @@ static char *corename[] = {
     "Piledriver",
     "Haswell",
     "Steamroller",
+    "Excavator",
 };
 
 char *gotoblas_corename(void) {
@@ -353,6 +413,7 @@ char *gotoblas_corename(void) {
   if (gotoblas == &gotoblas_PILEDRIVER)   return corename[19];
   if (gotoblas == &gotoblas_HASWELL)      return corename[20];
   if (gotoblas == &gotoblas_STEAMROLLER)  return corename[21];
+  if (gotoblas == &gotoblas_EXCAVATOR)    return corename[22];
 
   return corename[0];
 }
@@ -383,7 +444,7 @@ static gotoblas_t *force_coretype(char *coretype){
 
 	switch (found)
 	{
-
+		case 22: return (&gotoblas_EXCAVATOR);
 		case 21: return (&gotoblas_STEAMROLLER);
 		case 20: return (&gotoblas_HASWELL);
 		case 19: return (&gotoblas_PILEDRIVER);
diff --git a/driver/others/memory.c b/driver/others/memory.c
index 4010ec974..ba3dc8a23 100644
--- a/driver/others/memory.c
+++ b/driver/others/memory.c
@@ -90,7 +90,9 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #ifndef OS_WINDOWS
 #include <sys/mman.h>
+#ifndef NO_SYSV_IPC
 #include <sys/shm.h>
+#endif
 #include <sys/ipc.h>
 #endif
 
@@ -137,8 +139,16 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #define BITMASK(a, b, c) ((((a) >> (b)) & (c)))
 
+#if defined(_MSC_VER) && !defined(__clang__)
+#define CONSTRUCTOR __cdecl
+#define DESTRUCTOR __cdecl
+#elif defined(OS_DARWIN) && defined(C_GCC)
 #define CONSTRUCTOR	__attribute__ ((constructor))
 #define DESTRUCTOR	__attribute__ ((destructor))
+#else
+#define CONSTRUCTOR	__attribute__ ((constructor(101)))
+#define DESTRUCTOR	__attribute__ ((destructor(101)))
+#endif
 
 #ifdef DYNAMIC_ARCH
 gotoblas_t *gotoblas = NULL;
@@ -169,6 +179,14 @@ int get_num_procs(void) {
 #endif
 #endif
 
+#ifdef OS_ANDROID
+int get_num_procs(void) {
+  static int nums = 0;
+  if (!nums) nums = sysconf(_SC_NPROCESSORS_ONLN);
+  return nums;
+}
+#endif
+
 #ifdef OS_WINDOWS
 
 int get_num_procs(void) {
@@ -266,7 +284,7 @@ void openblas_fork_handler()
   //   http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60035
   // In the mean time build with USE_OPENMP=0 or link against another
   // implementation of OpenMP.
-#if !defined(OS_WINDOWS) && defined(SMP_SERVER)
+#if !(defined(OS_WINDOWS) || defined(OS_ANDROID)) && defined(SMP_SERVER)
   int err;
   err = pthread_atfork ((void (*)(void)) BLASFUNC(blas_thread_shutdown), NULL, NULL);
   if(err != 0)
@@ -276,7 +294,7 @@ void openblas_fork_handler()
 
 int blas_get_cpu_number(void){
   env_var_t p;
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN) || defined(OS_ANDROID)
   int max_num;
 #endif
   int blas_goto_num   = 0;
@@ -284,7 +302,7 @@ int blas_get_cpu_number(void){
 
   if (blas_num_threads) return blas_num_threads;
 
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN) || defined(OS_ANDROID)
   max_num = get_num_procs();
 #endif
 
@@ -308,7 +326,7 @@ int blas_get_cpu_number(void){
   else if (blas_omp_num > 0) blas_num_threads = blas_omp_num;
   else blas_num_threads = MAX_CPU_NUMBER;
 
-#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN)
+#if defined(OS_LINUX) || defined(OS_WINDOWS) || defined(OS_FREEBSD) || defined(OS_DARWIN) || defined(OS_ANDROID)
   if (blas_num_threads > max_num) blas_num_threads = max_num;
 #endif
 
@@ -709,8 +727,6 @@ static void *alloc_shm(void *address){
   return map_address;
 }
 
-#endif
-
 #if defined OS_LINUX  || defined OS_AIX  || defined __sun__  || defined OS_WINDOWS
 
 static void alloc_hugetlb_free(struct release_t *release){
@@ -787,12 +803,12 @@ static void *alloc_hugetlb(void *address){
   
   if (LookupPrivilegeValue(NULL, SE_LOCK_MEMORY_NAME, &tp.Privileges[0].Luid) != TRUE) {
       CloseHandle(hToken);
-      return -1;
+      return (void*)-1;
   }
 
   if (AdjustTokenPrivileges(hToken, FALSE, &tp, 0, NULL, NULL) != TRUE) {
       CloseHandle(hToken);
-      return -1;
+      return (void*)-1;
   }
 
   map_address  = (void *)VirtualAlloc(address,
@@ -817,6 +833,8 @@ static void *alloc_hugetlb(void *address){
 }
 #endif
 
+#endif
+
 #ifdef  ALLOC_HUGETLBFILE
 
 static int hugetlb_pid = 0;
@@ -917,12 +935,13 @@ void *blas_memory_alloc(int procpos){
 #ifdef ALLOC_DEVICEDRIVER
     alloc_devicedirver,
 #endif
-#if defined OS_LINUX  || defined OS_AIX  || defined __sun__  || defined OS_WINDOWS
-    alloc_hugetlb,
-#endif
+/* Hugetlb implicitly assumes ALLOC_SHM */
 #ifdef ALLOC_SHM
     alloc_shm,
 #endif
+#if ((defined ALLOC_SHM) && (defined OS_LINUX  || defined OS_AIX  || defined __sun__  || defined OS_WINDOWS))
+    alloc_hugetlb,
+#endif
 #ifdef ALLOC_MMAP
     alloc_mmap,
 #endif
@@ -1062,7 +1081,7 @@ void *blas_memory_alloc(int procpos){
 	}
 #endif
 
-#if defined OS_LINUX  || defined OS_AIX  || defined __sun__  || defined OS_WINDOWS
+#if (defined ALLOC_SHM) && (defined OS_LINUX  || defined OS_AIX  || defined __sun__  || defined OS_WINDOWS)
 	if ((*func == alloc_hugetlb) && (map_address != (void *)-1)) hugetlb_allocated = 1;
 #endif
 
@@ -1142,6 +1161,9 @@ void blas_memory_free(void *free_area){
   printf("  Position : %d\n", position);
 #endif
 
+  // arm: ensure all writes are finished before other thread takes this memory
+  WMB;
+
   memory[position].used = 0;
 
 #ifdef DEBUG
@@ -1161,6 +1183,16 @@ void blas_memory_free(void *free_area){
   return;
 }
 
+void *blas_memory_alloc_nolock(int unused) {
+  void *map_address;
+  map_address = (void *)malloc(BUFFER_SIZE + FIXED_PAGESIZE);
+  return map_address;
+}
+
+void blas_memory_free_nolock(void * map_address) {
+  free(map_address);
+}
+
 void blas_shutdown(void){
 
   int pos;
@@ -1378,6 +1410,28 @@ void DESTRUCTOR gotoblas_quit(void) {
 #endif
 }
 
+#if defined(_MSC_VER) && !defined(__clang__)
+BOOL APIENTRY DllMain(HMODULE hModule, DWORD  ul_reason_for_call, LPVOID lpReserved)
+{
+  switch (ul_reason_for_call)
+  {
+    case DLL_PROCESS_ATTACH:
+      gotoblas_init();
+      break;
+    case DLL_THREAD_ATTACH:
+      break;
+    case DLL_THREAD_DETACH:
+      break;
+    case DLL_PROCESS_DETACH:
+      gotoblas_quit();
+      break;
+    default:
+      break;
+  }
+  return TRUE;
+}
+#endif
+
 #if (defined(C_PGI) || (!defined(C_SUN) && defined(F_INTERFACE_SUN))) && (defined(ARCH_X86) || defined(ARCH_X86_64))
 /* Don't call me; this is just work around for PGI / Sun bug */
 void gotoblas_dummy_for_PGI(void) {
diff --git a/exports/Makefile b/exports/Makefile
index 1fdaf2213..177e975ea 100644
--- a/exports/Makefile
+++ b/exports/Makefile
@@ -100,8 +100,8 @@ else
 	$(OBJCONV) @objconv.def ../$(LIBNAME) ../$(LIBNAME).renamed
 $(LIBDYNNAME) : ../$(LIBNAME).renamed osx.def
 endif
-ifeq ($(NOFORTRAN), 2)
-#only build cblas without Fortran
+ifeq ($(NOFORTRAN), $(filter $(NOFORTRAN),1 2))
+#only build without Fortran
 	$(CC) $(CFLAGS) -all_load -headerpad_max_install_names -install_name $(CURDIR)/../$(LIBDYNNAME) -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
 else
 	$(FC) $(FFLAGS) -all_load -headerpad_max_install_names -install_name $(CURDIR)/../$(LIBDYNNAME) -dynamiclib -o ../$(LIBDYNNAME) $< -Wl,-exported_symbols_list,osx.def  $(FEXTRALIB)
diff --git a/f_check b/f_check
index 5719faff1..e7e46886f 100644
--- a/f_check
+++ b/f_check
@@ -3,11 +3,11 @@
 #
 # 1. Not specified
 #   1.1 Automatically detect, then check compiler
-#   1.2 If no fortran compiler is detected, g77 is default with NOFORTRAN definition
+#   1.2 If no fortran compiler is detected, gfortran is default with NOFORTRAN definition
 # 2. Specified
 #   2.1 If path is correct, check compiler
 #   2.2 If path is not correct, but still valid compiler name, force setting
-#     2.2.2 Path is not correct, invalid compiler name, then g77 is default with NOFORTRAN definition
+#     2.2.2 Path is not correct, invalid compiler name, then gfortran is default with NOFORTRAN definition
 #
 
 $makefile = shift(@ARGV);
@@ -25,7 +25,7 @@ $compiler = "" if $compiler eq "f77";
 
 if ($compiler eq "") {
 
-    @lists = ("g77", "g95", "gfortran", "frt", "fort", "openf90", "openf95",
+    @lists = ("gfortran", "g95", "frt", "fort", "openf90", "openf95",
 	      "sunf77", "sunf90", "sunf95",
               "xlf95", "xlf90", "xlf",
               "ppuf77", "ppuf95", "ppuf90", "ppuxlf",
@@ -38,6 +38,7 @@ OUTER:
         foreach $path (@path) {
             if (-x $path . "/" . $lists) {
                 $compiler = $lists;
+                $compiler_bin = $lists;
                 last OUTER;
             }
         }
@@ -48,8 +49,8 @@ OUTER:
 if ($compiler eq "") {
 
     $nofortran = 1;
-    $compiler = "g77";
-    $vendor = G77;
+    $compiler = "gfortran";
+    $vendor = GFORTRAN;
     $bu       = "_";
 
 } else {
@@ -196,8 +197,8 @@ if ($compiler eq "") {
 
 	if ($vendor eq "") {
 	    $nofortran = 1;
-	    $compiler = "g77";
-	    $vendor = G77;
+	    $compiler = "gfortran";
+	    $vendor = GFORTRAN;
 	    $bu       = "_";
 	    $openmp = "";
 	}
diff --git a/getarch.c b/getarch.c
index ee5f55fd1..0a49fd1b3 100644
--- a/getarch.c
+++ b/getarch.c
@@ -69,10 +69,14 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 /* or implied, of The University of Texas at Austin.                 */
 /*********************************************************************/
 
-#if defined(__WIN32__) || defined(__WIN64__) || defined(__CYGWIN32__) || defined(__CYGWIN64__)
+#if defined(__WIN32__) || defined(__WIN64__) || defined(__CYGWIN32__) || defined(__CYGWIN64__) || defined(_WIN32) || defined(_WIN64)
 #define OS_WINDOWS
 #endif
 
+#if defined(__i386__) || defined(__x86_64__) || defined(_M_IX86) || defined(_M_X64)
+#define INTEL_AMD
+#endif
+
 #include <stdio.h>
 #include <string.h>
 #ifdef OS_WINDOWS
@@ -116,6 +120,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 /* #define FORCE_POWER4		*/
 /* #define FORCE_POWER5		*/
 /* #define FORCE_POWER6		*/
+/* #define FORCE_POWER7		*/
+/* #define FORCE_POWER8		*/
 /* #define FORCE_PPCG4		*/
 /* #define FORCE_PPC970		*/
 /* #define FORCE_PPC970MP	*/
@@ -448,6 +454,22 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "STEAMROLLER"
 #endif
 
+#if defined (FORCE_EXCAVATOR)
+#define FORCE
+#define FORCE_INTEL
+#define ARCHITECTURE    "X86"
+#define SUBARCHITECTURE "EXCAVATOR"
+#define ARCHCONFIG   "-DEXCAVATOR " \
+		     "-DL1_DATA_SIZE=16384 -DL1_DATA_LINESIZE=64 " \
+		     "-DL2_SIZE=2097152 -DL2_LINESIZE=64  -DL3_SIZE=12582912 " \
+		     "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 " \
+		     "-DHAVE_MMX -DHAVE_SSE -DHAVE_SSE2 -DHAVE_SSE3 -DHAVE_SSE4_1 -DHAVE_SSE4_2 " \
+		     "-DHAVE_SSE4A -DHAVE_MISALIGNSSE -DHAVE_128BITFPU -DHAVE_FASTMOVU -DHAVE_CFLUSH " \
+                     "-DHAVE_AVX -DHAVE_FMA4 -DHAVE_FMA3"
+#define LIBNAME   "excavator"
+#define CORENAME  "EXCAVATOR"
+#endif
+
 
 #ifdef FORCE_SSE_GENERIC
 #define FORCE
@@ -530,7 +552,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define CORENAME  "POWER5"
 #endif
 
-#ifdef FORCE_POWER6
+#if defined(FORCE_POWER6) || defined(FORCE_POWER7) || defined(FORCE_POWER8)
 #define FORCE
 #define ARCHITECTURE    "POWER"
 #define SUBARCHITECTURE "POWER6"
@@ -732,7 +754,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define ARCHITECTURE    "ARM"
 #define SUBARCHITECTURE "CORTEXA9"
 #define SUBDIRNAME      "arm"
-#define ARCHCONFIG   "-DCORTEXA9 " \
+#define ARCHCONFIG   "-DCORTEXA9 -DARMV7 " \
        "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
        "-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
        "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 " \
@@ -747,7 +769,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define ARCHITECTURE    "ARM"
 #define SUBARCHITECTURE "CORTEXA15"
 #define SUBDIRNAME      "arm"
-#define ARCHCONFIG   "-DCORTEXA15 " \
+#define ARCHCONFIG   "-DCORTEXA15 -DARMV7 " \
        "-DL1_DATA_SIZE=32768 -DL1_DATA_LINESIZE=32 " \
        "-DL2_SIZE=1048576 -DL2_LINESIZE=32 " \
        "-DDTB_DEFAULT_ENTRIES=128 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 " \
@@ -780,8 +802,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define ARCHCONFIG   "-DARMV5 " \
        "-DL1_DATA_SIZE=65536 -DL1_DATA_LINESIZE=32 " \
        "-DL2_SIZE=512488 -DL2_LINESIZE=32 " \
-       "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 " \
-       "-DHAVE_VFP"
+       "-DDTB_DEFAULT_ENTRIES=64 -DDTB_SIZE=4096 -DL2_ASSOCIATIVE=4 "
 #define LIBNAME   "armv5"
 #define CORENAME  "ARMV5"
 #else
@@ -813,7 +834,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define OPENBLAS_SUPPORTED
 #endif
 
-#if defined(__i386__) || (__x86_64__)
+#ifdef INTEL_AMD
 #include "cpuid_x86.c"
 #define OPENBLAS_SUPPORTED
 #endif
@@ -908,7 +929,7 @@ int main(int argc, char *argv[]){
 #ifdef FORCE
     printf("CORE=%s\n", CORENAME);
 #else
-#if defined(__i386__) || defined(__x86_64__) || defined(POWER) || defined(__mips__) || defined(__arm__) || defined(__aarch64__)
+#if defined(INTEL_AMD) || defined(POWER) || defined(__mips__) || defined(__arm__) || defined(__aarch64__)
     printf("CORE=%s\n", get_corename());
 #endif
 #endif
@@ -928,7 +949,7 @@ int main(int argc, char *argv[]){
 #endif
 
 
-#if defined(__i386__) || defined(__x86_64__)
+#ifdef INTEL_AMD
 #ifndef FORCE
     get_sse();
 #else
@@ -1008,7 +1029,7 @@ int main(int argc, char *argv[]){
 #ifdef FORCE
     printf("#define CHAR_CORENAME \"%s\"\n", CORENAME);
 #else
-#if defined(__i386__) || defined(__x86_64__) || defined(POWER) || defined(__mips__) || defined(__arm__) || defined(__aarch64__)
+#if defined(INTEL_AMD) || defined(POWER) || defined(__mips__) || defined(__arm__) || defined(__aarch64__)
     printf("#define CHAR_CORENAME \"%s\"\n", get_corename());
 #endif
 #endif
diff --git a/interface/CMakeLists.txt b/interface/CMakeLists.txt
new file mode 100644
index 000000000..9ff924e5f
--- /dev/null
+++ b/interface/CMakeLists.txt
@@ -0,0 +1,166 @@
+
+include_directories(${CMAKE_SOURCE_DIR})
+
+
+set(BLAS1_SOURCES
+  copy.c
+  nrm2.c
+)
+
+set(BLAS1_REAL_ONLY_SOURCES
+  rotm.c rotmg.c # N.B. these do not have complex counterparts
+  rot.c
+  asum.c
+)
+
+# these will have 'z' prepended for the complex version
+set(BLAS1_MANGLED_SOURCES
+  axpy.c swap.c
+  scal.c
+  dot.c
+  rotg.c
+  axpby.c
+)
+
+# TODO: USE_NETLIB_GEMV shoudl switch gemv.c to netlib/*gemv.f
+# these all have 'z' sources for complex versions
+set(BLAS2_SOURCES
+  gemv.c ger.c
+  trsv.c trmv.c symv.c
+  syr.c syr2.c gbmv.c
+  sbmv.c spmv.c
+  spr.c spr2.c
+  tbsv.c tbmv.c
+  tpsv.c tpmv.c
+)
+
+set(BLAS2_COMPLEX_ONLY_MANGLED_SOURCES
+  hemv.c hbmv.c
+  her.c her2.c
+  hpmv.c hpr.c
+  hpr2.c
+)
+
+# these do not have separate 'z' sources
+set(BLAS3_SOURCES
+  gemm.c symm.c
+  trsm.c syrk.c syr2k.c
+)
+
+set(BLAS3_MANGLED_SOURCES
+  omatcopy.c imatcopy.c
+  geadd.c
+)
+
+# generate the BLAS objs once with and once without cblas
+set (CBLAS_FLAGS "")
+
+if (NOT DEFINED NO_FBLAS)
+  list(APPEND CBLAS_FLAGS 0)
+endif ()
+
+if (NOT DEFINED NO_CBLAS)
+  list(APPEND CBLAS_FLAGS 1)
+endif ()
+
+foreach (CBLAS_FLAG ${CBLAS_FLAGS})
+
+  # TODO: don't compile complex sources with cblas for now, the naming schemes are all different and they will have to be handled separately from SINGLE/DOUBLE
+  set(DISABLE_COMPLEX 0)
+  set(MANGLE_COMPLEX 3)
+  if (CBLAS_FLAG EQUAL 1)
+#    set(DISABLE_COMPLEX 1)
+#    set(MANGLE_COMPLEX 1)
+  endif ()
+  GenerateNamedObjects("${BLAS1_SOURCES}" "" "" ${CBLAS_FLAG} "" "" false ${DISABLE_COMPLEX})
+  GenerateNamedObjects("${BLAS1_REAL_ONLY_SOURCES}" "" "" ${CBLAS_FLAG} "" "" false 1)
+  GenerateNamedObjects("${BLAS1_MANGLED_SOURCES}" "" "" ${CBLAS_FLAG} "" "" false ${MANGLE_COMPLEX})
+  GenerateNamedObjects("${BLAS2_SOURCES}" "" "" ${CBLAS_FLAG} "" "" false ${MANGLE_COMPLEX})
+  GenerateNamedObjects("${BLAS2_COMPLEX_ONLY_MANGLED_SOURCES}" "" "" ${CBLAS_FLAG} "" "" false 4)
+  GenerateNamedObjects("${BLAS3_SOURCES}" "" "" ${CBLAS_FLAG} "" "" false ${DISABLE_COMPLEX})
+  GenerateNamedObjects("${BLAS3_MANGLED_SOURCES}" "" "" ${CBLAS_FLAG} "" "" false ${MANGLE_COMPLEX})
+
+  #sdsdot, dsdot
+  GenerateNamedObjects("sdsdot.c" "" "sdsdot" ${CBLAS_FLAG} "" "" true "SINGLE")
+  GenerateNamedObjects("dsdot.c" "" "dsdot" ${CBLAS_FLAG} "" "" true "SINGLE")
+
+  # trmm is trsm with a compiler flag set
+  GenerateNamedObjects("trsm.c" "TRMM" "trmm" ${CBLAS_FLAG})
+
+  # max and imax are compiled 4 times
+  GenerateNamedObjects("max.c" "" "" ${CBLAS_FLAG})
+  GenerateNamedObjects("max.c" "USE_ABS" "amax" ${CBLAS_FLAG})
+  GenerateNamedObjects("max.c" "USE_ABS;USE_MIN" "amin" ${CBLAS_FLAG})
+  GenerateNamedObjects("max.c" "USE_MIN" "min" ${CBLAS_FLAG})
+
+  GenerateNamedObjects("imax.c" "" "i*max" ${CBLAS_FLAG})
+  GenerateNamedObjects("imax.c" "USE_ABS" "i*amax" ${CBLAS_FLAG})
+  GenerateNamedObjects("imax.c" "USE_ABS;USE_MIN" "i*amin" ${CBLAS_FLAG})
+  GenerateNamedObjects("imax.c" "USE_MIN" "i*min" ${CBLAS_FLAG})
+
+
+# complex-specific sources
+foreach (float_type ${FLOAT_TYPES})
+
+  if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+    GenerateNamedObjects("zger.c" "" "geru" ${CBLAS_FLAG} "" "" false ${float_type})
+    GenerateNamedObjects("zger.c" "CONJ" "gerc" ${CBLAS_FLAG} "" "" false ${float_type})
+    GenerateNamedObjects("zdot.c" "CONJ" "dotc" ${CBLAS_FLAG} "" "" false ${float_type})
+    GenerateNamedObjects("zdot.c" "" "dotu" ${CBLAS_FLAG} "" "" false ${float_type})
+
+    GenerateNamedObjects("symm.c" "HEMM" "hemm" ${CBLAS_FLAG} "" "" false ${float_type})
+    GenerateNamedObjects("syrk.c" "HEMM" "herk" ${CBLAS_FLAG} "" "" false ${float_type})
+    GenerateNamedObjects("syr2k.c" "HEMM" "her2k" ${CBLAS_FLAG} "" "" false ${float_type})
+
+    if (USE_GEMM3M)
+      GenerateNamedObjects("gemm.c" "GEMM3M" "gemm3m" false "" "" false ${float_type})
+    endif()
+  endif ()
+  if (${float_type} STREQUAL "COMPLEX")
+    GenerateNamedObjects("zscal.c" "SSCAL" "sscal" ${CBLAS_FLAG} "" "" false "COMPLEX")
+    GenerateNamedObjects("nrm2.c" "" "scnrm2" ${CBLAS_FLAG} "" "" true "COMPLEX")
+    GenerateNamedObjects("zrot.c" "" "csrot" ${CBLAS_FLAG} "" "" true "COMPLEX")
+    GenerateNamedObjects("max.c" "USE_ABS;USE_MIN" "scamin" ${CBLAS_FLAG} "" "" true "COMPLEX")
+    GenerateNamedObjects("max.c" "USE_ABS" "scamax" ${CBLAS_FLAG} "" "" true "COMPLEX")
+    GenerateNamedObjects("asum.c" "" "scasum" ${CBLAS_FLAG} "" "" true "COMPLEX")
+  endif ()
+  if (${float_type} STREQUAL "ZCOMPLEX")
+    GenerateNamedObjects("zscal.c" "SSCAL" "dscal" ${CBLAS_FLAG} "" "" false "ZCOMPLEX")
+    GenerateNamedObjects("nrm2.c" "" "dznrm2" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
+    GenerateNamedObjects("zrot.c" "" "zdrot" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
+    GenerateNamedObjects("max.c" "USE_ABS;USE_MIN" "dzamin" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
+    GenerateNamedObjects("max.c" "USE_ABS" "dzamax" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
+    GenerateNamedObjects("asum.c" "" "dzasum" ${CBLAS_FLAG} "" "" true "ZCOMPLEX")
+  endif ()
+endforeach ()
+
+endforeach ()
+
+#Special functions for CBLAS
+if (NOT DEFINED NO_CBLAS)
+  foreach (float_type ${FLOAT_TYPES})
+  if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+    #cblas_dotc_sub cblas_dotu_sub
+    GenerateNamedObjects("zdot.c" "FORCE_USE_STACK" "dotu_sub" 1 "" "" false ${float_type})
+    GenerateNamedObjects("zdot.c" "FORCE_USE_STACK;CONJ" "dotc_sub" 1 "" "" false ${float_type})
+  endif()
+  endforeach ()
+endif()
+
+if (NOT DEFINED NO_LAPACK)
+  set(LAPACK_SOURCES
+    lapack/gesv.c
+  )
+
+  # prepend z for complex versions
+  set(LAPACK_MANGLED_SOURCES
+    lapack/getrf.c lapack/getrs.c lapack/potrf.c lapack/getf2.c
+    lapack/potf2.c lapack/laswp.c lapack/lauu2.c
+    lapack/lauum.c lapack/trti2.c lapack/trtri.c
+  )
+
+  GenerateNamedObjects("${LAPACK_SOURCES}")
+  GenerateNamedObjects("${LAPACK_MANGLED_SOURCES}" "" "" 0 "" "" 0 3)
+endif ()
+
+add_library(interface OBJECT ${OPENBLAS_SRC})
diff --git a/interface/gemm.c b/interface/gemm.c
index a5a2b4724..7253b0500 100644
--- a/interface/gemm.c
+++ b/interface/gemm.c
@@ -121,6 +121,9 @@ void NAME(char *TRANSA, char *TRANSB,
   FLOAT *sa, *sb;
 
 #ifdef SMP
+  int nthreads_max;
+  int nthreads_avail;
+  double MNK;
 #ifndef COMPLEX
 #ifdef XDOUBLE
   int mode  =  BLAS_XDOUBLE | BLAS_REAL;
@@ -237,6 +240,9 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANS
   XFLOAT *sa, *sb;
 
 #ifdef SMP
+  int nthreads_max;
+  int nthreads_avail;
+  double MNK;
 #ifndef COMPLEX
 #ifdef XDOUBLE
   int mode  =  BLAS_XDOUBLE | BLAS_REAL;
@@ -400,15 +406,15 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANS
   mode |= (transa << BLAS_TRANSA_SHIFT);
   mode |= (transb << BLAS_TRANSB_SHIFT);
 
-  int nthreads_max = num_cpu_avail(3);
-  int nthreads_avail = nthreads_max;
+  nthreads_max = num_cpu_avail(3);
+  nthreads_avail = nthreads_max;
 
 #ifndef COMPLEX
-  double MNK = (double) args.m * (double) args.n * (double) args.k;
+  MNK = (double) args.m * (double) args.n * (double) args.k;
   if ( MNK <= (65536.0  * (double) GEMM_MULTITHREAD_THRESHOLD)  )
 	nthreads_max = 1;
 #else
-  double MNK = (double) args.m * (double) args.n * (double) args.k;
+  MNK = (double) args.m * (double) args.n * (double) args.k;
   if ( MNK <= (8192.0  * (double) GEMM_MULTITHREAD_THRESHOLD)  )
 	nthreads_max = 1;
 #endif
diff --git a/interface/gemv.c b/interface/gemv.c
index f33973ef3..0a222a645 100644
--- a/interface/gemv.c
+++ b/interface/gemv.c
@@ -38,6 +38,7 @@
 
 #include <stdio.h>
 #include "common.h"
+#include "l1param.h"
 #ifdef FUNCTION_PROFILE
 #include "functable.h"
 #endif
@@ -80,6 +81,9 @@ void NAME(char *TRANS, blasint *M, blasint *N,
   FLOAT *buffer;
 #ifdef SMP
   int nthreads;
+  int nthreads_max;
+  int nthreads_avail;
+  double MNK;
 #endif
 
   int (*gemv[])(BLASLONG, BLASLONG, BLASLONG, FLOAT, FLOAT *, BLASLONG,  FLOAT * , BLASLONG, FLOAT *, BLASLONG, FLOAT *) = {
@@ -134,6 +138,9 @@ void CNAME(enum CBLAS_ORDER order,
   blasint info, t;
 #ifdef SMP
   int nthreads;
+  int nthreads_max;
+  int nthreads_avail;
+  double MNK;
 #endif
 
   int (*gemv[])(BLASLONG, BLASLONG, BLASLONG, FLOAT, FLOAT *, BLASLONG,  FLOAT * , BLASLONG, FLOAT *, BLASLONG, FLOAT *) = {
@@ -189,7 +196,7 @@ void CNAME(enum CBLAS_ORDER order,
   }
 
 #endif
-
+  //printf("m=%d, n=%d, trans=%d, incx=%d, incy=%d, alpha=%f, beta=%f\n", m, n, trans, incx, incy, alpha, beta);
   if ((m==0) || (n==0)) return;
 
   lenx = n;
@@ -211,24 +218,33 @@ void CNAME(enum CBLAS_ORDER order,
 #ifdef MAX_STACK_ALLOC
   // make it volatile because some gemv implementation (ex: dgemv_n.S)
   // do not restore all register
-  volatile int stack_alloc_size = m + n;
+  volatile int stack_alloc_size = 0;
+  //for gemv_n and gemv_t, try to allocate on stack
+  stack_alloc_size = m + n;
+#ifdef ALIGNED_ACCESS
+  stack_alloc_size += 3;
+#endif
   if(stack_alloc_size < 128)
-      //dgemv_n.S require a 128 bytes buffer
-      stack_alloc_size = 128;
+    //dgemv_n.S require a 128 bytes buffer
+    stack_alloc_size = 128;
+
   if(stack_alloc_size > MAX_STACK_ALLOC / sizeof(FLOAT))
-      stack_alloc_size = 0;
+    stack_alloc_size = 0;
+
   FLOAT stack_buffer[stack_alloc_size];
   buffer = stack_alloc_size ? stack_buffer : (FLOAT *)blas_memory_alloc(1);
+  //  printf("stack_alloc_size=%d\n", stack_alloc_size);
 #else
+  //Original OpenBLAS/GotoBLAS codes.
   buffer = (FLOAT *)blas_memory_alloc(1);
 #endif
 
 #ifdef SMP
 
-  int  nthreads_max = num_cpu_avail(2);
-  int  nthreads_avail = nthreads_max;
+  nthreads_max = num_cpu_avail(2);
+  nthreads_avail = nthreads_max;
 
-  double MNK = (double) m * (double) n;
+  MNK = (double) m * (double) n;
   if ( MNK <= (24.0 * 24.0  * (double) (GEMM_MULTITHREAD_THRESHOLD*GEMM_MULTITHREAD_THRESHOLD) )  )
         nthreads_max = 1;
 
@@ -251,10 +267,13 @@ void CNAME(enum CBLAS_ORDER order,
 #endif
 
 #ifdef MAX_STACK_ALLOC
-  if(!stack_alloc_size)
-#endif
+  if(!stack_alloc_size){
     blas_memory_free(buffer);
-
+  }
+#else
+    blas_memory_free(buffer);
+#endif
+  
   FUNCTION_PROFILE_END(1, m * n + m + n,  2 * m * n);
 
   IDEBUG_END;
diff --git a/interface/imatcopy.c b/interface/imatcopy.c
index 89f0ec823..f4309a85c 100644
--- a/interface/imatcopy.c
+++ b/interface/imatcopy.c
@@ -26,7 +26,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *****************************************************************************/
 
 /***********************************************************
- * 2014/06/10 Saar
+ * 2014-06-10 Saar
+ * 2015-09-07 grisuthedragon 
 ***********************************************************/
 
 #include <stdio.h>
@@ -50,6 +51,9 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #undef malloc
 #undef free
 
+/* Enables the New IMATCOPY code with inplace operation if lda == ldb   */
+#define NEW_IMATCOPY
+
 #ifndef CBLAS
 void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha, FLOAT *a, blasint *lda, blasint *ldb)
 {
@@ -75,7 +79,6 @@ void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha,
 #else 
 void CNAME( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, FLOAT calpha, FLOAT *a, blasint clda, blasint cldb)
 {
-	char Order, Trans;
 	int order=-1,trans=-1;
 	blasint info = -1;
 	FLOAT *b;
@@ -117,6 +120,34 @@ void CNAME( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows,
     		BLASFUNC(xerbla)(ERROR_NAME, &info, sizeof(ERROR_NAME));
     		return;
   	}
+#ifdef NEW_IMATCOPY
+    if ( *lda == *ldb ) {
+        if ( order == BlasColMajor )
+        {
+            if ( trans == BlasNoTrans )
+            {
+                IMATCOPY_K_CN(*rows, *cols, *alpha, a, *lda );
+            }
+            else
+            {
+                IMATCOPY_K_CT(*rows, *cols, *alpha, a, *lda );
+            }
+        }
+        else
+        {
+            if ( trans == BlasNoTrans )
+            {
+                IMATCOPY_K_RN(*rows, *cols, *alpha, a, *lda );
+            }
+            else
+            {
+                IMATCOPY_K_RT(*rows, *cols, *alpha, a, *lda );
+            }
+        }
+        return; 
+    }
+
+#endif
 
 	if ( *lda >  *ldb )
 		msize = (*lda) * (*ldb)  * sizeof(FLOAT);
diff --git a/interface/imax.c b/interface/imax.c
index 55ffa7c6e..4378f1e22 100644
--- a/interface/imax.c
+++ b/interface/imax.c
@@ -136,6 +136,8 @@ blasint NAME(blasint *N, FLOAT *x, blasint *INCX){
 
   ret = (blasint)MAX_K(n, x, incx);
 
+  if(ret > n) ret=n;
+
   FUNCTION_PROFILE_END(COMPSIZE, n, 0);
 
   IDEBUG_END;
@@ -159,6 +161,8 @@ CBLAS_INDEX CNAME(blasint n, FLOAT *x, blasint incx){
 
   ret = MAX_K(n, x, incx);
 
+  if (ret > n) ret=n;
+
   if (ret) ret --;
 
   FUNCTION_PROFILE_END(COMPSIZE, n, 0);
diff --git a/interface/rotg.c b/interface/rotg.c
index 49088ab02..a0e6efdab 100644
--- a/interface/rotg.c
+++ b/interface/rotg.c
@@ -14,8 +14,7 @@ void CNAME(FLOAT *DA, FLOAT *DB, FLOAT *C, FLOAT *S){
 
 #endif
 
-
-#if defined(__i386__) || defined(__x86_64__) || defined(__ia64__)
+#if defined(__i386__) || defined(__x86_64__) || defined(__ia64__) || defined(_M_X64) || defined(_M_IX86)
 
   long double da = *DA;
   long double db = *DB;
diff --git a/interface/trsm.c b/interface/trsm.c
index 266372988..3d4aed282 100644
--- a/interface/trsm.c
+++ b/interface/trsm.c
@@ -362,6 +362,12 @@ void CNAME(enum CBLAS_ORDER order,
   mode |= (side  << BLAS_RSIDE_SHIFT);
 
   args.nthreads = num_cpu_avail(3);
+  if ( args.m < 2*GEMM_MULTITHREAD_THRESHOLD )
+	args.nthreads = 1;
+  else
+	if ( args.n < 2*GEMM_MULTITHREAD_THRESHOLD )
+		args.nthreads = 1;
+		
 
   if (args.nthreads == 1) {
 #endif
diff --git a/interface/zaxpby.c b/interface/zaxpby.c
index 9e8324432..1abb24de9 100644
--- a/interface/zaxpby.c
+++ b/interface/zaxpby.c
@@ -53,13 +53,13 @@ void CNAME(blasint n, FLOAT *ALPHA, FLOAT *x, blasint incx, FLOAT *BETA, FLOAT *
 
 #endif
 
-  if (n <= 0) return;
-
   FLOAT alpha_r = *(ALPHA + 0);
   FLOAT alpha_i = *(ALPHA + 1);
   FLOAT beta_r  = *(BETA  + 0);
   FLOAT beta_i  = *(BETA  + 1);
 
+  if (n <= 0) return;
+
   FUNCTION_PROFILE_START();
 
   if (incx < 0) x -= (n - 1) * incx * 2;
diff --git a/interface/zdot.c b/interface/zdot.c
index 1380ce292..d4d0fab92 100644
--- a/interface/zdot.c
+++ b/interface/zdot.c
@@ -57,21 +57,25 @@
 #ifdef RETURN_BY_STRUCT
 MYTYPE         NAME(                        blasint *N, FLOAT *x, blasint *INCX, FLOAT *y, blasint *INCY) {
 #elif defined RETURN_BY_STACK
-void           NAME(FLOAT _Complex *result, blasint *N, FLOAT *x, blasint *INCX, FLOAT *y, blasint *INCY) {
+void           NAME(OPENBLAS_COMPLEX_FLOAT *result, blasint *N, FLOAT *x, blasint *INCX, FLOAT *y, blasint *INCY) {
 #else
-FLOAT _Complex NAME(                        blasint *N, FLOAT *x, blasint *INCX, FLOAT *y, blasint *INCY) {
+OPENBLAS_COMPLEX_FLOAT NAME(                        blasint *N, FLOAT *x, blasint *INCX, FLOAT *y, blasint *INCY) {
 #endif
 
   BLASLONG n    = *N;
   BLASLONG incx = *INCX;
   BLASLONG incy = *INCY;
 #ifndef RETURN_BY_STACK
-  FLOAT _Complex ret;
+  OPENBLAS_COMPLEX_FLOAT ret;
 #endif
 #ifdef RETURN_BY_STRUCT
   MYTYPE  myret;
 #endif
 
+#ifndef RETURN_BY_STRUCT
+  OPENBLAS_COMPLEX_FLOAT zero=OPENBLAS_MAKE_COMPLEX_FLOAT(0.0, 0.0);
+#endif
+
   PRINT_DEBUG_NAME;
 
   if (n <= 0) {
@@ -80,10 +84,10 @@ FLOAT _Complex NAME(                        blasint *N, FLOAT *x, blasint *INCX,
     myret.i = 0.;
     return myret;
 #elif defined RETURN_BY_STACK
-    *result = ZERO;
+    *result = zero;
     return;
 #else
-    return ZERO;
+    return zero;
 #endif
   }
 
@@ -144,21 +148,24 @@ FLOAT _Complex NAME(                        blasint *N, FLOAT *x, blasint *INCX,
 #else
 
 #ifdef FORCE_USE_STACK
-void           CNAME(blasint n, FLOAT *x, blasint incx, FLOAT *y, blasint incy, FLOAT _Complex *result){
+void           CNAME(blasint n, FLOAT *x, blasint incx, FLOAT *y, blasint incy, OPENBLAS_COMPLEX_FLOAT *result){
 #else
-FLOAT _Complex CNAME(blasint n, FLOAT *x, blasint incx, FLOAT *y, blasint incy){
+OPENBLAS_COMPLEX_FLOAT CNAME(blasint n, FLOAT *x, blasint incx, FLOAT *y, blasint incy){
 
-  FLOAT _Complex ret;
+  OPENBLAS_COMPLEX_FLOAT ret;
+  OPENBLAS_COMPLEX_FLOAT zero=OPENBLAS_MAKE_COMPLEX_FLOAT(0.0, 0.0);
 #endif
 
   PRINT_DEBUG_CNAME;
 
   if (n <= 0) {
 #ifdef FORCE_USE_STACK
-    *result = ZERO;
+    //*result = OPENBLAS_MAKE_COMPLEX_FLOAT(0.0, 0.0);
+	CREAL(*result) = 0.0;
+	CIMAG(*result) = 0.0;
     return;
 #else
-    return ZERO;
+    return zero;
 #endif
   }
 
diff --git a/interface/zgemv.c b/interface/zgemv.c
index 704034aaf..520136b45 100644
--- a/interface/zgemv.c
+++ b/interface/zgemv.c
@@ -79,6 +79,9 @@ void NAME(char *TRANS, blasint *M, blasint *N,
   FLOAT *buffer;
 #ifdef SMP
   int nthreads;
+  int nthreads_max;
+  int nthreads_avail;
+  double MNK;
 #endif
 
   int (*gemv[])(BLASLONG, BLASLONG, BLASLONG, FLOAT, FLOAT, FLOAT *, BLASLONG,
@@ -91,14 +94,14 @@ void NAME(char *TRANS, blasint *M, blasint *N,
   blasint    lenx, leny;
   blasint    i;
 
-  PRINT_DEBUG_NAME;
-
   FLOAT alpha_r = *(ALPHA + 0);
   FLOAT alpha_i = *(ALPHA + 1);
 
   FLOAT beta_r  = *(BETA + 0);
   FLOAT beta_i  = *(BETA + 1);
 
+  PRINT_DEBUG_NAME;
+
   TOUPPER(trans);
 
   info = 0;
@@ -145,6 +148,9 @@ void CNAME(enum CBLAS_ORDER order,
   blasint info, t;
 #ifdef SMP
   int nthreads;
+  int nthreads_max;
+  int nthreads_avail;
+  double MNK;
 #endif
 
   int (*gemv[])(BLASLONG, BLASLONG, BLASLONG, FLOAT, FLOAT, FLOAT *, BLASLONG,
@@ -153,14 +159,14 @@ void CNAME(enum CBLAS_ORDER order,
 	      GEMV_O, GEMV_U, GEMV_S, GEMV_D,
 	    };
 
-  PRINT_DEBUG_CNAME;
-
   FLOAT alpha_r = *(ALPHA + 0);
   FLOAT alpha_i = *(ALPHA + 1);
 
   FLOAT beta_r  = *(BETA + 0);
   FLOAT beta_i  = *(BETA + 1);
 
+  PRINT_DEBUG_CNAME;
+
   trans = -1;
   info  =  0;
 
@@ -234,10 +240,10 @@ void CNAME(enum CBLAS_ORDER order,
 
 #ifdef SMP
 
-  int  nthreads_max = num_cpu_avail(2);
-  int  nthreads_avail = nthreads_max;
+  nthreads_max = num_cpu_avail(2);
+  nthreads_avail = nthreads_max;
 
-  double MNK = (double) m * (double) n;
+  MNK = (double) m * (double) n;
   if ( MNK <= ( 256.0  * (double) (GEMM_MULTITHREAD_THRESHOLD * GEMM_MULTITHREAD_THRESHOLD)  ))
         nthreads_max = 1;
 
diff --git a/interface/zimatcopy.c b/interface/zimatcopy.c
index 3f273cf13..b1e1d15dc 100644
--- a/interface/zimatcopy.c
+++ b/interface/zimatcopy.c
@@ -26,7 +26,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *****************************************************************************/
 
 /***********************************************************
- * 2014/06/10 Saar
+ * 2014-06-10 Saar
+ * 2015-09-07 grisuthedragon 
 ***********************************************************/
 
 #include <stdio.h>
@@ -49,6 +50,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define BlasTransConj    2
 #define BlasConj         3
 
+#define NEW_IMATCOPY 
 
 #ifndef CBLAS
 void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha, FLOAT *a, blasint *lda, blasint *ldb)
@@ -124,6 +126,52 @@ void CNAME( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows,
     		return;
   	}
 
+#ifdef NEW_IMATCOPY
+    if (*lda == *ldb) {
+        if ( order == BlasColMajor )
+        {
+
+            if ( trans == BlasNoTrans )
+            {
+                IMATCOPY_K_CN(*rows, *cols, alpha[0], alpha[1], a, *lda );
+            }
+            if ( trans == BlasConj )
+            {
+                IMATCOPY_K_CNC(*rows, *cols, alpha[0], alpha[1], a, *lda );
+            }
+            if ( trans == BlasTrans )
+            {
+                IMATCOPY_K_CT(*rows, *cols, alpha[0], alpha[1], a, *lda );
+            }
+            if ( trans == BlasTransConj )
+            {
+                IMATCOPY_K_CTC(*rows, *cols, alpha[0], alpha[1], a, *lda );
+            }
+        }
+        else
+        {
+
+            if ( trans == BlasNoTrans )
+            {
+                IMATCOPY_K_RN(*rows, *cols, alpha[0], alpha[1], a, *lda );
+            }
+            if ( trans == BlasConj )
+            {
+                IMATCOPY_K_RNC(*rows, *cols, alpha[0], alpha[1], a, *lda );
+            }
+            if ( trans == BlasTrans )
+            {
+                IMATCOPY_K_RT(*rows, *cols, alpha[0], alpha[1], a, *lda );
+            }
+            if ( trans == BlasTransConj )
+            {
+                IMATCOPY_K_RTC(*rows, *cols, alpha[0], alpha[1], a, *lda );
+            }
+        }
+        return; 
+    }
+#endif
+
 	if ( *lda >  *ldb )
                 msize = (*lda) * (*ldb)  * sizeof(FLOAT) * 2;
         else
diff --git a/interface/zrotg.c b/interface/zrotg.c
index e9e8a11df..187343d41 100644
--- a/interface/zrotg.c
+++ b/interface/zrotg.c
@@ -6,13 +6,7 @@
 
 void NAME(FLOAT *DA, FLOAT *DB, FLOAT *C, FLOAT *S){
 
-  PRINT_DEBUG_NAME;
-
-  IDEBUG_START;
-
-  FUNCTION_PROFILE_START();
-
-#if defined(__i386__) || defined(__x86_64__) || defined(__ia64__)
+#if defined(__i386__) || defined(__x86_64__) || defined(__ia64__) || defined(_M_X64) || defined(_M_IX86)
 
   long double da_r = *(DA + 0);
   long double da_i = *(DA + 1);
@@ -22,6 +16,12 @@ void NAME(FLOAT *DA, FLOAT *DB, FLOAT *C, FLOAT *S){
 
   long double ada = fabs(da_r) + fabs(da_i);
 
+  PRINT_DEBUG_NAME;
+
+  IDEBUG_START;
+
+  FUNCTION_PROFILE_START();
+
   if (ada == ZERO) {
     *C        = ZERO;
     *(S  + 0) = ONE;
@@ -54,6 +54,12 @@ void NAME(FLOAT *DA, FLOAT *DB, FLOAT *C, FLOAT *S){
   FLOAT ada = fabs(da_r) + fabs(da_i);
   FLOAT adb;
 
+  PRINT_DEBUG_NAME;
+
+  IDEBUG_START;
+
+  FUNCTION_PROFILE_START();
+
   if (ada == ZERO) {
     *C        = ZERO;
     *(S  + 0) = ONE;
diff --git a/interface/zsyr.c b/interface/zsyr.c
index 5fe29cefa..09b1de578 100644
--- a/interface/zsyr.c
+++ b/interface/zsyr.c
@@ -121,6 +121,9 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_UPLO Uplo, int n, FLOAT alpha, FLO
   FLOAT *buffer;
   int trans, uplo;
   blasint info;
+  FLOAT * ALPHA = &alpha;
+  FLOAT alpha_r	= ALPHA[0];
+  FLOAT alpha_i	= ALPHA[1];
 #ifdef SMP
   int nthreads;
 #endif
diff --git a/kernel/CMakeLists.txt b/kernel/CMakeLists.txt
new file mode 100644
index 000000000..8a3b021cc
--- /dev/null
+++ b/kernel/CMakeLists.txt
@@ -0,0 +1,428 @@
+
+include_directories(${CMAKE_SOURCE_DIR})
+include("${CMAKE_SOURCE_DIR}/cmake/kernel.cmake")
+
+# Makefile
+
+if (DEFINED TARGET_CORE)
+  #override CFLAGS += -DBUILD_KERNEL -DTABLE_NAME=gotoblas_$(TARGET_CORE)
+  set(BUILD_KERNEL 1)
+  set(KDIR "")
+  set(TSUFFIX "_${TARGET_CORE}")
+else ()
+  set(TARGET_CORE ${CORE})
+  set(KDIR "")
+  set(TSUFFIX "")
+endif ()
+
+SetDefaultL1()
+SetDefaultL2()
+SetDefaultL3()
+ParseMakefileVars("${KERNELDIR}/KERNEL")
+ParseMakefileVars("${KERNELDIR}/KERNEL.${TARGET_CORE}")
+
+if (${ARCH} STREQUAL "x86")
+if (NOT MSVC)
+  GenerateNamedObjects("${KERNELDIR}/cpuid.S" "" "" false "" "" true)
+else()
+  GenerateNamedObjects("${KERNELDIR}/cpuid_win.c" "" "" false "" "" true)
+endif()
+endif ()
+
+# don't use float type name mangling here
+GenerateNamedObjects("${KERNELDIR}/${LSAME_KERNEL}" "F_INTERFACE" "lsame" false "" "" true)
+GenerateNamedObjects("${KERNELDIR}/${SCABS_KERNEL}" "COMPLEX;F_INTERFACE" "scabs1" false "" "" true)
+GenerateNamedObjects("${KERNELDIR}/${DCABS_KERNEL}" "DOUBLE;COMPLEX;F_INTERFACE" "dcabs1" false "" "" true)
+
+# Makefile.L1
+foreach (float_type ${FLOAT_TYPES})
+  # a bit of metaprogramming here to pull out the appropriate KERNEL var
+  string(SUBSTRING ${float_type} 0 1 float_char)
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}AMAXKERNEL}" "USE_ABS" "amax_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}AMINKERNEL}" "USE_ABS;USE_MIN" "amin_k" false "" "" false ${float_type})
+  if (DEFINED ${float_char}MAXKERNEL)
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}MAXKERNEL}" "" "max_k" false "" "" false ${float_type})
+  endif ()
+  if (DEFINED ${float_char}MINKERNEL)
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}MINKERNEL}" "" "min_k" false "" "" false ${float_type})
+  endif ()
+  GenerateNamedObjects("${KERNELDIR}/${I${float_char}AMAXKERNEL}" "USE_ABS" "i*amax_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${I${float_char}AMINKERNEL}" "USE_ABS;USE_MIN" "i*amin_k" false "" "" false ${float_type})
+  if (DEFINED I${float_char}MAXKERNEL)
+    GenerateNamedObjects("${KERNELDIR}/${I${float_char}MAXKERNEL}" "" "i*max_k" false "" "" false ${float_type})
+  endif ()
+  if (DEFINED I${float_char}MINKERNEL)
+    GenerateNamedObjects("${KERNELDIR}/${I${float_char}MINKERNEL}" "" "i*min_k" false "" "" false ${float_type})
+  endif ()
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}ASUMKERNEL}" "" "asum_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}AXPYKERNEL}" "" "axpy_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}COPYKERNEL}" "C_INTERFACE" "copy_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}NRM2KERNEL}" "" "nrm2_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}ROTKERNEL}" "" "rot_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}SCALKERNEL}" "" "scal_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}SWAPKERNEL}" "" "swap_k" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}AXPBYKERNEL}" "" "axpby_k" false "" "" false ${float_type})
+
+  if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}AXPYKERNEL}" "CONJ" "axpyc_k" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}DOTKERNEL}" "" "dotu_k" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}DOTKERNEL}" "CONJ" "dotc_k" false "" "" false ${float_type})
+  else ()
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}DOTKERNEL}" "" "dot_k" false "" "" false ${float_type})
+  endif ()
+
+  if (${float_type} STREQUAL "COMPLEX")
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}ROTKERNEL}" "" "srot_k" false "" "" false ${float_type})
+  endif()
+  if (${float_type} STREQUAL "ZCOMPLEX")
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}ROTKERNEL}" "" "drot_k" false "" "" false ${float_type})
+  endif()
+
+endforeach ()
+
+#dsdot,sdsdot
+GenerateNamedObjects("${KERNELDIR}/${DSDOTKERNEL}" "DSDOT" "d*dot_k" false "" "" false "SINGLE")
+GenerateNamedObjects("${KERNELDIR}/${DSDOTKERNEL}" "DSDOT" "dsdot_k" false "" "" false "SINGLE")
+
+# Makefile.L2
+GenerateCombinationObjects("generic/symv_k.c" "LOWER" "U" "" 1 "" "" 3)
+GenerateNamedObjects("generic/ger.c" "" "ger_k" false "" "" "" 3)
+foreach (float_type ${FLOAT_TYPES})
+  string(SUBSTRING ${float_type} 0 1 float_char)
+  if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GERUKERNEL}" "" "geru_k" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GERCKERNEL}" "CONJ" "gerc_k" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GERUKERNEL}" "XCONJ" "gerv_k" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GERCKERNEL}" "CONJ;XCONJ" "gerd_k" false "" "" false ${float_type})
+
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVNKERNEL}" "" "gemv_n" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVTKERNEL}" "TRANSA" "gemv_t" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVNKERNEL}" "CONJ" "gemv_r" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVTKERNEL}" "CONJ;TRANSA" "gemv_c" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVNKERNEL}" "XCONJ" "gemv_o" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVTKERNEL}" "XCONJ;TRANSA" "gemv_u" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVNKERNEL}" "XCONJ;CONJ" "gemv_s" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVTKERNEL}" "XCONJ;CONJ;TRANSA" "gemv_d" false "" "" false ${float_type})
+
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}HEMV_U_KERNEL}" "HEMV" "hemv_U" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}HEMV_L_KERNEL}" "HEMV;LOWER" "hemv_L" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}HEMV_V_KERNEL}" "HEMV;HEMVREV" "hemv_V" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}HEMV_M_KERNEL}" "HEMV;HEMVREV;LOWER" "hemv_M" false "" "" false ${float_type})
+
+  else ()
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVNKERNEL}" "" "gemv_n" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMVTKERNEL}" "TRANS" "gemv_t" false "" "" false ${float_type})
+  endif ()
+endforeach ()
+
+# Makefile.L3
+set(USE_TRMM false)
+
+if (${ARCH} STREQUAL "arm" OR ${ARCH} STREQUAL "arm64" OR "${TARGET}" STREQUAL "LONGSOON3B" OR "${TARGET}" STREQUAL "GENERIC" OR "${CORE}" STREQUAL "generic" OR "${TARGET}" STREQUAL "HASWELL" OR "${CORE}" STREQUAL "haswell")
+  set(USE_TRMM true)
+endif ()
+
+foreach (float_type ${FLOAT_TYPES})
+  string(SUBSTRING ${float_type} 0 1 float_char)
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMKERNEL}" "" "gemm_kernel" false "" "" false ${float_type})
+
+  if (${float_char}GEMMINCOPY)
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMINCOPY}" "${float_type}" "${${float_char}GEMMINCOPYOBJ}" false "" "" true ${float_type})
+  endif ()
+
+  if (${float_char}GEMMITCOPY)
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMITCOPY}" "${float_type}" "${${float_char}GEMMITCOPYOBJ}" false "" "" true ${float_type})
+  endif ()
+
+  if (${float_char}GEMMONCOPY)
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMONCOPY}" "${float_type}" "${${float_char}GEMMONCOPYOBJ}" false "" "" true ${float_type})
+  endif ()
+
+  if (${float_char}GEMMOTCOPY)
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMOTCOPY}" "${float_type}" "${${float_char}GEMMOTCOPYOBJ}" false "" "" true ${float_type})
+  endif ()
+
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMM_BETA}" "" "gemm_beta" false "" "" false ${float_type})
+
+  if (USE_TRMM)
+    set(TRMM_KERNEL "${${float_char}TRMMKERNEL}")
+  else ()
+    set(TRMM_KERNEL "${${float_char}GEMMKERNEL}")
+  endif ()
+
+  if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+
+    # just enumerate all these. there is an extra define for these indicating which side is a conjugate (e.g. CN NC NN) that I don't really want to work into GenerateCombinationObjects
+
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMKERNEL}" "NN" "gemm_kernel_n" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMKERNEL}" "CN" "gemm_kernel_l" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMKERNEL}" "NC" "gemm_kernel_r" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}GEMMKERNEL}" "CC" "gemm_kernel_b" false "" "" false ${float_type})
+
+    GenerateNamedObjects("${KERNELDIR}/${TRMM_KERNEL}" "TRMMKERNEL;LEFT;NN" "trmm_kernel_LN" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${TRMM_KERNEL}" "TRMMKERNEL;LEFT;TRANSA;NN" "trmm_kernel_LT" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${TRMM_KERNEL}" "TRMMKERNEL;LEFT;CONJ;CN" "trmm_kernel_LR" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${TRMM_KERNEL}" "TRMMKERNEL;LEFT;TRANSA;CONJ;CN" "trmm_kernel_LC" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${TRMM_KERNEL}" "TRMMKERNEL;NN" "trmm_kernel_RN" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${TRMM_KERNEL}" "TRMMKERNEL;TRANSA;NN" "trmm_kernel_RT" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${TRMM_KERNEL}" "TRMMKERNEL;CONJ;NC" "trmm_kernel_RR" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${TRMM_KERNEL}" "TRMMKERNEL;TRANSA;CONJ;NC" "trmm_kernel_RC" false "" "" false ${float_type})
+
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}TRSMKERNEL_LN}" "UPPER;LN;TRSMKERNEL;CONJ" "trsm_kernel_LR" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}TRSMKERNEL_LT}" "LT;TRSMKERNEL;CONJ" "trsm_kernel_LC" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}TRSMKERNEL_RN}" "UPPER;RN;TRSMKERNEL;CONJ" "trsm_kernel_RR" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}TRSMKERNEL_RT}" "RT;TRSMKERNEL;CONJ" "trsm_kernel_RC" false "" "" false ${float_type})
+
+
+    #hemm
+  GenerateNamedObjects("generic/zhemm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "" "hemm_iutcopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/zhemm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "hemm_iltcopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/zhemm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "hemm_outcopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/zhemm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "hemm_oltcopy" false "" "" false ${float_type})
+
+  # symm for c and z
+  GenerateNamedObjects("generic/zsymm_ucopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "symm_outcopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/zsymm_ucopy_${${float_char}GEMM_UNROLL_M}.c" "" "symm_iutcopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/zsymm_lcopy_${${float_char}GEMM_UNROLL_N}.c" "LOWER;OUTER" "symm_oltcopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/zsymm_lcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "symm_iltcopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/ztrmm_uncopy_${${float_char}GEMM_UNROLL_M}.c" "UNIT" "trmm_iunucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_uncopy_${${float_char}GEMM_UNROLL_M}.c" "" "trmm_iunncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_uncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;UNIT" "trmm_ounucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_uncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "trmm_ounncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/ztrmm_lncopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER;UNIT" "trmm_ilnucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_lncopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "trmm_ilnncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_lncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER;UNIT" "trmm_olnucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_lncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "trmm_olnncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/ztrmm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "UNIT" "trmm_iutucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "" "trmm_iutncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;UNIT" "trmm_outucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "trmm_outncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/ztrmm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER;UNIT" "trmm_iltucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "trmm_iltncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER;UNIT" "trmm_oltucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrmm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "trmm_oltncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/ztrsm_uncopy_${${float_char}GEMM_UNROLL_M}.c" "UNIT" "trsm_iunucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_uncopy_${${float_char}GEMM_UNROLL_M}.c" "" "trsm_iunncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_uncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;UNIT" "trsm_ounucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_uncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "trsm_ounncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/ztrsm_lncopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER;UNIT" "trsm_ilnucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_lncopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "trsm_ilnncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_lncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER;UNIT" "trsm_olnucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_lncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "trsm_olnncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/ztrsm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "UNIT" "trsm_iutucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "" "trsm_iutncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;UNIT" "trsm_outucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "trsm_outncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/ztrsm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER;UNIT" "trsm_iltucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "trsm_iltncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER;UNIT" "trsm_oltucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/ztrsm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "trsm_oltncopy" false "" "" false ${float_type})
+
+  else ()  #For real
+    GenerateCombinationObjects("${KERNELDIR}/${TRMM_KERNEL}" "LEFT;TRANSA" "R;N" "TRMMKERNEL" 2 "trmm_kernel" false ${float_type})
+
+    # symm for s and d
+    GenerateNamedObjects("generic/symm_ucopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "symm_outcopy" false "" "" false ${float_type})
+    GenerateNamedObjects("generic/symm_ucopy_${${float_char}GEMM_UNROLL_M}.c" "" "symm_iutcopy" false "" "" false ${float_type})
+
+    GenerateNamedObjects("generic/symm_lcopy_${${float_char}GEMM_UNROLL_N}.c" "LOWER;OUTER" "symm_oltcopy" false "" "" false ${float_type})
+    GenerateNamedObjects("generic/symm_lcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "symm_iltcopy" false "" "" false ${float_type})
+
+  # These don't use a scheme that is easy to iterate over - the filenames have part of the DEFINE codes in them, for UPPER/TRANS but not for UNIT/OUTER. Also TRANS is not passed in as a define.
+  # Could simplify it a bit by pairing up by -UUNIT/-DUNIT.
+
+  GenerateNamedObjects("generic/trmm_uncopy_${${float_char}GEMM_UNROLL_M}.c" "UNIT" "trmm_iunucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_uncopy_${${float_char}GEMM_UNROLL_M}.c" "" "trmm_iunncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_uncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;UNIT" "trmm_ounucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_uncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "trmm_ounncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/trmm_lncopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER;UNIT" "trmm_ilnucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_lncopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "trmm_ilnncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_lncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER;UNIT" "trmm_olnucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_lncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "trmm_olnncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/trmm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "UNIT" "trmm_iutucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "" "trmm_iutncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;UNIT" "trmm_outucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "trmm_outncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/trmm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER;UNIT" "trmm_iltucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "trmm_iltncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER;UNIT" "trmm_oltucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trmm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "trmm_oltncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/trsm_uncopy_${${float_char}GEMM_UNROLL_M}.c" "UNIT" "trsm_iunucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_uncopy_${${float_char}GEMM_UNROLL_M}.c" "" "trsm_iunncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_uncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;UNIT" "trsm_ounucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_uncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "trsm_ounncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/trsm_lncopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER;UNIT" "trsm_ilnucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_lncopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "trsm_ilnncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_lncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER;UNIT" "trsm_olnucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_lncopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "trsm_olnncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/trsm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "UNIT" "trsm_iutucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_utcopy_${${float_char}GEMM_UNROLL_M}.c" "" "trsm_iutncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;UNIT" "trsm_outucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_utcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER" "trsm_outncopy" false "" "" false ${float_type})
+
+  GenerateNamedObjects("generic/trsm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER;UNIT" "trsm_iltucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_ltcopy_${${float_char}GEMM_UNROLL_M}.c" "LOWER" "trsm_iltncopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER;UNIT" "trsm_oltucopy" false "" "" false ${float_type})
+  GenerateNamedObjects("generic/trsm_ltcopy_${${float_char}GEMM_UNROLL_N}.c" "OUTER;LOWER" "trsm_oltncopy" false "" "" false ${float_type})
+
+  endif ()
+
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}TRSMKERNEL_LN}" "UPPER;LN;TRSMKERNEL" "trsm_kernel_LN" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}TRSMKERNEL_LT}" "LT;TRSMKERNEL" "trsm_kernel_LT" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}TRSMKERNEL_RN}" "UPPER;RN;TRSMKERNEL" "trsm_kernel_RN" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}TRSMKERNEL_RT}" "RT;TRSMKERNEL" "trsm_kernel_RT" false "" "" false ${float_type})
+
+
+
+  if (NOT DEFINED ${float_char}OMATCOPY_CN)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}OMATCOPY_CN ../arm/zomatcopy_cn.c)
+    else ()
+      set(${float_char}OMATCOPY_CN ../arm/omatcopy_cn.c)
+    endif ()
+  endif ()
+  if (NOT DEFINED ${float_char}OMATCOPY_RN)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}OMATCOPY_RN ../arm/zomatcopy_rn.c)
+    else ()
+      set(${float_char}OMATCOPY_RN ../arm/omatcopy_rn.c)
+    endif  ()
+  endif ()
+  if (NOT DEFINED ${float_char}OMATCOPY_CT)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}OMATCOPY_CT ../arm/zomatcopy_ct.c)
+    else ()
+      set(${float_char}OMATCOPY_CT ../arm/omatcopy_ct.c)
+    endif ()
+  endif ()
+  if (NOT DEFINED ${float_char}OMATCOPY_RT)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}OMATCOPY_RT ../arm/zomatcopy_rt.c)
+    else ()
+      set(${float_char}OMATCOPY_RT ../arm/omatcopy_rt.c)
+    endif ()
+  endif ()
+
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}OMATCOPY_CN}" "" "omatcopy_k_cn" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}OMATCOPY_RN}" "ROWM" "omatcopy_k_rn" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}OMATCOPY_CT}" "" "omatcopy_k_ct" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}OMATCOPY_RT}" "ROWM" "omatcopy_k_rt" false "" "" false ${float_type})
+
+  if (NOT DEFINED ${float_char}OMATCOPY_CNC)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}OMATCOPY_CNC ../arm/zomatcopy_cnc.c)
+    endif ()
+  endif ()
+  if (NOT DEFINED ${float_char}OMATCOPY_RNC)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}OMATCOPY_RNC ../arm/zomatcopy_rnc.c)
+    endif  ()
+  endif ()
+  if (NOT DEFINED ${float_char}OMATCOPY_CTC)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}OMATCOPY_CTC ../arm/zomatcopy_ctc.c)
+    endif ()
+  endif ()
+  if (NOT DEFINED ${float_char}OMATCOPY_RTC)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}OMATCOPY_RTC ../arm/zomatcopy_rtc.c)
+    endif ()
+  endif ()
+
+  if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}OMATCOPY_CNC}" "CONJ" "omatcopy_k_cnc" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}OMATCOPY_RNC}" "CONJ;ROWM" "omatcopy_k_rnc" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}OMATCOPY_CTC}" "CONJ" "omatcopy_k_ctc" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}OMATCOPY_RTC}" "CONJ;ROWM" "omatcopy_k_rtc" false "" "" false ${float_type})
+  endif()
+
+  #imatcopy
+  if (NOT DEFINED ${float_char}IMATCOPY_CN)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}IMATCOPY_CN ../generic/zimatcopy_cn.c)
+    else ()
+      set(${float_char}IMATCOPY_CN ../generic/imatcopy_cn.c)
+    endif ()
+  endif ()
+
+  if (NOT DEFINED ${float_char}IMATCOPY_RN)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}IMATCOPY_RN ../generic/zimatcopy_rn.c)
+    else ()
+      set(${float_char}IMATCOPY_RN ../generic/imatcopy_rn.c)
+    endif ()
+  endif ()
+
+  if (NOT DEFINED ${float_char}IMATCOPY_CT)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}IMATCOPY_CT ../generic/zimatcopy_ct.c)
+    else ()
+      set(${float_char}IMATCOPY_CT ../generic/imatcopy_ct.c)
+    endif ()
+  endif ()
+
+  if (NOT DEFINED ${float_char}IMATCOPY_RT)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}IMATCOPY_RT ../generic/zimatcopy_rt.c)
+    else ()
+      set(${float_char}IMATCOPY_RT ../generic/imatcopy_rt.c)
+    endif ()
+  endif ()
+
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}IMATCOPY_CN}" "" "imatcopy_k_cn" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}IMATCOPY_RN}" "ROWM" "imatcopy_k_rn" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}IMATCOPY_CT}" "" "imatcopy_k_ct" false "" "" false ${float_type})
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}IMATCOPY_RT}" "ROWM" "imatcopy_k_rt" false "" "" false ${float_type})
+
+
+  if (NOT DEFINED ${float_char}IMATCOPY_CNC)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}IMATCOPY_CNC ../generic/zimatcopy_cnc.c)
+    endif ()
+  endif ()
+  if (NOT DEFINED ${float_char}IMATCOPY_RNC)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}IMATCOPY_RNC ../generic/zimatcopy_rnc.c)
+    endif  ()
+  endif ()
+  if (NOT DEFINED ${float_char}IMATCOPY_CTC)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}IMATCOPY_CTC ../generic/zimatcopy_ctc.c)
+    endif ()
+  endif ()
+  if (NOT DEFINED ${float_char}IMATCOPY_RTC)
+    if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+      set(${float_char}IMATCOPY_RTC ../generic/zimatcopy_rtc.c)
+    endif ()
+  endif ()
+
+  if (${float_char} STREQUAL "Z" OR ${float_char} STREQUAL "C")
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}IMATCOPY_CNC}" "CONJ" "imatcopy_k_cnc" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}IMATCOPY_RNC}" "CONJ;ROWM" "imatcopy_k_rnc" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}IMATCOPY_CTC}" "CONJ" "imatcopy_k_ctc" false "" "" false ${float_type})
+    GenerateNamedObjects("${KERNELDIR}/${${float_char}IMATCOPY_RTC}" "CONJ;ROWM" "imatcopy_k_rtc" false "" "" false ${float_type})
+  endif()
+
+  #geadd
+  GenerateNamedObjects("${KERNELDIR}/${${float_char}GEADD_KERNEL}" "" "geadd_k" false "" "" false ${float_type})
+endforeach ()
+
+# Makefile.LA
+#DBLASOBJS	+= dneg_tcopy$(TSUFFIX).$(SUFFIX) dlaswp_ncopy$(TSUFFIX).$(SUFFIX)
+
+add_library(kernel OBJECT ${OPENBLAS_SRC})
diff --git a/kernel/Makefile.L3 b/kernel/Makefile.L3
index fdbae2daa..63e675b8d 100644
--- a/kernel/Makefile.L3
+++ b/kernel/Makefile.L3
@@ -32,6 +32,10 @@ ifeq ($(TARGET), GENERIC)
 USE_TRMM = 1
 endif
 
+ifeq ($(CORE), HASWELL)
+USE_TRMM = 1
+endif
+
 
 
 SKERNELOBJS	+= \
@@ -330,11 +334,15 @@ endif
 SBLASOBJS += \
 	somatcopy_k_cn$(TSUFFIX).$(SUFFIX) somatcopy_k_rn$(TSUFFIX).$(SUFFIX) \
 	somatcopy_k_ct$(TSUFFIX).$(SUFFIX) somatcopy_k_rt$(TSUFFIX).$(SUFFIX) \
+	simatcopy_k_cn$(TSUFFIX).$(SUFFIX) simatcopy_k_rn$(TSUFFIX).$(SUFFIX) \
+	simatcopy_k_ct$(TSUFFIX).$(SUFFIX) simatcopy_k_rt$(TSUFFIX).$(SUFFIX) \
 	sgeadd_k$(TSUFFIX).$(SUFFIX) 
 
 DBLASOBJS += \
 	domatcopy_k_cn$(TSUFFIX).$(SUFFIX) domatcopy_k_rn$(TSUFFIX).$(SUFFIX) \
 	domatcopy_k_ct$(TSUFFIX).$(SUFFIX) domatcopy_k_rt$(TSUFFIX).$(SUFFIX) \
+	dimatcopy_k_cn$(TSUFFIX).$(SUFFIX) dimatcopy_k_rn$(TSUFFIX).$(SUFFIX) \
+	dimatcopy_k_ct$(TSUFFIX).$(SUFFIX) dimatcopy_k_rt$(TSUFFIX).$(SUFFIX) \
 	dgeadd_k$(TSUFFIX).$(SUFFIX) 
 
 CBLASOBJS += \
@@ -342,6 +350,10 @@ CBLASOBJS += \
 	comatcopy_k_ct$(TSUFFIX).$(SUFFIX) comatcopy_k_rt$(TSUFFIX).$(SUFFIX) \
 	comatcopy_k_cnc$(TSUFFIX).$(SUFFIX) comatcopy_k_rnc$(TSUFFIX).$(SUFFIX) \
 	comatcopy_k_ctc$(TSUFFIX).$(SUFFIX) comatcopy_k_rtc$(TSUFFIX).$(SUFFIX) \
+	cimatcopy_k_cn$(TSUFFIX).$(SUFFIX)  cimatcopy_k_rn$(TSUFFIX).$(SUFFIX) \
+	cimatcopy_k_ct$(TSUFFIX).$(SUFFIX)  cimatcopy_k_rt$(TSUFFIX).$(SUFFIX) \
+	cimatcopy_k_cnc$(TSUFFIX).$(SUFFIX) cimatcopy_k_rnc$(TSUFFIX).$(SUFFIX) \
+	cimatcopy_k_ctc$(TSUFFIX).$(SUFFIX) cimatcopy_k_rtc$(TSUFFIX).$(SUFFIX) \
 	cgeadd_k$(TSUFFIX).$(SUFFIX) 
 
 ZBLASOBJS += \
@@ -349,6 +361,10 @@ ZBLASOBJS += \
 	zomatcopy_k_ct$(TSUFFIX).$(SUFFIX) zomatcopy_k_rt$(TSUFFIX).$(SUFFIX) \
 	zomatcopy_k_cnc$(TSUFFIX).$(SUFFIX) zomatcopy_k_rnc$(TSUFFIX).$(SUFFIX) \
 	zomatcopy_k_ctc$(TSUFFIX).$(SUFFIX) zomatcopy_k_rtc$(TSUFFIX).$(SUFFIX) \
+	zimatcopy_k_cn$(TSUFFIX).$(SUFFIX)  zimatcopy_k_rn$(TSUFFIX).$(SUFFIX) \
+	zimatcopy_k_ct$(TSUFFIX).$(SUFFIX)  zimatcopy_k_rt$(TSUFFIX).$(SUFFIX) \
+	zimatcopy_k_cnc$(TSUFFIX).$(SUFFIX) zimatcopy_k_rnc$(TSUFFIX).$(SUFFIX) \
+	zimatcopy_k_ctc$(TSUFFIX).$(SUFFIX) zimatcopy_k_rtc$(TSUFFIX).$(SUFFIX) \
 	zgeadd_k$(TSUFFIX).$(SUFFIX) 
 
 
@@ -3301,6 +3317,34 @@ endif
 $(KDIR)domatcopy_k_rt$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(DOMATCOPY_RT)
 	$(CC) $(CFLAGS) -c -DDOUBLE -UCOMPLEX -DROWM $< -o $@
 
+ifndef DIMATCOPY_CN
+DIMATCOPY_CN = ../generic/imatcopy_cn.c
+endif
+
+$(KDIR)dimatcopy_k_cn$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(DIMATCOPY_CN)
+	$(CC) $(CFLAGS) -c -DDOUBLE -UCOMPLEX -UROWM $< -o $@
+
+ifndef DIMATCOPY_RN
+DIMATCOPY_RN = ../generic/imatcopy_rn.c
+endif
+
+$(KDIR)dimatcopy_k_rn$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(DIMATCOPY_RN)
+	$(CC) $(CFLAGS) -c -DDOUBLE -UCOMPLEX -DROWM $< -o $@
+
+ifndef DIMATCOPY_CT
+DIMATCOPY_CT = ../generic/imatcopy_ct.c
+endif
+
+$(KDIR)dimatcopy_k_ct$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(DIMATCOPY_CT)
+	$(CC) $(CFLAGS) -c -DDOUBLE -UCOMPLEX -UROWM $< -o $@
+
+ifndef DIMATCOPY_RT
+DIMATCOPY_RT = ../generic/imatcopy_rt.c
+endif
+
+$(KDIR)dimatcopy_k_rt$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(DIMATCOPY_RT)
+	$(CC) $(CFLAGS) -c -DDOUBLE -UCOMPLEX -DROWM $< -o $@
+
 ifndef SOMATCOPY_CN
 SOMATCOPY_CN = ../arm/omatcopy_cn.c
 endif
@@ -3329,6 +3373,34 @@ endif
 $(KDIR)somatcopy_k_rt$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(SOMATCOPY_RT)
 	$(CC) $(CFLAGS) -c -UDOUBLE -UCOMPLEX -DROWM $< -o $@
 
+ifndef SIMATCOPY_CN
+SIMATCOPY_CN = ../generic/imatcopy_cn.c
+endif
+
+$(KDIR)simatcopy_k_cn$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(SIMATCOPY_CN)
+	$(CC) $(CFLAGS) -c -UDOUBLE -UCOMPLEX -UROWM $< -o $@
+
+ifndef SIMATCOPY_RN
+SIMATCOPY_RN = ../generic/imatcopy_rn.c
+endif
+
+$(KDIR)simatcopy_k_rn$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(SIMATCOPY_RN)
+	$(CC) $(CFLAGS) -c -UDOUBLE -UCOMPLEX -DROWM $< -o $@
+
+ifndef SIMATCOPY_CT
+SIMATCOPY_CT = ../generic/imatcopy_ct.c
+endif
+
+$(KDIR)simatcopy_k_ct$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(SIMATCOPY_CT)
+	$(CC) $(CFLAGS) -c -UDOUBLE -UCOMPLEX -UROWM $< -o $@
+
+ifndef SIMATCOPY_RT
+SIMATCOPY_RT = ../generic/imatcopy_rt.c
+endif
+
+$(KDIR)simatcopy_k_rt$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(SIMATCOPY_RT)
+	$(CC) $(CFLAGS) -c -UDOUBLE -UCOMPLEX -DROWM $< -o $@
+
 
 ifndef COMATCOPY_CN
 COMATCOPY_CN = ../arm/zomatcopy_cn.c
@@ -3386,6 +3458,63 @@ endif
 $(KDIR)comatcopy_k_rtc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(COMATCOPY_RTC)
 	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -DROWM -DCONJ $< -o $@
 
+ifndef CIMATCOPY_CN
+CIMATCOPY_CN = ../generic/zimatcopy_cn.c
+endif
+
+$(KDIR)cimatcopy_k_cn$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CIMATCOPY_CN)
+	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -UROWM -UCONJ $< -o $@
+
+ifndef CIMATCOPY_RN
+CIMATCOPY_RN = ../generic/zimatcopy_rn.c
+endif
+
+$(KDIR)cimatcopy_k_rn$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CIMATCOPY_RN)
+	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -DROWM -UCONJ $< -o $@
+
+ifndef CIMATCOPY_CT
+CIMATCOPY_CT = ../generic/zimatcopy_ct.c
+endif
+
+$(KDIR)cimatcopy_k_ct$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CIMATCOPY_CT)
+	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -UROWM -UCONJ $< -o $@
+
+ifndef CIMATCOPY_RT
+CIMATCOPY_RT = ../generic/zimatcopy_rt.c
+endif
+
+$(KDIR)cimatcopy_k_rt$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CIMATCOPY_RT)
+	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -DROWM -UCONJ $< -o $@
+
+ifndef CIMATCOPY_CNC
+CIMATCOPY_CNC = ../generic/zimatcopy_cnc.c
+endif
+
+$(KDIR)cimatcopy_k_cnc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CIMATCOPY_CNC)
+	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -UROWM -DCONJ $< -o $@
+
+ifndef CIMATCOPY_RNC
+CIMATCOPY_RNC = ../generic/zimatcopy_rnc.c
+endif
+
+$(KDIR)cimatcopy_k_rnc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CIMATCOPY_RNC)
+	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -DROWM -DCONJ $< -o $@
+
+ifndef CIMATCOPY_CTC
+CIMATCOPY_CTC = ../generic/zimatcopy_ctc.c
+endif
+
+$(KDIR)cimatcopy_k_ctc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CIMATCOPY_CTC)
+	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -UROWM -DCONJ $< -o $@
+
+ifndef CIMATCOPY_RTC
+CIMATCOPY_RTC = ../generic/zimatcopy_rtc.c
+endif
+
+$(KDIR)cimatcopy_k_rtc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(CIMATCOPY_RTC)
+	$(CC) $(CFLAGS) -c -UDOUBLE -DCOMPLEX -DROWM -DCONJ $< -o $@
+
+
 
 ifndef ZOMATCOPY_CN
 ZOMATCOPY_CN = ../arm/zomatcopy_cn.c
@@ -3443,6 +3572,62 @@ endif
 $(KDIR)zomatcopy_k_rtc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZOMATCOPY_RTC)
 	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DROWM -DCONJ $< -o $@
 
+ifndef ZIMATCOPY_CN
+ZIMATCOPY_CN = ../generic/zimatcopy_cn.c
+endif
+
+$(KDIR)zimatcopy_k_cn$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZIMATCOPY_CN)
+	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -UROWM -UCONJ $< -o $@
+
+ifndef ZIMATCOPY_RN
+ZIMATCOPY_RN = ../generic/zimatcopy_rn.c
+endif
+
+$(KDIR)zimatcopy_k_rn$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZIMATCOPY_RN)
+	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DROWM -UCONJ $< -o $@
+
+ifndef ZIMATCOPY_CT
+ZIMATCOPY_CT = ../generic/zimatcopy_ct.c
+endif
+
+$(KDIR)zimatcopy_k_ct$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZIMATCOPY_CT)
+	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -UROWM -UCONJ $< -o $@
+
+ifndef ZIMATCOPY_RT
+ZIMATCOPY_RT = ../generic/zimatcopy_rt.c
+endif
+
+$(KDIR)zimatcopy_k_rt$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZIMATCOPY_RT)
+	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DROWM -UCONJ $< -o $@
+
+ifndef ZIMATCOPY_CNC
+ZIMATCOPY_CNC = ../generic/zimatcopy_cnc.c
+endif
+
+$(KDIR)zimatcopy_k_cnc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZIMATCOPY_CNC)
+	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -UROWM -DCONJ $< -o $@
+
+ifndef ZIMATCOPY_RNC
+ZIMATCOPY_RNC = ../generic/zimatcopy_rnc.c
+endif
+
+$(KDIR)zimatcopy_k_rnc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZIMATCOPY_RNC)
+	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DROWM -DCONJ $< -o $@
+
+ifndef ZIMATCOPY_CTC
+ZIMATCOPY_CTC = ../generic/zimatcopy_ctc.c
+endif
+
+$(KDIR)zimatcopy_k_ctc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZIMATCOPY_CTC)
+	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -UROWM -DCONJ $< -o $@
+
+ifndef ZIMATCOPY_RTC
+ZIMATCOPY_RTC = ../generic/zimatcopy_rtc.c
+endif
+
+$(KDIR)zimatcopy_k_rtc$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(ZIMATCOPY_RTC)
+	$(CC) $(CFLAGS) -c -DDOUBLE -DCOMPLEX -DROWM -DCONJ $< -o $@
+
 
 ifndef SGEADD_K
 SGEADD_K = ../generic/geadd.c
@@ -3455,7 +3640,7 @@ ifndef DGEADD_K
 DGEADD_K = ../generic/geadd.c
 endif
 
-$(KDIR)dgeadd_k$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(SGEADD_K)
+$(KDIR)dgeadd_k$(TSUFFIX).$(SUFFIX) : $(KERNELDIR)/$(DGEADD_K)
 	$(CC) $(CFLAGS) -c -DDOUBLE -UCOMPLEX -UROWM $< -o $@
 
 ifndef CGEADD_K
diff --git a/kernel/arm/scal.c b/kernel/arm/scal.c
index 4593e2279..91ca76569 100644
--- a/kernel/arm/scal.c
+++ b/kernel/arm/scal.c
@@ -52,7 +52,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 		j++;
 
 	}
-	return;
+	return 0;
 
 }
 
diff --git a/kernel/arm/zaxpby.c b/kernel/arm/zaxpby.c
index 2e0c2940d..d9948349d 100644
--- a/kernel/arm/zaxpby.c
+++ b/kernel/arm/zaxpby.c
@@ -38,13 +38,16 @@ int CNAME(BLASLONG n, FLOAT alpha_r, FLOAT alpha_i, FLOAT *x, BLASLONG inc_x, FL
 	BLASLONG ix,iy;
 	FLOAT temp;
 
+	BLASLONG inc_x2;
+	BLASLONG inc_y2;
+
 	if ( n < 0     )  return(0);
 
 	ix = 0;
 	iy = 0;
 
-	BLASLONG inc_x2 = 2 * inc_x;
-	BLASLONG inc_y2 = 2 * inc_y;
+	inc_x2 = 2 * inc_x;
+	inc_y2 = 2 * inc_y;
 
 	if ( beta_r == 0.0 && beta_i == 0.0)
 	{
diff --git a/kernel/arm/zaxpy.c b/kernel/arm/zaxpy.c
index 929ee8b54..1dcaeac27 100644
--- a/kernel/arm/zaxpy.c
+++ b/kernel/arm/zaxpy.c
@@ -41,6 +41,8 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da_r, FLOAT da_i,
 {
 	BLASLONG i=0;
 	BLASLONG ix,iy;
+	BLASLONG inc_x2;
+	BLASLONG inc_y2;
 
 	if ( n < 0     )  return(0);
 	if ( da_r == 0.0 && da_i == 0.0 ) return(0);
@@ -48,8 +50,8 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da_r, FLOAT da_i,
 	ix = 0;
 	iy = 0;
 
-	BLASLONG inc_x2 = 2 * inc_x;
-	BLASLONG inc_y2 = 2 * inc_y;
+	inc_x2 = 2 * inc_x;
+	inc_y2 = 2 * inc_y;
 
 	while(i < n)
 	{
diff --git a/kernel/arm/zcopy.c b/kernel/arm/zcopy.c
index f720d6ee5..07fe584c5 100644
--- a/kernel/arm/zcopy.c
+++ b/kernel/arm/zcopy.c
@@ -40,11 +40,13 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
 {
 	BLASLONG i=0;
 	BLASLONG ix=0,iy=0;
+	BLASLONG inc_x2;
+	BLASLONG inc_y2;
 
 	if ( n < 0     )  return(0);
 
-	BLASLONG inc_x2 = 2 * inc_x;
-	BLASLONG inc_y2 = 2 * inc_y;
+	inc_x2 = 2 * inc_x;
+	inc_y2 = 2 * inc_y;
 
 	while(i < n)
 	{
diff --git a/kernel/arm/zdot.c b/kernel/arm/zdot.c
index 469487531..57f47e58e 100644
--- a/kernel/arm/zdot.c
+++ b/kernel/arm/zdot.c
@@ -35,25 +35,31 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 **************************************************************************************/
 
 #include "common.h"
-#include <complex.h>
 
+#ifndef _MSC_VER
+#include <complex.h>
 FLOAT _Complex CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
+#else
+OPENBLAS_COMPLEX_FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
+#endif
 {
 	BLASLONG i=0;
 	BLASLONG ix=0,iy=0;
 	FLOAT dot[2];
-	FLOAT _Complex result;
+	OPENBLAS_COMPLEX_FLOAT result;
+	BLASLONG inc_x2;
+	BLASLONG inc_y2;
 
 	dot[0]=0.0;
 	dot[1]=0.0;
 
-	__real__ result = 0.0 ;
-	__imag__ result = 0.0 ;
+	CREAL(result) = 0.0 ;
+	CIMAG(result) = 0.0 ;
 
 	if ( n < 1 )  return(result);
 
-	BLASLONG inc_x2 = 2 * inc_x ;
-	BLASLONG inc_y2 = 2 * inc_y ;
+	inc_x2 = 2 * inc_x ;
+	inc_y2 = 2 * inc_y ;
 
 	while(i < n)
 	{
@@ -69,8 +75,8 @@ FLOAT _Complex CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG in
 		i++ ;
 
 	}
-	__real__ result = dot[0];
-	__imag__ result = dot[1];
+	CREAL(result) = dot[0];
+	CIMAG(result) = dot[1];
 	return(result);
 
 }
diff --git a/kernel/arm/zrot.c b/kernel/arm/zrot.c
index 356a4df72..98be68db8 100644
--- a/kernel/arm/zrot.c
+++ b/kernel/arm/zrot.c
@@ -41,11 +41,13 @@ int CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT
 	BLASLONG i=0;
 	BLASLONG ix=0,iy=0;
 	FLOAT temp[2];
+	BLASLONG inc_x2;
+	BLASLONG inc_y2;
 
 	if ( n <= 0     )  return(0);
 
-	BLASLONG inc_x2 = 2 * inc_x ;
-	BLASLONG inc_y2 = 2 * inc_y ;
+	inc_x2 = 2 * inc_x ;
+	inc_y2 = 2 * inc_y ;
 
 	while(i < n)
 	{
diff --git a/kernel/arm/zswap.c b/kernel/arm/zswap.c
index fcfb38506..ae4760ae0 100644
--- a/kernel/arm/zswap.c
+++ b/kernel/arm/zswap.c
@@ -42,11 +42,13 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT dummy3, FLOAT dumm
 	BLASLONG i=0;
 	BLASLONG ix=0,iy=0;
 	FLOAT temp[2];
+	BLASLONG inc_x2;
+	BLASLONG inc_y2;
 
 	if ( n < 0     )  return(0);
 
-	BLASLONG inc_x2 = 2 * inc_x;
-	BLASLONG inc_y2 = 2 * inc_y;
+	inc_x2 = 2 * inc_x;
+	inc_y2 = 2 * inc_y;
 
 	while(i < n)
 	{
diff --git a/kernel/arm64/KERNEL.XGENE1 b/kernel/arm64/KERNEL.XGENE1
new file mode 100644
index 000000000..6ee0c730c
--- /dev/null
+++ b/kernel/arm64/KERNEL.XGENE1
@@ -0,0 +1 @@
+include $(KERNELDIR)/KERNEL.ARMV8
\ No newline at end of file
diff --git a/kernel/generic/imatcopy_cn.c b/kernel/generic/imatcopy_cn.c
new file mode 100644
index 000000000..e63bc976c
--- /dev/null
+++ b/kernel/generic/imatcopy_cn.c
@@ -0,0 +1,67 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon 
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j;
+	FLOAT *aptr;
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+    if ( alpha == 1.0   )  return(0); 
+
+	aptr = a;
+	if ( alpha == 0.0 )
+	{
+		for ( i=0; i<cols ; i++ )
+		{
+			for(j=0; j<rows; j++)
+			{
+				aptr[j] = 0.0;
+			}
+			aptr += lda;
+		}
+		return(0);
+	}
+
+	for ( i=0; i<cols ; i++ )
+	{
+		for(j=0; j<rows; j++)
+		{
+			aptr[j] = alpha * aptr[j];
+		}
+		aptr += lda;
+	}
+	return(0);
+}
+
diff --git a/kernel/generic/imatcopy_ct.c b/kernel/generic/imatcopy_ct.c
new file mode 100644
index 000000000..eee74327f
--- /dev/null
+++ b/kernel/generic/imatcopy_ct.c
@@ -0,0 +1,91 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon 
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j;
+	FLOAT *aptr, *bptr;
+    FLOAT tmp; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+	aptr = a;
+
+	if ( alpha == 0.0 )
+	{
+		for ( i=0; i<cols ; i++ )
+		{
+			aptr = &a[i];
+			for(j=0; j<rows; j++)
+			{
+				aptr[j*lda] = 0.0;
+			}
+		}
+		return(0);
+	}
+
+	if ( alpha == 1.0 )
+	{
+		for ( i=0; i<cols ; i++ )
+		{
+			bptr = &a[i];
+			for(j=i; j<rows; j++)
+			{
+                tmp = bptr[j*lda]; 
+				bptr[j*lda] = aptr[j];
+                aptr[j] = tmp; 
+			}
+			aptr += lda;
+		}
+		return(0);
+	}
+
+	for ( i=0; i<cols ; i++ )
+	{
+		bptr = &a[i];
+        bptr[i*lda] = alpha * bptr[i*lda]; 
+		for(j=i+1; j<rows; j++)
+		{
+            tmp = bptr[j*lda]; 
+    		bptr[j*lda] = alpha*aptr[j];
+            aptr[j] = alpha*tmp; 
+		}
+		aptr += lda;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/imatcopy_rn.c b/kernel/generic/imatcopy_rn.c
new file mode 100644
index 000000000..d952b600c
--- /dev/null
+++ b/kernel/generic/imatcopy_rn.c
@@ -0,0 +1,72 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon 
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j;
+	FLOAT *aptr;
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+    if ( alpha == 1.0  )  return(0); 
+
+	aptr = a;
+
+	if ( alpha == 0.0 )
+	{
+		for ( i=0; i<rows ; i++ )
+		{
+			for(j=0; j<cols; j++)
+			{
+				aptr[j] = 0.0;
+			}
+			aptr += lda;
+		}
+		return(0);
+	}
+
+	
+	for ( i=0; i<rows ; i++ )
+	{
+		for(j=0; j<cols; j++)
+		{
+			aptr[j] = alpha * aptr[j];
+		}
+		aptr += lda;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/imatcopy_rt.c b/kernel/generic/imatcopy_rt.c
new file mode 100644
index 000000000..fc856a4b8
--- /dev/null
+++ b/kernel/generic/imatcopy_rt.c
@@ -0,0 +1,64 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon
+ *
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j;
+	FLOAT *aptr,*bptr;
+    FLOAT tmp; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+	aptr = a;
+
+	for ( i=0; i<rows ; i++ )
+	{
+		bptr = &a[i];
+        bptr[i*lda] = alpha * bptr[i*lda]; 
+		for(j=i+1; j<cols; j++)
+		{
+            tmp = bptr[j*lda]; 
+            bptr[j*lda] = alpha * aptr[j];
+            aptr[j] = alpha * tmp; 
+
+		}
+		aptr += lda;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/trmmkernel_4x8.c b/kernel/generic/trmmkernel_4x8.c
new file mode 100644
index 000000000..09c47f147
--- /dev/null
+++ b/kernel/generic/trmmkernel_4x8.c
@@ -0,0 +1,1402 @@
+#include "common.h"
+#include <stdbool.h>
+
+int CNAME(BLASLONG bm,BLASLONG bn,BLASLONG bk,FLOAT alpha,FLOAT* ba,FLOAT* bb,FLOAT* C,BLASLONG ldc ,BLASLONG offset)
+{
+
+   BLASLONG i,j,k;
+   FLOAT *C0,*C1,*C2,*C3,*C4,*C5,*C6,*C7,*ptrba,*ptrbb;
+
+   FLOAT res0_0;
+   FLOAT res0_1;
+   FLOAT res0_2;
+   FLOAT res0_3;
+
+   FLOAT res1_0;
+   FLOAT res1_1;
+   FLOAT res1_2;
+   FLOAT res1_3;
+
+   FLOAT res2_0;
+   FLOAT res2_1;
+   FLOAT res2_2;
+   FLOAT res2_3;
+
+   FLOAT res3_0;
+   FLOAT res3_1;
+   FLOAT res3_2;
+   FLOAT res3_3;
+
+   FLOAT res4_0;
+   FLOAT res4_1;
+   FLOAT res4_2;
+   FLOAT res4_3;
+
+   FLOAT res5_0;
+   FLOAT res5_1;
+   FLOAT res5_2;
+   FLOAT res5_3;
+
+   FLOAT res6_0;
+   FLOAT res6_1;
+   FLOAT res6_2;
+   FLOAT res6_3;
+
+   FLOAT res7_0;
+   FLOAT res7_1;
+   FLOAT res7_2;
+   FLOAT res7_3;
+
+   FLOAT a0;
+   FLOAT a1;
+
+   FLOAT b0;
+   FLOAT b1;
+   FLOAT b2;
+   FLOAT b3;
+   FLOAT b4;
+   FLOAT b5;
+   FLOAT b6;
+   FLOAT b7;
+
+   BLASLONG off, temp;
+
+   bool left;
+   bool transposed;
+   bool backwards;
+
+#ifdef LEFT
+   left = true;
+#else
+   left = false;
+#endif
+
+#ifdef TRANSA
+   transposed = true;
+#else
+   transposed = false;
+#endif
+
+   backwards = left != transposed;
+
+   if (!left) {
+      off = -offset;
+   }
+
+
+   for (j=0; j<bn/8; j+=1) // do blocks of the Mx8 loops 
+   {
+        C0 = C;
+        C1 = C0+ldc;
+        C2 = C1+ldc;
+        C3 = C2+ldc;
+        C4 = C3+ldc;
+        C5 = C4+ldc;
+        C6 = C5+ldc;
+        C7 = C6+ldc;
+
+
+        if (left) {
+            off = offset;
+        }
+
+        ptrba = ba;
+
+        for (i=0; i<bm/4; i+=1) // do blocks of 4x4
+	{
+
+		ptrbb = bb;
+                if (backwards)
+                {
+		   ptrba += off*4; // number of values in A
+		   ptrbb += off*8; // number of values in B
+                }
+
+		res0_0 = 0;
+		res0_1 = 0;
+		res0_2 = 0;
+		res0_3 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+		res1_2 = 0;
+		res1_3 = 0;
+
+		res2_0 = 0;
+		res2_1 = 0;
+		res2_2 = 0;
+		res2_3 = 0;
+
+		res3_0 = 0;
+		res3_1 = 0;
+		res3_2 = 0;
+		res3_3 = 0;
+
+		res4_0 = 0;
+		res4_1 = 0;
+		res4_2 = 0;
+		res4_3 = 0;
+
+		res5_0 = 0;
+		res5_1 = 0;
+		res5_2 = 0;
+		res5_3 = 0;
+
+		res6_0 = 0;
+		res6_1 = 0;
+		res6_2 = 0;
+		res6_3 = 0;
+
+		res7_0 = 0;
+		res7_1 = 0;
+		res7_2 = 0;
+		res7_3 = 0;
+
+                temp = backwards ? bk-off :
+                             left ? off + 4 : // number of values in A
+                                    off + 8;  // number of values in B
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+			b4 = ptrbb[4];
+			b5 = ptrbb[5];
+			b6 = ptrbb[6];
+			b7 = ptrbb[7];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+			res4_0 += a0*b4;
+			res5_0 += a0*b5;
+			res6_0 += a0*b6;
+			res7_0 += a0*b7;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+			res2_1 += a1*b2;
+			res3_1 += a1*b3;
+			res4_1 += a1*b4;
+			res5_1 += a1*b5;
+			res6_1 += a1*b6;
+			res7_1 += a1*b7;
+
+			a0 = ptrba[2];
+			res0_2 += a0*b0;
+			res1_2 += a0*b1;
+			res2_2 += a0*b2;
+			res3_2 += a0*b3;
+			res4_2 += a0*b4;
+			res5_2 += a0*b5;
+			res6_2 += a0*b6;
+			res7_2 += a0*b7;
+
+			a1 = ptrba[3];
+			res0_3 += a1*b0;
+			res1_3 += a1*b1;
+			res2_3 += a1*b2;
+			res3_3 += a1*b3;
+			res4_3 += a1*b4;
+			res5_3 += a1*b5;
+			res6_3 += a1*b6;
+			res7_3 += a1*b7;
+
+			ptrba = ptrba+4;
+			ptrbb = ptrbb+8;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+		res0_2 *= alpha;
+		res0_3 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+		res1_2 *= alpha;
+		res1_3 *= alpha;
+
+		res2_0 *= alpha;
+		res2_1 *= alpha;
+		res2_2 *= alpha;
+		res2_3 *= alpha;
+
+		res3_0 *= alpha;
+		res3_1 *= alpha;
+		res3_2 *= alpha;
+		res3_3 *= alpha;
+
+		res4_0 *= alpha;
+		res4_1 *= alpha;
+		res4_2 *= alpha;
+		res4_3 *= alpha;
+
+		res5_0 *= alpha;
+		res5_1 *= alpha;
+		res5_2 *= alpha;
+		res5_3 *= alpha;
+
+		res6_0 *= alpha;
+		res6_1 *= alpha;
+		res6_2 *= alpha;
+		res6_3 *= alpha;
+
+		res7_0 *= alpha;
+		res7_1 *= alpha;
+		res7_2 *= alpha;
+		res7_3 *= alpha;
+
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+		C0[2] = res0_2;
+		C0[3] = res0_3;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+		C1[2] = res1_2;
+		C1[3] = res1_3;
+
+		C2[0] = res2_0;
+		C2[1] = res2_1;
+		C2[2] = res2_2;
+		C2[3] = res2_3;
+
+		C3[0] = res3_0;
+		C3[1] = res3_1;
+		C3[2] = res3_2;
+		C3[3] = res3_3;
+
+		C4[0] = res4_0;
+		C4[1] = res4_1;
+		C4[2] = res4_2;
+		C4[3] = res4_3;
+
+		C5[0] = res5_0;
+		C5[1] = res5_1;
+		C5[2] = res5_2;
+		C5[3] = res5_3;
+
+		C6[0] = res6_0;
+		C6[1] = res6_1;
+		C6[2] = res6_2;
+		C6[3] = res6_3;
+
+		C7[0] = res7_0;
+		C7[1] = res7_1;
+		C7[2] = res7_2;
+		C7[3] = res7_3;
+
+		if (!backwards) {
+                    temp = bk-off;
+                    temp = left ? temp - 4 : // number of values in A
+                                  temp - 8;  // number of values in B
+
+                    ptrba += temp*4; // number of values in A
+		    ptrbb += temp*8; // number of values in B
+                }
+#ifdef LEFT
+		off += 4; // number of values in A
+#endif
+
+		C0 = C0+4;
+		C1 = C1+4;
+		C2 = C2+4;
+		C3 = C3+4;
+		C4 = C4+4;
+		C5 = C5+4;
+		C6 = C6+4;
+		C7 = C7+4;
+
+	}
+
+	if ( bm & 2 ) // do any 2x4 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*2;
+		ptrbb = bb + off*8;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+
+		res2_0 = 0;
+		res2_1 = 0;
+
+		res3_0 = 0;
+		res3_1 = 0;
+
+		res4_0 = 0;
+		res4_1 = 0;
+
+		res5_0 = 0;
+		res5_1 = 0;
+
+		res6_0 = 0;
+		res6_1 = 0;
+
+		res7_0 = 0;
+		res7_1 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+2;	// number of values in A
+#else
+		temp = off+8;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+			b4 = ptrbb[4];
+			b5 = ptrbb[5];
+			b6 = ptrbb[6];
+			b7 = ptrbb[7];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+			res4_0 += a0*b4;
+			res5_0 += a0*b5;
+			res6_0 += a0*b6;
+			res7_0 += a0*b7;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+			res2_1 += a1*b2;
+			res3_1 += a1*b3;
+			res4_1 += a1*b4;
+			res5_1 += a1*b5;
+			res6_1 += a1*b6;
+			res7_1 += a1*b7;
+
+			ptrba = ptrba+2;
+			ptrbb = ptrbb+8;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+
+		res2_0 *= alpha;
+		res2_1 *= alpha;
+
+		res3_0 *= alpha;
+		res3_1 *= alpha;
+
+		res4_0 *= alpha;
+		res4_1 *= alpha;
+
+		res5_0 *= alpha;
+		res5_1 *= alpha;
+
+		res6_0 *= alpha;
+		res6_1 *= alpha;
+
+		res7_0 *= alpha;
+		res7_1 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+
+		C2[0] = res2_0;
+		C2[1] = res2_1;
+
+		C3[0] = res3_0;
+		C3[1] = res3_1;
+
+		C4[0] = res4_0;
+		C4[1] = res4_1;
+
+		C5[0] = res5_0;
+		C5[1] = res5_1;
+
+		C6[0] = res6_0;
+		C6[1] = res6_1;
+
+		C7[0] = res7_0;
+		C7[1] = res7_1;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 2; // number of values in A
+#else
+		temp -= 8; // number of values in B
+#endif
+		ptrba += temp*2;
+		ptrbb += temp*8;
+#endif
+
+#ifdef LEFT
+		off += 2; // number of values in A
+#endif
+
+		C0 = C0+2;
+		C1 = C1+2;
+		C2 = C2+2;
+		C3 = C3+2;
+		C4 = C4+2;
+		C5 = C5+2;
+		C6 = C6+2;
+		C7 = C7+2;
+
+	}
+
+	if ( bm & 1 ) // do any 1x4 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*1;
+		ptrbb = bb + off*8;
+#endif
+
+		res0_0 = 0;
+		res1_0 = 0;
+		res2_0 = 0;
+		res3_0 = 0;
+		res4_0 = 0;
+		res5_0 = 0;
+		res6_0 = 0;
+		res7_0 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+1;	// number of values in A
+#else
+		temp = off+8;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+			b4 = ptrbb[4];
+			b5 = ptrbb[5];
+			b6 = ptrbb[6];
+			b7 = ptrbb[7];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+			res4_0 += a0*b4;
+			res5_0 += a0*b5;
+			res6_0 += a0*b6;
+			res7_0 += a0*b7;
+
+			ptrba = ptrba+1;
+			ptrbb = ptrbb+8;
+                }
+
+		res0_0 *= alpha;
+
+		res1_0 *= alpha;
+
+		res2_0 *= alpha;
+
+		res3_0 *= alpha;
+		res4_0 *= alpha;
+		res5_0 *= alpha;
+		res6_0 *= alpha;
+		res7_0 *= alpha;
+
+		C0[0] = res0_0;
+
+		C1[0] = res1_0;
+
+		C2[0] = res2_0;
+
+		C3[0] = res3_0;
+		C4[0] = res4_0;
+		C5[0] = res5_0;
+		C6[0] = res6_0;
+		C7[0] = res7_0;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 1; // number of values in A
+#else
+		temp -= 8; // number of values in B
+#endif
+		ptrba += temp*1;
+		ptrbb += temp*8;
+#endif
+
+#ifdef LEFT
+		off += 1; // number of values in A
+#endif
+
+		C0 = C0+1;
+		C1 = C1+1;
+		C2 = C2+1;
+		C3 = C3+1;
+		C4 = C4+1;
+		C5 = C5+1;
+		C6 = C6+1;
+		C7 = C7+1;
+
+	}
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+		off += 8;
+#endif
+
+        k = (bk<<3);
+        bb = bb+k;
+        i = (ldc<<3);
+        C = C+i;
+    }
+
+
+
+   for (j=0; j<(bn&4); j+=4) // do blocks of the Mx4 loops 
+   {
+        C0 = C;
+        C1 = C0+ldc;
+        C2 = C1+ldc;
+        C3 = C2+ldc;
+
+
+        if (left) {
+            off = offset;
+        }
+
+        ptrba = ba;
+
+        for (i=0; i<bm/4; i+=1) // do blocks of 4x4
+	{
+
+		ptrbb = bb;
+                if (backwards)
+                {
+		   ptrba += off*4; // number of values in A
+		   ptrbb += off*4; // number of values in B
+                }
+
+		res0_0 = 0;
+		res0_1 = 0;
+		res0_2 = 0;
+		res0_3 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+		res1_2 = 0;
+		res1_3 = 0;
+
+		res2_0 = 0;
+		res2_1 = 0;
+		res2_2 = 0;
+		res2_3 = 0;
+
+		res3_0 = 0;
+		res3_1 = 0;
+		res3_2 = 0;
+		res3_3 = 0;
+
+                temp = backwards ? bk-off :
+                             left ? off + 4 : // number of values in A
+                                    off + 4;  // number of values in B
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+			res2_1 += a1*b2;
+			res3_1 += a1*b3;
+
+			a0 = ptrba[2];
+			res0_2 += a0*b0;
+			res1_2 += a0*b1;
+			res2_2 += a0*b2;
+			res3_2 += a0*b3;
+
+			a1 = ptrba[3];
+			res0_3 += a1*b0;
+			res1_3 += a1*b1;
+			res2_3 += a1*b2;
+			res3_3 += a1*b3;
+
+			ptrba = ptrba+4;
+			ptrbb = ptrbb+4;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+		res0_2 *= alpha;
+		res0_3 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+		res1_2 *= alpha;
+		res1_3 *= alpha;
+
+		res2_0 *= alpha;
+		res2_1 *= alpha;
+		res2_2 *= alpha;
+		res2_3 *= alpha;
+
+		res3_0 *= alpha;
+		res3_1 *= alpha;
+		res3_2 *= alpha;
+		res3_3 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+		C0[2] = res0_2;
+		C0[3] = res0_3;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+		C1[2] = res1_2;
+		C1[3] = res1_3;
+
+		C2[0] = res2_0;
+		C2[1] = res2_1;
+		C2[2] = res2_2;
+		C2[3] = res2_3;
+
+		C3[0] = res3_0;
+		C3[1] = res3_1;
+		C3[2] = res3_2;
+		C3[3] = res3_3;
+
+		if (!backwards) {
+                    temp = bk-off;
+                    temp = left ? temp - 4 : // number of values in A
+                                  temp - 4;  // number of values in B
+
+                    ptrba += temp*4; // number of values in A
+		    ptrbb += temp*4; // number of values in B
+                }
+#ifdef LEFT
+		off += 4; // number of values in A
+#endif
+
+		C0 = C0+4;
+		C1 = C1+4;
+		C2 = C2+4;
+		C3 = C3+4;
+
+	}
+
+	if ( bm & 2 ) // do any 2x4 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*2;
+		ptrbb = bb + off*4;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+
+		res2_0 = 0;
+		res2_1 = 0;
+
+		res3_0 = 0;
+		res3_1 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+2;	// number of values in A
+#else
+		temp = off+4;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+			res2_1 += a1*b2;
+			res3_1 += a1*b3;
+
+			ptrba = ptrba+2;
+			ptrbb = ptrbb+4;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+
+		res2_0 *= alpha;
+		res2_1 *= alpha;
+
+		res3_0 *= alpha;
+		res3_1 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+
+		C2[0] = res2_0;
+		C2[1] = res2_1;
+
+		C3[0] = res3_0;
+		C3[1] = res3_1;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 2; // number of values in A
+#else
+		temp -= 4; // number of values in B
+#endif
+		ptrba += temp*2;
+		ptrbb += temp*4;
+#endif
+
+#ifdef LEFT
+		off += 2; // number of values in A
+#endif
+
+		C0 = C0+2;
+		C1 = C1+2;
+		C2 = C2+2;
+		C3 = C3+2;
+
+	}
+
+	if ( bm & 1 ) // do any 1x4 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*1;
+		ptrbb = bb + off*4;
+#endif
+
+		res0_0 = 0;
+		res1_0 = 0;
+		res2_0 = 0;
+		res3_0 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+1;	// number of values in A
+#else
+		temp = off+4;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+
+			ptrba = ptrba+1;
+			ptrbb = ptrbb+4;
+                }
+
+		res0_0 *= alpha;
+
+		res1_0 *= alpha;
+
+		res2_0 *= alpha;
+
+		res3_0 *= alpha;
+
+		C0[0] = res0_0;
+
+		C1[0] = res1_0;
+
+		C2[0] = res2_0;
+
+		C3[0] = res3_0;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 1; // number of values in A
+#else
+		temp -= 4; // number of values in B
+#endif
+		ptrba += temp*1;
+		ptrbb += temp*4;
+#endif
+
+#ifdef LEFT
+		off += 1; // number of values in A
+#endif
+
+		C0 = C0+1;
+		C1 = C1+1;
+		C2 = C2+1;
+		C3 = C3+1;
+
+	}
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+		off += 4;
+#endif
+
+        k = (bk<<2);
+        bb = bb+k;
+        i = (ldc<<2);
+        C = C+i;
+    }
+
+
+
+   for (j=0; j<(bn&2); j+=2) // do the Mx2 loops 
+   {
+        C0 = C;
+        C1 = C0+ldc;
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+		off = offset;
+#endif
+
+
+        ptrba = ba;
+
+        for (i=0; i<bm/4; i+=1) // do blocks of 4x2
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*4;
+		ptrbb = bb + off*2;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+		res0_2 = 0;
+		res0_3 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+		res1_2 = 0;
+		res1_3 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+4;	// number of values in A
+#else
+		temp = off+2;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+
+			a0 = ptrba[2];
+			res0_2 += a0*b0;
+			res1_2 += a0*b1;
+
+			a1 = ptrba[3];
+			res0_3 += a1*b0;
+			res1_3 += a1*b1;
+
+			ptrba = ptrba+4;
+			ptrbb = ptrbb+2;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+		res0_2 *= alpha;
+		res0_3 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+		res1_2 *= alpha;
+		res1_3 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+		C0[2] = res0_2;
+		C0[3] = res0_3;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+		C1[2] = res1_2;
+		C1[3] = res1_3;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 4; // number of values in A
+#else
+		temp -= 2; // number of values in B
+#endif
+		ptrba += temp*4;
+		ptrbb += temp*2;
+#endif
+
+#ifdef LEFT
+		off += 4; // number of values in A
+#endif
+
+		C0 = C0+4;
+		C1 = C1+4;
+
+	}
+
+	if ( bm & 2 ) // do any 2x2 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*2;
+		ptrbb = bb + off*2;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+2;	// number of values in A
+#else
+		temp = off+2;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+
+			ptrba = ptrba+2;
+			ptrbb = ptrbb+2;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 2; // number of values in A
+#else
+		temp -= 2; // number of values in B
+#endif
+		ptrba += temp*2;
+		ptrbb += temp*2;
+#endif
+
+#ifdef LEFT
+		off += 2; // number of values in A
+#endif
+
+		C0 = C0+2;
+		C1 = C1+2;
+
+	}
+
+	if ( bm & 1 ) // do any 1x2 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*1;
+		ptrbb = bb + off*2;
+#endif
+
+		res0_0 = 0;
+
+		res1_0 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+1;	// number of values in A
+#else
+		temp = off+2;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+
+			ptrba = ptrba+1;
+			ptrbb = ptrbb+2;
+                }
+
+		res0_0 *= alpha;
+
+		res1_0 *= alpha;
+
+		C0[0] = res0_0;
+
+		C1[0] = res1_0;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 1; // number of values in A
+#else
+		temp -= 2; // number of values in B
+#endif
+		ptrba += temp*1;
+		ptrbb += temp*2;
+#endif
+
+#ifdef LEFT
+		off += 1; // number of values in A
+#endif
+
+		C0 = C0+1;
+		C1 = C1+1;
+
+	}
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+		off += 2;
+#endif
+
+        k = (bk<<1);
+        bb = bb+k;
+        i = (ldc<<1);
+        C = C+i;
+    }
+
+
+
+
+
+
+
+   for (j=0; j<(bn&1); j+=1) // do the Mx1 loops
+   {
+        C0 = C;
+
+#if defined(TRMMKERNEL) &&  defined(LEFT)
+	off = offset;
+#endif
+
+        ptrba = ba;
+
+        for (i=0; i<bm/4; i+=1) // do blocks of 4x1 loops
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*4;
+		ptrbb = bb + off*1;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+		res0_2 = 0;
+		res0_3 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+4;	// number of values in A
+#else
+		temp = off+1;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+
+			a0 = ptrba[2];
+			res0_2 += a0*b0;
+
+			a1 = ptrba[3];
+			res0_3 += a1*b0;
+
+			ptrba = ptrba+4;
+			ptrbb = ptrbb+1;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+		res0_2 *= alpha;
+		res0_3 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+		C0[2] = res0_2;
+		C0[3] = res0_3;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 4; // number of values in A
+#else
+		temp -= 1; // number of values in B
+#endif
+		ptrba += temp*4;
+		ptrbb += temp*1;
+#endif
+
+#ifdef LEFT
+		off += 4; // number of values in A
+#endif
+
+		C0 = C0+4;
+
+	}
+
+	if ( bm & 2 ) // do any 2x1 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*2;
+		ptrbb = bb + off*1;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+2;	// number of values in A
+#else
+		temp = off+1;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+
+			ptrba = ptrba+2;
+			ptrbb = ptrbb+1;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 2; // number of values in A
+#else
+		temp -= 1; // number of values in B
+#endif
+		ptrba += temp*2;
+		ptrbb += temp*1;
+#endif
+
+#ifdef LEFT
+		off += 2; // number of values in A
+#endif
+
+		C0 = C0+2;
+
+	}
+
+	if ( bm & 1 ) // do any 1x1 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*1;
+		ptrbb = bb + off*1;
+#endif
+
+		res0_0 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+1;	// number of values in A
+#else
+		temp = off+1;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+
+			ptrba = ptrba+1;
+			ptrbb = ptrbb+1;
+                }
+
+		res0_0 *= alpha;
+
+		C0[0] = res0_0;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 1; // number of values in A
+#else
+		temp -= 1; // number of values in B
+#endif
+		ptrba += temp*1;
+		ptrbb += temp*1;
+#endif
+
+#ifdef LEFT
+		off += 1; // number of values in A
+#endif
+
+		C0 = C0+1;
+
+	}
+
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+		off += 1;
+#endif
+
+        k = (bk<<0);
+        bb = bb+k;
+        C = C+ldc;
+   }
+   return 0;
+}
diff --git a/kernel/generic/zgemm3mkernel_dump.c b/kernel/generic/zgemm3mkernel_dump.c
new file mode 100644
index 000000000..a59bb08ce
--- /dev/null
+++ b/kernel/generic/zgemm3mkernel_dump.c
@@ -0,0 +1,34 @@
+/***************************************************************************
+Copyright (c) 2011-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+int CNAME(BLASLONG bm, BLASLONG bn, BLASLONG bk, FLOAT alphar, FLOAT alphai, FLOAT * ba, FLOAT * bb, FLOAT * C, BLASLONG ldc)
+{
+  return 0;
+}
diff --git a/kernel/generic/zimatcopy_cn.c b/kernel/generic/zimatcopy_cn.c
new file mode 100644
index 000000000..3d108ac65
--- /dev/null
+++ b/kernel/generic/zimatcopy_cn.c
@@ -0,0 +1,67 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a , BLASLONG lda)
+{
+	BLASLONG i,j,ia;
+	FLOAT *aptr;
+    FLOAT a0, a1; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+    if ( alpha_r == 1.0 && alpha_i == 0.0) return(0); 
+
+	aptr = a;
+
+	lda *= 2;
+
+	for ( i=0; i<cols ; i++ )
+	{
+		ia = 0;
+
+		for(j=0; j<rows; j++)
+		{
+            a0 = aptr[ia]; 
+            a1 = aptr[ia+1]; 
+			aptr[ia]   = alpha_r * a0   - alpha_i * a1;
+			aptr[ia+1] = alpha_r * a1   + alpha_i * a0;
+			ia+=2;
+		}
+		aptr += lda;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/zimatcopy_cnc.c b/kernel/generic/zimatcopy_cnc.c
new file mode 100644
index 000000000..7fffe193a
--- /dev/null
+++ b/kernel/generic/zimatcopy_cnc.c
@@ -0,0 +1,67 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon
+ *
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j,ia;
+	FLOAT *aptr,*bptr; 
+    FLOAT a0, a1;
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+    if ( alpha_r == 1.0 && alpha_i == 0.0 ) return (0); 
+
+	aptr = a;
+	lda *= 2;
+
+	for ( i=0; i<cols ; i++ )
+	{
+		ia = 0;
+
+		for(j=0; j<rows; j++)
+		{
+            a0 = aptr[ia]; 
+            a1 = aptr[ia+1]; 
+			aptr[ia]   =   alpha_r * a0   + alpha_i * a1;
+			aptr[ia+1] = - alpha_r * a1   + alpha_i * a0;
+			ia += 2;
+		}
+		aptr += lda;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/zimatcopy_ct.c b/kernel/generic/zimatcopy_ct.c
new file mode 100644
index 000000000..99bff830d
--- /dev/null
+++ b/kernel/generic/zimatcopy_ct.c
@@ -0,0 +1,82 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon
+ *
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j,ia,ib;
+	FLOAT *aptr,*bptr;
+    FLOAT t0, t1; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+	aptr = a;
+
+	lda *= 2;
+	ib = 0;
+	for ( i=0; i<cols ; i++ )
+	{
+		bptr = &a[ib+i*lda];
+		ia = 2*i;
+        
+        /* Diagonal Element  */
+        t0 = bptr[0]; 
+        t1 = bptr[1]; 
+		bptr[0]   = alpha_r * t0   - alpha_i * t1;
+		bptr[1]   = alpha_r * t1   + alpha_i * t0;
+
+        bptr +=lda; 
+        ia += 2; 
+
+		for(j=i+1; j<rows; j++)
+		{
+            t0 = bptr[0]; 
+            t1 = bptr[1]; 
+			bptr[0]   = alpha_r * aptr[ia]   - alpha_i * aptr[ia+1];
+			bptr[1]   = alpha_r * aptr[ia+1] + alpha_i * aptr[ia];
+    		aptr[ia]   = alpha_r * t0   - alpha_i * t1;
+			aptr[ia+1] = alpha_r * t1   + alpha_i * t0;
+
+			ia += 2;
+			bptr += lda;
+		}
+		aptr += lda;
+		ib += 2;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/zimatcopy_ctc.c b/kernel/generic/zimatcopy_ctc.c
new file mode 100644
index 000000000..ca5c05533
--- /dev/null
+++ b/kernel/generic/zimatcopy_ctc.c
@@ -0,0 +1,85 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2014/06/09 Saar
+ *
+ * Order ColMajor
+ * Trans, conjugate
+ *
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j,ia,ib;
+	FLOAT *aptr,*bptr;
+    FLOAT t0, t1; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+	aptr = a;
+
+	lda *= 2;
+	ib = 0;
+	for ( i=0; i<cols ; i++ )
+	{
+        /* Start on the diagonal  */
+		bptr = &a[ib+i*lda];
+		ia = 2*i;
+
+        /* Diagonal Element  */
+        t0 = bptr[0]; 
+        t1 = bptr[1]; 
+		bptr[0]   =   alpha_r * t0   + alpha_i * t1;
+		bptr[1]   = - alpha_r * t1   + alpha_i * t0;
+        ia += 2; 
+        bptr += lda; 
+
+		for(j=i+1; j<rows; j++)
+		{
+            t0 = bptr[0]; 
+            t1 = bptr[1]; 
+			bptr[0]   =   alpha_r * aptr[ia]   + alpha_i * aptr[ia+1];
+			bptr[1]   = - alpha_r * aptr[ia+1] + alpha_i * aptr[ia];
+        	aptr[ia]   =   alpha_r * t0   + alpha_i * t1;
+			aptr[ia+1] = - alpha_r * t1   + alpha_i * t0;
+
+			ia += 2;
+			bptr += lda;
+		}
+		aptr += lda;
+		ib += 2;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/zimatcopy_rn.c b/kernel/generic/zimatcopy_rn.c
new file mode 100644
index 000000000..95819bdf6
--- /dev/null
+++ b/kernel/generic/zimatcopy_rn.c
@@ -0,0 +1,66 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon 
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j,ia;
+	FLOAT *aptr,*bptr;
+    FLOAT a0, a1; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+	aptr = a;
+
+	lda *=2;
+
+	for ( i=0; i<rows ; i++ )
+	{
+		ia = 0;
+
+		for(j=0; j<cols; j++)
+		{
+            a0 = aptr[ia]; 
+            a1 = aptr[ia+1];
+			aptr[ia]   = alpha_r * a0   - alpha_i * a1;
+			aptr[ia+1] = alpha_r * a1   + alpha_i * a0;
+			ia += 2;
+		}
+		aptr += lda;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/zimatcopy_rnc.c b/kernel/generic/zimatcopy_rnc.c
new file mode 100644
index 000000000..c95644987
--- /dev/null
+++ b/kernel/generic/zimatcopy_rnc.c
@@ -0,0 +1,65 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon 
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j,ia;
+	FLOAT *aptr,*bptr;
+    FLOAT a0, a1; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+	aptr = a;
+
+	lda *=2;
+
+	for ( i=0; i<rows ; i++ )
+	{
+		ia = 0;
+		for(j=0; j<cols; j++)
+		{
+            a0 = aptr[ia]; 
+            a1 = aptr[ia+1]; 
+			aptr[ia]   =   alpha_r * a0  + alpha_i * a1;
+			aptr[ia+1] = - alpha_r * a1  + alpha_i * a0;
+			ia += 2;
+		}
+		aptr += lda;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/zimatcopy_rt.c b/kernel/generic/zimatcopy_rt.c
new file mode 100644
index 000000000..f9fad04bc
--- /dev/null
+++ b/kernel/generic/zimatcopy_rt.c
@@ -0,0 +1,80 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon 
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j,ia,ib;
+	FLOAT *aptr,*bptr;
+    FLOAT t0, t1; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+    aptr = a;
+	lda *= 2;
+	ib = 0;
+	for ( i=0; i<rows ; i++ )
+	{
+		bptr = &a[ib+i*lda];
+		ia = 2*i;
+        
+        /* Diagonal Element  */
+        t0 = bptr[0]; 
+        t1 = bptr[1]; 
+		bptr[0]   = alpha_r * t0   - alpha_i * t1;
+		bptr[1]   = alpha_r * t1   + alpha_i * t0;
+
+        bptr +=lda; 
+        ia += 2; 
+
+		for(j=i+1; j<cols; j++)
+		{
+            t0 = bptr[0]; 
+            t1 = bptr[1]; 
+			bptr[0]   = alpha_r * aptr[ia]   - alpha_i * aptr[ia+1];
+			bptr[1]   = alpha_r * aptr[ia+1] + alpha_i * aptr[ia];
+    		aptr[ia]   = alpha_r * t0   - alpha_i * t1;
+			aptr[ia+1] = alpha_r * t1   + alpha_i * t0;
+
+			ia += 2;
+			bptr += lda;
+		}
+		aptr += lda;
+		ib += 2;
+	}
+
+	return(0);
+
+}
+
+
diff --git a/kernel/generic/zimatcopy_rtc.c b/kernel/generic/zimatcopy_rtc.c
new file mode 100644
index 000000000..a93ea5de7
--- /dev/null
+++ b/kernel/generic/zimatcopy_rtc.c
@@ -0,0 +1,82 @@
+/***************************************************************************
+Copyright (c) 2013, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+/*****************************************************
+ * 2015-09-07 grisuthedragon 
+ *
+******************************************************/
+
+int CNAME(BLASLONG rows, BLASLONG cols, FLOAT alpha_r, FLOAT alpha_i, FLOAT *a, BLASLONG lda)
+{
+	BLASLONG i,j,ia,ib;
+	FLOAT *aptr,*bptr;
+    FLOAT t0, t1; 
+
+	if ( rows <= 0     )  return(0);
+	if ( cols <= 0     )  return(0);
+
+    aptr = a;
+
+	lda *= 2;
+	ib = 0;
+	for ( i=0; i<rows ; i++ )
+	{
+        /* Start on the diagonal  */
+		bptr = &a[ib+i*lda];
+		ia = 2*i;
+
+        /* Diagonal Element  */
+        t0 = bptr[0]; 
+        t1 = bptr[1]; 
+		bptr[0]   =   alpha_r * t0   + alpha_i * t1;
+		bptr[1]   = - alpha_r * t1   + alpha_i * t0;
+        ia += 2; 
+        bptr += lda; 
+
+		for(j=i+1; j<cols; j++)
+		{
+            t0 = bptr[0]; 
+            t1 = bptr[1]; 
+			bptr[0]   =   alpha_r * aptr[ia]   + alpha_i * aptr[ia+1];
+			bptr[1]   = - alpha_r * aptr[ia+1] + alpha_i * aptr[ia];
+        	aptr[ia]   =   alpha_r * t0   + alpha_i * t1;
+			aptr[ia+1] = - alpha_r * t1   + alpha_i * t0;
+
+			ia += 2;
+			bptr += lda;
+		}
+		aptr += lda;
+		ib += 2;
+	}
+
+    return(0);
+
+}
+
+
diff --git a/kernel/power/axpy.S b/kernel/power/axpy.S
index 190f82d6b..fb9789da4 100644
--- a/kernel/power/axpy.S
+++ b/kernel/power/axpy.S
@@ -106,7 +106,7 @@
 	stfd	f24,   80(SP)
 
 #if (defined(_AIX) || defined(__APPLE__)) && !defined(__64BIT__) && defined(DOUBLE)
-	lwz	INCY,    56 + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 	fmr	ALPHA, f1
diff --git a/kernel/power/axpy_ppc440.S b/kernel/power/axpy_ppc440.S
index df3f25e5f..81a660e4d 100644
--- a/kernel/power/axpy_ppc440.S
+++ b/kernel/power/axpy_ppc440.S
@@ -104,7 +104,7 @@
 	stfd	f24,   80(SP)
 
 #if (defined(_AIX) || defined(__APPLE__)) && !defined(__64BIT__) && defined(DOUBLE)
-	lwz	INCY,    56 + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 	fmr	ALPHA, f1
diff --git a/kernel/power/gemm_beta.S b/kernel/power/gemm_beta.S
index 969f54c61..62d7761ec 100644
--- a/kernel/power/gemm_beta.S
+++ b/kernel/power/gemm_beta.S
@@ -64,24 +64,24 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	LDC,      8 + STACKSIZE(SP)
+	lwz	LDC,    FRAMESLOT(0) + STACKSIZE(SP)
 #else
-	ld	C,      112 + STACKSIZE(SP)
-	ld	LDC,    120 + STACKSIZE(SP)
+	ld	C,      FRAMESLOT(0) + STACKSIZE(SP)
+	ld	LDC,    FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	C,      112 + STACKSIZE(SP)
-	ld	LDC,    120 + STACKSIZE(SP)
+	ld	C,      FRAMESLOT(0) + STACKSIZE(SP)
+	ld	LDC,    FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	C,      FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC,    FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	C,       56 + STACKSIZE(SP)
-	lwz	LDC,     60 + STACKSIZE(SP)
+	lwz	C,      FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	LDC,    FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/gemm_kernel.S b/kernel/power/gemm_kernel.S
index cae2fabca..e5e9ec346 100644
--- a/kernel/power/gemm_kernel.S
+++ b/kernel/power/gemm_kernel.S
@@ -179,7 +179,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
@@ -187,17 +187,17 @@
 
 #if defined(TRMMKERNEL)
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,   FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -231,29 +231,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/gemm_kernel_altivec.S b/kernel/power/gemm_kernel_altivec.S
index 8a525ef22..6c7e78319 100644
--- a/kernel/power/gemm_kernel_altivec.S
+++ b/kernel/power/gemm_kernel_altivec.S
@@ -241,7 +241,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/gemm_kernel_altivec_cell.S b/kernel/power/gemm_kernel_altivec_cell.S
index ac750c2e8..b7445a1f6 100644
--- a/kernel/power/gemm_kernel_altivec_cell.S
+++ b/kernel/power/gemm_kernel_altivec_cell.S
@@ -247,7 +247,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/gemm_kernel_altivec_g4.S b/kernel/power/gemm_kernel_altivec_g4.S
index 26339afeb..548150143 100644
--- a/kernel/power/gemm_kernel_altivec_g4.S
+++ b/kernel/power/gemm_kernel_altivec_g4.S
@@ -241,7 +241,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/gemm_kernel_cell.S b/kernel/power/gemm_kernel_cell.S
index 1dbacc7f9..f3d3b8325 100644
--- a/kernel/power/gemm_kernel_cell.S
+++ b/kernel/power/gemm_kernel_cell.S
@@ -185,7 +185,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
@@ -193,17 +193,17 @@
 
 #if defined(TRMMKERNEL)
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET, FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -229,29 +229,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-xc	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+xc	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/gemm_kernel_g4.S b/kernel/power/gemm_kernel_g4.S
index b6c849965..259f04c4e 100644
--- a/kernel/power/gemm_kernel_g4.S
+++ b/kernel/power/gemm_kernel_g4.S
@@ -177,7 +177,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
@@ -185,17 +185,17 @@
 
 #if defined(TRMMKERNEL)
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET, FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/gemm_kernel_power3.S b/kernel/power/gemm_kernel_power3.S
index 6fe2def67..4a6b5da62 100644
--- a/kernel/power/gemm_kernel_power3.S
+++ b/kernel/power/gemm_kernel_power3.S
@@ -168,7 +168,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
@@ -190,29 +190,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/gemm_kernel_power6.S b/kernel/power/gemm_kernel_power6.S
index 5f8fa76ce..1a412c4fb 100644
--- a/kernel/power/gemm_kernel_power6.S
+++ b/kernel/power/gemm_kernel_power6.S
@@ -176,7 +176,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
@@ -184,17 +184,17 @@
 
 #if defined(TRMMKERNEL)
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/gemm_kernel_ppc440.S b/kernel/power/gemm_kernel_ppc440.S
index 2e86d5130..b128beb38 100644
--- a/kernel/power/gemm_kernel_ppc440.S
+++ b/kernel/power/gemm_kernel_ppc440.S
@@ -176,7 +176,7 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
@@ -184,17 +184,17 @@
 
 #if defined(TRMMKERNEL)
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET, FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/gemv_n.S b/kernel/power/gemv_n.S
index 2b19f0a4e..77587ecb1 100644
--- a/kernel/power/gemv_n.S
+++ b/kernel/power/gemv_n.S
@@ -248,31 +248,31 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	INCY,      8 + STACKSIZE(SP)
-	lwz	BUFFER,  12 + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	Y,       56 + STACKSIZE(SP)
-	lwz	INCY,    60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/gemv_n_ppc440.S b/kernel/power/gemv_n_ppc440.S
index baedebc2b..beb21200a 100644
--- a/kernel/power/gemv_n_ppc440.S
+++ b/kernel/power/gemv_n_ppc440.S
@@ -201,31 +201,31 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	INCY,      8 + STACKSIZE(SP)
-	lwz	BUFFER,  12 + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	Y,       56 + STACKSIZE(SP)
-	lwz	INCY,    60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/gemv_t.S b/kernel/power/gemv_t.S
index 005e5d56c..817a60b86 100644
--- a/kernel/power/gemv_t.S
+++ b/kernel/power/gemv_t.S
@@ -257,31 +257,31 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	INCY,	  8 + STACKSIZE(SP)
-	lwz	BUFFER,  12 + STACKSIZE(SP)
+	lwz	INCY,	 FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	Y,       56 + STACKSIZE(SP)
-	lwz	INCY,    60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/gemv_t_ppc440.S b/kernel/power/gemv_t_ppc440.S
index 62433af19..6e560db6c 100644
--- a/kernel/power/gemv_t_ppc440.S
+++ b/kernel/power/gemv_t_ppc440.S
@@ -192,31 +192,31 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	INCY,	  8 + STACKSIZE(SP)
-	lwz	BUFFER,  12 + STACKSIZE(SP)
+	lwz	INCY,	 FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	Y,       56 + STACKSIZE(SP)
-	lwz	INCY,    60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/ger.S b/kernel/power/ger.S
index bc10bf40d..fd397ce8c 100644
--- a/kernel/power/ger.S
+++ b/kernel/power/ger.S
@@ -226,31 +226,31 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	LDA,      8 + STACKSIZE(SP)
-	lwz	BUFFER,  12 + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	A,       112 + STACKSIZE(SP)
-	ld	LDA,     120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	A,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	LDA,     FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	INCY,    56 + STACKSIZE(SP)
-	lwz	A,       60 + STACKSIZE(SP)
-	lwz	LDA,     64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	A,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	A,       56 + STACKSIZE(SP)
-	lwz	LDA,     60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	A,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #else
-	ld	A,       112 + STACKSIZE(SP)
-	ld	LDA,     120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	A,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	LDA,     FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/swap.S b/kernel/power/swap.S
index f8b56d472..e862b17bb 100644
--- a/kernel/power/swap.S
+++ b/kernel/power/swap.S
@@ -115,7 +115,7 @@
 	stfd	f31,  136(SP)
 
 #if (defined(_AIX) || defined(__APPLE__)) && !defined(__64BIT__) && defined(DOUBLE)
-	lwz	INCY,    56 + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 	slwi	INCX, INCX, BASE_SHIFT
diff --git a/kernel/power/symv_L.S b/kernel/power/symv_L.S
index fbf735abd..f7d768c50 100644
--- a/kernel/power/symv_L.S
+++ b/kernel/power/symv_L.S
@@ -250,26 +250,26 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	BUFFER,   56 + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
-	ld	INCY,    112 + STACKSIZE(SP)
-	ld	BUFFER,  120 + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	Y,       56 + STACKSIZE(SP)
-	lwz	INCY,    60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	  INCY,  56 + STACKSIZE(SP)
-	lwz	BUFFER,  60 + STACKSIZE(SP)
+	lwz	  INCY,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #else
-	ld	INCY,    112 + STACKSIZE(SP)
-	ld	BUFFER,  120 + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/symv_U.S b/kernel/power/symv_U.S
index ec1aeea39..d8e082397 100644
--- a/kernel/power/symv_U.S
+++ b/kernel/power/symv_U.S
@@ -249,26 +249,26 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	BUFFER,   56 + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
-	ld	INCY,    112 + STACKSIZE(SP)
-	ld	BUFFER,  120 + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	Y,       56 + STACKSIZE(SP)
-	lwz	INCY,    60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	  INCY,  56 + STACKSIZE(SP)
-	lwz	BUFFER,  60 + STACKSIZE(SP)
+	lwz	  INCY,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #else
-	ld	INCY,    112 + STACKSIZE(SP)
-	ld	BUFFER,  120 + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/trsm_kernel_LN.S b/kernel/power/trsm_kernel_LN.S
index 0c13a25a4..7983c573b 100644
--- a/kernel/power/trsm_kernel_LN.S
+++ b/kernel/power/trsm_kernel_LN.S
@@ -174,24 +174,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -239,29 +239,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_LT.S b/kernel/power/trsm_kernel_LT.S
index 06481e5e9..c561fd014 100644
--- a/kernel/power/trsm_kernel_LT.S
+++ b/kernel/power/trsm_kernel_LT.S
@@ -174,24 +174,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -260,29 +260,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_RT.S b/kernel/power/trsm_kernel_RT.S
index 1777ba86d..07b88402c 100644
--- a/kernel/power/trsm_kernel_RT.S
+++ b/kernel/power/trsm_kernel_RT.S
@@ -174,24 +174,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -257,29 +257,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_cell_LN.S b/kernel/power/trsm_kernel_cell_LN.S
index b5ed925ed..803530cbb 100644
--- a/kernel/power/trsm_kernel_cell_LN.S
+++ b/kernel/power/trsm_kernel_cell_LN.S
@@ -174,24 +174,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -234,29 +234,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_cell_LT.S b/kernel/power/trsm_kernel_cell_LT.S
index cdc6f7514..105e7d43c 100644
--- a/kernel/power/trsm_kernel_cell_LT.S
+++ b/kernel/power/trsm_kernel_cell_LT.S
@@ -174,24 +174,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -260,29 +260,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_cell_RT.S b/kernel/power/trsm_kernel_cell_RT.S
index 731f52c19..a54a261cb 100644
--- a/kernel/power/trsm_kernel_cell_RT.S
+++ b/kernel/power/trsm_kernel_cell_RT.S
@@ -174,24 +174,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -234,29 +234,29 @@
 #ifdef linux
 #ifndef __64BIT__
 	mr	PREA,  r10
-	lwz	PREB,   8 + STACKSIZE(SP)
-	lwz	PREC,  12 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  112 + STACKSIZE(SP)
-	ld	PREB,  120 + STACKSIZE(SP)
-	ld	PREC,  128 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   60 + STACKSIZE(SP)
-	lwz	PREB,   64 + STACKSIZE(SP)
-	lwz	PREC,   68 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	PREA,   56 + STACKSIZE(SP)
-	lwz	PREB,   60 + STACKSIZE(SP)
-	lwz	PREC,   64 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_power6_LN.S b/kernel/power/trsm_kernel_power6_LN.S
index 2f85cd14b..937a6761a 100644
--- a/kernel/power/trsm_kernel_power6_LN.S
+++ b/kernel/power/trsm_kernel_power6_LN.S
@@ -173,24 +173,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_power6_LT.S b/kernel/power/trsm_kernel_power6_LT.S
index 6b3d21b14..924f00ec0 100644
--- a/kernel/power/trsm_kernel_power6_LT.S
+++ b/kernel/power/trsm_kernel_power6_LT.S
@@ -174,24 +174,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_power6_RT.S b/kernel/power/trsm_kernel_power6_RT.S
index f6b2e5cfb..40ee5e28d 100644
--- a/kernel/power/trsm_kernel_power6_RT.S
+++ b/kernel/power/trsm_kernel_power6_RT.S
@@ -173,24 +173,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_ppc440_LN.S b/kernel/power/trsm_kernel_ppc440_LN.S
index 265e79e0f..6b7312101 100644
--- a/kernel/power/trsm_kernel_ppc440_LN.S
+++ b/kernel/power/trsm_kernel_ppc440_LN.S
@@ -185,24 +185,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_ppc440_LT.S b/kernel/power/trsm_kernel_ppc440_LT.S
index de7ff7415..28b109b96 100644
--- a/kernel/power/trsm_kernel_ppc440_LT.S
+++ b/kernel/power/trsm_kernel_ppc440_LT.S
@@ -170,24 +170,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/trsm_kernel_ppc440_RT.S b/kernel/power/trsm_kernel_ppc440_RT.S
index e8d202d97..df80cd393 100644
--- a/kernel/power/trsm_kernel_ppc440_RT.S
+++ b/kernel/power/trsm_kernel_ppc440_RT.S
@@ -185,24 +185,24 @@
 
 #if defined(_AIX) || defined(__APPLE__)
 #if !defined(__64BIT__) && defined(DOUBLE)
-	lwz	LDC,    56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 	slwi	LDC, LDC, BASE_SHIFT
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,   112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  112 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   56 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zaxpy.S b/kernel/power/zaxpy.S
index 1acd729ae..ac5b249bb 100644
--- a/kernel/power/zaxpy.S
+++ b/kernel/power/zaxpy.S
@@ -124,19 +124,19 @@
 	stfd	f25,   88(SP)
 
 #if defined(linux) && defined(__64BIT__)
-        ld	INCY, 112 + STACKSIZE(SP)
+        ld	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	INCY,   112 + STACKSIZE(SP)
+	ld	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
+	lwz	INCX, FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	INCY,    56 + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zaxpy_ppc440.S b/kernel/power/zaxpy_ppc440.S
index 1ac232444..b5c604e91 100644
--- a/kernel/power/zaxpy_ppc440.S
+++ b/kernel/power/zaxpy_ppc440.S
@@ -113,19 +113,19 @@
 	stfd	f25,   88(SP)
 
 #if defined(linux) && defined(__64BIT__)
-        ld	INCY, 112 + STACKSIZE(SP)
+        ld	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	INCY,   112 + STACKSIZE(SP)
+	ld	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
+	lwz	INCX, FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	INCY,    56 + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_beta.S b/kernel/power/zgemm_beta.S
index 4a9cbd8bb..43b72ca15 100644
--- a/kernel/power/zgemm_beta.S
+++ b/kernel/power/zgemm_beta.S
@@ -64,24 +64,24 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	LDC,      8 + STACKSIZE(SP)
+	lwz	LDC,    FRAMESLOT(0) + STACKSIZE(SP)
 #else
-	ld	C,      120 + STACKSIZE(SP)
-	ld	LDC,    128 + STACKSIZE(SP)
+	ld	C,      FRAMESLOT(1) + STACKSIZE(SP)
+	ld	LDC,    FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	C,      120 + STACKSIZE(SP)
-	ld	LDC,    128 + STACKSIZE(SP)
+	ld	C,      FRAMESLOT(1) + STACKSIZE(SP)
+	ld	LDC,    FRAMESLOT(2) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	C,       68 + STACKSIZE(SP)
-	lwz	LDC,     72 + STACKSIZE(SP)
+	lwz	C,      FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	LDC,    FRAMESLOT(4) + STACKSIZE(SP)
 #else
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	C,      FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC,    FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel.S b/kernel/power/zgemm_kernel.S
index 3d6689531..8ec8b674a 100644
--- a/kernel/power/zgemm_kernel.S
+++ b/kernel/power/zgemm_kernel.S
@@ -171,37 +171,37 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #ifdef TRMMKERNEL
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -233,25 +233,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREA,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREA,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel_altivec.S b/kernel/power/zgemm_kernel_altivec.S
index 2267e975a..2b650cd02 100644
--- a/kernel/power/zgemm_kernel_altivec.S
+++ b/kernel/power/zgemm_kernel_altivec.S
@@ -240,20 +240,20 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC,   FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC,   FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,     FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,     FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC,   FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC,   FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -266,25 +266,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREB,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREB,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREB,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREB,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREB,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel_altivec_cell.S b/kernel/power/zgemm_kernel_altivec_cell.S
index 9a1407d6e..642d1f2e7 100644
--- a/kernel/power/zgemm_kernel_altivec_cell.S
+++ b/kernel/power/zgemm_kernel_altivec_cell.S
@@ -246,20 +246,20 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC,    FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC,   FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,     FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,     FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC,   FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC,   FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -272,25 +272,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREB,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREB,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREB,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREB,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREB,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREB,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREB,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel_altivec_g4.S b/kernel/power/zgemm_kernel_altivec_g4.S
index 4c774a1e3..0f7a6f9aa 100644
--- a/kernel/power/zgemm_kernel_altivec_g4.S
+++ b/kernel/power/zgemm_kernel_altivec_g4.S
@@ -240,20 +240,20 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC,  FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC,  FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel_cell.S b/kernel/power/zgemm_kernel_cell.S
index 5667b130d..8fd6b0afb 100644
--- a/kernel/power/zgemm_kernel_cell.S
+++ b/kernel/power/zgemm_kernel_cell.S
@@ -177,37 +177,37 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #ifdef TRMMKERNEL
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -232,25 +232,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREA,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREA,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel_g4.S b/kernel/power/zgemm_kernel_g4.S
index af6f88e99..bf6bf77e8 100644
--- a/kernel/power/zgemm_kernel_g4.S
+++ b/kernel/power/zgemm_kernel_g4.S
@@ -187,37 +187,37 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #ifdef TRMMKERNEL
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel_power3.S b/kernel/power/zgemm_kernel_power3.S
index d7d6e2aea..471d3b9ae 100644
--- a/kernel/power/zgemm_kernel_power3.S
+++ b/kernel/power/zgemm_kernel_power3.S
@@ -163,20 +163,20 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -204,25 +204,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREA,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREA,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel_power6.S b/kernel/power/zgemm_kernel_power6.S
index 3f79c0523..3c28649bc 100644
--- a/kernel/power/zgemm_kernel_power6.S
+++ b/kernel/power/zgemm_kernel_power6.S
@@ -201,37 +201,37 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #ifdef TRMMKERNEL
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemm_kernel_ppc440.S b/kernel/power/zgemm_kernel_ppc440.S
index 075fa2b4f..748b69a0c 100644
--- a/kernel/power/zgemm_kernel_ppc440.S
+++ b/kernel/power/zgemm_kernel_ppc440.S
@@ -184,37 +184,37 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #ifdef TRMMKERNEL
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zgemv_n.S b/kernel/power/zgemv_n.S
index ba4685dec..23e0177c0 100644
--- a/kernel/power/zgemv_n.S
+++ b/kernel/power/zgemv_n.S
@@ -247,31 +247,31 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	INCY,	  8 + STACKSIZE(SP)
+	lwz	INCY,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
-	ld	INCX,    112 + STACKSIZE(SP)
-	ld	Y,       120 + STACKSIZE(SP)
-	ld	INCY,    128 + STACKSIZE(SP)
+	ld	INCX,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	Y,     FRAMESLOT(1) + STACKSIZE(SP)
+	ld	INCY,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	LDA,     56 + STACKSIZE(SP)
-	lwz	X,       60 + STACKSIZE(SP)
-	lwz	INCX,    64 + STACKSIZE(SP)
-	lwz	Y,       68 + STACKSIZE(SP)
-	lwz	INCY,    72 + STACKSIZE(SP)
+	lwz	LDA,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	X,     FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCX,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	Y,     FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	INCY,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
+	lwz	INCX,  FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,     FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #else
-	ld	INCX,    112 + STACKSIZE(SP)
-	ld	Y,       120 + STACKSIZE(SP)
-	ld	INCY,    128 + STACKSIZE(SP)
+	ld	INCX,  FRAMESLOT(0) + STACKSIZE(SP)
+	ld	Y,     FRAMESLOT(1) + STACKSIZE(SP)
+	ld	INCY,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/zgemv_n_ppc440.S b/kernel/power/zgemv_n_ppc440.S
index 31e720261..55dd2d84f 100644
--- a/kernel/power/zgemv_n_ppc440.S
+++ b/kernel/power/zgemv_n_ppc440.S
@@ -225,36 +225,36 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	INCY,	  8 + STACKSIZE(SP)
-	lwz	BUFFER,  12 + STACKSIZE(SP)
+	lwz	INCY,	 FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	INCX,    112 + STACKSIZE(SP)
-	ld	Y,       120 + STACKSIZE(SP)
-	ld	INCY,    128 + STACKSIZE(SP)
-	ld	BUFFER,  136 + STACKSIZE(SP)
+	ld	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	LDA,     56 + STACKSIZE(SP)
-	lwz	X,       60 + STACKSIZE(SP)
-	lwz	INCX,    64 + STACKSIZE(SP)
-	lwz	Y,       68 + STACKSIZE(SP)
-	lwz	INCY,    72 + STACKSIZE(SP)
-	lwz	BUFFER,  76 + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	X,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #else
-	ld	INCX,    112 + STACKSIZE(SP)
-	ld	Y,       120 + STACKSIZE(SP)
-	ld	INCY,    128 + STACKSIZE(SP)
-	ld	BUFFER,  136 + STACKSIZE(SP)
+	ld	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/zgemv_t.S b/kernel/power/zgemv_t.S
index bd8ac4043..c0bad3152 100644
--- a/kernel/power/zgemv_t.S
+++ b/kernel/power/zgemv_t.S
@@ -222,36 +222,36 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	INCY,	  8 + STACKSIZE(SP)
-	lwz	BUFFER,  12 + STACKSIZE(SP)
+	lwz	INCY,	 FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	INCX,    112 + STACKSIZE(SP)
-	ld	Y,       120 + STACKSIZE(SP)
-	ld	INCY,    128 + STACKSIZE(SP)
-	ld	BUFFER,  136 + STACKSIZE(SP)
+	ld	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	LDA,     56 + STACKSIZE(SP)
-	lwz	X,       60 + STACKSIZE(SP)
-	lwz	INCX,    64 + STACKSIZE(SP)
-	lwz	Y,       68 + STACKSIZE(SP)
-	lwz	INCY,    72 + STACKSIZE(SP)
-	lwz	BUFFER,  76 + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	X,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #else
-	ld	INCX,    112 + STACKSIZE(SP)
-	ld	Y,       120 + STACKSIZE(SP)
-	ld	INCY,    128 + STACKSIZE(SP)
-	ld	BUFFER,  136 + STACKSIZE(SP)
+	ld	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/zgemv_t_ppc440.S b/kernel/power/zgemv_t_ppc440.S
index 043b9e37b..bfc039a0c 100644
--- a/kernel/power/zgemv_t_ppc440.S
+++ b/kernel/power/zgemv_t_ppc440.S
@@ -181,36 +181,36 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	INCY,	  8 + STACKSIZE(SP)
-	lwz	BUFFER,  12 + STACKSIZE(SP)
+	lwz	INCY,	 FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	INCX,    112 + STACKSIZE(SP)
-	ld	Y,       120 + STACKSIZE(SP)
-	ld	INCY,    128 + STACKSIZE(SP)
-	ld	BUFFER,  136 + STACKSIZE(SP)
+	ld	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	LDA,     56 + STACKSIZE(SP)
-	lwz	X,       60 + STACKSIZE(SP)
-	lwz	INCX,    64 + STACKSIZE(SP)
-	lwz	Y,       68 + STACKSIZE(SP)
-	lwz	INCY,    72 + STACKSIZE(SP)
-	lwz	BUFFER,  76 + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	X,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #else
-	ld	INCX,    112 + STACKSIZE(SP)
-	ld	Y,       120 + STACKSIZE(SP)
-	ld	INCY,    128 + STACKSIZE(SP)
-	ld	BUFFER,  136 + STACKSIZE(SP)
+	ld	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/zger.S b/kernel/power/zger.S
index 01cb90731..a9a607815 100644
--- a/kernel/power/zger.S
+++ b/kernel/power/zger.S
@@ -237,36 +237,36 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	LDA,       8 + STACKSIZE(SP)
-	lwz	BUFFER,   12 + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
-	ld	INCY,    112 + STACKSIZE(SP)
-	ld	A,       120 + STACKSIZE(SP)
-	ld	LDA,     128 + STACKSIZE(SP)
-	ld	BUFFER,  136 + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	A,       FRAMESLOT(1) + STACKSIZE(SP)
+	ld	LDA,     FRAMESLOT(2) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
-	lwz	A,       68 + STACKSIZE(SP)
-	lwz	LDA,     72 + STACKSIZE(SP)
-	lwz	BUFFER,  76 + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	A,       FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	INCY,    56 + STACKSIZE(SP)
-	lwz	A,       60 + STACKSIZE(SP)
-	lwz	LDA,     64 + STACKSIZE(SP)
-	lwz	BUFFER,  68 + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	A,       FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDA,     FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #else
-	ld	INCY,    112 + STACKSIZE(SP)
-	ld	A,       120 + STACKSIZE(SP)
-	ld	LDA,     128 + STACKSIZE(SP)
-	ld	BUFFER,  136 + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(0) + STACKSIZE(SP)
+	ld	A,       FRAMESLOT(1) + STACKSIZE(SP)
+	ld	LDA,     FRAMESLOT(2) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(3) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/zswap.S b/kernel/power/zswap.S
index 048e8ac5f..8befadca2 100644
--- a/kernel/power/zswap.S
+++ b/kernel/power/zswap.S
@@ -118,19 +118,19 @@
 	stfd	f31,  136(SP)
 
 #if defined(linux) && defined(__64BIT__)
-	ld	INCY, 112 + STACKSIZE(SP)
+	ld	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	INCY,   112 + STACKSIZE(SP)
+	ld	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	INCX,    56 + STACKSIZE(SP)
-	lwz	Y,       60 + STACKSIZE(SP)
-	lwz	INCY,    64 + STACKSIZE(SP)
+	lwz	INCX, FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	Y,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	INCY,    56 + STACKSIZE(SP)
+	lwz	INCY, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/zsymv_L.S b/kernel/power/zsymv_L.S
index ad4a8cd5c..b348e328f 100644
--- a/kernel/power/zsymv_L.S
+++ b/kernel/power/zsymv_L.S
@@ -261,31 +261,31 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	BUFFER,   56 + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	X,       56 + STACKSIZE(SP)
-	lwz	INCX,    60 + STACKSIZE(SP)
-	lwz	Y,       64 + STACKSIZE(SP)
-	lwz	INCY,    68 + STACKSIZE(SP)
-	lwz	BUFFER,  72 + STACKSIZE(SP)
+	lwz	X,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
-	lwz	Y,       56 + STACKSIZE(SP)
-	lwz	INCY,    60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/zsymv_U.S b/kernel/power/zsymv_U.S
index 4032b66bb..b631cbe35 100644
--- a/kernel/power/zsymv_U.S
+++ b/kernel/power/zsymv_U.S
@@ -258,31 +258,31 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	BUFFER,   56 + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(0) + STACKSIZE(SP)
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifndef __64BIT__
 #ifdef DOUBLE
-	lwz	X,       56 + STACKSIZE(SP)
-	lwz	INCX,    60 + STACKSIZE(SP)
-	lwz	Y,       64 + STACKSIZE(SP)
-	lwz	INCY,    68 + STACKSIZE(SP)
-	lwz	BUFFER,  72 + STACKSIZE(SP)
+	lwz	X,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCX,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
-	lwz	Y,       56 + STACKSIZE(SP)
-	lwz	INCY,    60 + STACKSIZE(SP)
-	lwz	BUFFER,  64 + STACKSIZE(SP)
+	lwz	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #else
-	ld	Y,       112 + STACKSIZE(SP)
-	ld	INCY,    120 + STACKSIZE(SP)
-	ld	BUFFER,  128 + STACKSIZE(SP)
+	ld	Y,       FRAMESLOT(0) + STACKSIZE(SP)
+	ld	INCY,    FRAMESLOT(1) + STACKSIZE(SP)
+	ld	BUFFER,  FRAMESLOT(2) + STACKSIZE(SP)
 #endif
 #endif
 
diff --git a/kernel/power/ztrsm_kernel_LN.S b/kernel/power/ztrsm_kernel_LN.S
index 64fb96823..87473b45d 100644
--- a/kernel/power/ztrsm_kernel_LN.S
+++ b/kernel/power/ztrsm_kernel_LN.S
@@ -168,36 +168,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -246,25 +246,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREA,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREA,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_LT.S b/kernel/power/ztrsm_kernel_LT.S
index ae4615cf5..db0860124 100644
--- a/kernel/power/ztrsm_kernel_LT.S
+++ b/kernel/power/ztrsm_kernel_LT.S
@@ -168,36 +168,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -249,25 +249,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREA,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREA,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_RT.S b/kernel/power/ztrsm_kernel_RT.S
index f756dda77..c50ab86df 100644
--- a/kernel/power/ztrsm_kernel_RT.S
+++ b/kernel/power/ztrsm_kernel_RT.S
@@ -168,36 +168,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -249,25 +249,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREA,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREA,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_cell_LN.S b/kernel/power/ztrsm_kernel_cell_LN.S
index 2427a4ddd..884a3e864 100644
--- a/kernel/power/ztrsm_kernel_cell_LN.S
+++ b/kernel/power/ztrsm_kernel_cell_LN.S
@@ -174,36 +174,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_cell_LT.S b/kernel/power/ztrsm_kernel_cell_LT.S
index 0d88ded9a..388dfe3c2 100644
--- a/kernel/power/ztrsm_kernel_cell_LT.S
+++ b/kernel/power/ztrsm_kernel_cell_LT.S
@@ -174,36 +174,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
@@ -248,25 +248,25 @@
 
 #ifdef linux
 #ifndef __64BIT__
-	lwz	PREA,   16 + STACKSIZE(SP)
-	lwz	PREC,   20 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(2) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	PREA,  136 + STACKSIZE(SP)
-	ld	PREC,  144 + STACKSIZE(SP)
+	ld	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	ld	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	PREA,   72 + STACKSIZE(SP)
-	lwz	PREC,   76 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(4) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(5) + STACKSIZE(SP)
 #else
-	lwz	PREA,   68 + STACKSIZE(SP)
-	lwz	PREC,   72 + STACKSIZE(SP)
+	lwz	PREA,  FRAMESLOT(3) + STACKSIZE(SP)
+	lwz	PREC,  FRAMESLOT(4) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_cell_RT.S b/kernel/power/ztrsm_kernel_cell_RT.S
index 84f2089fa..00b50fe04 100644
--- a/kernel/power/ztrsm_kernel_cell_RT.S
+++ b/kernel/power/ztrsm_kernel_cell_RT.S
@@ -174,36 +174,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_power6_LN.S b/kernel/power/ztrsm_kernel_power6_LN.S
index 42239bb55..65b8077db 100644
--- a/kernel/power/ztrsm_kernel_power6_LN.S
+++ b/kernel/power/ztrsm_kernel_power6_LN.S
@@ -186,36 +186,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_power6_LT.S b/kernel/power/ztrsm_kernel_power6_LT.S
index dfae4d60b..c27170604 100644
--- a/kernel/power/ztrsm_kernel_power6_LT.S
+++ b/kernel/power/ztrsm_kernel_power6_LT.S
@@ -186,36 +186,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_power6_RT.S b/kernel/power/ztrsm_kernel_power6_RT.S
index 79f8b70b8..ff0338cdc 100644
--- a/kernel/power/ztrsm_kernel_power6_RT.S
+++ b/kernel/power/ztrsm_kernel_power6_RT.S
@@ -186,36 +186,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_ppc440_LN.S b/kernel/power/ztrsm_kernel_ppc440_LN.S
index 51db71903..d33522456 100644
--- a/kernel/power/ztrsm_kernel_ppc440_LN.S
+++ b/kernel/power/ztrsm_kernel_ppc440_LN.S
@@ -179,36 +179,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_ppc440_LT.S b/kernel/power/ztrsm_kernel_ppc440_LT.S
index b5e23b3c6..a9e7b891f 100644
--- a/kernel/power/ztrsm_kernel_ppc440_LT.S
+++ b/kernel/power/ztrsm_kernel_ppc440_LT.S
@@ -179,36 +179,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/power/ztrsm_kernel_ppc440_RT.S b/kernel/power/ztrsm_kernel_ppc440_RT.S
index 2bb374d22..43f4b07cb 100644
--- a/kernel/power/ztrsm_kernel_ppc440_RT.S
+++ b/kernel/power/ztrsm_kernel_ppc440_RT.S
@@ -179,36 +179,36 @@
 
 #ifdef linux
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	LDC,    112 + STACKSIZE(SP)
+	ld	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	B,       56 + STACKSIZE(SP)
-	lwz	C,       60 + STACKSIZE(SP)
-	lwz	LDC,     64 + STACKSIZE(SP)
+	lwz	B,   FRAMESLOT(0) + STACKSIZE(SP)
+	lwz	C,   FRAMESLOT(1) + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(2) + STACKSIZE(SP)
 #else
-	lwz	LDC,     56 + STACKSIZE(SP)
+	lwz	LDC, FRAMESLOT(0) + STACKSIZE(SP)
 #endif
 #endif
 #endif
 
 #if defined(linux) && defined(__64BIT__)
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 
 #if defined(_AIX) || defined(__APPLE__)
 #ifdef __64BIT__
-	ld	OFFSET,  120 + STACKSIZE(SP)
+	ld	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #else
 #ifdef DOUBLE
-	lwz	OFFSET,   68 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(3) + STACKSIZE(SP)
 #else
-	lwz	OFFSET,   60 + STACKSIZE(SP)
+	lwz	OFFSET,  FRAMESLOT(1) + STACKSIZE(SP)
 #endif
 #endif
 #endif
diff --git a/kernel/setparam-ref.c b/kernel/setparam-ref.c
index 0eeac2e1f..a4d1486fc 100644
--- a/kernel/setparam-ref.c
+++ b/kernel/setparam-ref.c
@@ -550,6 +550,13 @@ gotoblas_t TABLE_NAME = {
   zomatcopy_k_cnTS, zomatcopy_k_ctTS, zomatcopy_k_rnTS, zomatcopy_k_rtTS,
   zomatcopy_k_cncTS, zomatcopy_k_ctcTS, zomatcopy_k_rncTS, zomatcopy_k_rtcTS,
 
+  simatcopy_k_cnTS, simatcopy_k_ctTS, simatcopy_k_rnTS, simatcopy_k_rtTS,
+  dimatcopy_k_cnTS, dimatcopy_k_ctTS, dimatcopy_k_rnTS, dimatcopy_k_rtTS,
+  cimatcopy_k_cnTS, cimatcopy_k_ctTS, cimatcopy_k_rnTS, cimatcopy_k_rtTS,
+  cimatcopy_k_cncTS, cimatcopy_k_ctcTS, cimatcopy_k_rncTS, cimatcopy_k_rtcTS,
+  zimatcopy_k_cnTS, zimatcopy_k_ctTS, zimatcopy_k_rnTS, zimatcopy_k_rtTS,
+  zimatcopy_k_cncTS, zimatcopy_k_ctcTS, zimatcopy_k_rncTS, zimatcopy_k_rtcTS,
+
   sgeadd_kTS, dgeadd_kTS, cgeadd_kTS, zgeadd_kTS
 
 };
diff --git a/kernel/x86/KERNEL b/kernel/x86/KERNEL
index 8b30355ec..39be2ef80 100644
--- a/kernel/x86/KERNEL
+++ b/kernel/x86/KERNEL
@@ -98,6 +98,23 @@ ifndef XAXPYKERNEL
 XAXPYKERNEL = xaxpy.S
 endif
 
+#Use C kernel for sgemv and dgemv
+ifndef SGEMVNKERNEL
+SGEMVNKERNEL = ../arm/gemv_n.c
+endif
+
+ifndef SGEMVTKERNEL
+SGEMVTKERNEL = ../arm/gemv_t.c
+endif
+
+ifndef DGEMVNKERNEL
+DGEMVNKERNEL = ../arm/gemv_n.c
+endif
+
+ifndef DGEMVTKERNEL
+DGEMVTKERNEL = ../arm/gemv_t.c
+endif
+
 ifndef QGEMVNKERNEL
 QGEMVNKERNEL = qgemv_n.S
 endif
diff --git a/kernel/x86/KERNEL.generic b/kernel/x86/KERNEL.generic
new file mode 100644
index 000000000..672edb069
--- /dev/null
+++ b/kernel/x86/KERNEL.generic
@@ -0,0 +1,159 @@
+SGEMM_BETA = ../generic/gemm_beta.c
+DGEMM_BETA = ../generic/gemm_beta.c
+CGEMM_BETA = ../generic/zgemm_beta.c
+ZGEMM_BETA = ../generic/zgemm_beta.c
+
+STRMMKERNEL	= ../generic/trmmkernel_2x2.c
+DTRMMKERNEL	= ../generic/trmmkernel_2x2.c
+CTRMMKERNEL	= ../generic/ztrmmkernel_2x2.c
+ZTRMMKERNEL	= ../generic/ztrmmkernel_2x2.c
+
+SGEMMKERNEL    =  ../generic/gemmkernel_2x2.c
+SGEMMONCOPY    =  ../generic/gemm_ncopy_2.c
+SGEMMOTCOPY    =  ../generic/gemm_tcopy_2.c
+SGEMMONCOPYOBJ =  sgemm_oncopy.o
+SGEMMOTCOPYOBJ =  sgemm_otcopy.o
+
+DGEMMKERNEL    =  ../generic/gemmkernel_2x2.c
+DGEMMONCOPY    = ../generic/gemm_ncopy_2.c
+DGEMMOTCOPY    = ../generic/gemm_tcopy_2.c
+DGEMMONCOPYOBJ = dgemm_oncopy.o
+DGEMMOTCOPYOBJ = dgemm_otcopy.o
+
+CGEMMKERNEL    = ../generic/zgemmkernel_2x2.c
+CGEMMONCOPY    = ../generic/zgemm_ncopy_2.c
+CGEMMOTCOPY    = ../generic/zgemm_tcopy_2.c
+CGEMMONCOPYOBJ =  cgemm_oncopy.o
+CGEMMOTCOPYOBJ =  cgemm_otcopy.o
+
+ZGEMMKERNEL    = ../generic/zgemmkernel_2x2.c
+ZGEMMONCOPY    = ../generic/zgemm_ncopy_2.c
+ZGEMMOTCOPY    = ../generic/zgemm_tcopy_2.c
+ZGEMMONCOPYOBJ =  zgemm_oncopy.o
+ZGEMMOTCOPYOBJ =  zgemm_otcopy.o
+
+STRSMKERNEL_LN	=  ../generic/trsm_kernel_LN.c
+STRSMKERNEL_LT	=  ../generic/trsm_kernel_LT.c
+STRSMKERNEL_RN	=  ../generic/trsm_kernel_RN.c
+STRSMKERNEL_RT	=  ../generic/trsm_kernel_RT.c
+
+DTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+DTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+DTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+DTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+CTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+CTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+CTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+CTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+ZTRSMKERNEL_LN	= ../generic/trsm_kernel_LN.c
+ZTRSMKERNEL_LT	= ../generic/trsm_kernel_LT.c
+ZTRSMKERNEL_RN	= ../generic/trsm_kernel_RN.c
+ZTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
+
+#Todo: CGEMM3MKERNEL should be 4x4 blocksizes.
+CGEMM3MKERNEL    =  zgemm3m_kernel_8x4_sse3.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_4x4_sse3.S
+
+#Pure C for other kernels
+SAMAXKERNEL  = ../arm/amax.c
+DAMAXKERNEL  = ../arm/amax.c
+CAMAXKERNEL  = ../arm/zamax.c
+ZAMAXKERNEL  = ../arm/zamax.c
+
+SAMINKERNEL  = ../arm/amin.c
+DAMINKERNEL  = ../arm/amin.c
+CAMINKERNEL  = ../arm/zamin.c
+ZAMINKERNEL  = ../arm/zamin.c
+
+SMAXKERNEL   = ../arm/max.c
+DMAXKERNEL   = ../arm/max.c
+
+SMINKERNEL   = ../arm/min.c
+DMINKERNEL   = ../arm/min.c
+
+ISAMAXKERNEL = ../arm/iamax.c
+IDAMAXKERNEL = ../arm/iamax.c
+ICAMAXKERNEL = ../arm/izamax.c
+IZAMAXKERNEL = ../arm/izamax.c
+
+ISAMINKERNEL = ../arm/iamin.c
+IDAMINKERNEL = ../arm/iamin.c
+ICAMINKERNEL = ../arm/izamin.c
+IZAMINKERNEL = ../arm/izamin.c
+
+ISMAXKERNEL  = ../arm/imax.c
+IDMAXKERNEL  = ../arm/imax.c
+
+ISMINKERNEL  = ../arm/imin.c
+IDMINKERNEL  = ../arm/imin.c
+
+SASUMKERNEL  = ../arm/asum.c
+DASUMKERNEL  = ../arm/asum.c
+CASUMKERNEL  = ../arm/zasum.c
+ZASUMKERNEL  = ../arm/zasum.c
+
+SAXPYKERNEL  = ../arm/axpy.c
+DAXPYKERNEL  = ../arm/axpy.c
+CAXPYKERNEL  = ../arm/zaxpy.c
+ZAXPYKERNEL  = ../arm/zaxpy.c
+
+SCOPYKERNEL  = ../arm/copy.c
+DCOPYKERNEL  = ../arm/copy.c
+CCOPYKERNEL  = ../arm/zcopy.c
+ZCOPYKERNEL  = ../arm/zcopy.c
+
+SDOTKERNEL   = ../arm/dot.c
+DDOTKERNEL   = ../arm/dot.c
+CDOTKERNEL   = ../arm/zdot.c
+ZDOTKERNEL   = ../arm/zdot.c
+
+SNRM2KERNEL  = ../arm/nrm2.c
+DNRM2KERNEL  = ../arm/nrm2.c
+CNRM2KERNEL  = ../arm/znrm2.c
+ZNRM2KERNEL  = ../arm/znrm2.c
+
+SROTKERNEL   = ../arm/rot.c
+DROTKERNEL   = ../arm/rot.c
+CROTKERNEL   = ../arm/zrot.c
+ZROTKERNEL   = ../arm/zrot.c
+
+SSCALKERNEL  = ../arm/scal.c
+DSCALKERNEL  = ../arm/scal.c
+CSCALKERNEL  = ../arm/zscal.c
+ZSCALKERNEL  = ../arm/zscal.c
+
+SSWAPKERNEL  = ../arm/swap.c
+DSWAPKERNEL  = ../arm/swap.c
+CSWAPKERNEL  = ../arm/zswap.c
+ZSWAPKERNEL  = ../arm/zswap.c
+
+SGEMVNKERNEL = ../arm/gemv_n.c
+DGEMVNKERNEL = ../arm/gemv_n.c
+CGEMVNKERNEL = ../arm/zgemv_n.c
+ZGEMVNKERNEL = ../arm/zgemv_n.c
+
+SGEMVTKERNEL = ../arm/gemv_t.c
+DGEMVTKERNEL = ../arm/gemv_t.c
+CGEMVTKERNEL = ../arm/zgemv_t.c
+ZGEMVTKERNEL = ../arm/zgemv_t.c
+
+SSYMV_U_KERNEL =  ../generic/symv_k.c
+SSYMV_L_KERNEL =  ../generic/symv_k.c
+DSYMV_U_KERNEL =  ../generic/symv_k.c
+DSYMV_L_KERNEL =  ../generic/symv_k.c
+QSYMV_U_KERNEL =  ../generic/symv_k.c
+QSYMV_L_KERNEL =  ../generic/symv_k.c
+CSYMV_U_KERNEL =  ../generic/zsymv_k.c
+CSYMV_L_KERNEL =  ../generic/zsymv_k.c
+ZSYMV_U_KERNEL =  ../generic/zsymv_k.c
+ZSYMV_L_KERNEL =  ../generic/zsymv_k.c
+XSYMV_U_KERNEL =  ../generic/zsymv_k.c
+XSYMV_L_KERNEL =  ../generic/zsymv_k.c
+
+ZHEMV_U_KERNEL =  ../generic/zhemv_k.c
+ZHEMV_L_KERNEL =  ../generic/zhemv_k.c
+
+CGEMM3MKERNEL    = ../generic/zgemm3mkernel_dump.c
+ZGEMM3MKERNEL    = ../generic/zgemm3mkernel_dump.c
diff --git a/kernel/x86/cpuid_win.c b/kernel/x86/cpuid_win.c
new file mode 100644
index 000000000..a1b00016b
--- /dev/null
+++ b/kernel/x86/cpuid_win.c
@@ -0,0 +1,41 @@
+/***************************************************************************
+Copyright (c) 2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#if defined(_MSC_VER) && !defined(__clang__)
+
+#include<intrin.h>
+
+void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx)
+{
+  int cpuInfo[4] = {-1};
+  __cpuid(cpuInfo, op);
+  *eax = cpuInfo[0];
+  *ebx = cpuInfo[1];
+  *ecx = cpuInfo[2];
+  *edx = cpuInfo[3];
+}
+#endif
diff --git a/kernel/x86_64/KERNEL b/kernel/x86_64/KERNEL
index 3508753ee..2dcc8658b 100644
--- a/kernel/x86_64/KERNEL
+++ b/kernel/x86_64/KERNEL
@@ -119,11 +119,11 @@ XCOPYKERNEL = zcopy.S
 endif
 
 ifndef SDOTKERNEL
-SDOTKERNEL = ../generic/dot.c 
+SDOTKERNEL = ../generic/dot.c
 endif
 
 ifndef DSDOTKERNEL
-DSDOTKERNEL = ../generic/dot.c 
+DSDOTKERNEL = ../generic/dot.c
 endif
 
 ifndef DDOTKERNEL
@@ -440,10 +440,18 @@ XGEMMITCOPYOBJ =
 XGEMMONCOPYOBJ =  xgemm_oncopy$(TSUFFIX).$(SUFFIX)
 XGEMMOTCOPYOBJ =  xgemm_otcopy$(TSUFFIX).$(SUFFIX)
 
+ifndef SGEMM_BETA
 SGEMM_BETA = gemm_beta.S
+endif
+ifndef DGEMM_BETA
 DGEMM_BETA = gemm_beta.S
+endif
+ifndef CGEMM_BETA
 CGEMM_BETA = zgemm_beta.S
+endif
+ifndef ZGEMM_BETA
 ZGEMM_BETA = zgemm_beta.S
+endif
 QGEMM_BETA = ../generic/gemm_beta.c
 XGEMM_BETA = ../generic/zgemm_beta.c
 
diff --git a/kernel/x86_64/KERNEL.BULLDOZER b/kernel/x86_64/KERNEL.BULLDOZER
index 289529772..9f124c97f 100644
--- a/kernel/x86_64/KERNEL.BULLDOZER
+++ b/kernel/x86_64/KERNEL.BULLDOZER
@@ -1,9 +1,14 @@
+DSCALKERNEL = dscal.c
+CSCALKERNEL = cscal.c
+ZSCALKERNEL = zscal.c
+
 DAXPYKERNEL = daxpy.c
 CAXPYKERNEL = caxpy.c
 ZAXPYKERNEL = zaxpy.c
 
 SDOTKERNEL     = sdot.c
-DDOTKERNEL     = ddot.c
+CDOTKERNEL     = cdot.c
+ZDOTKERNEL     = zdot.c
 
 DSYMV_U_KERNEL = dsymv_U.c
 DSYMV_L_KERNEL = dsymv_L.c
@@ -26,11 +31,11 @@ SGEMMINCOPY    =  ../generic/gemm_ncopy_16.c
 SGEMMITCOPY    =  ../generic/gemm_tcopy_16.c
 SGEMMONCOPY    =  gemm_ncopy_2_bulldozer.S
 SGEMMOTCOPY    =  gemm_tcopy_2_bulldozer.S
-
 SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX)
 SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX)
 SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
 SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
 DGEMMKERNEL    =  dgemm_kernel_8x2_bulldozer.S
 DGEMMINCOPY    =  dgemm_ncopy_8_bulldozer.S
 DGEMMITCOPY    =  dgemm_tcopy_8_bulldozer.S
@@ -40,6 +45,7 @@ DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
 DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
 DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
 DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
 CGEMMKERNEL    =  cgemm_kernel_4x2_bulldozer.S
 CGEMMINCOPY    =  ../generic/zgemm_ncopy_4.c
 CGEMMITCOPY    =  ../generic/zgemm_tcopy_4.c
@@ -49,6 +55,7 @@ CGEMMINCOPYOBJ =  cgemm_incopy$(TSUFFIX).$(SUFFIX)
 CGEMMITCOPYOBJ =  cgemm_itcopy$(TSUFFIX).$(SUFFIX)
 CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
 CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
 ZGEMMKERNEL    =  zgemm_kernel_2x2_bulldozer.S
 ZGEMMINCOPY    =
 ZGEMMITCOPY    =
diff --git a/kernel/x86_64/KERNEL.EXCAVATOR b/kernel/x86_64/KERNEL.EXCAVATOR
new file mode 100644
index 000000000..dbdd1fe9b
--- /dev/null
+++ b/kernel/x86_64/KERNEL.EXCAVATOR
@@ -0,0 +1,92 @@
+SAXPYKERNEL = saxpy.c
+DAXPYKERNEL = daxpy.c
+CAXPYKERNEL = caxpy.c
+ZAXPYKERNEL = zaxpy.c
+
+SDOTKERNEL     = sdot.c
+DDOTKERNEL     = ddot.c
+CDOTKERNEL     = cdot.c
+ZDOTKERNEL     = zdot.c
+
+
+DSYMV_U_KERNEL = dsymv_U.c
+DSYMV_L_KERNEL = dsymv_L.c
+SSYMV_U_KERNEL = ssymv_U.c
+SSYMV_L_KERNEL = ssymv_L.c
+
+SGEMVNKERNEL = sgemv_n_4.c
+SGEMVTKERNEL = sgemv_t_4.c
+
+DGEMVNKERNEL = dgemv_n_4.c
+DGEMVTKERNEL = dgemv_t_4.c
+
+ZGEMVNKERNEL = zgemv_n_dup.S
+ZGEMVTKERNEL = zgemv_t_4.c
+
+DCOPYKERNEL  = dcopy_bulldozer.S
+
+
+SGEMMKERNEL    =  sgemm_kernel_16x2_piledriver.S
+SGEMMINCOPY    =  ../generic/gemm_ncopy_16.c
+SGEMMITCOPY    =  ../generic/gemm_tcopy_16.c
+SGEMMONCOPY    =  gemm_ncopy_2_bulldozer.S
+SGEMMOTCOPY    =  gemm_tcopy_2_bulldozer.S
+SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX)
+SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX)
+SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
+SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+DGEMMKERNEL    =  dgemm_kernel_8x2_piledriver.S
+DGEMMINCOPY    =  ../generic/gemm_ncopy_8.c
+DGEMMITCOPY    =  ../generic/gemm_tcopy_8.c
+DGEMMONCOPY    =  gemm_ncopy_2_bulldozer.S
+DGEMMOTCOPY    =  gemm_tcopy_2_bulldozer.S
+DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
+DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
+DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
+DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+CGEMMKERNEL    =  cgemm_kernel_4x2_piledriver.S
+CGEMMINCOPY    =  ../generic/zgemm_ncopy_4.c
+CGEMMITCOPY    =  ../generic/zgemm_tcopy_4.c
+CGEMMONCOPY    =  ../generic/zgemm_ncopy_2.c
+CGEMMOTCOPY    =  ../generic/zgemm_tcopy_2.c
+CGEMMINCOPYOBJ =  cgemm_incopy$(TSUFFIX).$(SUFFIX)
+CGEMMITCOPYOBJ =  cgemm_itcopy$(TSUFFIX).$(SUFFIX)
+CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
+CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
+ZGEMMKERNEL    =  zgemm_kernel_2x2_piledriver.S
+ZGEMMINCOPY    =
+ZGEMMITCOPY    =
+ZGEMMONCOPY    =  ../generic/zgemm_ncopy_2.c
+ZGEMMOTCOPY    =  ../generic/zgemm_tcopy_2.c
+ZGEMMINCOPYOBJ =
+ZGEMMITCOPYOBJ =
+ZGEMMONCOPYOBJ =  zgemm_oncopy$(TSUFFIX).$(SUFFIX)
+ZGEMMOTCOPYOBJ =  zgemm_otcopy$(TSUFFIX).$(SUFFIX)
+
+CGEMM3MKERNEL    =  zgemm3m_kernel_8x4_barcelona.S
+ZGEMM3MKERNEL    =  zgemm3m_kernel_4x4_barcelona.S
+
+STRSMKERNEL_LN  =  ../generic/trsm_kernel_LN.c
+STRSMKERNEL_LT  =  ../generic/trsm_kernel_LT.c
+STRSMKERNEL_RN  =  ../generic/trsm_kernel_RN.c
+STRSMKERNEL_RT  =  ../generic/trsm_kernel_RT.c
+
+
+DTRSMKERNEL_LN  = ../generic/trsm_kernel_LN.c
+DTRSMKERNEL_LT  = dtrsm_kernel_LT_8x2_bulldozer.S
+DTRSMKERNEL_RN  = dtrsm_kernel_RN_8x2_bulldozer.S
+DTRSMKERNEL_RT  = ../generic/trsm_kernel_RT.c
+
+CTRSMKERNEL_LN  = ../generic/trsm_kernel_LN.c
+CTRSMKERNEL_LT  = ../generic/trsm_kernel_LT.c
+CTRSMKERNEL_RN  = ../generic/trsm_kernel_RN.c
+CTRSMKERNEL_RT  = ../generic/trsm_kernel_RT.c
+
+ZTRSMKERNEL_LN  = ../generic/trsm_kernel_LN.c
+ZTRSMKERNEL_LT  = ../generic/trsm_kernel_LT.c
+ZTRSMKERNEL_RN  = ../generic/trsm_kernel_RN.c
+ZTRSMKERNEL_RT  = ../generic/trsm_kernel_RT.c
+
+
diff --git a/kernel/x86_64/KERNEL.HASWELL b/kernel/x86_64/KERNEL.HASWELL
index a621b4484..a4686debb 100644
--- a/kernel/x86_64/KERNEL.HASWELL
+++ b/kernel/x86_64/KERNEL.HASWELL
@@ -1,3 +1,7 @@
+DSCALKERNEL = dscal.c
+CSCALKERNEL = cscal.c
+ZSCALKERNEL = zscal.c
+
 SGEMVNKERNEL = sgemv_n_4.c
 SGEMVTKERNEL = sgemv_t_4.c
 
@@ -10,6 +14,22 @@ ZGEMVTKERNEL = zgemv_t_4.c
 CGEMVNKERNEL = cgemv_n_4.c
 CGEMVTKERNEL = cgemv_t_4.c
 
+SSYMV_L_KERNEL = ssymv_L.c
+SSYMV_U_KERNEL = ssymv_U.c
+DSYMV_L_KERNEL = dsymv_L.c
+DSYMV_U_KERNEL = dsymv_U.c
+
+SDOTKERNEL = sdot.c
+DDOTKERNEL = ddot.c
+CDOTKERNEL = cdot.c
+ZDOTKERNEL = zdot.c
+
+SAXPYKERNEL = saxpy.c
+DAXPYKERNEL = daxpy.c
+CAXPYKERNEL = caxpy.c
+ZAXPYKERNEL = zaxpy.c
+
+STRMMKERNEL    =  sgemm_kernel_16x4_haswell.S
 SGEMMKERNEL    =  sgemm_kernel_16x4_haswell.S
 SGEMMINCOPY    =  ../generic/gemm_ncopy_16.c
 SGEMMITCOPY    =  ../generic/gemm_tcopy_16.c
@@ -20,16 +40,18 @@ SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX)
 SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
 SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)
 
-DGEMMKERNEL    =  dgemm_kernel_4x4_haswell.S
-DGEMMINCOPY    =
-DGEMMITCOPY    =
-DGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
-DGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
-DGEMMINCOPYOBJ =
-DGEMMITCOPYOBJ =
+DTRMMKERNEL    =  dtrmm_kernel_4x8_haswell.c
+DGEMMKERNEL    =  dgemm_kernel_4x8_haswell.S
+DGEMMINCOPY    =  ../generic/gemm_ncopy_4.c
+DGEMMITCOPY    =  ../generic/gemm_tcopy_4.c
+DGEMMONCOPY    =  ../generic/gemm_ncopy_8.c
+DGEMMOTCOPY    =  ../generic/gemm_tcopy_8.c
+DGEMMINCOPYOBJ =  dgemm_incopy$(TSUFFIX).$(SUFFIX)
+DGEMMITCOPYOBJ =  dgemm_itcopy$(TSUFFIX).$(SUFFIX)
 DGEMMONCOPYOBJ =  dgemm_oncopy$(TSUFFIX).$(SUFFIX)
 DGEMMOTCOPYOBJ =  dgemm_otcopy$(TSUFFIX).$(SUFFIX)
 
+CTRMMKERNEL    =  cgemm_kernel_8x2_haswell.S
 CGEMMKERNEL    =  cgemm_kernel_8x2_haswell.S
 CGEMMINCOPY    =  ../generic/zgemm_ncopy_8.c
 CGEMMITCOPY    =  ../generic/zgemm_tcopy_8.c
@@ -40,6 +62,7 @@ CGEMMITCOPYOBJ =  cgemm_itcopy$(TSUFFIX).$(SUFFIX)
 CGEMMONCOPYOBJ =  cgemm_oncopy$(TSUFFIX).$(SUFFIX)
 CGEMMOTCOPYOBJ =  cgemm_otcopy$(TSUFFIX).$(SUFFIX)
 
+ZTRMMKERNEL    =  zgemm_kernel_4x2_haswell.S
 ZGEMMKERNEL    =  zgemm_kernel_4x2_haswell.S
 ZGEMMINCOPY    =  ../generic/zgemm_ncopy_4.c
 ZGEMMITCOPY    =  ../generic/zgemm_tcopy_4.c
diff --git a/kernel/x86_64/KERNEL.PILEDRIVER b/kernel/x86_64/KERNEL.PILEDRIVER
index 55285e3d3..5d3c7a2af 100644
--- a/kernel/x86_64/KERNEL.PILEDRIVER
+++ b/kernel/x86_64/KERNEL.PILEDRIVER
@@ -1,3 +1,13 @@
+DSCALKERNEL = dscal.c
+CSCALKERNEL = cscal.c
+ZSCALKERNEL = zscal.c
+
+
+SAXPYKERNEL = saxpy.c
+DAXPYKERNEL = daxpy.c
+CAXPYKERNEL = caxpy.c
+ZAXPYKERNEL = zaxpy.c
+
 SGEMVNKERNEL = sgemv_n_4.c
 SGEMVTKERNEL = sgemv_t_4.c
 
@@ -7,7 +17,11 @@ ZGEMVTKERNEL = zgemv_t_4.c
 DGEMVNKERNEL = dgemv_n_bulldozer.S
 DGEMVTKERNEL = dgemv_t_bulldozer.S
 
-DDOTKERNEL   = ddot_bulldozer.S
+SDOTKERNEL   = sdot.c
+DDOTKERNEL   = ddot.c
+CDOTKERNEL   = cdot.c
+ZDOTKERNEL   = zdot.c
+
 DCOPYKERNEL  = dcopy_bulldozer.S
 
 SGEMMKERNEL    =  sgemm_kernel_16x2_piledriver.S
diff --git a/kernel/x86_64/KERNEL.SANDYBRIDGE b/kernel/x86_64/KERNEL.SANDYBRIDGE
index ff96cd011..355d1e2f1 100644
--- a/kernel/x86_64/KERNEL.SANDYBRIDGE
+++ b/kernel/x86_64/KERNEL.SANDYBRIDGE
@@ -1,8 +1,28 @@
+DSCALKERNEL = dscal.c
+CSCALKERNEL = cscal.c
+
+SGERKERNEL = sger.c
+DGERKERNEL = dger.c
+
 SGEMVNKERNEL = sgemv_n_4.c
 SGEMVTKERNEL = sgemv_t_4.c
 
 ZGEMVNKERNEL = zgemv_n_4.c
 
+SDOTKERNEL = sdot.c
+DDOTKERNEL = ddot.c
+CDOTKERNEL = cdot.c
+ZDOTKERNEL = zdot.c
+
+SSYMV_L_KERNEL = ssymv_L.c
+SSYMV_U_KERNEL = ssymv_U.c
+DSYMV_L_KERNEL = dsymv_L.c
+DSYMV_U_KERNEL = dsymv_U.c
+
+SAXPYKERNEL = saxpy.c
+DAXPYKERNEL = daxpy.c
+CAXPYKERNEL = caxpy.c
+ZAXPYKERNEL = zaxpy.c
 
 SGEMMKERNEL    =  sgemm_kernel_16x4_sandy.S
 SGEMMINCOPY    =  ../generic/gemm_ncopy_16.c
diff --git a/kernel/x86_64/KERNEL.STEAMROLLER b/kernel/x86_64/KERNEL.STEAMROLLER
index f5b5cb942..51e6d616a 100644
--- a/kernel/x86_64/KERNEL.STEAMROLLER
+++ b/kernel/x86_64/KERNEL.STEAMROLLER
@@ -1,9 +1,17 @@
+DSCALKERNEL = dscal.c
+CSCALKERNEL = cscal.c
+ZSCALKERNEL = zscal.c
+
+SAXPYKERNEL = saxpy.c
 DAXPYKERNEL = daxpy.c
 CAXPYKERNEL = caxpy.c
 ZAXPYKERNEL = zaxpy.c
 
 SDOTKERNEL     = sdot.c
 DDOTKERNEL     = ddot.c
+CDOTKERNEL     = cdot.c
+ZDOTKERNEL     = zdot.c
+
 
 DSYMV_U_KERNEL = dsymv_U.c
 DSYMV_L_KERNEL = dsymv_L.c
diff --git a/kernel/x86_64/KERNEL.generic b/kernel/x86_64/KERNEL.generic
index 2bcd83636..a23e59f3f 100644
--- a/kernel/x86_64/KERNEL.generic
+++ b/kernel/x86_64/KERNEL.generic
@@ -1,3 +1,8 @@
+SGEMM_BETA = ../generic/gemm_beta.c
+DGEMM_BETA = ../generic/gemm_beta.c
+CGEMM_BETA = ../generic/zgemm_beta.c
+ZGEMM_BETA = ../generic/zgemm_beta.c
+
 STRMMKERNEL	= ../generic/trmmkernel_2x2.c
 DTRMMKERNEL	= ../generic/trmmkernel_2x2.c
 CTRMMKERNEL	= ../generic/ztrmmkernel_2x2.c
@@ -50,3 +55,111 @@ ZTRSMKERNEL_RT	= ../generic/trsm_kernel_RT.c
 #Todo: CGEMM3MKERNEL should be 4x4 blocksizes.
 CGEMM3MKERNEL    =  zgemm3m_kernel_8x4_sse3.S
 ZGEMM3MKERNEL    =  zgemm3m_kernel_4x4_sse3.S
+
+#Pure C for other kernels
+SAMAXKERNEL  = ../arm/amax.c
+DAMAXKERNEL  = ../arm/amax.c
+CAMAXKERNEL  = ../arm/zamax.c
+ZAMAXKERNEL  = ../arm/zamax.c
+
+SAMINKERNEL  = ../arm/amin.c
+DAMINKERNEL  = ../arm/amin.c
+CAMINKERNEL  = ../arm/zamin.c
+ZAMINKERNEL  = ../arm/zamin.c
+
+SMAXKERNEL   = ../arm/max.c
+DMAXKERNEL   = ../arm/max.c
+
+SMINKERNEL   = ../arm/min.c
+DMINKERNEL   = ../arm/min.c
+
+ISAMAXKERNEL = ../arm/iamax.c
+IDAMAXKERNEL = ../arm/iamax.c
+ICAMAXKERNEL = ../arm/izamax.c
+IZAMAXKERNEL = ../arm/izamax.c
+
+ISAMINKERNEL = ../arm/iamin.c
+IDAMINKERNEL = ../arm/iamin.c
+ICAMINKERNEL = ../arm/izamin.c
+IZAMINKERNEL = ../arm/izamin.c
+
+ISMAXKERNEL  = ../arm/imax.c
+IDMAXKERNEL  = ../arm/imax.c
+
+ISMINKERNEL  = ../arm/imin.c
+IDMINKERNEL  = ../arm/imin.c
+
+SASUMKERNEL  = ../arm/asum.c
+DASUMKERNEL  = ../arm/asum.c
+CASUMKERNEL  = ../arm/zasum.c
+ZASUMKERNEL  = ../arm/zasum.c
+
+SAXPYKERNEL  = ../arm/axpy.c
+DAXPYKERNEL  = ../arm/axpy.c
+CAXPYKERNEL  = ../arm/zaxpy.c
+ZAXPYKERNEL  = ../arm/zaxpy.c
+
+SCOPYKERNEL  = ../arm/copy.c
+DCOPYKERNEL  = ../arm/copy.c
+CCOPYKERNEL  = ../arm/zcopy.c
+ZCOPYKERNEL  = ../arm/zcopy.c
+
+SDOTKERNEL   = ../arm/dot.c
+DDOTKERNEL   = ../arm/dot.c
+CDOTKERNEL   = ../arm/zdot.c
+ZDOTKERNEL   = ../arm/zdot.c
+
+SNRM2KERNEL  = ../arm/nrm2.c
+DNRM2KERNEL  = ../arm/nrm2.c
+CNRM2KERNEL  = ../arm/znrm2.c
+ZNRM2KERNEL  = ../arm/znrm2.c
+
+SROTKERNEL   = ../arm/rot.c
+DROTKERNEL   = ../arm/rot.c
+CROTKERNEL   = ../arm/zrot.c
+ZROTKERNEL   = ../arm/zrot.c
+
+SSCALKERNEL  = ../arm/scal.c
+DSCALKERNEL  = ../arm/scal.c
+CSCALKERNEL  = ../arm/zscal.c
+ZSCALKERNEL  = ../arm/zscal.c
+
+SSWAPKERNEL  = ../arm/swap.c
+DSWAPKERNEL  = ../arm/swap.c
+CSWAPKERNEL  = ../arm/zswap.c
+ZSWAPKERNEL  = ../arm/zswap.c
+
+SGEMVNKERNEL = ../arm/gemv_n.c
+DGEMVNKERNEL = ../arm/gemv_n.c
+CGEMVNKERNEL = ../arm/zgemv_n.c
+ZGEMVNKERNEL = ../arm/zgemv_n.c
+
+SGEMVTKERNEL = ../arm/gemv_t.c
+DGEMVTKERNEL = ../arm/gemv_t.c
+CGEMVTKERNEL = ../arm/zgemv_t.c
+ZGEMVTKERNEL = ../arm/zgemv_t.c
+
+SSYMV_U_KERNEL =  ../generic/symv_k.c
+SSYMV_L_KERNEL =  ../generic/symv_k.c
+DSYMV_U_KERNEL =  ../generic/symv_k.c
+DSYMV_L_KERNEL =  ../generic/symv_k.c
+QSYMV_U_KERNEL =  ../generic/symv_k.c
+QSYMV_L_KERNEL =  ../generic/symv_k.c
+CSYMV_U_KERNEL =  ../generic/zsymv_k.c
+CSYMV_L_KERNEL =  ../generic/zsymv_k.c
+ZSYMV_U_KERNEL =  ../generic/zsymv_k.c
+ZSYMV_L_KERNEL =  ../generic/zsymv_k.c
+XSYMV_U_KERNEL =  ../generic/zsymv_k.c
+XSYMV_L_KERNEL =  ../generic/zsymv_k.c
+
+ZHEMV_U_KERNEL =  ../generic/zhemv_k.c
+ZHEMV_L_KERNEL =  ../generic/zhemv_k.c
+
+LSAME_KERNEL = ../generic/lsame.c
+SCABS_KERNEL	= ../generic/cabs.c
+DCABS_KERNEL	= ../generic/cabs.c
+QCABS_KERNEL	= ../generic/cabs.c
+
+#Dump kernel
+CGEMM3MKERNEL    = ../generic/zgemm3mkernel_dump.c
+ZGEMM3MKERNEL    = ../generic/zgemm3mkernel_dump.c
diff --git a/kernel/x86_64/caxpy.c b/kernel/x86_64/caxpy.c
index be945a441..1ee0499a7 100644
--- a/kernel/x86_64/caxpy.c
+++ b/kernel/x86_64/caxpy.c
@@ -29,8 +29,14 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 
 
-#if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)
+#if defined(PILEDRIVER) || defined(STEAMROLLER)
+#include "caxpy_microk_steamroller-2.c"
+#elif defined(BULLDOZER)
 #include "caxpy_microk_bulldozer-2.c"
+#elif defined(HASWELL)
+#include "caxpy_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "caxpy_microk_sandy-2.c"
 #endif
 
 
@@ -78,15 +84,16 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da_r, FLOAT da_i,
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-		int n1 = n & -8;
+		BLASLONG n1 = n & -32;
 
 		if ( n1 )
 		{
 			da[0] = da_r;
 			da[1] = da_i;
-			caxpy_kernel_8(n1, x, y , &da );
+			caxpy_kernel_8(n1, x, y , da );
 			ix = 2 * n1;
 		}
+
 		i = n1;
 		while(i < n)
 		{
diff --git a/kernel/x86_64/caxpy_microk_bulldozer-2.c b/kernel/x86_64/caxpy_microk_bulldozer-2.c
index 63575c374..33bda0943 100644
--- a/kernel/x86_64/caxpy_microk_bulldozer-2.c
+++ b/kernel/x86_64/caxpy_microk_bulldozer-2.c
@@ -31,89 +31,87 @@ static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __att
 static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
 {
 
+#if !defined(CONJ)
+	FLOAT mvec[4] = { -1.0, 1.0, -1.0, 1.0 };
+#else
+	FLOAT mvec[4] = { 1.0, -1.0, 1.0, -1.0 };
+#endif
 
 	BLASLONG register i = 0;
 
+  if ( n < 640 )
+  {
+
 	__asm__  __volatile__
 	(
+	"vzeroupper					    \n\t"
 	"vbroadcastss		(%4), %%xmm0		    \n\t"  // real part of alpha
 	"vbroadcastss	       4(%4), %%xmm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulps		(%5), %%xmm1 , %%xmm1		    \n\t"
+#else
+	"vmulps		(%5), %%xmm0 , %%xmm0		    \n\t"
+#endif
 
 	".align 16				            \n\t"
 	"1:				            \n\t"
 
-	"prefetcht0  768(%2,%0,4)                           \n\t"
 	"vmovups        (%2,%0,4), %%xmm5                   \n\t" // 2 complex values from x
+	".align 2					    \n\t"
 	"vmovups      16(%2,%0,4), %%xmm7                   \n\t" // 2 complex values from x
 	"vmovups      32(%2,%0,4), %%xmm9                   \n\t" // 2 complex values from x
 	"vmovups      48(%2,%0,4), %%xmm11                  \n\t" // 2 complex values from x
-	"prefetcht0  768(%3,%0,4)                           \n\t"
 
-#if !defined(CONJ)
-	"vfmaddps    (%3,%0,4), %%xmm0 , %%xmm5, %%xmm12    \n\t"
+	"vmovups      64(%2,%0,4), %%xmm12                  \n\t" // 2 complex values from x
+	"vmovups      80(%2,%0,4), %%xmm13                  \n\t" // 2 complex values from x
+	"vmovups      96(%2,%0,4), %%xmm14                  \n\t" // 2 complex values from x
+	"vmovups     112(%2,%0,4), %%xmm15                  \n\t" // 2 complex values from x
+
 	"vpermilps	$0xb1 , %%xmm5 , %%xmm4 	    \n\t"  // exchange real and imag part
-	"vmulps         %%xmm1, %%xmm4 , %%xmm4             \n\t"
-
-	"vfmaddps  16(%3,%0,4), %%xmm0 , %%xmm7, %%xmm13    \n\t"
 	"vpermilps	$0xb1 , %%xmm7 , %%xmm6 	    \n\t"  // exchange real and imag part
-	"vmulps         %%xmm1, %%xmm6 , %%xmm6             \n\t"
-
-	"vfmaddps  32(%3,%0,4), %%xmm0 , %%xmm9, %%xmm14    \n\t"
 	"vpermilps	$0xb1 , %%xmm9 , %%xmm8 	    \n\t"  // exchange real and imag part
-	"vmulps         %%xmm1, %%xmm8 , %%xmm8             \n\t"
-
-	"vfmaddps  48(%3,%0,4), %%xmm0 , %%xmm11,%%xmm15    \n\t"
 	"vpermilps	$0xb1 , %%xmm11, %%xmm10 	    \n\t"  // exchange real and imag part
-	"vmulps         %%xmm1, %%xmm10, %%xmm10            \n\t"
 
-        "vaddsubps      %%xmm4, %%xmm12, %%xmm12            \n\t"
-        "vaddsubps      %%xmm6, %%xmm13, %%xmm13            \n\t"
-        "vaddsubps      %%xmm8, %%xmm14, %%xmm14            \n\t"
-        "vaddsubps      %%xmm10,%%xmm15, %%xmm15            \n\t"
+	"vfmaddps    (%3,%0,4), %%xmm0 , %%xmm5, %%xmm5     \n\t"
+	".align 2					    \n\t"
+	"vfmaddps  16(%3,%0,4), %%xmm0 , %%xmm7, %%xmm7     \n\t"
+	"vfmaddps  32(%3,%0,4), %%xmm0 , %%xmm9, %%xmm9     \n\t"
+	"vfmaddps  48(%3,%0,4), %%xmm0 , %%xmm11,%%xmm11    \n\t"
 
-#else
+	"vfmaddps	%%xmm5 , %%xmm1 , %%xmm4 , %%xmm5   \n\t"
+	"vfmaddps	%%xmm7 , %%xmm1 , %%xmm6 , %%xmm7   \n\t"
+	"vfmaddps	%%xmm9 , %%xmm1 , %%xmm8 , %%xmm9   \n\t"
+	"vfmaddps	%%xmm11, %%xmm1 , %%xmm10, %%xmm11  \n\t"
 
-	"vmulps		%%xmm0, %%xmm5, %%xmm4		    \n\t" // a_r*x_r, a_r*x_i
-	"vmulps		%%xmm1, %%xmm5, %%xmm5 		    \n\t" // a_i*x_r, a_i*x_i
-	"vmulps		%%xmm0, %%xmm7, %%xmm6		    \n\t" // a_r*x_r, a_r*x_i
-	"vmulps		%%xmm1, %%xmm7, %%xmm7 		    \n\t" // a_i*x_r, a_i*x_i
-	"vmulps		%%xmm0, %%xmm9, %%xmm8		    \n\t" // a_r*x_r, a_r*x_i
-	"vmulps		%%xmm1, %%xmm9, %%xmm9 		    \n\t" // a_i*x_r, a_i*x_i
-	"vmulps		%%xmm0, %%xmm11, %%xmm10	    \n\t" // a_r*x_r, a_r*x_i
-	"vmulps		%%xmm1, %%xmm11, %%xmm11	    \n\t" // a_i*x_r, a_i*x_i
+	"vpermilps	$0xb1 , %%xmm12, %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm13, %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm14, %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm15, %%xmm10 	    \n\t"  // exchange real and imag part
 
-	"vpermilps	$0xb1 , %%xmm4 , %%xmm4 	    \n\t"  // exchange real and imag part
-	"vaddsubps	%%xmm4 ,%%xmm5 , %%xmm4 	    \n\t"
-	"vpermilps	$0xb1 , %%xmm4 , %%xmm4 	    \n\t"  // exchange real and imag part
+	"vfmaddps  64(%3,%0,4), %%xmm0 , %%xmm12, %%xmm12   \n\t"
+	"vfmaddps  80(%3,%0,4), %%xmm0 , %%xmm13, %%xmm13   \n\t"
+	"vfmaddps  96(%3,%0,4), %%xmm0 , %%xmm14, %%xmm14   \n\t"
+	"vfmaddps 112(%3,%0,4), %%xmm0 , %%xmm15, %%xmm15   \n\t"
 
-	"vpermilps	$0xb1 , %%xmm6 , %%xmm6 	    \n\t"  // exchange real and imag part
-	"vaddsubps	%%xmm6 ,%%xmm7 , %%xmm6 	    \n\t"
-	"vpermilps	$0xb1 , %%xmm6 , %%xmm6 	    \n\t"  // exchange real and imag part
+	"vfmaddps	%%xmm12, %%xmm1 , %%xmm4 , %%xmm12  \n\t"
+	"vfmaddps	%%xmm13, %%xmm1 , %%xmm6 , %%xmm13  \n\t"
+	"vfmaddps	%%xmm14, %%xmm1 , %%xmm8 , %%xmm14  \n\t"
+	"vfmaddps	%%xmm15, %%xmm1 , %%xmm10, %%xmm15  \n\t"
 
-	"vpermilps	$0xb1 , %%xmm8 , %%xmm8 	    \n\t"  // exchange real and imag part
-	"vaddsubps	%%xmm8 ,%%xmm9 , %%xmm8 	    \n\t"
-	"vpermilps	$0xb1 , %%xmm8 , %%xmm8 	    \n\t"  // exchange real and imag part
+	"vmovups	%%xmm5 ,   (%3,%0,4)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%xmm7 , 16(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm9 , 32(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm12, 64(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm13, 80(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm14, 96(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm15,112(%3,%0,4)		    \n\t"
 
-	"vpermilps	$0xb1 , %%xmm10, %%xmm10	    \n\t"  // exchange real and imag part
-	"vaddsubps	%%xmm10,%%xmm11, %%xmm10	    \n\t"
-	"vpermilps	$0xb1 , %%xmm10, %%xmm10	    \n\t"  // exchange real and imag part
-
-	"vaddps	     (%3,%0,4) ,%%xmm4 , %%xmm12	    \n\t"
-	"vaddps	   16(%3,%0,4) ,%%xmm6 , %%xmm13	    \n\t"
-	"vaddps	   32(%3,%0,4) ,%%xmm8 , %%xmm14	    \n\t"
-	"vaddps	   48(%3,%0,4) ,%%xmm10, %%xmm15	    \n\t"
-
-
-#endif
-
-	"vmovups	%%xmm12,   (%3,%0,4)		    \n\t"
-	"vmovups	%%xmm13, 16(%3,%0,4)		    \n\t"
-	"vmovups	%%xmm14, 32(%3,%0,4)		    \n\t"
-	"vmovups	%%xmm15, 48(%3,%0,4)		    \n\t"
-
-	"addq		$16, %0	  	 	             \n\t"
-	"subq	        $8 , %1			             \n\t"		
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
 	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
 
 	:
         : 
@@ -121,7 +119,8 @@ static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
 	  "r" (n),  	// 1
           "r" (x),      // 2
           "r" (y),      // 3
-          "r" (alpha)   // 4
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
 	: "cc", 
 	  "%xmm0", "%xmm1",
 	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
@@ -129,7 +128,73 @@ static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
 	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
 	  "memory"
 	);
+	return;
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vbroadcastss		(%4), %%xmm0		    \n\t"  // real part of alpha
+	"vbroadcastss	       4(%4), %%xmm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulps		(%5), %%xmm1 , %%xmm1		    \n\t"
+#else
+	"vmulps		(%5), %%xmm0 , %%xmm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"prefetcht0	512(%2,%0,4)			    \n\t"
+	"vmovups        (%2,%0,4), %%xmm5                   \n\t" // 2 complex values from x
+	".align 2					    \n\t"
+	"vmovups      16(%2,%0,4), %%xmm7                   \n\t" // 2 complex values from x
+	"vmovups      32(%2,%0,4), %%xmm9                   \n\t" // 2 complex values from x
+	"vmovups      48(%2,%0,4), %%xmm11                  \n\t" // 2 complex values from x
+
+	"vpermilps	$0xb1 , %%xmm5 , %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm7 , %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm9 , %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm11, %%xmm10 	    \n\t"  // exchange real and imag part
+
+	"prefetcht0	512(%3,%0,4)			    \n\t"
+	"vfmaddps    (%3,%0,4), %%xmm0 , %%xmm5, %%xmm5     \n\t"
+	".align 2					    \n\t"
+	"vfmaddps  16(%3,%0,4), %%xmm0 , %%xmm7, %%xmm7     \n\t"
+	"vfmaddps  32(%3,%0,4), %%xmm0 , %%xmm9, %%xmm9     \n\t"
+	"vfmaddps  48(%3,%0,4), %%xmm0 , %%xmm11,%%xmm11    \n\t"
+
+	"vfmaddps	%%xmm5 , %%xmm1 , %%xmm4 , %%xmm5   \n\t"
+	"vfmaddps	%%xmm7 , %%xmm1 , %%xmm6 , %%xmm7   \n\t"
+	"vfmaddps	%%xmm9 , %%xmm1 , %%xmm8 , %%xmm9   \n\t"
+	"vfmaddps	%%xmm11, %%xmm1 , %%xmm10, %%xmm11  \n\t"
+
+	"vmovups	%%xmm5 ,   (%3,%0,4)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%xmm7 , 16(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm9 , 32(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,4)		    \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $8, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "memory"
+	);
+
 
 } 
 
-
diff --git a/kernel/x86_64/caxpy_microk_haswell-2.c b/kernel/x86_64/caxpy_microk_haswell-2.c
new file mode 100644
index 000000000..7a9fc1b95
--- /dev/null
+++ b/kernel/x86_64/caxpy_microk_haswell-2.c
@@ -0,0 +1,132 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+#if !defined(CONJ)
+	FLOAT mvec[8] = { -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0 };
+#else
+	FLOAT mvec[8] = { 1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0 };
+#endif
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vbroadcastss		(%4), %%ymm0		    \n\t"  // real part of alpha
+	"vbroadcastss	       4(%4), %%ymm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulps		(%5), %%ymm1 , %%ymm1		    \n\t"
+#else
+	"vmulps		(%5), %%ymm0 , %%ymm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmovups        (%2,%0,4), %%ymm5                   \n\t" // 4 complex values from x
+	".align 2					    \n\t"
+	"vmovups      32(%2,%0,4), %%ymm7                   \n\t" // 4 complex values from x
+	"vmovups      64(%2,%0,4), %%ymm9                   \n\t" // 4 complex values from x
+	"vmovups      96(%2,%0,4), %%ymm11                  \n\t" // 4 complex values from x
+
+	"vmovups     128(%2,%0,4), %%ymm12                  \n\t" // 4 complex values from x
+	"vmovups     160(%2,%0,4), %%ymm13                  \n\t" // 4 complex values from x
+	"vmovups     192(%2,%0,4), %%ymm14                  \n\t" // 4 complex values from x
+	"vmovups     224(%2,%0,4), %%ymm15                  \n\t" // 4 complex values from x
+
+	"vpermilps	$0xb1 , %%ymm5 , %%ymm4 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm7 , %%ymm6 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm9 , %%ymm8 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm11, %%ymm10 	    \n\t"  // exchange real and imag part
+
+	"vfmadd213ps    (%3,%0,4), %%ymm0 , %%ymm5          \n\t"
+	".align 2					    \n\t"
+	"vfmadd213ps  32(%3,%0,4), %%ymm0 , %%ymm7          \n\t"
+	"vfmadd213ps  64(%3,%0,4), %%ymm0 , %%ymm9          \n\t"
+	"vfmadd213ps  96(%3,%0,4), %%ymm0 , %%ymm11         \n\t"
+
+	"vfmadd231ps	%%ymm1 , %%ymm4 , %%ymm5   \n\t"
+	"vfmadd231ps	%%ymm1 , %%ymm6 , %%ymm7   \n\t"
+	"vfmadd231ps	%%ymm1 , %%ymm8 , %%ymm9   \n\t"
+	"vfmadd231ps	%%ymm1 , %%ymm10, %%ymm11  \n\t"
+
+	"vpermilps	$0xb1 , %%ymm12, %%ymm4 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm13, %%ymm6 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm14, %%ymm8 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm15, %%ymm10 	    \n\t"  // exchange real and imag part
+
+	"vfmadd213ps 128(%3,%0,4), %%ymm0 , %%ymm12         \n\t"
+	"vfmadd213ps 160(%3,%0,4), %%ymm0 , %%ymm13         \n\t"
+	"vfmadd213ps 192(%3,%0,4), %%ymm0 , %%ymm14         \n\t"
+	"vfmadd213ps 224(%3,%0,4), %%ymm0 , %%ymm15         \n\t"
+
+	"vfmadd231ps	%%ymm1 , %%ymm4 , %%ymm12  \n\t"
+	"vfmadd231ps	%%ymm1 , %%ymm6 , %%ymm13  \n\t"
+	"vfmadd231ps	%%ymm1 , %%ymm8 , %%ymm14  \n\t"
+	"vfmadd231ps	%%ymm1 , %%ymm10, %%ymm15  \n\t"
+
+	"vmovups	%%ymm5 ,   (%3,%0,4)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%ymm7 , 32(%3,%0,4)		    \n\t"
+	"vmovups	%%ymm9 , 64(%3,%0,4)		    \n\t"
+	"vmovups	%%ymm11, 96(%3,%0,4)		    \n\t"
+
+	"vmovups	%%ymm12,128(%3,%0,4)		    \n\t"
+	"vmovups	%%ymm13,160(%3,%0,4)		    \n\t"
+	"vmovups	%%ymm14,192(%3,%0,4)		    \n\t"
+	"vmovups	%%ymm15,224(%3,%0,4)		    \n\t"
+
+	"addq		$64, %0	  	 	             \n\t"
+	"subq	        $32, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15", 
+	  "memory"
+	);
+
+
+} 
+
diff --git a/kernel/x86_64/caxpy_microk_sandy-2.c b/kernel/x86_64/caxpy_microk_sandy-2.c
new file mode 100644
index 000000000..dbfce208f
--- /dev/null
+++ b/kernel/x86_64/caxpy_microk_sandy-2.c
@@ -0,0 +1,116 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+#if !defined(CONJ)
+	FLOAT mvec[8] = { -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0 };
+#else
+	FLOAT mvec[8] = { 1.0, -1.0, 1.0, -1.0, 1.0, -1.0, 1.0, -1.0 };
+#endif
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vbroadcastss		(%4), %%ymm0		    \n\t"  // real part of alpha
+	"vbroadcastss	       4(%4), %%ymm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulps		(%5), %%ymm1 , %%ymm1		    \n\t"
+#else
+	"vmulps		(%5), %%ymm0 , %%ymm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmovups        (%2,%0,4), %%ymm5                   \n\t" // 4 complex values from x
+	".align 2					    \n\t"
+	"vmovups      32(%2,%0,4), %%ymm7                   \n\t" // 4 complex values from x
+	"vmovups      64(%2,%0,4), %%ymm9                   \n\t" // 4 complex values from x
+	"vmovups      96(%2,%0,4), %%ymm11                  \n\t" // 4 complex values from x
+
+	"vpermilps	$0xb1 , %%ymm5 , %%ymm4 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm7 , %%ymm6 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm9 , %%ymm8 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%ymm11, %%ymm10 	    \n\t"  // exchange real and imag part
+
+	"vmulps	        %%ymm5 , %%ymm0 , %%ymm5	    \n\t"
+	"vmulps	        %%ymm7 , %%ymm0 , %%ymm7	    \n\t"
+	"vmulps	        %%ymm9 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulps	        %%ymm11, %%ymm0 , %%ymm11           \n\t"
+
+	"vaddps	        (%3,%0,4), %%ymm5 , %%ymm5          \n\t"
+	"vaddps	      32(%3,%0,4), %%ymm7 , %%ymm7          \n\t"
+	"vaddps	      64(%3,%0,4), %%ymm9 , %%ymm9          \n\t"
+	"vaddps	      96(%3,%0,4), %%ymm11, %%ymm11         \n\t"
+
+	"vmulps	        %%ymm4 , %%ymm1 , %%ymm4	    \n\t"
+	"vmulps	        %%ymm6 , %%ymm1 , %%ymm6	    \n\t"
+	"vmulps	        %%ymm8 , %%ymm1 , %%ymm8	    \n\t"
+	"vmulps	        %%ymm10, %%ymm1 , %%ymm10           \n\t"
+
+	"vaddps         %%ymm4 , %%ymm5 , %%ymm5            \n\t"
+	"vaddps         %%ymm6 , %%ymm7 , %%ymm7            \n\t"
+	"vaddps         %%ymm8 , %%ymm9 , %%ymm9            \n\t"
+	"vaddps         %%ymm10, %%ymm11, %%ymm11           \n\t"
+
+	"vmovups	%%ymm5 ,   (%3,%0,4)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%ymm7 , 32(%3,%0,4)		    \n\t"
+	"vmovups	%%ymm9 , 64(%3,%0,4)		    \n\t"
+	"vmovups	%%ymm11, 96(%3,%0,4)		    \n\t"
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15", 
+	  "memory"
+	);
+
+
+} 
+
diff --git a/kernel/x86_64/caxpy_microk_steamroller-2.c b/kernel/x86_64/caxpy_microk_steamroller-2.c
new file mode 100644
index 000000000..87370b032
--- /dev/null
+++ b/kernel/x86_64/caxpy_microk_steamroller-2.c
@@ -0,0 +1,200 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void caxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+#if !defined(CONJ)
+	FLOAT mvec[4] = { -1.0, 1.0, -1.0, 1.0 };
+#else
+	FLOAT mvec[4] = { 1.0, -1.0, 1.0, -1.0 };
+#endif
+
+	BLASLONG register i = 0;
+
+  if ( n <= 2048 )
+  {
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vbroadcastss		(%4), %%xmm0		    \n\t"  // real part of alpha
+	"vbroadcastss	       4(%4), %%xmm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulps		(%5), %%xmm1 , %%xmm1		    \n\t"
+#else
+	"vmulps		(%5), %%xmm0 , %%xmm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmovups        (%2,%0,4), %%xmm5                   \n\t" // 2 complex values from x
+	".align 2					    \n\t"
+	"vmovups      16(%2,%0,4), %%xmm7                   \n\t" // 2 complex values from x
+	"vmovups      32(%2,%0,4), %%xmm9                   \n\t" // 2 complex values from x
+	"vmovups      48(%2,%0,4), %%xmm11                  \n\t" // 2 complex values from x
+
+	"vmovups      64(%2,%0,4), %%xmm12                  \n\t" // 2 complex values from x
+	"vmovups      80(%2,%0,4), %%xmm13                  \n\t" // 2 complex values from x
+	"vmovups      96(%2,%0,4), %%xmm14                  \n\t" // 2 complex values from x
+	"vmovups     112(%2,%0,4), %%xmm15                  \n\t" // 2 complex values from x
+
+	"vpermilps	$0xb1 , %%xmm5 , %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm7 , %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm9 , %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm11, %%xmm10 	    \n\t"  // exchange real and imag part
+
+	"vfmadd213ps    (%3,%0,4), %%xmm0 , %%xmm5          \n\t"
+	".align 2					    \n\t"
+	"vfmadd213ps  16(%3,%0,4), %%xmm0 , %%xmm7          \n\t"
+	"vfmadd213ps  32(%3,%0,4), %%xmm0 , %%xmm9          \n\t"
+	"vfmadd213ps  48(%3,%0,4), %%xmm0 , %%xmm11         \n\t"
+
+	"vfmadd231ps	%%xmm1 , %%xmm4 , %%xmm5   \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm6 , %%xmm7   \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm8 , %%xmm9   \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm10, %%xmm11  \n\t"
+
+	"vpermilps	$0xb1 , %%xmm12, %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm13, %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm14, %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm15, %%xmm10 	    \n\t"  // exchange real and imag part
+
+	"vfmadd213ps  64(%3,%0,4), %%xmm0 , %%xmm12   \n\t"
+	"vfmadd213ps  80(%3,%0,4), %%xmm0 , %%xmm13   \n\t"
+	"vfmadd213ps  96(%3,%0,4), %%xmm0 , %%xmm14   \n\t"
+	"vfmadd213ps 112(%3,%0,4), %%xmm0 , %%xmm15   \n\t"
+
+	"vfmadd231ps	%%xmm1 , %%xmm4 , %%xmm12  \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm6 , %%xmm13  \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm8 , %%xmm14  \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm10, %%xmm15  \n\t"
+
+	"vmovups	%%xmm5 ,   (%3,%0,4)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%xmm7 , 16(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm9 , 32(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm12, 64(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm13, 80(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm14, 96(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm15,112(%3,%0,4)		    \n\t"
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vbroadcastss		(%4), %%xmm0		    \n\t"  // real part of alpha
+	"vbroadcastss	       4(%4), %%xmm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulps		(%5), %%xmm1 , %%xmm1		    \n\t"
+#else
+	"vmulps		(%5), %%xmm0 , %%xmm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"prefetcht0	512(%2,%0,4)			    \n\t"
+	"vmovups        (%2,%0,4), %%xmm5                   \n\t" // 2 complex values from x
+	".align 2					    \n\t"
+	"vmovups      16(%2,%0,4), %%xmm7                   \n\t" // 2 complex values from x
+	"vmovups      32(%2,%0,4), %%xmm9                   \n\t" // 2 complex values from x
+	"vmovups      48(%2,%0,4), %%xmm11                  \n\t" // 2 complex values from x
+
+	"vpermilps	$0xb1 , %%xmm5 , %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm7 , %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm9 , %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilps	$0xb1 , %%xmm11, %%xmm10 	    \n\t"  // exchange real and imag part
+
+	"prefetcht0	512(%3,%0,4)			    \n\t"
+	"vfmadd213ps    (%3,%0,4), %%xmm0 , %%xmm5          \n\t"
+	".align 2					    \n\t"
+	"vfmadd213ps  16(%3,%0,4), %%xmm0 , %%xmm7          \n\t"
+	"vfmadd213ps  32(%3,%0,4), %%xmm0 , %%xmm9          \n\t"
+	"vfmadd213ps  48(%3,%0,4), %%xmm0 , %%xmm11         \n\t"
+
+	"vfmadd231ps	%%xmm1 , %%xmm4 , %%xmm5   \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm6 , %%xmm7   \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm8 , %%xmm9   \n\t"
+	"vfmadd231ps	%%xmm1 , %%xmm10, %%xmm11  \n\t"
+
+	"vmovups	%%xmm5 ,   (%3,%0,4)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%xmm7 , 16(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm9 , 32(%3,%0,4)		    \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,4)		    \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $8, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "memory"
+	);
+
+
+} 
+
diff --git a/kernel/x86_64/cdot.c b/kernel/x86_64/cdot.c
new file mode 100644
index 000000000..2b2c4ff7a
--- /dev/null
+++ b/kernel/x86_64/cdot.c
@@ -0,0 +1,176 @@
+/***************************************************************************
+Copyright (c) 2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+
+#include "common.h"
+#include <complex.h>
+
+
+#if defined(BULLDOZER)
+#include "cdot_microk_bulldozer-2.c"
+#elif defined(STEAMROLLER) || defined(PILEDRIVER)
+#include "cdot_microk_steamroller-2.c"
+#elif defined(HASWELL)
+#include "cdot_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "cdot_microk_sandy-2.c"
+#endif
+
+
+#ifndef HAVE_KERNEL_16
+
+static void cdot_kernel_16(BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *d) __attribute__ ((noinline));
+
+static void cdot_kernel_16(BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *d)
+{
+	BLASLONG register i = 0;
+	FLOAT dot[8] = { 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0 };
+	BLASLONG j=0;
+
+	while( i < n )
+        {
+
+            dot[0] += x[j]   * y[j]   ;
+            dot[1] += x[j+1] * y[j+1] ;
+            dot[4] += x[j]   * y[j+1] ;
+            dot[5] += x[j+1] * y[j]   ;
+
+            dot[2] += x[j+2] * y[j+2] ;
+            dot[3] += x[j+3] * y[j+3] ;
+            dot[6] += x[j+2] * y[j+3] ;
+            dot[7] += x[j+3] * y[j+2] ;
+
+            dot[0] += x[j+4] * y[j+4] ;
+            dot[1] += x[j+5] * y[j+5] ;
+            dot[4] += x[j+4] * y[j+5] ;
+            dot[5] += x[j+5] * y[j+4] ;
+
+            dot[2] += x[j+6] * y[j+6] ;
+            dot[3] += x[j+7] * y[j+7] ;
+            dot[6] += x[j+6] * y[j+7] ;
+            dot[7] += x[j+7] * y[j+6] ;
+
+	    j+=8;
+            i+=4;
+
+        }
+	d[0] = dot[0];
+	d[1] = dot[1];
+	d[2] = dot[2];
+	d[3] = dot[3];
+	d[4] = dot[4];
+	d[5] = dot[5];
+	d[6] = dot[6];
+	d[7] = dot[7];
+
+}
+
+#endif
+
+FLOAT _Complex CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
+{
+	BLASLONG i;
+	BLASLONG ix,iy;
+	FLOAT _Complex result;
+	FLOAT  dot[8] = { 0.0, 0.0, 0.0 , 0.0, 0.0, 0.0, 0.0, 0.0 } ; 
+
+	if ( n <= 0 ) 
+	{
+	        __real__ result = 0.0 ;
+        	__imag__ result = 0.0 ;
+		return(result);
+
+	}
+
+	if ( (inc_x == 1) && (inc_y == 1) )
+	{
+
+		BLASLONG n1 = n & -16;
+
+		if ( n1 )
+		{
+			cdot_kernel_16(n1, x, y , dot );
+			dot[0] += dot[2];
+			dot[1] += dot[3];
+			dot[4] += dot[6];
+			dot[5] += dot[7];
+		}
+
+		i = n1;
+		BLASLONG j = i * 2;
+
+		while( i < n )
+		{
+
+			dot[0] += x[j]   * y[j]   ;
+			dot[1] += x[j+1] * y[j+1] ;
+			dot[4] += x[j]   * y[j+1] ;
+			dot[5] += x[j+1] * y[j]   ;
+
+			j+=2;
+			i++ ;
+
+		}
+
+
+	}
+	else
+	{
+		i=0;
+		ix=0;
+		iy=0;
+		inc_x <<= 1;
+		inc_y <<= 1;
+		while(i < n)
+		{
+
+			dot[0] += x[ix]   * y[iy]   ;
+			dot[1] += x[ix+1] * y[iy+1] ;
+			dot[4] += x[ix]   * y[iy+1] ;
+			dot[5] += x[ix+1] * y[iy]   ;
+
+			ix  += inc_x ;
+			iy  += inc_y ;
+			i++ ;
+
+		}
+	}
+
+#if !defined(CONJ)
+	__real__ result = dot[0] - dot[1];
+	__imag__ result = dot[4] + dot[5];
+#else
+	__real__ result = dot[0] + dot[1];
+	__imag__ result = dot[4] - dot[5];
+
+#endif
+
+	return(result);
+
+}
+
+
diff --git a/kernel/x86_64/cdot_microk_bulldozer-2.c b/kernel/x86_64/cdot_microk_bulldozer-2.c
new file mode 100644
index 000000000..f587aa036
--- /dev/null
+++ b/kernel/x86_64/cdot_microk_bulldozer-2.c
@@ -0,0 +1,196 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void cdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void cdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n <=1024 )
+  { 
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorps		%%xmm0, %%xmm0, %%xmm0	             \n\t"
+	"vxorps		%%xmm1, %%xmm1, %%xmm1	             \n\t"
+	"vxorps		%%xmm2, %%xmm2, %%xmm2	             \n\t"
+	"vxorps		%%xmm3, %%xmm3, %%xmm3	             \n\t"
+	"vxorps		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorps		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorps		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorps		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,4), %%xmm8          \n\t"  // 2 * x
+        "vmovups                16(%2,%0,4), %%xmm9          \n\t"  // 2 * x
+
+        "vmovups                  (%3,%0,4), %%xmm12         \n\t"  // 2 * y
+        "vmovups                16(%3,%0,4), %%xmm13         \n\t"  // 2 * y
+
+        "vmovups                32(%2,%0,4), %%xmm10         \n\t"  // 2 * x
+        "vmovups                48(%2,%0,4), %%xmm11         \n\t"  // 2 * x
+
+        "vmovups                32(%3,%0,4), %%xmm14         \n\t"  // 2 * y
+        "vmovups                48(%3,%0,4), %%xmm15         \n\t"  // 2 * y
+
+	"vfmaddps       %%xmm0, %%xmm8 , %%xmm12, %%xmm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmaddps       %%xmm1, %%xmm9 , %%xmm13, %%xmm1     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilps      $0xb1 , %%xmm12, %%xmm12               \n\t" // exchange real and imag part
+	"vpermilps      $0xb1 , %%xmm13, %%xmm13               \n\t"
+
+	"vfmaddps       %%xmm2, %%xmm10, %%xmm14, %%xmm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmaddps       %%xmm3, %%xmm11, %%xmm15, %%xmm3     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilps      $0xb1 , %%xmm14, %%xmm14               \n\t"
+	"vpermilps      $0xb1 , %%xmm15, %%xmm15               \n\t"
+
+	"vfmaddps       %%xmm4, %%xmm8 , %%xmm12, %%xmm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$16 , %0  	 	             \n\t"
+	"vfmaddps       %%xmm5, %%xmm9 , %%xmm13, %%xmm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmaddps       %%xmm6, %%xmm10, %%xmm14, %%xmm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $8 , %1			             \n\t"		
+	"vfmaddps       %%xmm7, %%xmm11, %%xmm15, %%xmm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddps        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddps        %%xmm2, %%xmm3, %%xmm2	\n\t"
+	"vaddps        %%xmm0, %%xmm2, %%xmm0	\n\t"
+
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddps        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddps        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vmovups        %%xmm0,    (%4)		\n\t"
+	"vmovups        %%xmm4,  16(%4)		\n\t"
+	"vzeroupper		        	\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+	return;
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorps		%%xmm0, %%xmm0, %%xmm0	             \n\t"
+	"vxorps		%%xmm1, %%xmm1, %%xmm1	             \n\t"
+	"vxorps		%%xmm2, %%xmm2, %%xmm2	             \n\t"
+	"vxorps		%%xmm3, %%xmm3, %%xmm3	             \n\t"
+	"vxorps		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorps		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorps		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorps		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+	"prefetcht0            384(%2,%0,4)		     \n\t"
+        "vmovups                  (%2,%0,4), %%xmm8          \n\t"  // 2 * x
+        "vmovups                16(%2,%0,4), %%xmm9          \n\t"  // 2 * x
+
+	"prefetcht0            384(%3,%0,4)		     \n\t"
+        "vmovups                  (%3,%0,4), %%xmm12         \n\t"  // 2 * y
+        "vmovups                16(%3,%0,4), %%xmm13         \n\t"  // 2 * y
+
+        "vmovups                32(%2,%0,4), %%xmm10         \n\t"  // 2 * x
+        "vmovups                48(%2,%0,4), %%xmm11         \n\t"  // 2 * x
+
+        "vmovups                32(%3,%0,4), %%xmm14         \n\t"  // 2 * y
+        "vmovups                48(%3,%0,4), %%xmm15         \n\t"  // 2 * y
+
+	"vfmaddps       %%xmm0, %%xmm8 , %%xmm12, %%xmm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmaddps       %%xmm1, %%xmm9 , %%xmm13, %%xmm1     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilps      $0xb1 , %%xmm12, %%xmm12               \n\t" // exchange real and imag part
+	"vpermilps      $0xb1 , %%xmm13, %%xmm13               \n\t"
+
+	"vfmaddps       %%xmm2, %%xmm10, %%xmm14, %%xmm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmaddps       %%xmm3, %%xmm11, %%xmm15, %%xmm3     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilps      $0xb1 , %%xmm14, %%xmm14               \n\t"
+	"vpermilps      $0xb1 , %%xmm15, %%xmm15               \n\t"
+
+	"vfmaddps       %%xmm4, %%xmm8 , %%xmm12, %%xmm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$16 , %0  	 	             \n\t"
+	"vfmaddps       %%xmm5, %%xmm9 , %%xmm13, %%xmm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmaddps       %%xmm6, %%xmm10, %%xmm14, %%xmm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $8 , %1			             \n\t"		
+	"vfmaddps       %%xmm7, %%xmm11, %%xmm15, %%xmm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddps        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddps        %%xmm2, %%xmm3, %%xmm2	\n\t"
+	"vaddps        %%xmm0, %%xmm2, %%xmm0	\n\t"
+
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddps        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddps        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vmovups        %%xmm0,    (%4)		\n\t"
+	"vmovups        %%xmm4,  16(%4)		\n\t"
+	"vzeroupper		        	\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/cdot_microk_haswell-2.c b/kernel/x86_64/cdot_microk_haswell-2.c
new file mode 100644
index 000000000..fc76b138a
--- /dev/null
+++ b/kernel/x86_64/cdot_microk_haswell-2.c
@@ -0,0 +1,119 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void cdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void cdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorps		%%ymm0, %%ymm0, %%ymm0	             \n\t"
+	"vxorps		%%ymm1, %%ymm1, %%ymm1	             \n\t"
+	"vxorps		%%ymm2, %%ymm2, %%ymm2	             \n\t"
+	"vxorps		%%ymm3, %%ymm3, %%ymm3	             \n\t"
+	"vxorps		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorps		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorps		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorps		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16			             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,4), %%ymm8          \n\t"  // 2 * x
+        "vmovups                32(%2,%0,4), %%ymm9          \n\t"  // 2 * x
+
+        "vmovups                  (%3,%0,4), %%ymm12         \n\t"  // 2 * y
+        "vmovups                32(%3,%0,4), %%ymm13         \n\t"  // 2 * y
+
+        "vmovups                64(%2,%0,4), %%ymm10         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,4), %%ymm11         \n\t"  // 2 * x
+
+        "vmovups                64(%3,%0,4), %%ymm14         \n\t"  // 2 * y
+        "vmovups                96(%3,%0,4), %%ymm15         \n\t"  // 2 * y
+
+	"vfmadd231ps       %%ymm8 , %%ymm12, %%ymm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231ps       %%ymm9 , %%ymm13, %%ymm1     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermilps      $0xb1 , %%ymm12, %%ymm12               \n\t"
+	"vpermilps      $0xb1 , %%ymm13, %%ymm13               \n\t"
+
+	"vfmadd231ps       %%ymm10, %%ymm14, %%ymm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231ps       %%ymm11, %%ymm15, %%ymm3     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermilps      $0xb1 , %%ymm14, %%ymm14               \n\t"
+	"vpermilps      $0xb1 , %%ymm15, %%ymm15               \n\t"
+
+	"vfmadd231ps       %%ymm8 , %%ymm12, %%ymm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$32 , %0	  	 	             \n\t"
+	"vfmadd231ps       %%ymm9 , %%ymm13, %%ymm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmadd231ps       %%ymm10, %%ymm14, %%ymm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $16 , %1			             \n\t"		
+	"vfmadd231ps       %%ymm11, %%ymm15, %%ymm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddps        %%ymm0, %%ymm1, %%ymm0	\n\t"
+	"vaddps        %%ymm2, %%ymm3, %%ymm2	\n\t"
+	"vaddps        %%ymm0, %%ymm2, %%ymm0	\n\t"
+
+	"vaddps        %%ymm4, %%ymm5, %%ymm4	\n\t"
+	"vaddps        %%ymm6, %%ymm7, %%ymm6	\n\t"
+	"vaddps        %%ymm4, %%ymm6, %%ymm4	\n\t"
+
+	"vextractf128 $1 , %%ymm0 , %%xmm1	\n\t"
+	"vextractf128 $1 , %%ymm4 , %%xmm5	\n\t"
+
+	"vaddps        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+
+} 
+
+
diff --git a/kernel/x86_64/cdot_microk_sandy-2.c b/kernel/x86_64/cdot_microk_sandy-2.c
new file mode 100644
index 000000000..22cd79e2e
--- /dev/null
+++ b/kernel/x86_64/cdot_microk_sandy-2.c
@@ -0,0 +1,127 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void cdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void cdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorps		%%ymm0, %%ymm0, %%ymm0	             \n\t"
+	"vxorps		%%ymm1, %%ymm1, %%ymm1	             \n\t"
+	"vxorps		%%ymm2, %%ymm2, %%ymm2	             \n\t"
+	"vxorps		%%ymm3, %%ymm3, %%ymm3	             \n\t"
+	"vxorps		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorps		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorps		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorps		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16			             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,4), %%ymm8          \n\t"  // 2 * x
+        "vmovups                32(%2,%0,4), %%ymm9          \n\t"  // 2 * x
+
+        "vmovups                  (%3,%0,4), %%ymm12         \n\t"  // 2 * y
+        "vmovups                32(%3,%0,4), %%ymm13         \n\t"  // 2 * y
+
+        "vmovups                64(%3,%0,4), %%ymm14         \n\t"  // 2 * y
+        "vmovups                96(%3,%0,4), %%ymm15         \n\t"  // 2 * y
+
+	"vmulps		   %%ymm8 , %%ymm12, %%ymm10	\n\t"
+	"vmulps		   %%ymm9 , %%ymm13, %%ymm11	\n\t"
+	"vpermilps      $0xb1 , %%ymm12, %%ymm12               \n\t"
+	"vpermilps      $0xb1 , %%ymm13, %%ymm13               \n\t"
+	"vaddps		   %%ymm0 , %%ymm10, %%ymm0	\n\t"
+	"vaddps		   %%ymm1 , %%ymm11, %%ymm1	\n\t"
+	"vmulps		   %%ymm8 , %%ymm12, %%ymm10	\n\t"
+	"vmulps		   %%ymm9 , %%ymm13, %%ymm11	\n\t"
+
+        "vmovups                64(%2,%0,4), %%ymm8         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,4), %%ymm9         \n\t"  // 2 * x
+
+	"vaddps		   %%ymm4 , %%ymm10, %%ymm4	\n\t"
+	"vaddps		   %%ymm5 , %%ymm11, %%ymm5	\n\t"
+
+	"vmulps		   %%ymm8 , %%ymm14, %%ymm10	\n\t"
+	"vmulps		   %%ymm9 , %%ymm15, %%ymm11	\n\t"
+	"vpermilps      $0xb1 , %%ymm14, %%ymm14               \n\t"
+	"vpermilps      $0xb1 , %%ymm15, %%ymm15               \n\t"
+	"vaddps		   %%ymm2 , %%ymm10, %%ymm2	\n\t"
+	"vaddps		   %%ymm3 , %%ymm11, %%ymm3	\n\t"
+	"vmulps		   %%ymm8 , %%ymm14, %%ymm10	\n\t"
+	"vmulps		   %%ymm9 , %%ymm15, %%ymm11	\n\t"
+	"addq		$32 , %0	  	 	             \n\t"
+	"subq	        $16 , %1			             \n\t"		
+	"vaddps		   %%ymm6 , %%ymm10, %%ymm6	\n\t"
+	"vaddps		   %%ymm7 , %%ymm11, %%ymm7	\n\t"
+
+	"jnz		1b		             \n\t"
+
+	"vaddps        %%ymm0, %%ymm1, %%ymm0	\n\t"
+	"vaddps        %%ymm2, %%ymm3, %%ymm2	\n\t"
+	"vaddps        %%ymm0, %%ymm2, %%ymm0	\n\t"
+
+	"vaddps        %%ymm4, %%ymm5, %%ymm4	\n\t"
+	"vaddps        %%ymm6, %%ymm7, %%ymm6	\n\t"
+	"vaddps        %%ymm4, %%ymm6, %%ymm4	\n\t"
+
+	"vextractf128 $1 , %%ymm0 , %%xmm1	\n\t"
+	"vextractf128 $1 , %%ymm4 , %%xmm5	\n\t"
+
+	"vaddps        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+
+} 
+
+
diff --git a/kernel/x86_64/cdot_microk_steamroller-2.c b/kernel/x86_64/cdot_microk_steamroller-2.c
new file mode 100644
index 000000000..76a3aa0eb
--- /dev/null
+++ b/kernel/x86_64/cdot_microk_steamroller-2.c
@@ -0,0 +1,196 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void cdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void cdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n < 1280 )
+  { 
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorps		%%xmm0, %%xmm0, %%xmm0	             \n\t"
+	"vxorps		%%xmm1, %%xmm1, %%xmm1	             \n\t"
+	"vxorps		%%xmm2, %%xmm2, %%xmm2	             \n\t"
+	"vxorps		%%xmm3, %%xmm3, %%xmm3	             \n\t"
+	"vxorps		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorps		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorps		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorps		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,4), %%xmm8          \n\t"  // 2 * x
+        "vmovups                16(%2,%0,4), %%xmm9          \n\t"  // 2 * x
+
+        "vmovups                  (%3,%0,4), %%xmm12         \n\t"  // 2 * y
+        "vmovups                16(%3,%0,4), %%xmm13         \n\t"  // 2 * y
+
+        "vmovups                32(%2,%0,4), %%xmm10         \n\t"  // 2 * x
+        "vmovups                48(%2,%0,4), %%xmm11         \n\t"  // 2 * x
+
+        "vmovups                32(%3,%0,4), %%xmm14         \n\t"  // 2 * y
+        "vmovups                48(%3,%0,4), %%xmm15         \n\t"  // 2 * y
+
+	"vfmadd231ps       %%xmm8 , %%xmm12, %%xmm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231ps       %%xmm9 , %%xmm13, %%xmm1     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilps      $0xb1 , %%xmm12, %%xmm12               \n\t" // exchange real and imag part
+	"vpermilps      $0xb1 , %%xmm13, %%xmm13               \n\t"
+
+	"vfmadd231ps       %%xmm10, %%xmm14, %%xmm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231ps       %%xmm11, %%xmm15, %%xmm3     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilps      $0xb1 , %%xmm14, %%xmm14               \n\t"
+	"vpermilps      $0xb1 , %%xmm15, %%xmm15               \n\t"
+
+	"vfmadd231ps       %%xmm8 , %%xmm12, %%xmm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$16 , %0  	 	             \n\t"
+	"vfmadd231ps       %%xmm9 , %%xmm13, %%xmm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmadd231ps       %%xmm10, %%xmm14, %%xmm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $8 , %1			             \n\t"		
+	"vfmadd231ps       %%xmm11, %%xmm15, %%xmm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddps        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddps        %%xmm2, %%xmm3, %%xmm2	\n\t"
+	"vaddps        %%xmm0, %%xmm2, %%xmm0	\n\t"
+
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddps        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddps        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vmovups        %%xmm0,    (%4)		\n\t"
+	"vmovups        %%xmm4,  16(%4)		\n\t"
+	"vzeroupper		        	\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+	return;
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorps		%%xmm0, %%xmm0, %%xmm0	             \n\t"
+	"vxorps		%%xmm1, %%xmm1, %%xmm1	             \n\t"
+	"vxorps		%%xmm2, %%xmm2, %%xmm2	             \n\t"
+	"vxorps		%%xmm3, %%xmm3, %%xmm3	             \n\t"
+	"vxorps		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorps		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorps		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorps		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+	"prefetcht0            512(%2,%0,4)		     \n\t"
+        "vmovups                  (%2,%0,4), %%xmm8          \n\t"  // 2 * x
+        "vmovups                16(%2,%0,4), %%xmm9          \n\t"  // 2 * x
+
+	"prefetcht0            512(%3,%0,4)		     \n\t"
+        "vmovups                  (%3,%0,4), %%xmm12         \n\t"  // 2 * y
+        "vmovups                16(%3,%0,4), %%xmm13         \n\t"  // 2 * y
+
+        "vmovups                32(%2,%0,4), %%xmm10         \n\t"  // 2 * x
+        "vmovups                48(%2,%0,4), %%xmm11         \n\t"  // 2 * x
+
+        "vmovups                32(%3,%0,4), %%xmm14         \n\t"  // 2 * y
+        "vmovups                48(%3,%0,4), %%xmm15         \n\t"  // 2 * y
+
+	"vfmadd231ps       %%xmm8 , %%xmm12, %%xmm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231ps       %%xmm9 , %%xmm13, %%xmm1     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilps      $0xb1 , %%xmm12, %%xmm12               \n\t" // exchange real and imag part
+	"vpermilps      $0xb1 , %%xmm13, %%xmm13               \n\t"
+
+	"vfmadd231ps       %%xmm10, %%xmm14, %%xmm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231ps       %%xmm11, %%xmm15, %%xmm3     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilps      $0xb1 , %%xmm14, %%xmm14               \n\t"
+	"vpermilps      $0xb1 , %%xmm15, %%xmm15               \n\t"
+
+	"vfmadd231ps       %%xmm8 , %%xmm12, %%xmm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$16 , %0  	 	             \n\t"
+	"vfmadd231ps       %%xmm9 , %%xmm13, %%xmm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmadd231ps       %%xmm10, %%xmm14, %%xmm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $8 , %1			             \n\t"		
+	"vfmadd231ps       %%xmm11, %%xmm15, %%xmm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddps        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddps        %%xmm2, %%xmm3, %%xmm2	\n\t"
+	"vaddps        %%xmm0, %%xmm2, %%xmm0	\n\t"
+
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddps        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddps        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vmovups        %%xmm0,    (%4)		\n\t"
+	"vmovups        %%xmm4,  16(%4)		\n\t"
+	"vzeroupper		        	\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/cscal.c b/kernel/x86_64/cscal.c
new file mode 100644
index 000000000..5d86b1929
--- /dev/null
+++ b/kernel/x86_64/cscal.c
@@ -0,0 +1,452 @@
+/***************************************************************************
+Copyright (c) 2013 - 2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+
+#if defined(HASWELL)
+#include "cscal_microk_haswell-2.c"
+#elif defined(BULLDOZER)  || defined(PILEDRIVER)
+#include "cscal_microk_bulldozer-2.c"
+#elif defined(STEAMROLLER)
+#include "cscal_microk_steamroller-2.c"
+#elif defined(SANDYBRIDGE)
+#include "cscal_microk_bulldozer-2.c"
+#endif
+
+
+#if !defined(HAVE_KERNEL_16)
+
+static void cscal_kernel_16( BLASLONG n, FLOAT *alpha , FLOAT *x )  __attribute__ ((noinline));
+static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha , FLOAT *x )  __attribute__ ((noinline));
+static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha , FLOAT *x )  __attribute__ ((noinline));
+static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha , FLOAT *x )  __attribute__ ((noinline));
+
+static void cscal_kernel_16( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	FLOAT da_r = alpha[0];
+	FLOAT da_i = alpha[1];
+	FLOAT t0,t1,t2,t3;
+
+	for( i=0; i<n; i+=4 )
+	{
+		t0 = da_r *x[0] - da_i *x[1];	
+		t1 = da_r *x[2] - da_i *x[3];	
+		t2 = da_r *x[4] - da_i *x[5];	
+		t3 = da_r *x[6] - da_i *x[7];	
+
+		x[1] = da_r * x[1] + da_i * x[0];
+		x[3] = da_r * x[3] + da_i * x[2];
+		x[5] = da_r * x[5] + da_i * x[4];
+		x[7] = da_r * x[7] + da_i * x[6];
+		
+		x[0] = t0;
+		x[2] = t1;
+		x[4] = t2;
+		x[6] = t3;
+
+		x+=8;
+	}
+
+}
+
+
+static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	FLOAT da_i = alpha[1];
+	FLOAT t0,t1,t2,t3;
+
+	for( i=0; i<n; i+=4 )
+	{
+		t0 =  - da_i *x[1];	
+		t1 =  - da_i *x[3];	
+		t2 =  - da_i *x[5];	
+		t3 =  - da_i *x[7];	
+
+		x[1] =  da_i * x[0];
+		x[3] =  da_i * x[2];
+		x[5] =  da_i * x[4];
+		x[7] =  da_i * x[6];
+		
+		x[0] = t0;
+		x[2] = t1;
+		x[4] = t2;
+		x[6] = t3;
+
+		x+=8;
+	}
+
+}
+
+
+static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	FLOAT da_r = alpha[0];
+	FLOAT t0,t1,t2,t3;
+
+	for( i=0; i<n; i+=4 )
+	{
+		t0 = da_r *x[0];	
+		t1 = da_r *x[2];	
+		t2 = da_r *x[4];	
+		t3 = da_r *x[6];	
+
+		x[1] = da_r * x[1];
+		x[3] = da_r * x[3];
+		x[5] = da_r * x[5];
+		x[7] = da_r * x[7];
+		
+		x[0] = t0;
+		x[2] = t1;
+		x[4] = t2;
+		x[6] = t3;
+
+		x+=8;
+	}
+
+}
+
+
+
+static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	for( i=0; i<n; i+=4 )
+	{
+		x[0] = 0.0;	
+		x[1] = 0.0;	
+		x[2] = 0.0;	
+		x[3] = 0.0;	
+		x[4] = 0.0;	
+		x[5] = 0.0;	
+		x[6] = 0.0;	
+		x[7] = 0.0;	
+		x+=8;
+	}
+
+}
+
+#endif
+
+static void cscal_kernel_inc_8(BLASLONG n, FLOAT *alpha, FLOAT *x, BLASLONG inc_x)  __attribute__ ((noinline));
+
+static void cscal_kernel_inc_8(BLASLONG n, FLOAT *alpha, FLOAT *x, BLASLONG inc_x)
+{
+
+	BLASLONG i;
+	BLASLONG inc_x2 = 2 * inc_x;
+	BLASLONG inc_x3 = inc_x2 + inc_x;
+	FLOAT t0,t1,t2,t3;
+	FLOAT da_r = alpha[0];
+	FLOAT da_i = alpha[1];
+
+	for ( i=0; i<n; i+=4 )
+	{
+		t0 = da_r * x[0]         - da_i *x[1];	
+		t1 = da_r * x[inc_x]     - da_i *x[inc_x  + 1];	
+		t2 = da_r * x[inc_x2]    - da_i *x[inc_x2 + 1];	
+		t3 = da_r * x[inc_x3]    - da_i *x[inc_x3 + 1];	
+
+		x[1]               = da_i * x[0]       + da_r * x[1];
+		x[inc_x  +1]       = da_i * x[inc_x]   + da_r * x[inc_x  +1];
+		x[inc_x2 +1]       = da_i * x[inc_x2]  + da_r * x[inc_x2 +1];
+		x[inc_x3 +1]       = da_i * x[inc_x3]  + da_r * x[inc_x3 +1];
+
+		x[0]        = t0;
+		x[inc_x]    = t1;
+		x[inc_x2]   = t2;
+		x[inc_x3]   = t3;
+
+		x+=4*inc_x;
+
+	}
+
+
+}
+
+
+int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da_r, FLOAT da_i, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT *dummy, BLASLONG dummy2)
+{
+	BLASLONG i=0,j=0;
+	FLOAT temp0;
+	FLOAT temp1;
+	FLOAT alpha[2];
+
+	if ( inc_x != 1 )
+	{
+		inc_x <<= 1;
+
+		if ( da_r == 0.0 )
+		{
+
+			BLASLONG n1 = n & -2;
+
+			if ( da_i == 0.0 )
+			{
+
+				while(j < n1)
+				{
+			
+					x[i]=0.0;
+					x[i+1]=0.0;
+					x[i+inc_x]=0.0;
+					x[i+1+inc_x]=0.0;
+					i += 2*inc_x ;
+					j+=2;
+
+				}
+
+				while(j < n)
+				{
+			
+					x[i]=0.0;
+					x[i+1]=0.0;
+					i += inc_x ;
+					j++;
+
+				}
+
+			}
+			else
+			{
+
+				while(j < n1)
+				{
+			
+					temp0        = -da_i * x[i+1];
+					x[i+1]       =  da_i * x[i];
+					x[i]         =  temp0;
+					temp1        = -da_i * x[i+1+inc_x];
+					x[i+1+inc_x] =  da_i * x[i+inc_x];
+					x[i+inc_x]   =  temp1;
+					i += 2*inc_x ;
+					j+=2;
+
+				}
+
+				while(j < n)
+				{
+			
+					temp0        = -da_i * x[i+1];
+					x[i+1]       =  da_i * x[i];
+					x[i]         =  temp0;
+					i += inc_x ;
+					j++;
+
+				}
+
+
+
+			}
+
+		}
+		else
+		{
+
+
+			if ( da_i == 0.0 )
+			{
+				BLASLONG n1 = n & -2;
+
+				while(j < n1)
+				{
+			
+					temp0        =  da_r * x[i];
+					x[i+1]       =  da_r * x[i+1];
+					x[i]         =  temp0;
+					temp1        =  da_r * x[i+inc_x];
+					x[i+1+inc_x] =  da_r * x[i+1+inc_x];
+					x[i+inc_x]   =  temp1;
+					i += 2*inc_x ;
+					j+=2;
+
+				}
+
+				while(j < n)
+				{
+			
+					temp0        =  da_r * x[i];
+					x[i+1]       =  da_r * x[i+1];
+					x[i]         =  temp0;
+					i += inc_x ;
+					j++;
+
+				}
+
+			}
+			else
+			{
+
+				BLASLONG n1 = n & -8;
+				if ( n1 > 0 )
+				{
+					alpha[0] = da_r;
+			                alpha[1] = da_i;
+					cscal_kernel_inc_8(n1, alpha, x, inc_x);
+					j = n1 ;
+					i = n1 * inc_x;
+				}
+
+				while(j < n)
+				{
+
+					temp0        =  da_r * x[i]   - da_i * x[i+1];
+					x[i+1]       =  da_r * x[i+1] + da_i * x[i];
+					x[i]         =  temp0;
+					i += inc_x ;
+					j++;
+
+				}
+
+			}
+
+		}
+
+		return(0);
+	}
+
+
+	BLASLONG n1 = n & -16;
+	if ( n1 > 0 )
+	{
+
+		alpha[0] = da_r;
+		alpha[1] = da_i;
+	
+		if ( da_r == 0.0 )
+			if ( da_i == 0 )
+				cscal_kernel_16_zero(n1 , alpha , x);
+			else
+				cscal_kernel_16_zero_r(n1 , alpha , x);
+		else
+			if ( da_i == 0 )
+				cscal_kernel_16_zero_i(n1 , alpha , x);
+			else
+				cscal_kernel_16(n1 , alpha , x);
+
+		i = n1 << 1;
+		j = n1;
+	}
+
+
+	if ( da_r == 0.0 )
+	{
+
+		if ( da_i == 0.0 )
+		{
+
+			while(j < n)
+			{
+		
+					x[i]=0.0;
+					x[i+1]=0.0;
+					i += 2 ;
+					j++;
+
+			}
+
+		}
+		else
+		{
+
+			while(j < n)
+			{
+			
+				temp0        = -da_i * x[i+1];
+				x[i+1]       =  da_i * x[i];
+				x[i]         =  temp0;
+				i += 2 ;
+				j++;
+
+			}
+
+		}
+
+	}
+	else
+	{
+
+		if ( da_i == 0.0 )
+		{
+
+			while(j < n)
+			{
+			
+					temp0        =  da_r * x[i];
+					x[i+1]       =  da_r * x[i+1];
+					x[i]         =  temp0;
+					i += 2 ;
+					j++;
+
+			}
+
+		}
+		else
+		{
+
+			BLASLONG n2 = n & -2;
+
+			while(j < n2)
+			{
+
+				temp0        =  da_r * x[i]   - da_i * x[i+1];
+				temp1        =  da_r * x[i+2] - da_i * x[i+3];
+				x[i+1]       =  da_r * x[i+1] + da_i * x[i];
+				x[i+3]       =  da_r * x[i+3] + da_i * x[i+2];
+				x[i]         =  temp0;
+				x[i+2]       =  temp1;
+				i += 4 ;
+				j+=2;
+
+			}
+
+			while(j < n)
+			{
+
+				temp0        =  da_r * x[i]   - da_i * x[i+1];
+				x[i+1]       =  da_r * x[i+1] + da_i * x[i];
+				x[i]         =  temp0;
+				i += 2 ;
+				j++;
+
+			}
+
+		}		
+
+	}
+
+	return(0);
+}
+
+
diff --git a/kernel/x86_64/cscal_microk_bulldozer-2.c b/kernel/x86_64/cscal_microk_bulldozer-2.c
new file mode 100644
index 000000000..f470cf843
--- /dev/null
+++ b/kernel/x86_64/cscal_microk_bulldozer-2.c
@@ -0,0 +1,348 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+
+#define HAVE_KERNEL_16 1
+
+static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss    		(%2), %%xmm0		    \n\t"  // da_r	
+	"vbroadcastss              4(%2), %%xmm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%xmm4		    \n\t"
+	"vmovups	-112(%1), %%xmm5		    \n\t"
+	"vmovups	 -96(%1), %%xmm6		    \n\t"
+	"vmovups	 -80(%1), %%xmm7		    \n\t"
+
+	"vpermilps	$0xb1 , %%xmm4, %%xmm12		    \n\t"
+	"vpermilps	$0xb1 , %%xmm5, %%xmm13		    \n\t"
+	"vpermilps	$0xb1 , %%xmm6, %%xmm14 	    \n\t"
+	"vpermilps	$0xb1 , %%xmm7, %%xmm15		    \n\t"
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	"prefetcht0     320(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulps		%%xmm0, %%xmm4 , %%xmm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	 -64(%1), %%xmm4		    \n\t"
+	"vmulps		%%xmm0, %%xmm5 , %%xmm9		    \n\t"
+	"vmovups	 -48(%1), %%xmm5		    \n\t"
+	"vmulps		%%xmm0, %%xmm6 , %%xmm10	    \n\t" 
+	"vmovups	 -32(%1), %%xmm6		    \n\t"
+	"vmulps		%%xmm0, %%xmm7 , %%xmm11	    \n\t" 
+	"vmovups	 -16(%1), %%xmm7		    \n\t"
+
+	"vmulps		%%xmm1, %%xmm12, %%xmm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%xmm12 , %%xmm8 , %%xmm8	    \n\t"
+	"vmulps		%%xmm1, %%xmm13, %%xmm13	    \n\t" 
+	"vaddsubps	%%xmm13 , %%xmm9 , %%xmm9	    \n\t"
+	"vmulps		%%xmm1, %%xmm14, %%xmm14	    \n\t" 
+	"vaddsubps	%%xmm14 , %%xmm10, %%xmm10	    \n\t"
+	"vmulps		%%xmm1, %%xmm15, %%xmm15	    \n\t" 
+	"vaddsubps	%%xmm15 , %%xmm11, %%xmm11	    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%xmm4, %%xmm12		    \n\t"
+	"vpermilps	$0xb1 , %%xmm5, %%xmm13		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%xmm6, %%xmm14 	    \n\t"
+	"vpermilps	$0xb1 , %%xmm7, %%xmm15		    \n\t"
+
+	"addq		$64  ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulps		%%xmm0, %%xmm4 , %%xmm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulps		%%xmm0, %%xmm5 , %%xmm9		    \n\t"
+	"vmulps		%%xmm0, %%xmm6 , %%xmm10	    \n\t" 
+	"vmulps		%%xmm0, %%xmm7 , %%xmm11	    \n\t" 
+
+	"vmulps		%%xmm1, %%xmm12, %%xmm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%xmm12 , %%xmm8 , %%xmm8	    \n\t"
+	"vmulps		%%xmm1, %%xmm13, %%xmm13	    \n\t" 
+	"vaddsubps	%%xmm13 , %%xmm9 , %%xmm9	    \n\t"
+	"vmulps		%%xmm1, %%xmm14, %%xmm14	    \n\t" 
+	"vaddsubps	%%xmm14 , %%xmm10, %%xmm10	    \n\t"
+	"vmulps		%%xmm1, %%xmm15, %%xmm15	    \n\t" 
+	"vaddsubps	%%xmm15 , %%xmm11, %%xmm11	    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorps	           %%xmm0, %%xmm0, %%xmm0	    \n\t"
+	"vbroadcastss              4(%2), %%xmm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%xmm4		    \n\t"
+	"vmovups	-112(%1), %%xmm5		    \n\t"
+	"vmovups	 -96(%1), %%xmm6		    \n\t"
+	"vmovups	 -80(%1), %%xmm7		    \n\t"
+
+	"vpermilps	$0xb1 , %%xmm4, %%xmm12		    \n\t"
+	"vpermilps	$0xb1 , %%xmm5, %%xmm13		    \n\t"
+	"vpermilps	$0xb1 , %%xmm6, %%xmm14 	    \n\t"
+	"vpermilps	$0xb1 , %%xmm7, %%xmm15		    \n\t"
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	 -64(%1), %%xmm4		    \n\t"
+	"vmovups	 -48(%1), %%xmm5		    \n\t"
+	"vmovups	 -32(%1), %%xmm6		    \n\t"
+	"vmovups	 -16(%1), %%xmm7		    \n\t"
+
+	"vmulps		%%xmm1, %%xmm12, %%xmm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%xmm12 , %%xmm0 , %%xmm8	    \n\t"
+	"vmulps		%%xmm1, %%xmm13, %%xmm13	    \n\t" 
+	"vaddsubps	%%xmm13 , %%xmm0 , %%xmm9	    \n\t"
+	"vmulps		%%xmm1, %%xmm14, %%xmm14	    \n\t" 
+	"vaddsubps	%%xmm14 , %%xmm0 , %%xmm10	    \n\t"
+	"vmulps		%%xmm1, %%xmm15, %%xmm15	    \n\t" 
+	"vaddsubps	%%xmm15 , %%xmm0 , %%xmm11	    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%xmm4, %%xmm12		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%xmm5, %%xmm13		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%xmm6, %%xmm14 	    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%xmm7, %%xmm15		    \n\t"
+
+	"addq		$64  ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"vmulps		%%xmm1, %%xmm12, %%xmm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%xmm12 , %%xmm0 , %%xmm8	    \n\t"
+	"vmulps		%%xmm1, %%xmm13, %%xmm13	    \n\t" 
+	"vaddsubps	%%xmm13 , %%xmm0 , %%xmm9	    \n\t"
+	"vmulps		%%xmm1, %%xmm14, %%xmm14	    \n\t" 
+	"vaddsubps	%%xmm14 , %%xmm0 , %%xmm10	    \n\t"
+	"vmulps		%%xmm1, %%xmm15, %%xmm15	    \n\t" 
+	"vaddsubps	%%xmm15 , %%xmm0 , %%xmm11	    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
+static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss    		(%2), %%xmm0		    \n\t"  // da_r	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%xmm4		    \n\t"
+	"vmovups	-112(%1), %%xmm5		    \n\t"
+	"vmovups	 -96(%1), %%xmm6		    \n\t"
+	"vmovups	 -80(%1), %%xmm7		    \n\t"
+
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulps		%%xmm0, %%xmm4 , %%xmm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	 -64(%1), %%xmm4		    \n\t"
+	"vmulps		%%xmm0, %%xmm5 , %%xmm9		    \n\t"
+	"vmovups	 -48(%1), %%xmm5		    \n\t"
+	"vmulps		%%xmm0, %%xmm6 , %%xmm10	    \n\t" 
+	"vmovups	 -32(%1), %%xmm6		    \n\t"
+	"vmulps		%%xmm0, %%xmm7 , %%xmm11	    \n\t" 
+	"vmovups	 -16(%1), %%xmm7		    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+
+	"addq		$64  ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulps		%%xmm0, %%xmm4 , %%xmm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulps		%%xmm0, %%xmm5 , %%xmm9		    \n\t"
+	"vmulps		%%xmm0, %%xmm6 , %%xmm10	    \n\t" 
+	"vmulps		%%xmm0, %%xmm7 , %%xmm11	    \n\t" 
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorps	           %%xmm0, %%xmm0, %%xmm0	    \n\t"
+
+	"addq	$128, %1				    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	%%xmm0 , -128(%1)		    \n\t"
+	"vmovups	%%xmm0 , -112(%1)		    \n\t"
+	"vmovups	%%xmm0 ,  -96(%1)		    \n\t"
+	"vmovups	%%xmm0 ,  -80(%1)		    \n\t"
+
+	"addq		$64  ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
diff --git a/kernel/x86_64/cscal_microk_haswell-2.c b/kernel/x86_64/cscal_microk_haswell-2.c
new file mode 100644
index 000000000..0424de3a5
--- /dev/null
+++ b/kernel/x86_64/cscal_microk_haswell-2.c
@@ -0,0 +1,348 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+
+#define HAVE_KERNEL_16 1
+
+static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss		(%2), %%ymm0		    \n\t"  // da_r	
+	"vbroadcastss          4(%2), %%ymm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+	"vpermilps	$0xb1 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilps	$0xb1 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilps	$0xb1 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilps	$0xb1 , %%ymm7, %%ymm15		    \n\t"
+
+	"subq	        $16, %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulps		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmulps		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmulps		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmulps		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmulps		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%ymm12 , %%ymm8 , %%ymm8	    \n\t"
+	"vmulps		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubps	%%ymm13 , %%ymm9 , %%ymm9	    \n\t"
+	"vmulps		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubps	%%ymm14 , %%ymm10, %%ymm10	    \n\t"
+	"vmulps		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubps	%%ymm15 , %%ymm11, %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm4, %%ymm12		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm5, %%ymm13		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm6, %%ymm14 	    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm7, %%ymm15		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $16, %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulps		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulps		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmulps		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmulps		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+
+	"vmulps		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%ymm12 , %%ymm8 , %%ymm8	    \n\t"
+	"vmulps		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubps	%%ymm13 , %%ymm9 , %%ymm9	    \n\t"
+	"vmulps		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubps	%%ymm14 , %%ymm10, %%ymm10	    \n\t"
+	"vmulps		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubps	%%ymm15 , %%ymm11, %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "0", "1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorps	           %%ymm0, %%ymm0, %%ymm0	    \n\t"
+	"vbroadcastss          4(%2), %%ymm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+	"vpermilps	$0xb1 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilps	$0xb1 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilps	$0xb1 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilps	$0xb1 , %%ymm7, %%ymm15		    \n\t"
+
+	"subq	        $16, %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmulps		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%ymm12 , %%ymm0 , %%ymm8	    \n\t"
+	"vmulps		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubps	%%ymm13 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulps		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubps	%%ymm14 , %%ymm0 , %%ymm10	    \n\t"
+	"vmulps		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubps	%%ymm15 , %%ymm0 , %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm4, %%ymm12		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm5, %%ymm13		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm6, %%ymm14 	    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm7, %%ymm15		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $16, %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"vmulps		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%ymm12 , %%ymm0 , %%ymm8	    \n\t"
+	"vmulps		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubps	%%ymm13 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulps		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubps	%%ymm14 , %%ymm0 , %%ymm10	    \n\t"
+	"vmulps		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubps	%%ymm15 , %%ymm0 , %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "0", "1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
+static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss		(%2), %%ymm0		    \n\t"  // da_r	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+
+	"subq	        $16, %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulps		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmulps		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmulps		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmulps		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $16, %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulps		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulps		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmulps		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmulps		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorps	           %%ymm0, %%ymm0, %%ymm0	    \n\t"
+
+	"addq	$128, %1				    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	%%ymm0 , -128(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -64(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -32(%1)		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $16, %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "0", "1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
diff --git a/kernel/x86_64/cscal_microk_steamroller-2.c b/kernel/x86_64/cscal_microk_steamroller-2.c
new file mode 100644
index 000000000..763e7add4
--- /dev/null
+++ b/kernel/x86_64/cscal_microk_steamroller-2.c
@@ -0,0 +1,349 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+
+#define HAVE_KERNEL_16 1
+
+static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss		(%2), %%ymm0		    \n\t"  // da_r	
+	"vbroadcastss          4(%2), %%ymm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+	"vpermilps	$0xb1 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilps	$0xb1 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilps	$0xb1 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilps	$0xb1 , %%ymm7, %%ymm15		    \n\t"
+
+	"subq	        $16, %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	"prefetcht0     512(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulps		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmulps		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"prefetcht0     768(%1)				    \n\t"
+	"vmulps		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmulps		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmulps		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vmulps		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubps	%%ymm12 , %%ymm8 , %%ymm8	    \n\t"
+	"vaddsubps	%%ymm13 , %%ymm9 , %%ymm9	    \n\t"
+	"vmulps		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vmulps		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubps	%%ymm14 , %%ymm10, %%ymm10	    \n\t"
+	"vaddsubps	%%ymm15 , %%ymm11, %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilps	$0xb1 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilps	$0xb1 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilps	$0xb1 , %%ymm7, %%ymm15		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $16, %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulps		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulps		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmulps		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmulps		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+
+	"vmulps		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%ymm12 , %%ymm8 , %%ymm8	    \n\t"
+	"vmulps		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubps	%%ymm13 , %%ymm9 , %%ymm9	    \n\t"
+	"vmulps		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubps	%%ymm14 , %%ymm10, %%ymm10	    \n\t"
+	"vmulps		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubps	%%ymm15 , %%ymm11, %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "0", "1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorps	           %%ymm0, %%ymm0, %%ymm0	    \n\t"
+	"vbroadcastss          4(%2), %%ymm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+	"vpermilps	$0xb1 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilps	$0xb1 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilps	$0xb1 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilps	$0xb1 , %%ymm7, %%ymm15		    \n\t"
+
+	"subq	        $16, %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmulps		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%ymm12 , %%ymm0 , %%ymm8	    \n\t"
+	"vmulps		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubps	%%ymm13 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulps		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubps	%%ymm14 , %%ymm0 , %%ymm10	    \n\t"
+	"vmulps		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubps	%%ymm15 , %%ymm0 , %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm4, %%ymm12		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm5, %%ymm13		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm6, %%ymm14 	    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+	"vpermilps	$0xb1 , %%ymm7, %%ymm15		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $16, %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"vmulps		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubps	%%ymm12 , %%ymm0 , %%ymm8	    \n\t"
+	"vmulps		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubps	%%ymm13 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulps		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubps	%%ymm14 , %%ymm0 , %%ymm10	    \n\t"
+	"vmulps		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubps	%%ymm15 , %%ymm0 , %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "0", "1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
+static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss		(%2), %%ymm0		    \n\t"  // da_r	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+
+	"subq	        $16, %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulps		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmulps		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmulps		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmulps		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $16, %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulps		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulps		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmulps		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmulps		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorps	           %%ymm0, %%ymm0, %%ymm0	    \n\t"
+
+	"addq	$128, %1				    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	%%ymm0 , -128(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -64(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -32(%1)		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $16, %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "0", "1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
diff --git a/kernel/x86_64/daxpy.c b/kernel/x86_64/daxpy.c
index fd5343eba..56d323cbe 100644
--- a/kernel/x86_64/daxpy.c
+++ b/kernel/x86_64/daxpy.c
@@ -31,8 +31,16 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(NEHALEM)
 #include "daxpy_microk_nehalem-2.c"
-#elif defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)
+#elif defined(BULLDOZER)
 #include "daxpy_microk_bulldozer-2.c"
+#elif defined(STEAMROLLER)
+#include "daxpy_microk_steamroller-2.c"
+#elif defined(PILEDRIVER)
+#include "daxpy_microk_piledriver-2.c"
+#elif defined(HASWELL)
+#include "daxpy_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "daxpy_microk_sandy-2.c"
 #endif
 
 
@@ -71,7 +79,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-		int n1 = n & -8;
+		BLASLONG n1 = n & -16;
 
 		if ( n1 )
 			daxpy_kernel_8(n1, x, y , &da );
@@ -89,6 +97,27 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 
 	}
 
+	BLASLONG n1 = n & -4;
+
+	while(i < n1)
+	{
+
+		FLOAT m1      = da * x[ix] ;
+		FLOAT m2      = da * x[ix+inc_x] ;
+		FLOAT m3      = da * x[ix+2*inc_x] ;
+		FLOAT m4      = da * x[ix+3*inc_x] ;
+
+		y[iy]         += m1 ;
+		y[iy+inc_y]   += m2 ;
+		y[iy+2*inc_y] += m3 ;
+		y[iy+3*inc_y] += m4 ;
+
+		ix  += inc_x*4 ;
+		iy  += inc_y*4 ;
+		i+=4 ;
+
+	}
+
 	while(i < n)
 	{
 
diff --git a/kernel/x86_64/daxpy_microk_haswell-2.c b/kernel/x86_64/daxpy_microk_haswell-2.c
new file mode 100644
index 000000000..db117a8ba
--- /dev/null
+++ b/kernel/x86_64/daxpy_microk_haswell-2.c
@@ -0,0 +1,78 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void daxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void daxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vbroadcastsd		(%4), %%ymm0		    \n\t"  // alpha	
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+        "vmovups                  (%3,%0,8), %%ymm12         \n\t"  // 4 * y
+        "vmovups                32(%3,%0,8), %%ymm13         \n\t"  // 4 * y
+        "vmovups                64(%3,%0,8), %%ymm14         \n\t"  // 4 * y
+        "vmovups                96(%3,%0,8), %%ymm15         \n\t"  // 4 * y
+	"vfmadd231pd       (%2,%0,8), %%ymm0  , %%ymm12  	     \n\t"   // y += alpha * x
+	"vfmadd231pd     32(%2,%0,8), %%ymm0  , %%ymm13  	     \n\t"   // y += alpha * x
+	"vfmadd231pd     64(%2,%0,8), %%ymm0  , %%ymm14  	     \n\t"   // y += alpha * x
+	"vfmadd231pd     96(%2,%0,8), %%ymm0  , %%ymm15  	     \n\t"   // y += alpha * x
+	"vmovups	%%ymm12,   (%3,%0,8)		     \n\t"
+	"vmovups	%%ymm13, 32(%3,%0,8)		     \n\t"
+	"vmovups	%%ymm14, 64(%3,%0,8)		     \n\t"
+	"vmovups	%%ymm15, 96(%3,%0,8)		     \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/daxpy_microk_piledriver-2.c b/kernel/x86_64/daxpy_microk_piledriver-2.c
new file mode 100644
index 000000000..95eb953b4
--- /dev/null
+++ b/kernel/x86_64/daxpy_microk_piledriver-2.c
@@ -0,0 +1,160 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void daxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void daxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n < 640 )
+  {
+
+	__asm__  __volatile__
+	(
+	"vmovddup		(%4), %%xmm0		    		\n\t"  // alpha	
+
+	".align 16				            		\n\t"
+	"1:				            			\n\t"
+
+        "vmovups                  (%3,%0,8), %%xmm8          		\n\t"  // 2  y
+        "vmovups                16(%3,%0,8), %%xmm9         		\n\t"  // 2  y
+        "vmovups                32(%3,%0,8), %%xmm10         		\n\t"  // 2  y
+        "vmovups                48(%3,%0,8), %%xmm11         		\n\t"  // 2  y
+
+        "vmovups                64(%3,%0,8), %%xmm12         		\n\t"  // 2  y
+        "vmovups                80(%3,%0,8), %%xmm13        		\n\t"  // 2  y
+        "vmovups                96(%3,%0,8), %%xmm14         		\n\t"  // 2  y
+        "vmovups               112(%3,%0,8), %%xmm15         		\n\t"  // 2  y
+
+	"vfmadd231pd       (%2,%0,8), %%xmm0 , %%xmm8  			\n\t"   // y += alpha * x
+	"vfmadd231pd     16(%2,%0,8), %%xmm0 , %%xmm9 		  	\n\t"   // y += alpha * x
+	"vfmadd231pd     32(%2,%0,8), %%xmm0 , %%xmm10 			\n\t"   // y += alpha * x
+	"vfmadd231pd     48(%2,%0,8), %%xmm0 , %%xmm11 			\n\t"   // y += alpha * x
+
+	"vfmadd231pd     64(%2,%0,8), %%xmm0 , %%xmm12 			\n\t"   // y += alpha * x
+	"vfmadd231pd     80(%2,%0,8), %%xmm0 , %%xmm13		  	\n\t"   // y += alpha * x
+	"vfmadd231pd     96(%2,%0,8), %%xmm0 , %%xmm14 			\n\t"   // y += alpha * x
+	"vfmadd231pd    112(%2,%0,8), %%xmm0 , %%xmm15 			\n\t"   // y += alpha * x
+
+	"vmovups		%%xmm8 ,   (%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm9 , 16(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm10, 32(%3,%0,8)           		\n\t"
+	"vmovups		%%xmm11, 48(%3,%0,8)                 	\n\t"
+
+	"vmovups		%%xmm12, 64(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm13, 80(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm14, 96(%3,%0,8)           		\n\t"
+	"vmovups		%%xmm15,112(%3,%0,8)                 	\n\t"
+
+	"addq		$16, %0	  	 	             		\n\t"
+	"subq	        $16, %1			             		\n\t"		
+	"jnz		1b		             			\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+  }
+
+
+	__asm__  __volatile__
+	(
+	"vmovddup		(%4), %%xmm0		    		\n\t"  // alpha	
+
+	".align 16				            		\n\t"
+	"1:				            			\n\t"
+
+	"prefetcht0	       512(%3,%0,8)				\n\t"
+        "vmovups                  (%3,%0,8), %%xmm8          		\n\t"  // 2  y
+        "vmovups                16(%3,%0,8), %%xmm9         		\n\t"  // 2  y
+        "vmovups                32(%3,%0,8), %%xmm10         		\n\t"  // 2  y
+        "vmovups                48(%3,%0,8), %%xmm11         		\n\t"  // 2  y
+
+	"prefetcht0	       576(%3,%0,8)				\n\t"
+        "vmovups                64(%3,%0,8), %%xmm12         		\n\t"  // 2  y
+        "vmovups                80(%3,%0,8), %%xmm13        		\n\t"  // 2  y
+        "vmovups                96(%3,%0,8), %%xmm14         		\n\t"  // 2  y
+        "vmovups               112(%3,%0,8), %%xmm15         		\n\t"  // 2  y
+
+	"prefetcht0	       512(%2,%0,8)				\n\t"
+	"vfmadd231pd       (%2,%0,8), %%xmm0 , %%xmm8  			\n\t"   // y += alpha * x
+	"vfmadd231pd     16(%2,%0,8), %%xmm0 , %%xmm9 		  	\n\t"   // y += alpha * x
+	"vfmadd231pd     32(%2,%0,8), %%xmm0 , %%xmm10 			\n\t"   // y += alpha * x
+	"vfmadd231pd     48(%2,%0,8), %%xmm0 , %%xmm11 			\n\t"   // y += alpha * x
+
+	"prefetcht0	       576(%2,%0,8)				\n\t"
+	"vfmadd231pd     64(%2,%0,8), %%xmm0 , %%xmm12 			\n\t"   // y += alpha * x
+	"vfmadd231pd     80(%2,%0,8), %%xmm0 , %%xmm13		  	\n\t"   // y += alpha * x
+	"vfmadd231pd     96(%2,%0,8), %%xmm0 , %%xmm14 			\n\t"   // y += alpha * x
+	"vfmadd231pd    112(%2,%0,8), %%xmm0 , %%xmm15 			\n\t"   // y += alpha * x
+
+	"vmovups		%%xmm8 ,   (%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm9 , 16(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm10, 32(%3,%0,8)           		\n\t"
+	"vmovups		%%xmm11, 48(%3,%0,8)                 	\n\t"
+
+	"vmovups		%%xmm12, 64(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm13, 80(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm14, 96(%3,%0,8)           		\n\t"
+	"vmovups		%%xmm15,112(%3,%0,8)                 	\n\t"
+
+	"addq		$16, %0	  	 	             		\n\t"
+	"subq	        $16, %1			             		\n\t"		
+	"jnz		1b		             			\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/daxpy_microk_sandy-2.c b/kernel/x86_64/daxpy_microk_sandy-2.c
new file mode 100644
index 000000000..522e084dc
--- /dev/null
+++ b/kernel/x86_64/daxpy_microk_sandy-2.c
@@ -0,0 +1,119 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void daxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void daxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vbroadcastsd		(%4), %%ymm0		    \n\t"  // alpha	
+	"vmovups	  (%3,%0,8), %%ymm8         \n\t"
+	"vmovups	32(%3,%0,8), %%ymm9         \n\t"
+	"vmovups	64(%3,%0,8), %%ymm10        \n\t"
+	"vmovups	96(%3,%0,8), %%ymm11        \n\t"
+	"vmovups	  (%2,%0,8), %%ymm4         \n\t"
+	"vmovups	32(%2,%0,8), %%ymm5         \n\t"
+	"vmovups	64(%2,%0,8), %%ymm6         \n\t"
+	"vmovups	96(%2,%0,8), %%ymm7        \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
+	"jz		2f		             \n\t"
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmulpd		%%ymm4, %%ymm0, %%ymm4		\n\t"
+	"vaddpd		%%ymm8 , %%ymm4, %%ymm12	     \n\t"
+	"vmulpd		%%ymm5, %%ymm0, %%ymm5		\n\t"
+	"vaddpd		%%ymm9 , %%ymm5, %%ymm13	     \n\t"
+	"vmulpd		%%ymm6, %%ymm0, %%ymm6		\n\t"
+	"vaddpd		%%ymm10, %%ymm6, %%ymm14	     \n\t"
+	"vmulpd		%%ymm7, %%ymm0, %%ymm7		\n\t"
+	"vaddpd		%%ymm11, %%ymm7, %%ymm15	     \n\t"
+
+	"vmovups	  (%3,%0,8), %%ymm8         \n\t"
+	"vmovups	32(%3,%0,8), %%ymm9         \n\t"
+	"vmovups	64(%3,%0,8), %%ymm10        \n\t"
+	"vmovups	96(%3,%0,8), %%ymm11        \n\t"
+
+	"vmovups	  (%2,%0,8), %%ymm4         \n\t"
+	"vmovups	32(%2,%0,8), %%ymm5         \n\t"
+	"vmovups	64(%2,%0,8), %%ymm6         \n\t"
+	"vmovups	96(%2,%0,8), %%ymm7        \n\t"
+
+	"vmovups	%%ymm12, -128(%3,%0,8)		     \n\t"
+	"vmovups	%%ymm13,  -96(%3,%0,8)		     \n\t"
+	"vmovups	%%ymm14,  -64(%3,%0,8)		     \n\t"
+	"vmovups	%%ymm15,  -32(%3,%0,8)		     \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+
+	"2:				            \n\t"
+	"vmulpd		%%ymm4, %%ymm0, %%ymm4		\n\t"
+	"vmulpd		%%ymm5, %%ymm0, %%ymm5		\n\t"
+	"vmulpd		%%ymm6, %%ymm0, %%ymm6		\n\t"
+	"vmulpd		%%ymm7, %%ymm0, %%ymm7		\n\t"
+
+	"vaddpd		%%ymm8 , %%ymm4, %%ymm12	     \n\t"
+	"vaddpd		%%ymm9 , %%ymm5, %%ymm13	     \n\t"
+	"vaddpd		%%ymm10, %%ymm6, %%ymm14	     \n\t"
+	"vaddpd		%%ymm11, %%ymm7, %%ymm15	     \n\t"
+
+	"vmovups	%%ymm12, -128(%3,%0,8)		     \n\t"
+	"vmovups	%%ymm13,  -96(%3,%0,8)		     \n\t"
+	"vmovups	%%ymm14,  -64(%3,%0,8)		     \n\t"
+	"vmovups	%%ymm15,  -32(%3,%0,8)		     \n\t"
+
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/daxpy_microk_steamroller-2.c b/kernel/x86_64/daxpy_microk_steamroller-2.c
new file mode 100644
index 000000000..e40009037
--- /dev/null
+++ b/kernel/x86_64/daxpy_microk_steamroller-2.c
@@ -0,0 +1,160 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void daxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void daxpy_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n < 2048 )
+  {
+
+	__asm__  __volatile__
+	(
+	"vmovddup		(%4), %%xmm0		    		\n\t"  // alpha	
+
+	".align 16				            		\n\t"
+	"1:				            			\n\t"
+
+        "vmovups                  (%3,%0,8), %%xmm8          		\n\t"  // 2  y
+        "vmovups                16(%3,%0,8), %%xmm9         		\n\t"  // 2  y
+        "vmovups                32(%3,%0,8), %%xmm10         		\n\t"  // 2  y
+        "vmovups                48(%3,%0,8), %%xmm11         		\n\t"  // 2  y
+
+        "vmovups                64(%3,%0,8), %%xmm12         		\n\t"  // 2  y
+        "vmovups                80(%3,%0,8), %%xmm13        		\n\t"  // 2  y
+        "vmovups                96(%3,%0,8), %%xmm14         		\n\t"  // 2  y
+        "vmovups               112(%3,%0,8), %%xmm15         		\n\t"  // 2  y
+
+	"vfmadd231pd       (%2,%0,8), %%xmm0 , %%xmm8  			\n\t"   // y += alpha * x
+	"vfmadd231pd     16(%2,%0,8), %%xmm0 , %%xmm9 		  	\n\t"   // y += alpha * x
+	"vfmadd231pd     32(%2,%0,8), %%xmm0 , %%xmm10 			\n\t"   // y += alpha * x
+	"vfmadd231pd     48(%2,%0,8), %%xmm0 , %%xmm11 			\n\t"   // y += alpha * x
+
+	"vfmadd231pd     64(%2,%0,8), %%xmm0 , %%xmm12 			\n\t"   // y += alpha * x
+	"vfmadd231pd     80(%2,%0,8), %%xmm0 , %%xmm13		  	\n\t"   // y += alpha * x
+	"vfmadd231pd     96(%2,%0,8), %%xmm0 , %%xmm14 			\n\t"   // y += alpha * x
+	"vfmadd231pd    112(%2,%0,8), %%xmm0 , %%xmm15 			\n\t"   // y += alpha * x
+
+	"vmovups		%%xmm8 ,   (%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm9 , 16(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm10, 32(%3,%0,8)           		\n\t"
+	"vmovups		%%xmm11, 48(%3,%0,8)                 	\n\t"
+
+	"vmovups		%%xmm12, 64(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm13, 80(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm14, 96(%3,%0,8)           		\n\t"
+	"vmovups		%%xmm15,112(%3,%0,8)                 	\n\t"
+
+	"addq		$16, %0	  	 	             		\n\t"
+	"subq	        $16, %1			             		\n\t"		
+	"jnz		1b		             			\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+  }
+
+
+	__asm__  __volatile__
+	(
+	"vmovddup		(%4), %%xmm0		    		\n\t"  // alpha	
+
+	".align 16				            		\n\t"
+	"1:				            			\n\t"
+
+	"prefetcht0	       512(%3,%0,8)				\n\t"
+        "vmovups                  (%3,%0,8), %%xmm8          		\n\t"  // 2  y
+        "vmovups                16(%3,%0,8), %%xmm9         		\n\t"  // 2  y
+        "vmovups                32(%3,%0,8), %%xmm10         		\n\t"  // 2  y
+        "vmovups                48(%3,%0,8), %%xmm11         		\n\t"  // 2  y
+
+	"prefetcht0	       576(%3,%0,8)				\n\t"
+        "vmovups                64(%3,%0,8), %%xmm12         		\n\t"  // 2  y
+        "vmovups                80(%3,%0,8), %%xmm13        		\n\t"  // 2  y
+        "vmovups                96(%3,%0,8), %%xmm14         		\n\t"  // 2  y
+        "vmovups               112(%3,%0,8), %%xmm15         		\n\t"  // 2  y
+
+	"prefetcht0	       512(%2,%0,8)				\n\t"
+	"vfmadd231pd       (%2,%0,8), %%xmm0 , %%xmm8  			\n\t"   // y += alpha * x
+	"vfmadd231pd     16(%2,%0,8), %%xmm0 , %%xmm9 		  	\n\t"   // y += alpha * x
+	"vfmadd231pd     32(%2,%0,8), %%xmm0 , %%xmm10 			\n\t"   // y += alpha * x
+	"vfmadd231pd     48(%2,%0,8), %%xmm0 , %%xmm11 			\n\t"   // y += alpha * x
+
+	"prefetcht0	       576(%2,%0,8)				\n\t"
+	"vfmadd231pd     64(%2,%0,8), %%xmm0 , %%xmm12 			\n\t"   // y += alpha * x
+	"vfmadd231pd     80(%2,%0,8), %%xmm0 , %%xmm13		  	\n\t"   // y += alpha * x
+	"vfmadd231pd     96(%2,%0,8), %%xmm0 , %%xmm14 			\n\t"   // y += alpha * x
+	"vfmadd231pd    112(%2,%0,8), %%xmm0 , %%xmm15 			\n\t"   // y += alpha * x
+
+	"vmovups		%%xmm8 ,   (%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm9 , 16(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm10, 32(%3,%0,8)           		\n\t"
+	"vmovups		%%xmm11, 48(%3,%0,8)                 	\n\t"
+
+	"vmovups		%%xmm12, 64(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm13, 80(%3,%0,8)                 	\n\t"
+	"vmovups		%%xmm14, 96(%3,%0,8)           		\n\t"
+	"vmovups		%%xmm15,112(%3,%0,8)                 	\n\t"
+
+	"addq		$16, %0	  	 	             		\n\t"
+	"subq	        $16, %1			             		\n\t"		
+	"jnz		1b		             			\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/ddot.c b/kernel/x86_64/ddot.c
index d501c2f68..4bf8082c9 100644
--- a/kernel/x86_64/ddot.c
+++ b/kernel/x86_64/ddot.c
@@ -29,10 +29,18 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 
 
-#if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)
+#if defined(BULLDOZER) 
 #include "ddot_microk_bulldozer-2.c"
-#elif defined(NEHALEM)
+#elif defined(STEAMROLLER)
+#include "ddot_microk_steamroller-2.c"
+#elif defined(PILEDRIVER)
+#include "ddot_microk_piledriver-2.c"
+#elif defined(NEHALEM) 
 #include "ddot_microk_nehalem-2.c"
+#elif defined(HASWELL)
+#include "ddot_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "ddot_microk_sandy-2.c"
 #endif
 
 
@@ -75,12 +83,11 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-		int n1 = n & -8;
+		BLASLONG n1 = n & -16;
 
 		if ( n1 )
 			ddot_kernel_8(n1, x, y , &dot );
 
-
 		i = n1;
 		while(i < n)
 		{
@@ -94,15 +101,40 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
 
 	}
 
+	FLOAT temp1 = 0.0;
+	FLOAT temp2 = 0.0;
+
+        BLASLONG n1 = n & -4;	
+
+	while(i < n1)
+	{
+
+		FLOAT m1 = y[iy]       * x[ix] ;
+		FLOAT m2 = y[iy+inc_y] * x[ix+inc_x] ;
+
+		FLOAT m3 = y[iy+2*inc_y] * x[ix+2*inc_x] ;
+		FLOAT m4 = y[iy+3*inc_y] * x[ix+3*inc_x] ;
+
+		ix  += inc_x*4 ;
+		iy  += inc_y*4 ;
+
+		temp1 += m1+m3;
+		temp2 += m2+m4;
+
+		i+=4 ;
+
+	}
+
 	while(i < n)
 	{
 
-		dot += y[iy] * x[ix] ;
+		temp1 += y[iy] * x[ix] ;
 		ix  += inc_x ;
 		iy  += inc_y ;
 		i++ ;
 
 	}
+	dot = temp1 + temp2;
 	return(dot);
 
 }
diff --git a/kernel/x86_64/ddot_microk_haswell-2.c b/kernel/x86_64/ddot_microk_haswell-2.c
new file mode 100644
index 000000000..d36577af3
--- /dev/null
+++ b/kernel/x86_64/ddot_microk_haswell-2.c
@@ -0,0 +1,95 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void ddot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void ddot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vxorpd		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorpd		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorpd		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorpd		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,8), %%ymm12         \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%ymm13         \n\t"  // 2 * x
+        "vmovups                64(%2,%0,8), %%ymm14         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,8), %%ymm15         \n\t"  // 2 * x
+
+	"vfmadd231pd      (%3,%0,8), %%ymm12, %%ymm4 \n\t"  // 2 * y
+	"vfmadd231pd    32(%3,%0,8), %%ymm13, %%ymm5 \n\t"  // 2 * y
+	"vfmadd231pd    64(%3,%0,8), %%ymm14, %%ymm6 \n\t"  // 2 * y
+	"vfmadd231pd    96(%3,%0,8), %%ymm15, %%ymm7 \n\t"  // 2 * y
+
+	"addq		$16 , %0	  	     \n\t"
+	"subq	        $16 , %1		     \n\t"		
+	"jnz		1b		             \n\t"
+
+	"vextractf128	$1 , %%ymm4 , %%xmm12	     \n\t"
+	"vextractf128	$1 , %%ymm5 , %%xmm13	     \n\t"
+	"vextractf128	$1 , %%ymm6 , %%xmm14	     \n\t"
+	"vextractf128	$1 , %%ymm7 , %%xmm15	     \n\t"
+
+	"vaddpd        %%xmm4, %%xmm12, %%xmm4	\n\t"
+	"vaddpd        %%xmm5, %%xmm13, %%xmm5	\n\t"
+	"vaddpd        %%xmm6, %%xmm14, %%xmm6	\n\t"
+	"vaddpd        %%xmm7, %%xmm15, %%xmm7	\n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddpd        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovsd		%%xmm4,    (%4)		\n\t"
+	"vzeroupper				\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm4", "%xmm5", 
+	  "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/ddot_microk_piledriver-2.c b/kernel/x86_64/ddot_microk_piledriver-2.c
new file mode 100644
index 000000000..ac950885c
--- /dev/null
+++ b/kernel/x86_64/ddot_microk_piledriver-2.c
@@ -0,0 +1,165 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void ddot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void ddot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n < 1408 )
+  {
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorpd		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorpd		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorpd		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,8), %%xmm12         \n\t"  // 2 * x
+        "vmovups                16(%2,%0,8), %%xmm13         \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%xmm14         \n\t"  // 2 * x
+        "vmovups                48(%2,%0,8), %%xmm15         \n\t"  // 2 * x
+
+	"vfmadd231pd         (%3,%0,8), %%xmm12, %%xmm4 \n\t"  // 2 * y
+        "vmovups                64(%2,%0,8), %%xmm0         \n\t"  // 2 * x
+        "vmovups                80(%2,%0,8), %%xmm1         \n\t"  // 2 * x
+	"vfmadd231pd       16(%3,%0,8), %%xmm13, %%xmm5 \n\t"  // 2 * y
+        "vmovups                96(%2,%0,8), %%xmm2         \n\t"  // 2 * x
+        "vmovups               112(%2,%0,8), %%xmm3         \n\t"  // 2 * x
+	"vfmadd231pd       32(%3,%0,8), %%xmm14, %%xmm6 \n\t"  // 2 * y
+	"vfmadd231pd       48(%3,%0,8), %%xmm15, %%xmm7 \n\t"  // 2 * y
+
+
+	"vfmadd231pd       64(%3,%0,8), %%xmm0 , %%xmm4 \n\t"  // 2 * y
+	"vfmadd231pd       80(%3,%0,8), %%xmm1 , %%xmm5 \n\t"  // 2 * y
+	"vfmadd231pd       96(%3,%0,8), %%xmm2 , %%xmm6 \n\t"  // 2 * y
+	"vfmadd231pd      112(%3,%0,8), %%xmm3 , %%xmm7 \n\t"  // 2 * y
+
+	"addq		$16 , %0	  	 	             \n\t"
+	"subq	        $16 , %1			             \n\t"		
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddpd        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovsd		%%xmm4,    (%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorpd		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorpd		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorpd		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+	"prefetcht0	768(%2,%0,8)			     \n\t"
+	"prefetcht0	832(%2,%0,8)			     \n\t"
+        "vmovups                  (%2,%0,8), %%xmm12         \n\t"  // 2 * x
+        "vmovups                16(%2,%0,8), %%xmm13         \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%xmm14         \n\t"  // 2 * x
+        "vmovups                48(%2,%0,8), %%xmm15         \n\t"  // 2 * x
+
+	"prefetcht0	768(%3,%0,8)			     \n\t"
+	"prefetcht0	832(%3,%0,8)			     \n\t"
+	"vfmadd231pd         (%3,%0,8), %%xmm12, %%xmm4 \n\t"  // 2 * y
+        "vmovups                64(%2,%0,8), %%xmm0         \n\t"  // 2 * x
+        "vmovups                80(%2,%0,8), %%xmm1         \n\t"  // 2 * x
+	"vfmadd231pd       16(%3,%0,8), %%xmm13, %%xmm5 \n\t"  // 2 * y
+        "vmovups                96(%2,%0,8), %%xmm2         \n\t"  // 2 * x
+        "vmovups               112(%2,%0,8), %%xmm3         \n\t"  // 2 * x
+	"vfmadd231pd       32(%3,%0,8), %%xmm14, %%xmm6 \n\t"  // 2 * y
+	"vfmadd231pd       48(%3,%0,8), %%xmm15, %%xmm7 \n\t"  // 2 * y
+
+
+	"vfmadd231pd       64(%3,%0,8), %%xmm0 , %%xmm4 \n\t"  // 2 * y
+	"vfmadd231pd       80(%3,%0,8), %%xmm1 , %%xmm5 \n\t"  // 2 * y
+	"vfmadd231pd       96(%3,%0,8), %%xmm2 , %%xmm6 \n\t"  // 2 * y
+	"vfmadd231pd      112(%3,%0,8), %%xmm3 , %%xmm7 \n\t"  // 2 * y
+
+	"addq		$16 , %0	  	 	             \n\t"
+	"subq	        $16 , %1			             \n\t"		
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddpd        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovsd		%%xmm4,    (%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/ddot_microk_sandy-2.c b/kernel/x86_64/ddot_microk_sandy-2.c
new file mode 100644
index 000000000..e2e6701c7
--- /dev/null
+++ b/kernel/x86_64/ddot_microk_sandy-2.c
@@ -0,0 +1,100 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void ddot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void ddot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vxorpd		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorpd		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorpd		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorpd		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,8), %%ymm12         \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%ymm13         \n\t"  // 2 * x
+        "vmovups                64(%2,%0,8), %%ymm14         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,8), %%ymm15         \n\t"  // 2 * x
+
+	"vmulpd      (%3,%0,8), %%ymm12, %%ymm12 \n\t"  // 2 * y
+	"vmulpd    32(%3,%0,8), %%ymm13, %%ymm13 \n\t"  // 2 * y
+	"vmulpd    64(%3,%0,8), %%ymm14, %%ymm14 \n\t"  // 2 * y
+	"vmulpd    96(%3,%0,8), %%ymm15, %%ymm15 \n\t"  // 2 * y
+
+	"vaddpd      %%ymm4 , %%ymm12, %%ymm4 \n\t"  // 2 * y
+	"vaddpd      %%ymm5 , %%ymm13, %%ymm5 \n\t"  // 2 * y
+	"vaddpd      %%ymm6 , %%ymm14, %%ymm6 \n\t"  // 2 * y
+	"vaddpd      %%ymm7 , %%ymm15, %%ymm7 \n\t"  // 2 * y
+
+	"addq		$16 , %0	  	     \n\t"
+	"subq	        $16 , %1		     \n\t"		
+	"jnz		1b		             \n\t"
+
+	"vextractf128	$1 , %%ymm4 , %%xmm12	     \n\t"
+	"vextractf128	$1 , %%ymm5 , %%xmm13	     \n\t"
+	"vextractf128	$1 , %%ymm6 , %%xmm14	     \n\t"
+	"vextractf128	$1 , %%ymm7 , %%xmm15	     \n\t"
+
+	"vaddpd        %%xmm4, %%xmm12, %%xmm4	\n\t"
+	"vaddpd        %%xmm5, %%xmm13, %%xmm5	\n\t"
+	"vaddpd        %%xmm6, %%xmm14, %%xmm6	\n\t"
+	"vaddpd        %%xmm7, %%xmm15, %%xmm7	\n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddpd        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovsd		%%xmm4,    (%4)		\n\t"
+	"vzeroupper				\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm4", "%xmm5", 
+	  "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/ddot_microk_steamroller-2.c b/kernel/x86_64/ddot_microk_steamroller-2.c
new file mode 100644
index 000000000..5ce20b5de
--- /dev/null
+++ b/kernel/x86_64/ddot_microk_steamroller-2.c
@@ -0,0 +1,97 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void ddot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void ddot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorpd		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorpd		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorpd		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,8), %%xmm12         \n\t"  // 2 * x
+        "vmovups                16(%2,%0,8), %%xmm13         \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%xmm14         \n\t"  // 2 * x
+        "vmovups                48(%2,%0,8), %%xmm15         \n\t"  // 2 * x
+
+	"vfmadd231pd         (%3,%0,8), %%xmm12, %%xmm4 \n\t"  // 2 * y
+        "vmovups                64(%2,%0,8), %%xmm0         \n\t"  // 2 * x
+        "vmovups                80(%2,%0,8), %%xmm1         \n\t"  // 2 * x
+	"vfmadd231pd       16(%3,%0,8), %%xmm13, %%xmm5 \n\t"  // 2 * y
+        "vmovups                96(%2,%0,8), %%xmm2         \n\t"  // 2 * x
+        "vmovups               112(%2,%0,8), %%xmm3         \n\t"  // 2 * x
+	"vfmadd231pd       32(%3,%0,8), %%xmm14, %%xmm6 \n\t"  // 2 * y
+	"vfmadd231pd       48(%3,%0,8), %%xmm15, %%xmm7 \n\t"  // 2 * y
+
+
+	"vfmadd231pd       64(%3,%0,8), %%xmm0 , %%xmm4 \n\t"  // 2 * y
+	"vfmadd231pd       80(%3,%0,8), %%xmm1 , %%xmm5 \n\t"  // 2 * y
+	"vfmadd231pd       96(%3,%0,8), %%xmm2 , %%xmm6 \n\t"  // 2 * y
+	"vfmadd231pd      112(%3,%0,8), %%xmm3 , %%xmm7 \n\t"  // 2 * y
+
+	"addq		$16 , %0	  	 	             \n\t"
+	"subq	        $16 , %1			             \n\t"		
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddpd        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovsd		%%xmm4,    (%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dgemm_kernel_4x8_haswell.S b/kernel/x86_64/dgemm_kernel_4x8_haswell.S
new file mode 100644
index 000000000..c84b599ce
--- /dev/null
+++ b/kernel/x86_64/dgemm_kernel_4x8_haswell.S
@@ -0,0 +1,4753 @@
+/*********************************************************************************
+Copyright (c) 2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+**********************************************************************************/
+
+
+#define ASSEMBLER
+#include "common.h"
+ 
+#define OLD_M	%rdi
+#define OLD_N	%rsi
+#define M	%r13
+#define J	%r14
+#define OLD_K	%rdx
+
+#define A	%rcx
+#define B	%r8
+#define C	%r9
+#define LDC	%r10
+	
+#define I	%r11
+#define AO	%rdi
+#define BO	%rsi
+#define	CO1	%r15
+#define K	%r12
+#define	SP	%rbx
+
+#define BO1	%rdi
+#define BO2	%r15
+#define BO3	%rbp
+
+#ifndef WINDOWS_ABI
+
+#define STACKSIZE 96
+#define L_BUFFER_SIZE 256*8*12+4096
+
+#else
+
+#define STACKSIZE 256
+#define L_BUFFER_SIZE 128*8*12+512
+
+#define OLD_A		40 + STACKSIZE(%rsp)
+#define OLD_B		48 + STACKSIZE(%rsp)
+#define OLD_C		56 + STACKSIZE(%rsp)
+#define OLD_LDC		64 + STACKSIZE(%rsp)
+#define OLD_OFFSET	72 + STACKSIZE(%rsp)
+
+#endif
+
+
+#define Ndiv12	 24(%rsp)
+#define Nmod12	 32(%rsp)
+#define N	 40(%rsp)
+#define ALPHA	 48(%rsp)
+#define OFFSET	 56(%rsp)
+#define KK	 64(%rsp)
+#define KKK	 72(%rsp)
+#define BUFFER1	           128(%rsp)
+
+#if defined(OS_WINDOWS)
+#if   L_BUFFER_SIZE > 16384
+#define STACK_TOUCH \
+        movl    $ 0,  4096 * 4(%rsp);\
+        movl    $ 0,  4096 * 3(%rsp);\
+        movl    $ 0,  4096 * 2(%rsp);\
+        movl    $ 0,  4096 * 1(%rsp);
+#elif L_BUFFER_SIZE > 12288
+#define STACK_TOUCH \
+        movl    $ 0,  4096 * 3(%rsp);\
+        movl    $ 0,  4096 * 2(%rsp);\
+        movl    $ 0,  4096 * 1(%rsp);
+#elif L_BUFFER_SIZE > 8192
+#define STACK_TOUCH \
+        movl    $ 0,  4096 * 2(%rsp);\
+        movl    $ 0,  4096 * 1(%rsp);
+#elif L_BUFFER_SIZE > 4096
+#define STACK_TOUCH \
+        movl    $ 0,  4096 * 1(%rsp);
+#else
+#define STACK_TOUCH
+#endif
+#else
+#define STACK_TOUCH
+#endif
+
+#define	A_PR1	512
+#define	B_PR1	512
+
+/*******************************************************************************************
+* Macro definitions
+*******************************************************************************************/
+
+.macro INIT4x12
+
+	vxorpd		%ymm4 , %ymm4 , %ymm4
+	vxorpd		%ymm5 , %ymm5 , %ymm5
+	vxorpd		%ymm6 , %ymm6 , %ymm6
+	vxorpd		%ymm7 , %ymm7 , %ymm7
+	vxorpd		%ymm8 , %ymm8 , %ymm8
+	vxorpd		%ymm9 , %ymm9 , %ymm9
+	vxorpd		%ymm10, %ymm10, %ymm10
+	vxorpd		%ymm11, %ymm11, %ymm11
+	vxorpd		%ymm12, %ymm12, %ymm12
+	vxorpd		%ymm13, %ymm13, %ymm13
+	vxorpd		%ymm14, %ymm14, %ymm14
+	vxorpd		%ymm15, %ymm15, %ymm15
+
+.endm
+
+.macro KERNEL4x12_I
+	prefetcht0	A_PR1(AO)
+	vmovups		-12 * SIZE(BO), %ymm1
+	prefetcht0	B_PR1(BO)
+	vmovups 	-16 * SIZE(AO), %ymm0
+	prefetcht0	B_PR1+64(BO)
+	vmovups		 -8 * SIZE(BO), %ymm2
+	prefetcht0	B_PR1+128(BO)
+	vmovups		 -4 * SIZE(BO), %ymm3
+	vmulpd  	%ymm0 ,%ymm1  , %ymm4
+	prefetcht0	B_PR1+192(BO)
+	vmulpd  	%ymm0 ,%ymm2  , %ymm8
+	vmulpd  	%ymm0 ,%ymm3  , %ymm12
+	prefetcht0	B_PR1+256(BO)
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm5
+	vmulpd  	%ymm0 ,%ymm2  , %ymm9
+	vmulpd  	%ymm0 ,%ymm3  , %ymm13
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm6
+	vmulpd  	%ymm0 ,%ymm2  , %ymm10
+
+	addq		$ 12*SIZE, BO
+	vmulpd  	%ymm0 ,%ymm3  , %ymm14
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		-12 * SIZE(BO), %ymm1
+	vmulpd  	%ymm0 ,%ymm2  , %ymm11
+	vmovups		 -8 * SIZE(BO), %ymm2
+	vmulpd  	%ymm0 ,%ymm3  , %ymm15
+	vmovups		 -4 * SIZE(BO), %ymm3
+
+.endm
+
+.macro KERNEL4x12_M1
+	prefetcht0	A_PR1(AO)
+	vmovups 	-16 * SIZE(AO), %ymm0
+	prefetcht0	B_PR1(BO)
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	prefetcht0	B_PR1+64(BO)
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm8
+	prefetcht0	B_PR1+128(BO)
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm12
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm9
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm13
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm10
+
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm14
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		-12 * SIZE(BO), %ymm1
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm11
+	vmovups		 -8 * SIZE(BO), %ymm2
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm15
+	vmovups		 -4 * SIZE(BO), %ymm3
+
+.endm
+
+.macro KERNEL4x12_M2
+	vmovups 	-12 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm8
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm12
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm9
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm13
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm10
+
+	addq		$ 8*SIZE, AO
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm14
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		  0 * SIZE(BO), %ymm1
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm11
+	vmovups		  4 * SIZE(BO), %ymm2
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm15
+	vmovups		  8 * SIZE(BO), %ymm3
+	addq		$ 24*SIZE, BO
+.endm
+
+
+.macro KERNEL4x12_E
+	vmovups 	-12 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm8
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm12
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm9
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm13
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm10
+
+	addq		$ 8*SIZE, AO
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm14
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm11
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm15
+	addq		$ 12*SIZE, BO
+.endm
+
+.macro KERNEL4x12_SUB
+	vmovups		-12 * SIZE(BO), %ymm1
+	vmovups 	-16 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vmovups		 -8 * SIZE(BO), %ymm2
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm8
+	vmovups		 -4 * SIZE(BO), %ymm3
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm12
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm9
+	addq		$ 12*SIZE, BO
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm13
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm10
+	addq		$ 4*SIZE, AO
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm14
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm11
+	vfmadd231pd  	%ymm0 ,%ymm3  , %ymm15
+
+.endm
+
+
+.macro SAVE4x12
+
+	vbroadcastsd	ALPHA, %ymm0
+
+	vmulpd	%ymm0 , %ymm4 , %ymm4
+	vmulpd	%ymm0 , %ymm5 , %ymm5
+	vmulpd	%ymm0 , %ymm6 , %ymm6
+	vmulpd	%ymm0 , %ymm7 , %ymm7
+
+	vmulpd	%ymm0 , %ymm8 , %ymm8
+	vmulpd	%ymm0 , %ymm9 , %ymm9
+	vmulpd	%ymm0 , %ymm10, %ymm10
+	vmulpd	%ymm0 , %ymm11, %ymm11
+
+	vmulpd	%ymm0 , %ymm12, %ymm12
+	vmulpd	%ymm0 , %ymm13, %ymm13
+	vmulpd	%ymm0 , %ymm14, %ymm14
+	vmulpd	%ymm0 , %ymm15, %ymm15
+
+	vpermpd $ 0xb1 , %ymm5, %ymm5
+	vpermpd $ 0xb1 , %ymm7, %ymm7
+
+	vblendpd $ 0x0a, %ymm5, %ymm4, %ymm0
+	vblendpd $ 0x05, %ymm5, %ymm4, %ymm1
+	vblendpd $ 0x0a, %ymm7, %ymm6, %ymm2
+	vblendpd $ 0x05, %ymm7, %ymm6, %ymm3
+
+	vpermpd $ 0x1b , %ymm2, %ymm2
+	vpermpd $ 0x1b , %ymm3, %ymm3
+	vpermpd $ 0xb1 , %ymm2, %ymm2
+	vpermpd $ 0xb1 , %ymm3, %ymm3
+
+	vblendpd $ 0x03, %ymm0, %ymm2 , %ymm4
+	vblendpd $ 0x03, %ymm1, %ymm3 , %ymm5
+	vblendpd $ 0x03, %ymm2, %ymm0 , %ymm6
+	vblendpd $ 0x03, %ymm3, %ymm1 , %ymm7
+
+        leaq    (CO1, LDC, 2), %rax     
+	
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1), %ymm4, %ymm4
+	vaddpd 	           (CO1, LDC), %ymm5, %ymm5
+	vaddpd 	               (%rax), %ymm6, %ymm6
+	vaddpd 	          (%rax, LDC), %ymm7, %ymm7
+
+#endif
+
+	vmovups	%ymm4 ,  	(CO1)
+	vmovups	%ymm5 ,  	(CO1, LDC)
+	vmovups	%ymm6 ,  	(%rax)
+	vmovups	%ymm7 ,  	(%rax, LDC)
+
+	prefetcht0	32(CO1)
+	prefetcht0	32(CO1,LDC)
+	prefetcht0	32(%rax)
+	prefetcht0	32(%rax,LDC)
+
+	vpermpd $ 0xb1 , %ymm9 , %ymm9
+	vpermpd $ 0xb1 , %ymm11, %ymm11
+
+	vblendpd $ 0x0a, %ymm9 , %ymm8 , %ymm0
+	vblendpd $ 0x05, %ymm9 , %ymm8 , %ymm1
+	vblendpd $ 0x0a, %ymm11, %ymm10, %ymm2
+	vblendpd $ 0x05, %ymm11, %ymm10, %ymm3
+
+	vpermpd $ 0x1b , %ymm2, %ymm2
+	vpermpd $ 0x1b , %ymm3, %ymm3
+	vpermpd $ 0xb1 , %ymm2, %ymm2
+	vpermpd $ 0xb1 , %ymm3, %ymm3
+
+	vblendpd $ 0x03, %ymm0, %ymm2 , %ymm4
+	vblendpd $ 0x03, %ymm1, %ymm3 , %ymm5
+	vblendpd $ 0x03, %ymm2, %ymm0 , %ymm6
+	vblendpd $ 0x03, %ymm3, %ymm1 , %ymm7
+
+
+	leaq	(%rax, LDC, 2), %rax
+	leaq	(%rax, LDC, 2), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (%rax), %ymm4, %ymm4
+	vaddpd 	           (%rax, LDC), %ymm5, %ymm5
+	vaddpd 	                (%rbp), %ymm6, %ymm6
+	vaddpd 	           (%rbp, LDC), %ymm7, %ymm7
+
+#endif
+
+	vmovups	%ymm4 ,  	(%rax)
+	vmovups	%ymm5 ,  	(%rax, LDC)
+	vmovups	%ymm6 ,  	(%rbp)
+	vmovups	%ymm7 ,  	(%rbp, LDC)
+
+	prefetcht0	32(%rax)
+	prefetcht0	32(%rax,LDC)
+	prefetcht0	32(%rbp)
+	prefetcht0	32(%rbp,LDC)
+
+	vpermpd $ 0xb1 , %ymm13, %ymm13
+	vpermpd $ 0xb1 , %ymm15, %ymm15
+
+	vblendpd $ 0x0a, %ymm13, %ymm12, %ymm0
+	vblendpd $ 0x05, %ymm13, %ymm12, %ymm1
+	vblendpd $ 0x0a, %ymm15, %ymm14, %ymm2
+	vblendpd $ 0x05, %ymm15, %ymm14, %ymm3
+
+	vpermpd $ 0x1b , %ymm2, %ymm2
+	vpermpd $ 0x1b , %ymm3, %ymm3
+	vpermpd $ 0xb1 , %ymm2, %ymm2
+	vpermpd $ 0xb1 , %ymm3, %ymm3
+
+	vblendpd $ 0x03, %ymm0, %ymm2 , %ymm4
+	vblendpd $ 0x03, %ymm1, %ymm3 , %ymm5
+	vblendpd $ 0x03, %ymm2, %ymm0 , %ymm6
+	vblendpd $ 0x03, %ymm3, %ymm1 , %ymm7
+
+
+	leaq	(%rax, LDC, 4), %rax
+	leaq	(%rbp, LDC, 4), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (%rax), %ymm4, %ymm4
+	vaddpd 	           (%rax, LDC), %ymm5, %ymm5
+	vaddpd 	                (%rbp), %ymm6, %ymm6
+	vaddpd 	           (%rbp, LDC), %ymm7, %ymm7
+
+#endif
+
+	vmovups	%ymm4 ,  	(%rax)
+	vmovups	%ymm5 ,  	(%rax, LDC)
+	vmovups	%ymm6 ,  	(%rbp)
+	vmovups	%ymm7 ,  	(%rbp, LDC)
+
+	prefetcht0	32(%rax)
+	prefetcht0	32(%rax,LDC)
+	prefetcht0	32(%rbp)
+	prefetcht0	32(%rbp,LDC)
+
+	addq	$ 4*SIZE, CO1
+.endm
+
+/******************************************************************************************/
+
+.macro INIT2x12
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm5 , %xmm5 , %xmm5
+	vxorpd		%xmm6 , %xmm6 , %xmm6
+	vxorpd		%xmm7 , %xmm7 , %xmm7
+	vxorpd		%xmm8 , %xmm8 , %xmm8
+	vxorpd		%xmm9 , %xmm9 , %xmm9
+	vxorpd		%xmm10, %xmm10, %xmm10
+	vxorpd		%xmm11, %xmm11, %xmm11
+	vxorpd		%xmm12, %xmm12, %xmm12
+	vxorpd		%xmm13, %xmm13, %xmm13
+	vxorpd		%xmm14, %xmm14, %xmm14
+	vxorpd		%xmm15, %xmm15, %xmm15
+
+.endm
+
+.macro KERNEL2x12_SUB
+	vmovups 	-16 * SIZE(AO), %xmm0
+	vmovddup	-12 * SIZE(BO), %xmm1
+	vmovddup	-11 * SIZE(BO), %xmm2
+	vmovddup	-10 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm1  , %xmm4
+	vmovddup	 -9 * SIZE(BO), %xmm1
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm5
+	vmovddup	 -8 * SIZE(BO), %xmm2
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm6
+	vmovddup	 -7 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm1  , %xmm7
+	vmovddup	 -6 * SIZE(BO), %xmm1
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm8
+	vmovddup	 -5 * SIZE(BO), %xmm2
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm9
+	vmovddup	 -4 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm1  , %xmm10
+	vmovddup	 -3 * SIZE(BO), %xmm1
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm11
+	vmovddup	 -2 * SIZE(BO), %xmm2
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm12
+	vmovddup	 -1 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm1  , %xmm13
+	addq		$ 12*SIZE, BO
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm14
+	addq		$ 2*SIZE, AO
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm15
+
+.endm
+
+.macro SAVE2x12
+
+	vmovddup	ALPHA, %xmm0
+
+	vmulpd	%xmm0 , %xmm4 , %xmm4
+	vmulpd	%xmm0 , %xmm5 , %xmm5
+	vmulpd	%xmm0 , %xmm6 , %xmm6
+	vmulpd	%xmm0 , %xmm7 , %xmm7
+
+	vmulpd	%xmm0 , %xmm8 , %xmm8
+	vmulpd	%xmm0 , %xmm9 , %xmm9
+	vmulpd	%xmm0 , %xmm10, %xmm10
+	vmulpd	%xmm0 , %xmm11, %xmm11
+
+	vmulpd	%xmm0 , %xmm12, %xmm12
+	vmulpd	%xmm0 , %xmm13, %xmm13
+	vmulpd	%xmm0 , %xmm14, %xmm14
+	vmulpd	%xmm0 , %xmm15, %xmm15
+
+
+        leaq    (CO1, LDC, 2), %rax     
+	
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1), %xmm4, %xmm4
+	vaddpd 	           (CO1, LDC), %xmm5, %xmm5
+	vaddpd 	               (%rax), %xmm6, %xmm6
+	vaddpd 	          (%rax, LDC), %xmm7, %xmm7
+
+#endif
+
+	vmovups	%xmm4 ,  	(CO1)
+	vmovups	%xmm5 ,  	(CO1, LDC)
+	vmovups	%xmm6 ,  	(%rax)
+	vmovups	%xmm7 ,  	(%rax, LDC)
+
+
+	leaq	(%rax, LDC, 2), %rax
+	leaq	(%rax, LDC, 2), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (%rax), %xmm8 , %xmm4
+	vaddpd 	           (%rax, LDC), %xmm9 , %xmm5
+	vaddpd 	                (%rbp), %xmm10, %xmm6
+	vaddpd 	           (%rbp, LDC), %xmm11, %xmm7
+
+#endif
+
+	vmovups	%xmm4 ,  	(%rax)
+	vmovups	%xmm5 ,  	(%rax, LDC)
+	vmovups	%xmm6 ,  	(%rbp)
+	vmovups	%xmm7 ,  	(%rbp, LDC)
+
+
+	leaq	(%rax, LDC, 4), %rax
+	leaq	(%rbp, LDC, 4), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (%rax), %xmm12, %xmm4
+	vaddpd 	           (%rax, LDC), %xmm13, %xmm5
+	vaddpd 	                (%rbp), %xmm14, %xmm6
+	vaddpd 	           (%rbp, LDC), %xmm15, %xmm7
+
+#endif
+
+	vmovups	%xmm4 ,  	(%rax)
+	vmovups	%xmm5 ,  	(%rax, LDC)
+	vmovups	%xmm6 ,  	(%rbp)
+	vmovups	%xmm7 ,  	(%rbp, LDC)
+
+	addq	$ 2*SIZE, CO1
+.endm
+
+
+/******************************************************************************************/
+
+.macro INIT1x12
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm5 , %xmm5 , %xmm5
+	vxorpd		%xmm6 , %xmm6 , %xmm6
+	vxorpd		%xmm7 , %xmm7 , %xmm7
+	vxorpd		%xmm8 , %xmm8 , %xmm8
+	vxorpd		%xmm9 , %xmm9 , %xmm9
+	vxorpd		%xmm10, %xmm10, %xmm10
+	vxorpd		%xmm11, %xmm11, %xmm11
+	vxorpd		%xmm12, %xmm12, %xmm12
+	vxorpd		%xmm13, %xmm13, %xmm13
+	vxorpd		%xmm14, %xmm14, %xmm14
+	vxorpd		%xmm15, %xmm15, %xmm15
+
+.endm
+
+.macro KERNEL1x12_SUB
+	vmovsd 	-16 * SIZE(AO), %xmm0
+	vmovsd	-12 * SIZE(BO), %xmm1
+	vmovsd	-11 * SIZE(BO), %xmm2
+	vmovsd	-10 * SIZE(BO), %xmm3
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm4
+	vmovsd	 -9 * SIZE(BO), %xmm1
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm5
+	vmovsd	 -8 * SIZE(BO), %xmm2
+	vfmadd231sd  	%xmm0 ,%xmm3  , %xmm6
+	vmovsd	 -7 * SIZE(BO), %xmm3
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm7
+	vmovsd	 -6 * SIZE(BO), %xmm1
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm8
+	vmovsd	 -5 * SIZE(BO), %xmm2
+	vfmadd231sd  	%xmm0 ,%xmm3  , %xmm9
+	vmovsd	 -4 * SIZE(BO), %xmm3
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm10
+	vmovsd	 -3 * SIZE(BO), %xmm1
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm11
+	vmovsd	 -2 * SIZE(BO), %xmm2
+	vfmadd231sd  	%xmm0 ,%xmm3  , %xmm12
+	vmovsd	 -1 * SIZE(BO), %xmm3
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm13
+	addq		$ 12*SIZE, BO
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm14
+	addq		$ 1*SIZE, AO
+	vfmadd231sd  	%xmm0 ,%xmm3  , %xmm15
+
+.endm
+
+.macro SAVE1x12
+
+	vmovsd	ALPHA, %xmm0
+
+	vmulsd	%xmm0 , %xmm4 , %xmm4
+	vmulsd	%xmm0 , %xmm5 , %xmm5
+	vmulsd	%xmm0 , %xmm6 , %xmm6
+	vmulsd	%xmm0 , %xmm7 , %xmm7
+
+	vmulsd	%xmm0 , %xmm8 , %xmm8
+	vmulsd	%xmm0 , %xmm9 , %xmm9
+	vmulsd	%xmm0 , %xmm10, %xmm10
+	vmulsd	%xmm0 , %xmm11, %xmm11
+
+	vmulsd	%xmm0 , %xmm12, %xmm12
+	vmulsd	%xmm0 , %xmm13, %xmm13
+	vmulsd	%xmm0 , %xmm14, %xmm14
+	vmulsd	%xmm0 , %xmm15, %xmm15
+
+
+        leaq    (CO1, LDC, 2), %rax     
+	
+
+#if !defined(TRMMKERNEL)
+
+	vaddsd 	                (CO1), %xmm4, %xmm4
+	vaddsd 	           (CO1, LDC), %xmm5, %xmm5
+	vaddsd 	               (%rax), %xmm6, %xmm6
+	vaddsd 	          (%rax, LDC), %xmm7, %xmm7
+
+#endif
+
+	vmovsd	%xmm4 ,  	(CO1)
+	vmovsd	%xmm5 ,  	(CO1, LDC)
+	vmovsd	%xmm6 ,  	(%rax)
+	vmovsd	%xmm7 ,  	(%rax, LDC)
+
+
+	leaq	(%rax, LDC, 2), %rax
+	leaq	(%rax, LDC, 2), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddsd 	                (%rax), %xmm8 , %xmm4
+	vaddsd 	           (%rax, LDC), %xmm9 , %xmm5
+	vaddsd 	                (%rbp), %xmm10, %xmm6
+	vaddsd 	           (%rbp, LDC), %xmm11, %xmm7
+
+#endif
+
+	vmovsd	%xmm4 ,  	(%rax)
+	vmovsd	%xmm5 ,  	(%rax, LDC)
+	vmovsd	%xmm6 ,  	(%rbp)
+	vmovsd	%xmm7 ,  	(%rbp, LDC)
+
+
+	leaq	(%rax, LDC, 4), %rax
+	leaq	(%rbp, LDC, 4), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddsd 	                (%rax), %xmm12, %xmm4
+	vaddsd 	           (%rax, LDC), %xmm13, %xmm5
+	vaddsd 	                (%rbp), %xmm14, %xmm6
+	vaddsd 	           (%rbp, LDC), %xmm15, %xmm7
+
+#endif
+
+	vmovsd	%xmm4 ,  	(%rax)
+	vmovsd	%xmm5 ,  	(%rax, LDC)
+	vmovsd	%xmm6 ,  	(%rbp)
+	vmovsd	%xmm7 ,  	(%rbp, LDC)
+
+	addq	$ 1*SIZE, CO1
+.endm
+
+
+
+
+/******************************************************************************************/
+
+
+.macro INIT4x8
+
+	vxorpd		%ymm4 , %ymm4 , %ymm4
+	vxorpd		%ymm5 , %ymm5 , %ymm5
+	vxorpd		%ymm6 , %ymm6 , %ymm6
+	vxorpd		%ymm7 , %ymm7 , %ymm7
+	vxorpd		%ymm8 , %ymm8 , %ymm8
+	vxorpd		%ymm9 , %ymm9 , %ymm9
+	vxorpd		%ymm10, %ymm10, %ymm10
+	vxorpd		%ymm11, %ymm11, %ymm11
+
+.endm
+
+.macro KERNEL4x8_I
+	vmovups		-12 * SIZE(BO), %ymm1
+	vmovups 	-16 * SIZE(AO), %ymm0
+	vmovups		 -8 * SIZE(BO), %ymm2
+	vmulpd  	%ymm0 ,%ymm1  , %ymm4
+	vmulpd  	%ymm0 ,%ymm2  , %ymm8
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm5
+	vmulpd  	%ymm0 ,%ymm2  , %ymm9
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm6
+	vmulpd  	%ymm0 ,%ymm2  , %ymm10
+
+	addq		$  8*SIZE, BO
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		-12 * SIZE(BO), %ymm1
+	vmulpd  	%ymm0 ,%ymm2  , %ymm11
+	vmovups		 -8 * SIZE(BO), %ymm2
+
+.endm
+
+.macro KERNEL4x8_M1
+	prefetcht0	A_PR1(AO)
+	vmovups 	-16 * SIZE(AO), %ymm0
+	prefetcht0	B_PR1(BO)
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	prefetcht0	B_PR1+64(BO)
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm8
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm9
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm10
+
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		-12 * SIZE(BO), %ymm1
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm11
+	vmovups		 -8 * SIZE(BO), %ymm2
+
+.endm
+
+.macro KERNEL4x8_M2
+	vmovups 	-12 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm8
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm9
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm10
+
+	addq		$ 8*SIZE, AO
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		 -4 * SIZE(BO), %ymm1
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm11
+	vmovups		  0 * SIZE(BO), %ymm2
+	addq		$ 16*SIZE, BO
+.endm
+
+
+.macro KERNEL4x8_E
+	vmovups 	-12 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm8
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm9
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm10
+
+	addq		$ 8*SIZE, AO
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm11
+	addq		$  8*SIZE, BO
+.endm
+
+.macro KERNEL4x8_SUB
+	vmovups		-12 * SIZE(BO), %ymm1
+	vmovups 	-16 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vmovups		 -8 * SIZE(BO), %ymm2
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm8
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm9
+	addq		$  8*SIZE, BO
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm10
+	addq		$ 4*SIZE, AO
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm11
+
+.endm
+
+
+.macro SAVE4x8
+
+	vbroadcastsd	ALPHA, %ymm0
+
+	vmulpd	%ymm0 , %ymm4 , %ymm4
+	vmulpd	%ymm0 , %ymm5 , %ymm5
+	vmulpd	%ymm0 , %ymm6 , %ymm6
+	vmulpd	%ymm0 , %ymm7 , %ymm7
+
+	vmulpd	%ymm0 , %ymm8 , %ymm8
+	vmulpd	%ymm0 , %ymm9 , %ymm9
+	vmulpd	%ymm0 , %ymm10, %ymm10
+	vmulpd	%ymm0 , %ymm11, %ymm11
+
+	vpermpd $ 0xb1 , %ymm5, %ymm5
+	vpermpd $ 0xb1 , %ymm7, %ymm7
+
+	vblendpd $ 0x0a, %ymm5, %ymm4, %ymm0
+	vblendpd $ 0x05, %ymm5, %ymm4, %ymm1
+	vblendpd $ 0x0a, %ymm7, %ymm6, %ymm2
+	vblendpd $ 0x05, %ymm7, %ymm6, %ymm3
+
+	vpermpd $ 0x1b , %ymm2, %ymm2
+	vpermpd $ 0x1b , %ymm3, %ymm3
+	vpermpd $ 0xb1 , %ymm2, %ymm2
+	vpermpd $ 0xb1 , %ymm3, %ymm3
+
+	vblendpd $ 0x03, %ymm0, %ymm2 , %ymm4
+	vblendpd $ 0x03, %ymm1, %ymm3 , %ymm5
+	vblendpd $ 0x03, %ymm2, %ymm0 , %ymm6
+	vblendpd $ 0x03, %ymm3, %ymm1 , %ymm7
+
+        leaq    (CO1, LDC, 2), %rax     
+	
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1), %ymm4, %ymm4
+	vaddpd 	           (CO1, LDC), %ymm5, %ymm5
+	vaddpd 	               (%rax), %ymm6, %ymm6
+	vaddpd 	          (%rax, LDC), %ymm7, %ymm7
+
+#endif
+
+	vmovups	%ymm4 ,  	(CO1)
+	vmovups	%ymm5 ,  	(CO1, LDC)
+	vmovups	%ymm6 ,  	(%rax)
+	vmovups	%ymm7 ,  	(%rax, LDC)
+
+	prefetcht0	32(CO1)
+	prefetcht0	32(CO1,LDC)
+	prefetcht0	32(%rax)
+	prefetcht0	32(%rax,LDC)
+
+	vpermpd $ 0xb1 , %ymm9 , %ymm9
+	vpermpd $ 0xb1 , %ymm11, %ymm11
+
+	vblendpd $ 0x0a, %ymm9 , %ymm8 , %ymm0
+	vblendpd $ 0x05, %ymm9 , %ymm8 , %ymm1
+	vblendpd $ 0x0a, %ymm11, %ymm10, %ymm2
+	vblendpd $ 0x05, %ymm11, %ymm10, %ymm3
+
+	vpermpd $ 0x1b , %ymm2, %ymm2
+	vpermpd $ 0x1b , %ymm3, %ymm3
+	vpermpd $ 0xb1 , %ymm2, %ymm2
+	vpermpd $ 0xb1 , %ymm3, %ymm3
+
+	vblendpd $ 0x03, %ymm0, %ymm2 , %ymm4
+	vblendpd $ 0x03, %ymm1, %ymm3 , %ymm5
+	vblendpd $ 0x03, %ymm2, %ymm0 , %ymm6
+	vblendpd $ 0x03, %ymm3, %ymm1 , %ymm7
+
+
+	leaq	(%rax, LDC, 2), %rax
+	leaq	(%rax, LDC, 2), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (%rax), %ymm4, %ymm4
+	vaddpd 	           (%rax, LDC), %ymm5, %ymm5
+	vaddpd 	                (%rbp), %ymm6, %ymm6
+	vaddpd 	           (%rbp, LDC), %ymm7, %ymm7
+
+#endif
+
+	vmovups	%ymm4 ,  	(%rax)
+	vmovups	%ymm5 ,  	(%rax, LDC)
+	vmovups	%ymm6 ,  	(%rbp)
+	vmovups	%ymm7 ,  	(%rbp, LDC)
+
+	prefetcht0	32(%rax)
+	prefetcht0	32(%rax,LDC)
+	prefetcht0	32(%rbp)
+	prefetcht0	32(%rbp,LDC)
+
+	addq	$ 4*SIZE, CO1
+.endm
+
+/******************************************************************************************/
+
+.macro INIT2x8
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm5 , %xmm5 , %xmm5
+	vxorpd		%xmm6 , %xmm6 , %xmm6
+	vxorpd		%xmm7 , %xmm7 , %xmm7
+	vxorpd		%xmm8 , %xmm8 , %xmm8
+	vxorpd		%xmm9 , %xmm9 , %xmm9
+	vxorpd		%xmm10, %xmm10, %xmm10
+	vxorpd		%xmm11, %xmm11, %xmm11
+
+.endm
+
+.macro KERNEL2x8_SUB
+	vmovups 	-16 * SIZE(AO), %xmm0
+	vmovddup	-12 * SIZE(BO), %xmm1
+	vmovddup	-11 * SIZE(BO), %xmm2
+	vmovddup	-10 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm1  , %xmm4
+	vmovddup	 -9 * SIZE(BO), %xmm1
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm5
+	vmovddup	 -8 * SIZE(BO), %xmm2
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm6
+	vmovddup	 -7 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm1  , %xmm7
+	vmovddup	 -6 * SIZE(BO), %xmm1
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm8
+	vmovddup	 -5 * SIZE(BO), %xmm2
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm9
+	vfmadd231pd  	%xmm0 ,%xmm1  , %xmm10
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm11
+	addq		$  8*SIZE, BO
+	addq		$ 2*SIZE, AO
+
+.endm
+
+.macro SAVE2x8
+
+	vmovddup	ALPHA, %xmm0
+
+	vmulpd	%xmm0 , %xmm4 , %xmm4
+	vmulpd	%xmm0 , %xmm5 , %xmm5
+	vmulpd	%xmm0 , %xmm6 , %xmm6
+	vmulpd	%xmm0 , %xmm7 , %xmm7
+
+	vmulpd	%xmm0 , %xmm8 , %xmm8
+	vmulpd	%xmm0 , %xmm9 , %xmm9
+	vmulpd	%xmm0 , %xmm10, %xmm10
+	vmulpd	%xmm0 , %xmm11, %xmm11
+
+        leaq    (CO1, LDC, 2), %rax     
+	
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1), %xmm4, %xmm4
+	vaddpd 	           (CO1, LDC), %xmm5, %xmm5
+	vaddpd 	               (%rax), %xmm6, %xmm6
+	vaddpd 	          (%rax, LDC), %xmm7, %xmm7
+
+#endif
+
+	vmovups	%xmm4 ,  	(CO1)
+	vmovups	%xmm5 ,  	(CO1, LDC)
+	vmovups	%xmm6 ,  	(%rax)
+	vmovups	%xmm7 ,  	(%rax, LDC)
+
+
+	leaq	(%rax, LDC, 2), %rax
+	leaq	(%rax, LDC, 2), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (%rax), %xmm8 , %xmm4
+	vaddpd 	           (%rax, LDC), %xmm9 , %xmm5
+	vaddpd 	                (%rbp), %xmm10, %xmm6
+	vaddpd 	           (%rbp, LDC), %xmm11, %xmm7
+
+#endif
+
+	vmovups	%xmm4 ,  	(%rax)
+	vmovups	%xmm5 ,  	(%rax, LDC)
+	vmovups	%xmm6 ,  	(%rbp)
+	vmovups	%xmm7 ,  	(%rbp, LDC)
+
+	addq	$ 2*SIZE, CO1
+.endm
+
+
+/******************************************************************************************/
+
+.macro INIT1x8
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm5 , %xmm5 , %xmm5
+	vxorpd		%xmm6 , %xmm6 , %xmm6
+	vxorpd		%xmm7 , %xmm7 , %xmm7
+	vxorpd		%xmm8 , %xmm8 , %xmm8
+	vxorpd		%xmm9 , %xmm9 , %xmm9
+	vxorpd		%xmm10, %xmm10, %xmm10
+	vxorpd		%xmm11, %xmm11, %xmm11
+
+.endm
+
+.macro KERNEL1x8_SUB
+	vmovsd 	-16 * SIZE(AO), %xmm0
+	vmovsd	-12 * SIZE(BO), %xmm1
+	vmovsd	-11 * SIZE(BO), %xmm2
+	vmovsd	-10 * SIZE(BO), %xmm3
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm4
+	vmovsd	 -9 * SIZE(BO), %xmm1
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm5
+	vmovsd	 -8 * SIZE(BO), %xmm2
+	vfmadd231sd  	%xmm0 ,%xmm3  , %xmm6
+	vmovsd	 -7 * SIZE(BO), %xmm3
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm7
+	vmovsd	 -6 * SIZE(BO), %xmm1
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm8
+	vmovsd	 -5 * SIZE(BO), %xmm2
+	vfmadd231sd  	%xmm0 ,%xmm3  , %xmm9
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm10
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm11
+	addq		$  8*SIZE, BO
+	addq		$ 1*SIZE, AO
+
+.endm
+
+.macro SAVE1x8
+
+	vmovsd	ALPHA, %xmm0
+
+	vmulsd	%xmm0 , %xmm4 , %xmm4
+	vmulsd	%xmm0 , %xmm5 , %xmm5
+	vmulsd	%xmm0 , %xmm6 , %xmm6
+	vmulsd	%xmm0 , %xmm7 , %xmm7
+
+	vmulsd	%xmm0 , %xmm8 , %xmm8
+	vmulsd	%xmm0 , %xmm9 , %xmm9
+	vmulsd	%xmm0 , %xmm10, %xmm10
+	vmulsd	%xmm0 , %xmm11, %xmm11
+
+        leaq    (CO1, LDC, 2), %rax     
+	
+
+#if !defined(TRMMKERNEL)
+
+	vaddsd 	                (CO1), %xmm4, %xmm4
+	vaddsd 	           (CO1, LDC), %xmm5, %xmm5
+	vaddsd 	               (%rax), %xmm6, %xmm6
+	vaddsd 	          (%rax, LDC), %xmm7, %xmm7
+
+#endif
+
+	vmovsd	%xmm4 ,  	(CO1)
+	vmovsd	%xmm5 ,  	(CO1, LDC)
+	vmovsd	%xmm6 ,  	(%rax)
+	vmovsd	%xmm7 ,  	(%rax, LDC)
+
+
+	leaq	(%rax, LDC, 2), %rax
+	leaq	(%rax, LDC, 2), %rbp
+
+#if !defined(TRMMKERNEL)
+
+	vaddsd 	                (%rax), %xmm8 , %xmm4
+	vaddsd 	           (%rax, LDC), %xmm9 , %xmm5
+	vaddsd 	                (%rbp), %xmm10, %xmm6
+	vaddsd 	           (%rbp, LDC), %xmm11, %xmm7
+
+#endif
+
+	vmovsd	%xmm4 ,  	(%rax)
+	vmovsd	%xmm5 ,  	(%rax, LDC)
+	vmovsd	%xmm6 ,  	(%rbp)
+	vmovsd	%xmm7 ,  	(%rbp, LDC)
+
+	addq	$ 1*SIZE, CO1
+.endm
+
+
+
+
+
+/******************************************************************************************/
+
+.macro INIT4x4
+
+	vxorpd		%ymm4 , %ymm4 , %ymm4
+	vxorpd		%ymm5 , %ymm5 , %ymm5
+	vxorpd		%ymm6 , %ymm6 , %ymm6
+	vxorpd		%ymm7 , %ymm7 , %ymm7
+
+.endm
+
+.macro KERNEL4x4_I
+	prefetcht0	A_PR1(AO)
+	vmovups		-12 * SIZE(BO), %ymm1
+	vmovups 	-16 * SIZE(AO), %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm4
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm5
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm6
+
+	addq		$ 4*SIZE, BO
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vmulpd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		-12 * SIZE(BO), %ymm1
+
+.endm
+
+.macro KERNEL4x4_M1
+	prefetcht0	A_PR1(AO)
+	vmovups 	-16 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		-12 * SIZE(BO), %ymm1
+
+.endm
+
+.macro KERNEL4x4_M2
+	vmovups 	-12 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+
+	addq		$ 8*SIZE, AO
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	vmovups		 -8 * SIZE(BO), %ymm1
+	addq		$ 8*SIZE, BO
+.endm
+
+
+.macro KERNEL4x4_E
+	vmovups 	-12 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+
+	addq		$ 8*SIZE, AO
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+	addq		$ 4*SIZE, BO
+.endm
+
+.macro KERNEL4x4_SUB
+	vmovups		-12 * SIZE(BO), %ymm1
+	vmovups 	-16 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm4
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm5
+	addq		$ 4*SIZE, BO
+	vpermpd		$ 0x1b, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm6
+	addq		$ 4*SIZE, AO
+	vpermpd		$ 0xb1, %ymm0  , %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm1  , %ymm7
+
+.endm
+
+.macro SAVE4x4
+
+	vbroadcastsd	ALPHA, %ymm0
+
+	vmulpd	%ymm0 , %ymm4 , %ymm4
+	vmulpd	%ymm0 , %ymm7 , %ymm7
+	vmulpd	%ymm0 , %ymm5 , %ymm5
+	vmulpd	%ymm0 , %ymm6 , %ymm6
+
+	vpermpd $ 0xb1 , %ymm5, %ymm5
+	vpermpd $ 0xb1 , %ymm7, %ymm7
+
+	vblendpd $ 0x0a, %ymm5, %ymm4, %ymm0
+	vblendpd $ 0x05, %ymm5, %ymm4, %ymm1
+	vblendpd $ 0x0a, %ymm7, %ymm6, %ymm2
+	vblendpd $ 0x05, %ymm7, %ymm6, %ymm3
+
+	vpermpd $ 0x1b , %ymm2, %ymm2
+	vpermpd $ 0x1b , %ymm3, %ymm3
+	vpermpd $ 0xb1 , %ymm2, %ymm2
+	vpermpd $ 0xb1 , %ymm3, %ymm3
+
+	vblendpd $ 0x03, %ymm0, %ymm2 , %ymm4
+	vblendpd $ 0x03, %ymm1, %ymm3 , %ymm5
+	vblendpd $ 0x03, %ymm2, %ymm0 , %ymm6
+	vblendpd $ 0x03, %ymm3, %ymm1 , %ymm7
+
+        leaq    (CO1, LDC, 2), %rax     
+	
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1), %ymm4, %ymm4
+	vaddpd 	           (CO1, LDC), %ymm5, %ymm5
+	vaddpd 	               (%rax), %ymm6, %ymm6
+	vaddpd 	          (%rax, LDC), %ymm7, %ymm7
+
+#endif
+
+	vmovups	%ymm4 ,  	(CO1)
+	vmovups	%ymm5 ,  	(CO1, LDC)
+	vmovups	%ymm6 ,  	(%rax)
+	vmovups	%ymm7 ,  	(%rax, LDC)
+
+	addq	$ 4*SIZE, CO1
+.endm
+
+/******************************************************************************************/
+/******************************************************************************************/
+
+.macro INIT2x4
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm5 , %xmm5 , %xmm5
+	vxorpd		%xmm6 , %xmm6 , %xmm6
+	vxorpd		%xmm7 , %xmm7 , %xmm7
+
+.endm
+
+
+.macro KERNEL2x4_SUB
+	vmovddup	-12 * SIZE(BO), %xmm1
+	vmovups 	-16 * SIZE(AO), %xmm0
+	vmovddup	-11 * SIZE(BO), %xmm2
+	vfmadd231pd  	%xmm0 ,%xmm1  , %xmm4
+	vmovddup	-10 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm5
+	vmovddup	 -9 * SIZE(BO), %xmm8
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm6
+	addq		$ 4*SIZE, BO
+	vfmadd231pd  	%xmm0 ,%xmm8  , %xmm7
+	addq		$ 2*SIZE, AO
+
+.endm
+
+
+.macro SAVE2x4
+
+	vmovddup	ALPHA, %xmm0
+
+	vmulpd	%xmm0 , %xmm4 , %xmm4
+	vmulpd	%xmm0 , %xmm5 , %xmm5
+	vmulpd	%xmm0 , %xmm6 , %xmm6
+	vmulpd	%xmm0 , %xmm7 , %xmm7
+
+        leaq    (CO1, LDC, 2), %rax     
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1), %xmm4, %xmm4
+	vaddpd 	           (CO1, LDC), %xmm5, %xmm5
+	vaddpd 	               (%rax), %xmm6, %xmm6
+	vaddpd 	          (%rax, LDC), %xmm7, %xmm7
+
+#endif
+
+	vmovups	%xmm4 ,  	(CO1)
+	vmovups	%xmm5 ,  	(CO1, LDC)
+	vmovups	%xmm6 ,  	(%rax)
+	vmovups	%xmm7 ,  	(%rax, LDC)
+
+	addq	$ 2*SIZE, CO1
+.endm
+
+/******************************************************************************************/
+/******************************************************************************************/
+
+.macro INIT1x4
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm5 , %xmm5 , %xmm5
+	vxorpd		%xmm6 , %xmm6 , %xmm6
+	vxorpd		%xmm7 , %xmm7 , %xmm7
+
+.endm
+
+
+.macro KERNEL1x4_SUB
+	vmovsd	-12 * SIZE(BO), %xmm1
+	vmovsd 	-16 * SIZE(AO), %xmm0
+	vmovsd	-11 * SIZE(BO), %xmm2
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm4
+	vmovsd	-10 * SIZE(BO), %xmm3
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm5
+	vmovsd	 -9 * SIZE(BO), %xmm8
+	vfmadd231sd  	%xmm0 ,%xmm3  , %xmm6
+	addq		$ 4*SIZE, BO
+	vfmadd231sd  	%xmm0 ,%xmm8  , %xmm7
+	addq		$ 1*SIZE, AO
+
+.endm
+
+
+.macro SAVE1x4
+
+	vmovsd	ALPHA, %xmm0
+
+	vmulsd	%xmm0 , %xmm4 , %xmm4
+	vmulsd	%xmm0 , %xmm5 , %xmm5
+	vmulsd	%xmm0 , %xmm6 , %xmm6
+	vmulsd	%xmm0 , %xmm7 , %xmm7
+
+        leaq    (CO1, LDC, 2), %rax     
+
+#if !defined(TRMMKERNEL)
+
+	vaddsd 	                (CO1), %xmm4, %xmm4
+	vaddsd 	           (CO1, LDC), %xmm5, %xmm5
+	vaddsd 	               (%rax), %xmm6, %xmm6
+	vaddsd 	          (%rax, LDC), %xmm7, %xmm7
+
+#endif
+
+	vmovsd	%xmm4 ,  	(CO1)
+	vmovsd	%xmm5 ,  	(CO1, LDC)
+	vmovsd	%xmm6 ,  	(%rax)
+	vmovsd	%xmm7 ,  	(%rax, LDC)
+
+	addq	$ 1*SIZE, CO1
+.endm
+
+
+/******************************************************************************************/
+/******************************************************************************************/
+
+.macro INIT4x2
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm5 , %xmm5 , %xmm5
+	vxorpd		%xmm6 , %xmm6 , %xmm6
+	vxorpd		%xmm7 , %xmm7 , %xmm7
+
+.endm
+
+
+.macro KERNEL4x2_SUB
+	vmovddup	-12 * SIZE(BO), %xmm2
+	vmovups 	-16 * SIZE(AO), %xmm0
+	vmovups 	-14 * SIZE(AO), %xmm1
+	vmovddup	-11 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm4
+	vfmadd231pd  	%xmm1 ,%xmm2  , %xmm5
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm6
+	vfmadd231pd  	%xmm1 ,%xmm3  , %xmm7
+	addq		$ 2*SIZE, BO
+	addq		$ 4*SIZE, AO
+
+.endm
+
+
+.macro SAVE4x2
+
+	vmovddup	ALPHA, %xmm0
+
+	vmulpd	%xmm0 , %xmm4 , %xmm4
+	vmulpd	%xmm0 , %xmm5 , %xmm5
+	vmulpd	%xmm0 , %xmm6 , %xmm6
+	vmulpd	%xmm0 , %xmm7 , %xmm7
+
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1)     , %xmm4, %xmm4
+	vaddpd 	        2 * SIZE(CO1)     , %xmm5, %xmm5
+	vaddpd 	                (CO1, LDC), %xmm6, %xmm6
+	vaddpd 	        2 * SIZE(CO1, LDC), %xmm7, %xmm7
+
+#endif
+
+	vmovups	%xmm4 ,  	(CO1)
+	vmovups	%xmm5 , 2 * SIZE(CO1)
+	vmovups	%xmm6 ,  	(CO1, LDC)
+	vmovups	%xmm7 , 2 * SIZE(CO1, LDC)
+
+	addq	$ 4*SIZE, CO1
+.endm
+
+
+/******************************************************************************************/
+/******************************************************************************************/
+
+.macro INIT2x2
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm6 , %xmm6 , %xmm6
+
+.endm
+
+
+.macro KERNEL2x2_SUB
+	vmovddup	-12 * SIZE(BO), %xmm2
+	vmovups 	-16 * SIZE(AO), %xmm0
+	vmovddup	-11 * SIZE(BO), %xmm3
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm4
+	vfmadd231pd  	%xmm0 ,%xmm3  , %xmm6
+	addq		$ 2*SIZE, BO
+	addq		$ 2*SIZE, AO
+
+.endm
+
+
+.macro SAVE2x2
+
+	vmovddup	ALPHA, %xmm0
+
+	vmulpd	%xmm0 , %xmm4 , %xmm4
+	vmulpd	%xmm0 , %xmm6 , %xmm6
+
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1)     , %xmm4, %xmm4
+	vaddpd 	                (CO1, LDC), %xmm6, %xmm6
+
+#endif
+
+	vmovups	%xmm4 ,  	(CO1)
+	vmovups	%xmm6 ,  	(CO1, LDC)
+
+	addq	$ 2*SIZE, CO1
+.endm
+
+/******************************************************************************************/
+/******************************************************************************************/
+
+.macro INIT1x2
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+	vxorpd		%xmm5 , %xmm5 , %xmm5
+
+.endm
+
+
+.macro KERNEL1x2_SUB
+	vmovsd	-12 * SIZE(BO), %xmm1
+	vmovsd 	-16 * SIZE(AO), %xmm0
+	vmovsd	-11 * SIZE(BO), %xmm2
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm4
+	vfmadd231sd  	%xmm0 ,%xmm2  , %xmm5
+	addq		$ 2*SIZE, BO
+	addq		$ 1*SIZE, AO
+
+.endm
+
+
+.macro SAVE1x2
+
+	vmovsd	ALPHA, %xmm0
+
+	vmulsd	%xmm0 , %xmm4 , %xmm4
+	vmulsd	%xmm0 , %xmm5 , %xmm5
+
+
+#if !defined(TRMMKERNEL)
+
+	vaddsd 	                (CO1), %xmm4, %xmm4
+	vaddsd 	           (CO1, LDC), %xmm5, %xmm5
+
+#endif
+
+	vmovsd	%xmm4 ,  	(CO1)
+	vmovsd	%xmm5 ,  	(CO1, LDC)
+
+	addq	$ 1*SIZE, CO1
+.endm
+
+
+/******************************************************************************************/
+/******************************************************************************************/
+
+.macro INIT4x1
+
+	vxorpd		%ymm4 , %ymm4 , %ymm4
+	vxorpd		%ymm5 , %ymm5 , %ymm5
+	vxorpd		%ymm6 , %ymm6 , %ymm6
+	vxorpd		%ymm7 , %ymm7 , %ymm7
+
+.endm
+
+
+.macro KERNEL4x1
+
+	vbroadcastsd	-12 * SIZE(BO), %ymm0
+	vbroadcastsd	-11 * SIZE(BO), %ymm1
+	vbroadcastsd	-10 * SIZE(BO), %ymm2
+	vbroadcastsd	-9  * SIZE(BO), %ymm3
+
+	vfmadd231pd  	-16 * SIZE(AO) ,%ymm0  , %ymm4
+	vfmadd231pd  	-12 * SIZE(AO) ,%ymm1  , %ymm5
+
+	vbroadcastsd	-8  * SIZE(BO), %ymm0
+	vbroadcastsd	-7  * SIZE(BO), %ymm1
+
+	vfmadd231pd  	-8  * SIZE(AO) ,%ymm2  , %ymm6
+	vfmadd231pd  	-4  * SIZE(AO) ,%ymm3  , %ymm7
+
+	vbroadcastsd	-6  * SIZE(BO), %ymm2
+	vbroadcastsd	-5  * SIZE(BO), %ymm3
+
+	vfmadd231pd  	 0  * SIZE(AO) ,%ymm0  , %ymm4
+	vfmadd231pd  	 4  * SIZE(AO) ,%ymm1  , %ymm5
+	vfmadd231pd  	 8  * SIZE(AO) ,%ymm2  , %ymm6
+	vfmadd231pd  	 12 * SIZE(AO) ,%ymm3  , %ymm7
+
+	addq		$ 8 *SIZE, BO
+	addq		$ 32*SIZE, AO
+
+.endm
+
+
+.macro KERNEL4x1_SUB
+	vbroadcastsd	-12 * SIZE(BO), %ymm2
+	vmovups 	-16 * SIZE(AO), %ymm0
+	vfmadd231pd  	%ymm0 ,%ymm2  , %ymm4
+	addq		$ 1*SIZE, BO
+	addq		$ 4*SIZE, AO
+
+.endm
+
+
+.macro SAVE4x1
+
+	vbroadcastsd	ALPHA, %ymm0
+
+	vaddpd	%ymm4,%ymm5, %ymm4 
+	vaddpd	%ymm6,%ymm7, %ymm6 
+	vaddpd	%ymm4,%ymm6, %ymm4 
+
+	vmulpd	%ymm0 , %ymm4 , %ymm4
+
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1)     , %ymm4, %ymm4
+
+#endif
+
+	vmovups	%ymm4 ,  	(CO1)
+
+	addq	$ 4*SIZE, CO1
+.endm
+
+
+/******************************************************************************************/
+/******************************************************************************************/
+
+.macro INIT2x1
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+
+.endm
+
+
+.macro KERNEL2x1_SUB
+	vmovddup	-12 * SIZE(BO), %xmm2
+	vmovups 	-16 * SIZE(AO), %xmm0
+	vfmadd231pd  	%xmm0 ,%xmm2  , %xmm4
+	addq		$ 1*SIZE, BO
+	addq		$ 2*SIZE, AO
+
+.endm
+
+
+.macro SAVE2x1
+
+	vmovddup	ALPHA, %xmm0
+
+	vmulpd	%xmm0 , %xmm4 , %xmm4
+
+
+#if !defined(TRMMKERNEL)
+
+	vaddpd 	                (CO1)     , %xmm4, %xmm4
+
+#endif
+
+	vmovups	%xmm4 ,  	(CO1)
+
+	addq	$ 2*SIZE, CO1
+.endm
+
+
+/******************************************************************************************/
+/******************************************************************************************/
+
+.macro INIT1x1
+
+	vxorpd		%xmm4 , %xmm4 , %xmm4
+
+.endm
+
+
+.macro KERNEL1x1_SUB
+	vmovsd	-12 * SIZE(BO), %xmm1
+	vmovsd 	-16 * SIZE(AO), %xmm0
+	vfmadd231sd  	%xmm0 ,%xmm1  , %xmm4
+	addq		$ 1*SIZE, BO
+	addq		$ 1*SIZE, AO
+
+.endm
+
+
+.macro SAVE1x1
+
+	vmovsd	ALPHA, %xmm0
+
+	vmulsd	%xmm0 , %xmm4 , %xmm4
+
+
+#if !defined(TRMMKERNEL)
+
+	vaddsd 	                (CO1), %xmm4, %xmm4
+
+#endif
+
+	vmovsd	%xmm4 ,  	(CO1)
+
+	addq	$ 1*SIZE, CO1
+.endm
+
+
+/*******************************************************************************************/
+
+#if !defined(TRMMKERNEL)
+
+
+	PROLOGUE
+	PROFCODE
+	
+	subq	$STACKSIZE, %rsp
+	movq	%rbx,   (%rsp)
+	movq	%rbp,  8(%rsp)
+	movq	%r12, 16(%rsp)
+	movq	%r13, 24(%rsp)
+	movq	%r14, 32(%rsp)
+	movq	%r15, 40(%rsp)
+
+	vzeroupper
+
+#ifdef WINDOWS_ABI
+	movq	%rdi,    48(%rsp)
+	movq	%rsi,    56(%rsp)
+	vmovups	%xmm6,   64(%rsp)
+	vmovups	%xmm7,   80(%rsp)
+	vmovups	%xmm8,   96(%rsp)
+	vmovups	%xmm9,  112(%rsp)
+	vmovups	%xmm10, 128(%rsp)
+	vmovups	%xmm11, 144(%rsp)
+	vmovups	%xmm12, 160(%rsp)
+	vmovups	%xmm13, 176(%rsp)
+	vmovups	%xmm14, 192(%rsp)
+	vmovups	%xmm15, 208(%rsp)
+
+	movq	ARG1,      OLD_M
+	movq	ARG2,      OLD_N
+	movq	ARG3,      OLD_K
+	movq	OLD_A,     A
+	movq	OLD_B,     B
+	movq	OLD_C,     C
+	movq	OLD_LDC,   LDC
+
+	vmovups	%xmm3, %xmm0
+
+#else
+	movq	STACKSIZE +  8(%rsp), LDC
+
+#endif
+
+	movq    %rsp, SP      # save old stack
+        subq    $128 + L_BUFFER_SIZE, %rsp
+        andq    $-4096, %rsp    # align stack
+
+        STACK_TOUCH
+
+	cmpq	$ 0, OLD_M
+	je	.L999
+
+	cmpq	$ 0, OLD_N
+	je	.L999
+
+	cmpq	$ 0, OLD_K
+	je	.L999
+
+	movq	OLD_M, M
+	movq	OLD_N, N
+	movq	OLD_K, K
+
+	vmovsd	 %xmm0, ALPHA
+
+	salq	$BASE_SHIFT, LDC
+
+	movq    N, %rax
+        xorq    %rdx, %rdx
+        movq    $24,  %rdi
+        divq    %rdi                     //    N / 24
+        movq    %rax, Ndiv12             //    N / 24
+        movq    %rdx, Nmod12             //    N % 24
+
+
+	movq	Ndiv12,  J
+	cmpq	$ 0, J
+	je	.L8_0
+	ALIGN_4
+
+.L12_01:
+        // copy to sub buffer
+        movq    K, %rax
+        salq    $3,%rax                 // K * 8 ; read 8 values from BO1
+        movq    B, BO1
+        leaq    (B,%rax, SIZE), BO2     // next offset to BO2
+	movq	BO2 , B			
+
+        leaq    BUFFER1, BO             // first buffer to BO
+        movq    K, %rax
+
+        ALIGN_4
+
+.L12_02b:
+
+	vmovups	0 * SIZE(BO1), %ymm1
+	vmovups	4 * SIZE(BO1), %ymm2
+	vmovups	0 * SIZE(BO2), %ymm3
+	vmovups	%ymm1, 0 * SIZE(BO)
+	vmovups	%ymm2, 4 * SIZE(BO)
+	vmovups	%ymm3, 8 * SIZE(BO)
+	addq	$ 8*SIZE,BO1
+	addq	$ 8*SIZE,BO2
+	addq	$ 12*SIZE,BO
+	decq	%rax
+	jnz	.L12_02b
+
+.L12_03c:
+
+
+.L12_10:
+	movq	C, CO1
+	leaq	(C, LDC, 8), C		 
+	leaq	(C, LDC, 4), C		// c += 12 * ldc
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L12_20
+
+	ALIGN_4
+
+.L12_11:
+        leaq    BUFFER1, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+        movq    K, %rax
+
+	sarq $3, %rax			//  K / 8
+	cmpq $2, %rax
+
+	jl	.L12_13
+
+
+	KERNEL4x12_I
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	subq $2, %rax
+	je	.L12_12a
+
+	ALIGN_5
+.L12_12:
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	dec	%rax
+	jne	.L12_12
+
+.L12_12a:
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_E
+
+	jmp .L12_16
+
+
+.L12_13:
+
+	test $1, %rax
+	jz .L12_14
+
+	KERNEL4x12_I
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_E
+
+	jmp .L12_16
+
+
+.L12_14:
+
+	INIT4x12
+
+
+.L12_16:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L12_19
+
+	ALIGN_4
+
+.L12_17:
+
+	KERNEL4x12_SUB
+
+	dec	%rax
+	jne	.L12_17
+	ALIGN_4
+
+
+.L12_19:
+
+	SAVE4x12
+
+	decq	I			# i --
+	jne	.L12_11
+	ALIGN_4	
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L12_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L12_100			// to next 16 lines of N
+
+
+.L12_30:
+	testq	$2, M		
+	jz	.L12_40
+
+	ALIGN_4
+
+.L12_31:
+        leaq    BUFFER1, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT2x12
+
+        movq    K, %rax
+
+	sarq	$3, %rax
+	je	.L12_36
+	ALIGN_4
+
+.L12_32:
+
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+
+	dec %rax
+	jne	.L12_32
+	ALIGN_4
+
+.L12_36:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L12_39
+
+	ALIGN_4
+
+.L12_37:
+
+	KERNEL2x12_SUB
+
+	dec %rax
+	jne	.L12_37
+	ALIGN_4
+
+
+.L12_39:
+
+	SAVE2x12
+
+	ALIGN_4
+
+.L12_40:
+	testq	$1, M		
+	jz	.L12_100		// to next 3 lines of N
+
+	ALIGN_4
+
+.L12_41:
+        leaq    BUFFER1, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT1x12
+
+        movq    K, %rax
+
+	sarq	$3,%rax
+	je	.L12_46
+
+	ALIGN_4
+
+.L12_42:
+
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+
+
+	dec %rax
+	jne	.L12_42
+	ALIGN_4
+
+.L12_46:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L12_49
+
+	ALIGN_4
+
+.L12_47:
+
+	KERNEL1x12_SUB
+
+	dec	%rax
+	jne	.L12_47
+	ALIGN_4
+
+
+.L12_49:
+
+	SAVE1x12
+
+	ALIGN_4
+	
+.L12_100:
+
+
+
+/**************************************************************************************************/
+
+.L13_01:
+        // copy to sub buffer
+        movq    K, %rax
+        salq    $3,%rax                 // K * 8 ; read 8 values
+        movq    B, BO2
+        leaq    (B,%rax, SIZE), BO3     // next offset to BO2
+        leaq    (BO3,%rax, SIZE), B     // next offset to B
+
+
+        leaq    BUFFER1, BO             // first buffer to BO
+        movq    K, %rax
+
+        ALIGN_4
+
+
+.L13_02b:
+
+	vmovups	4 * SIZE(BO2), %ymm1
+	vmovups	0 * SIZE(BO3), %ymm2
+	vmovups	4 * SIZE(BO3), %ymm3
+	vmovups	%ymm1, 0 * SIZE(BO)
+	vmovups	%ymm2, 4 * SIZE(BO)
+	vmovups	%ymm3, 8 * SIZE(BO)
+	addq	$ 8*SIZE,BO2
+	addq	$ 8*SIZE,BO3
+	addq	$ 12*SIZE,BO
+	decq	%rax
+	jnz	.L13_02b
+
+
+
+.L13_10:
+	movq	C, CO1
+	leaq	(C, LDC, 8), C		 
+	leaq	(C, LDC, 4), C		// c += 12 * ldc
+
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L13_20
+
+	ALIGN_4
+
+.L13_11:
+        leaq    BUFFER1, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+        movq    K, %rax
+
+	sarq $3, %rax			//  K / 8
+	cmpq $2, %rax
+
+	jl	.L13_13
+
+
+	KERNEL4x12_I
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	subq $2, %rax
+	je	.L13_12a
+
+	ALIGN_5
+.L13_12:
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	dec	%rax
+	jne	.L13_12
+
+.L13_12a:
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_E
+
+	jmp .L13_16
+
+
+.L13_13:
+
+	test $1, %rax
+	jz .L13_14
+
+	KERNEL4x12_I
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+
+	KERNEL4x12_M1
+	KERNEL4x12_M2
+	KERNEL4x12_M1
+	KERNEL4x12_E
+
+	jmp .L13_16
+
+
+.L13_14:
+
+	INIT4x12
+
+
+.L13_16:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L13_19
+
+	ALIGN_4
+
+.L13_17:
+
+	KERNEL4x12_SUB
+
+	dec	%rax
+	jne	.L13_17
+	ALIGN_4
+
+
+.L13_19:
+
+	SAVE4x12
+
+	decq	I			# i --
+	jne	.L13_11
+	ALIGN_4	
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L13_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L13_100			// to next 16 lines of N
+
+
+.L13_30:
+	testq	$2, M		
+	jz	.L13_40
+
+	ALIGN_4
+
+.L13_31:
+        leaq    BUFFER1, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT2x12
+
+        movq    K, %rax
+
+	sarq	$3, %rax
+	je	.L13_36
+	ALIGN_4
+
+.L13_32:
+
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+	KERNEL2x12_SUB
+
+	dec %rax
+	jne	.L13_32
+	ALIGN_4
+
+.L13_36:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L13_39
+
+	ALIGN_4
+
+.L13_37:
+
+	KERNEL2x12_SUB
+
+	dec %rax
+	jne	.L13_37
+	ALIGN_4
+
+
+.L13_39:
+
+	SAVE2x12
+
+	ALIGN_4
+
+.L13_40:
+	testq	$1, M		
+	jz	.L13_100		// to next 3 lines of N
+
+	ALIGN_4
+
+.L13_41:
+        leaq    BUFFER1, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT1x12
+
+        movq    K, %rax
+
+	sarq	$3,%rax
+	je	.L13_46
+
+	ALIGN_4
+
+.L13_42:
+
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+	KERNEL1x12_SUB
+
+
+	dec %rax
+	jne	.L13_42
+	ALIGN_4
+
+.L13_46:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L13_49
+
+	ALIGN_4
+
+.L13_47:
+
+	KERNEL1x12_SUB
+
+	dec	%rax
+	jne	.L13_47
+	ALIGN_4
+
+
+.L13_49:
+
+	SAVE1x12
+
+	ALIGN_4
+	
+.L13_100:
+
+	decq	J			// j --
+	jg	.L12_01
+
+
+
+
+/**************************************************************************************************/
+
+.L8_0:
+
+	cmpq	$ 0, Nmod12		// N % 12 == 0
+	je	.L999
+
+	movq	Nmod12, J		
+	sarq	$3, J			// j = j / 8
+	je	.L4_0
+
+.L8_10:
+	movq	C, CO1
+	leaq	(C, LDC, 8), C		// c += 4 * ldc
+
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L8_20
+
+	ALIGN_4
+
+.L8_11:
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+
+        movq    K, %rax
+
+	sarq	$3, %rax			//  K / 8
+	cmpq    $2, %rax
+	jl	.L8_13
+
+
+	KERNEL4x8_I
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	subq $2, %rax
+	je	.L8_12a
+
+	ALIGN_5
+
+.L8_12:
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	dec	%rax
+	jne	.L8_12
+
+.L8_12a:
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_E
+
+	jmp .L8_16
+
+
+.L8_13:
+
+	test $1, %rax
+	jz .L8_14
+
+	KERNEL4x8_I
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_E
+
+	jmp .L8_16
+
+
+.L8_14:
+
+	INIT4x8
+
+
+.L8_16:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L8_19
+
+	ALIGN_4
+
+.L8_17:
+
+	KERNEL4x8_SUB
+
+	dec	%rax
+	jne	.L8_17
+	ALIGN_4
+
+
+.L8_19:
+
+	SAVE4x8
+
+	decq	I			# i --
+	jg	.L8_11
+	ALIGN_4	
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L8_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L8_100			// to next 16 lines of N
+
+
+.L8_30:
+	testq	$2, M		
+	jz	.L8_40
+
+	ALIGN_4
+
+.L8_31:
+        movq    B, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT2x8
+
+        movq    K, %rax
+
+	sarq	$3, %rax
+	je	.L8_36
+	ALIGN_4
+
+.L8_32:
+
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+
+	dec %rax
+	jne	.L8_32
+	ALIGN_4
+
+.L8_36:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L8_39
+
+	ALIGN_4
+
+.L8_37:
+
+	KERNEL2x8_SUB
+
+	dec %rax
+	jne	.L8_37
+
+
+.L8_39:
+
+	SAVE2x8
+
+.L8_40:
+	testq	$1, M		
+	jz	.L8_100		// to next 3 lines of N
+
+	ALIGN_4
+
+.L8_41:
+        movq    B, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT1x8
+
+        movq    K, %rax
+
+	sarq	$3,%rax
+	je	.L8_46
+
+	ALIGN_4
+
+.L8_42:
+
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+
+	dec %rax
+	jne	.L8_42
+	ALIGN_4
+
+.L8_46:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L8_49
+
+	ALIGN_4
+
+.L8_47:
+
+	KERNEL1x8_SUB
+
+	dec	%rax
+	jne	.L8_47
+	ALIGN_4
+
+
+.L8_49:
+
+	SAVE1x8
+
+	ALIGN_4
+	
+.L8_100:
+
+	movq	K, %rax
+	salq	$3, %rax		// * 8
+	leaq	(B , %rax, SIZE), B
+	decq	J			// j --
+	jg	.L8_10
+
+
+
+/**************************************************************************************************/
+
+.L4_0:
+
+	cmpq	$ 0, Nmod12		// N % 12 == 0
+	je	.L999
+
+	movq	Nmod12, J		
+	testq   $4, J			// j = j / 4
+	je	.L2_0
+
+.L4_10:
+	movq	C, CO1
+	leaq	(C, LDC, 4), C		// c += 4 * ldc
+
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L4_20
+
+	ALIGN_4
+
+.L4_11:
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+
+        movq    K, %rax
+
+	sarq	$3, %rax			//  K / 8
+	cmpq    $2, %rax
+	jl	.L4_13
+
+
+	KERNEL4x4_I
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	subq $2, %rax
+	je	.L4_12a
+
+	ALIGN_5
+
+.L4_12:
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	dec	%rax
+	jne	.L4_12
+
+.L4_12a:
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_E
+
+	jmp .L4_16
+
+
+.L4_13:
+
+	test $1, %rax
+	jz .L4_14
+
+	KERNEL4x4_I
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_E
+
+	jmp .L4_16
+
+
+.L4_14:
+
+	INIT4x4
+
+
+.L4_16:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L4_19
+
+	ALIGN_4
+
+.L4_17:
+
+	KERNEL4x4_SUB
+
+	dec	%rax
+	jne	.L4_17
+	ALIGN_4
+
+
+.L4_19:
+
+	SAVE4x4
+
+	decq	I			# i --
+	jg	.L4_11
+
+	ALIGN_4	
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L4_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L4_100			// to next 16 lines of N
+
+
+.L4_30:
+	testq	$2, M		
+	jz	.L4_40
+
+	ALIGN_4
+
+.L4_31:
+        movq    B, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT2x4
+
+        movq    K, %rax
+
+	sarq	$3, %rax
+	je	.L4_36
+	ALIGN_4
+
+.L4_32:
+
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+
+	dec %rax
+	jne	.L4_32
+	ALIGN_4
+
+.L4_36:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L4_39
+
+	ALIGN_4
+
+.L4_37:
+
+	KERNEL2x4_SUB
+
+	dec %rax
+	jne	.L4_37
+
+
+.L4_39:
+
+	SAVE2x4
+
+.L4_40:
+	testq	$1, M		
+	jz	.L4_100		// to next 3 lines of N
+
+	ALIGN_4
+
+.L4_41:
+        movq    B, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT1x4
+
+        movq    K, %rax
+
+	sarq	$3,%rax
+	je	.L4_46
+
+	ALIGN_4
+
+.L4_42:
+
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+
+	dec %rax
+	jne	.L4_42
+	ALIGN_4
+
+.L4_46:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L4_49
+
+	ALIGN_4
+
+.L4_47:
+
+	KERNEL1x4_SUB
+
+	dec	%rax
+	jne	.L4_47
+	ALIGN_4
+
+
+.L4_49:
+
+	SAVE1x4
+
+	ALIGN_4
+	
+.L4_100:
+
+	movq	K, %rax
+	salq	$2, %rax		// * 4
+	leaq	(B , %rax, SIZE), B
+
+
+
+
+/***************************************************************************************************************/
+
+.L2_0:
+
+	movq	Nmod12, J		
+	testq	$2, J
+	je	.L1_0
+
+.L2_10:
+	movq	C, CO1
+	leaq	(C, LDC, 2), C		// c += 2 * ldc
+
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L2_20
+
+	ALIGN_4
+
+.L2_11:
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+
+	INIT4x2
+
+        movq    K, %rax
+	sarq $3, %rax			//  K / 8
+
+	je	.L2_16
+
+	ALIGN_5
+
+.L2_12:
+
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+
+	dec	%rax
+	jne	.L2_12
+
+
+.L2_16:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L2_19
+
+	ALIGN_4
+
+.L2_17:
+
+	KERNEL4x2_SUB
+
+	dec	%rax
+	jne	.L2_17
+	ALIGN_4
+
+
+.L2_19:
+
+	SAVE4x2
+
+	decq	I			# i --
+	jg	.L2_11
+	ALIGN_4	
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L2_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L2_100			// to next 16 lines of N
+
+
+.L2_30:
+	testq	$2, M		
+	jz	.L2_40
+
+	ALIGN_4
+
+.L2_31:
+        movq    B, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT2x2
+
+        movq    K, %rax
+
+	sarq	$3, %rax
+	je	.L2_36
+	ALIGN_4
+
+.L2_32:
+
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+
+	dec %rax
+	jne	.L2_32
+
+.L2_36:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L2_39
+
+	ALIGN_4
+
+.L2_37:
+
+	KERNEL2x2_SUB
+
+	dec %rax
+	jne	.L2_37
+
+
+.L2_39:
+
+	SAVE2x2
+
+.L2_40:
+	testq	$1, M		
+	jz	.L2_100		// to next 3 lines of N
+
+.L2_41:
+        movq    B, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT1x2
+
+        movq    K, %rax
+
+	sarq	$3,%rax
+	je	.L2_46
+
+	ALIGN_4
+
+.L2_42:
+
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+
+	dec %rax
+	jne	.L2_42
+
+.L2_46:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L2_49
+
+	ALIGN_4
+
+.L2_47:
+
+	KERNEL1x2_SUB
+
+	dec	%rax
+	jne	.L2_47
+
+.L2_49:
+
+	SAVE1x2
+
+.L2_100:
+
+	movq	K, %rax
+	salq	$1, %rax		// * 2
+	leaq	(B , %rax, SIZE), B
+
+/***************************************************************************************************************/
+
+.L1_0:
+
+	movq	Nmod12, J		
+	testq	$1, J
+	je	.L999
+
+.L1_10:
+	movq	C, CO1
+	leaq	(C, LDC, 1), C		// c += 1 * ldc
+
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L1_20
+
+	ALIGN_4
+
+.L1_11:
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+
+	INIT4x1
+
+        movq    K, %rax
+
+	sarq	$3, %rax			//  K / 8
+	je	.L1_16
+
+	ALIGN_5
+
+.L1_12:
+
+	KERNEL4x1
+
+	dec	%rax
+	jne	.L1_12
+
+
+.L1_16:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L1_19
+
+	ALIGN_4
+
+.L1_17:
+
+	KERNEL4x1_SUB
+
+	dec	%rax
+	jne	.L1_17
+	ALIGN_4
+
+
+.L1_19:
+
+	SAVE4x1
+
+	decq	I			# i --
+	jg	.L1_11
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L1_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L1_100	
+
+
+.L1_30:
+	testq	$2, M		
+	jz	.L1_40
+
+	ALIGN_4
+
+.L1_31:
+        movq    B, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT2x1
+
+        movq    K, %rax
+
+	sarq	$3, %rax
+	je	.L1_36
+	ALIGN_4
+
+.L1_32:
+
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+
+
+	dec %rax
+	jne	.L1_32
+
+.L1_36:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L1_39
+
+	ALIGN_4
+
+.L1_37:
+
+	KERNEL2x1_SUB
+
+	dec %rax
+	jne	.L1_37
+
+.L1_39:
+
+	SAVE2x1
+
+.L1_40:
+	testq	$1, M		
+	jz	.L1_100		// to next 3 lines of N
+
+
+.L1_41:
+        movq    B, BO             // first buffer to BO
+        addq    $12 * SIZE, BO
+
+	INIT1x1
+
+        movq    K, %rax
+
+	sarq	$3,%rax
+	je	.L1_46
+
+	ALIGN_4
+
+.L1_42:
+
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+
+	dec %rax
+	jne	.L1_42
+
+.L1_46:
+        movq    K, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L1_49
+
+	ALIGN_4
+
+.L1_47:
+
+	KERNEL1x1_SUB
+
+	dec	%rax
+	jne	.L1_47
+
+
+.L1_49:
+
+	SAVE1x1
+
+.L1_100:
+
+
+
+
+.L999:
+	vzeroupper
+
+	movq   		SP, %rsp
+	movq	   (%rsp), %rbx
+	movq	  8(%rsp), %rbp
+	movq	 16(%rsp), %r12
+	movq	 24(%rsp), %r13
+	movq	 32(%rsp), %r14
+	movq	 40(%rsp), %r15
+
+#ifdef WINDOWS_ABI
+	movq	 48(%rsp), %rdi
+	movq	 56(%rsp), %rsi
+	vmovups	 64(%rsp), %xmm6
+	vmovups	 80(%rsp), %xmm7
+	vmovups	 96(%rsp), %xmm8
+	vmovups	112(%rsp), %xmm9
+	vmovups	128(%rsp), %xmm10
+	vmovups	144(%rsp), %xmm11
+	vmovups	160(%rsp), %xmm12
+	vmovups	176(%rsp), %xmm13
+	vmovups	192(%rsp), %xmm14
+	vmovups	208(%rsp), %xmm15
+#endif
+
+	addq	$STACKSIZE, %rsp
+	ret
+
+	EPILOGUE
+
+
+#else
+/*************************************************************************************
+* TRMM Kernel
+*************************************************************************************/
+
+
+	PROLOGUE
+	PROFCODE
+	
+	subq	$STACKSIZE, %rsp
+	movq	%rbx,   (%rsp)
+	movq	%rbp,  8(%rsp)
+	movq	%r12, 16(%rsp)
+	movq	%r13, 24(%rsp)
+	movq	%r14, 32(%rsp)
+	movq	%r15, 40(%rsp)
+
+	vzeroupper
+
+#ifdef WINDOWS_ABI
+	movq	%rdi,    48(%rsp)
+	movq	%rsi,    56(%rsp)
+	vmovups	%xmm6,   64(%rsp)
+	vmovups	%xmm7,   80(%rsp)
+	vmovups	%xmm8,   96(%rsp)
+	vmovups	%xmm9,  112(%rsp)
+	vmovups	%xmm10, 128(%rsp)
+	vmovups	%xmm11, 144(%rsp)
+	vmovups	%xmm12, 160(%rsp)
+	vmovups	%xmm13, 176(%rsp)
+	vmovups	%xmm14, 192(%rsp)
+	vmovups	%xmm15, 208(%rsp)
+
+	movq	ARG1,      OLD_M
+	movq	ARG2,      OLD_N
+	movq	ARG3,      OLD_K
+	movq	OLD_A,     A
+	movq	OLD_B,     B
+	movq	OLD_C,     C
+	movq	OLD_LDC,   LDC
+#ifdef TRMMKERNEL
+	vmovsd	OLD_OFFSET, %xmm12
+#endif
+	vmovups	%xmm3, %xmm0
+
+#else
+	movq	STACKSIZE +  8(%rsp), LDC
+#ifdef TRMMKERNEL
+	vmovsd	STACKSIZE + 16(%rsp), %xmm12
+#endif
+
+#endif
+
+	movq    %rsp, SP      # save old stack
+        subq    $128 + L_BUFFER_SIZE, %rsp
+        andq    $-4096, %rsp    # align stack
+
+        STACK_TOUCH
+
+	cmpq	$ 0, OLD_M
+	je	.L999
+
+	cmpq	$ 0, OLD_N
+	je	.L999
+
+	cmpq	$ 0, OLD_K
+	je	.L999
+
+	movq	OLD_M, M
+	movq	OLD_N, N
+	movq	OLD_K, K
+
+	vmovsd	 %xmm0, ALPHA
+
+	salq	$BASE_SHIFT, LDC
+
+	movq    N, %rax
+        xorq    %rdx, %rdx
+        movq    $8,  %rdi
+        divq    %rdi                     //    N / 8
+        movq    %rax, Ndiv12             //    N / 8
+        movq    %rdx, Nmod12             //    N % 8
+
+#ifdef TRMMKERNEL
+        vmovsd  %xmm12, OFFSET
+        vmovsd  %xmm12, KK
+#ifndef LEFT
+        negq    KK
+#endif  
+#endif
+
+/*************************************************************************************************/
+.L8_0:
+	movq	Ndiv12,  J
+	cmpq	$ 0, J
+	je	.L4_0
+	ALIGN_4
+
+.L8_10:
+	movq	C, CO1
+	leaq	(C, LDC, 8), C		// c += 8 * ldc
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        movq    OFFSET, %rax
+        movq    %rax, KK
+#endif
+
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L8_20
+
+	ALIGN_4
+
+.L8_11:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,8), BO		// add number of values in B
+	leaq	(AO,%rax,4), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $4, %rax        // number of values in AO
+#else
+        addq    $8, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	sarq	$3, %rax			//  K / 8
+	cmpq    $2, %rax
+	jl	.L8_13
+
+
+	KERNEL4x8_I
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	subq $2, %rax
+	je	.L8_12a
+
+	ALIGN_5
+
+.L8_12:
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	dec	%rax
+	jne	.L8_12
+
+.L8_12a:
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_E
+
+	jmp .L8_16
+
+
+.L8_13:
+
+	test $1, %rax
+	jz .L8_14
+
+	KERNEL4x8_I
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+
+	KERNEL4x8_M1
+	KERNEL4x8_M2
+	KERNEL4x8_M1
+	KERNEL4x8_E
+
+	jmp .L8_16
+
+
+.L8_14:
+
+	INIT4x8
+
+
+.L8_16:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L8_19
+
+	ALIGN_4
+
+.L8_17:
+
+	KERNEL4x8_SUB
+
+	dec	%rax
+	jne	.L8_17
+	ALIGN_4
+
+
+.L8_19:
+
+	SAVE4x8
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax + SIZE
+        leaq    (BO, %rax, 8), BO		// number of values in B
+        leaq    (AO, %rax, 4), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $4, KK				// number of values in A
+#endif
+
+	decq	I			# i --
+	jg	.L8_11
+	ALIGN_4	
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L8_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L8_100			// to next 16 lines of N
+
+
+.L8_30:
+	testq	$2, M		
+	jz	.L8_40
+
+	ALIGN_4
+
+.L8_31:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,8), BO		// add number of values in B
+	leaq	(AO,%rax,2), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $2, %rax        // number of values in AO
+#else
+        addq    $8, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT2x8
+
+	sarq	$3, %rax
+	je	.L8_36
+	ALIGN_4
+
+.L8_32:
+
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+	KERNEL2x8_SUB
+
+	dec %rax
+	jne	.L8_32
+	ALIGN_4
+
+.L8_36:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L8_39
+
+	ALIGN_4
+
+.L8_37:
+
+	KERNEL2x8_SUB
+
+	dec %rax
+	jne	.L8_37
+
+
+.L8_39:
+
+	SAVE2x8
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax + SIZE
+        leaq    (BO, %rax, 8), BO		// number of values in B
+        leaq    (AO, %rax, 2), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $2, KK				// number of values in A
+#endif
+
+
+.L8_40:
+	testq	$1, M		
+	jz	.L8_100		// to next 3 lines of N
+
+	ALIGN_4
+
+.L8_41:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,8), BO		// add number of values in B
+	leaq	(AO,%rax,1), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $1, %rax        // number of values in AO
+#else
+        addq    $8, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT1x8
+
+	sarq	$3,%rax
+	je	.L8_46
+
+	ALIGN_4
+
+.L8_42:
+
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+	KERNEL1x8_SUB
+
+	dec %rax
+	jne	.L8_42
+	ALIGN_4
+
+.L8_46:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L8_49
+
+	ALIGN_4
+
+.L8_47:
+
+	KERNEL1x8_SUB
+
+	dec	%rax
+	jne	.L8_47
+	ALIGN_4
+
+
+.L8_49:
+
+	SAVE1x8
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax + SIZE
+        leaq    (BO, %rax, 8), BO		// number of values in B
+        leaq    (AO, %rax, 1), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $1, KK				// number of values in A
+#endif
+
+.L8_100:
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+        addq    $8, KK				// number of values in B
+#endif
+
+
+	decq	J			// j --
+	jg	.L8_10
+
+
+
+
+
+/*************************************************************************************************/
+.L4_0:
+	movq	Nmod12, J		
+	testq	$4, J
+	je	.L2_0
+	ALIGN_4
+
+.L4_10:
+	movq	C, CO1
+	leaq	(C, LDC, 4), C		// c += 4 * ldc
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        movq    OFFSET, %rax
+        movq    %rax, KK
+#endif
+
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L4_20
+
+	ALIGN_4
+
+.L4_11:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,4), BO		// add number of values in B
+	leaq	(AO,%rax,4), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $4, %rax        // number of values in AO
+#else
+        addq    $4, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	sarq	$3, %rax			//  K / 8
+	cmpq    $2, %rax
+	jl	.L4_13
+
+
+	KERNEL4x4_I
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	subq $2, %rax
+	je	.L4_12a
+
+	ALIGN_5
+
+.L4_12:
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	dec	%rax
+	jne	.L4_12
+
+.L4_12a:
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_E
+
+	jmp .L4_16
+
+
+.L4_13:
+
+	test $1, %rax
+	jz .L4_14
+
+	KERNEL4x4_I
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+
+	KERNEL4x4_M1
+	KERNEL4x4_M2
+	KERNEL4x4_M1
+	KERNEL4x4_E
+
+	jmp .L4_16
+
+
+.L4_14:
+
+	INIT4x4
+
+
+.L4_16:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L4_19
+
+	ALIGN_4
+
+.L4_17:
+
+	KERNEL4x4_SUB
+
+	dec	%rax
+	jne	.L4_17
+	ALIGN_4
+
+
+.L4_19:
+
+	SAVE4x4
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax + SIZE
+        leaq    (BO, %rax, 4), BO		// number of values in B
+        leaq    (AO, %rax, 4), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $4, KK				// number of values in A
+#endif
+
+	decq	I			# i --
+	jg	.L4_11
+	ALIGN_4	
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L4_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L4_100			// to next 16 lines of N
+
+
+.L4_30:
+	testq	$2, M		
+	jz	.L4_40
+
+	ALIGN_4
+
+.L4_31:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,4), BO		// add number of values in B
+	leaq	(AO,%rax,2), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $2, %rax        // number of values in AO
+#else
+        addq    $4, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT2x4
+
+	sarq	$3, %rax
+	je	.L4_36
+	ALIGN_4
+
+.L4_32:
+
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+	KERNEL2x4_SUB
+
+	dec %rax
+	jne	.L4_32
+	ALIGN_4
+
+.L4_36:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L4_39
+
+	ALIGN_4
+
+.L4_37:
+
+	KERNEL2x4_SUB
+
+	dec %rax
+	jne	.L4_37
+
+
+.L4_39:
+
+	SAVE2x4
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax + SIZE
+        leaq    (BO, %rax, 4), BO		// number of values in B
+        leaq    (AO, %rax, 2), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $2, KK				// number of values in A
+#endif
+
+
+.L4_40:
+	testq	$1, M		
+	jz	.L4_100		// to next 3 lines of N
+
+	ALIGN_4
+
+.L4_41:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,4), BO		// add number of values in B
+	leaq	(AO,%rax,1), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $1, %rax        // number of values in AO
+#else
+        addq    $4, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT1x4
+
+	sarq	$3,%rax
+	je	.L4_46
+
+	ALIGN_4
+
+.L4_42:
+
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+	KERNEL1x4_SUB
+
+	dec %rax
+	jne	.L4_42
+	ALIGN_4
+
+.L4_46:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L4_49
+
+	ALIGN_4
+
+.L4_47:
+
+	KERNEL1x4_SUB
+
+	dec	%rax
+	jne	.L4_47
+	ALIGN_4
+
+
+.L4_49:
+
+	SAVE1x4
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax + SIZE
+        leaq    (BO, %rax, 4), BO		// number of values in B
+        leaq    (AO, %rax, 1), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $1, KK				// number of values in A
+#endif
+
+.L4_100:
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+        addq    $4, KK				// number of values in B
+#endif
+
+
+	movq	K, %rax
+	salq	$2, %rax		// * 4
+	leaq	(B , %rax, SIZE), B
+
+
+
+
+/***************************************************************************************************************/
+
+.L2_0:
+
+	movq	Nmod12, J		
+	testq	$2, J
+	je	.L1_0
+
+.L2_10:
+	movq	C, CO1
+	leaq	(C, LDC, 2), C		// c += 2 * ldc
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        movq    OFFSET, %rax
+        movq    %rax, KK
+#endif
+
+
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L2_20
+
+	ALIGN_4
+
+.L2_11:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,2), BO		// add number of values in B
+	leaq	(AO,%rax,4), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $4, %rax        // number of values in AO
+#else
+        addq    $2, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT4x2
+
+	sarq $3, %rax			//  K / 8
+
+	je	.L2_16
+
+	ALIGN_5
+
+.L2_12:
+
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+	KERNEL4x2_SUB
+
+	dec	%rax
+	jne	.L2_12
+
+
+.L2_16:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L2_19
+
+	ALIGN_4
+
+.L2_17:
+
+	KERNEL4x2_SUB
+
+	dec	%rax
+	jne	.L2_17
+	ALIGN_4
+
+
+.L2_19:
+
+	SAVE4x2
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax + SIZE
+        leaq    (BO, %rax, 2), BO		// number of values in B
+        leaq    (AO, %rax, 4), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $4, KK				// number of values in A
+#endif
+
+
+	decq	I			# i --
+	jg	.L2_11
+	ALIGN_4	
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L2_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L2_100			// to next 16 lines of N
+
+
+.L2_30:
+	testq	$2, M		
+	jz	.L2_40
+
+	ALIGN_4
+
+.L2_31:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,2), BO		// add number of values in B
+	leaq	(AO,%rax,2), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $2, %rax        // number of values in AO
+#else
+        addq    $2, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT2x2
+
+	sarq	$3, %rax
+	je	.L2_36
+	ALIGN_4
+
+.L2_32:
+
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+	KERNEL2x2_SUB
+
+	dec %rax
+	jne	.L2_32
+
+.L2_36:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L2_39
+
+	ALIGN_4
+
+.L2_37:
+
+	KERNEL2x2_SUB
+
+	dec %rax
+	jne	.L2_37
+
+
+.L2_39:
+
+	SAVE2x2
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax + SIZE
+        leaq    (BO, %rax, 2), BO		// number of values in B
+        leaq    (AO, %rax, 2), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $2, KK				// number of values in A
+#endif
+
+
+.L2_40:
+	testq	$1, M		
+	jz	.L2_100		// to next 3 lines of N
+
+.L2_41:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,2), BO		// add number of values in B
+	leaq	(AO,%rax,1), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $1, %rax        // number of values in AO
+#else
+        addq    $2, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT1x2
+
+	sarq	$3,%rax
+	je	.L2_46
+
+	ALIGN_4
+
+.L2_42:
+
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+	KERNEL1x2_SUB
+
+	dec %rax
+	jne	.L2_42
+
+.L2_46:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L2_49
+
+	ALIGN_4
+
+.L2_47:
+
+	KERNEL1x2_SUB
+
+	dec	%rax
+	jne	.L2_47
+
+.L2_49:
+
+	SAVE1x2
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax * SIZE
+        leaq    (BO, %rax, 2), BO		// number of values in B
+        leaq    (AO, %rax, 1), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $1, KK				// number of values in A
+#endif
+
+
+.L2_100:
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+        addq    $2, KK				// number of values in B
+#endif
+
+	movq	K, %rax
+	salq	$1, %rax		// * 2
+	leaq	(B , %rax, SIZE), B
+
+/***************************************************************************************************************/
+
+.L1_0:
+
+	movq	Nmod12, J		
+	testq	$1, J
+	je	.L999
+
+.L1_10:
+	movq	C, CO1
+	leaq	(C, LDC, 1), C		// c += 1 * ldc
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        movq    OFFSET, %rax
+        movq    %rax, KK
+#endif
+	
+	movq	A, AO		 	// aoffset = a
+	addq	$16 * SIZE, AO
+
+	movq	M,  I
+	sarq	$2, I			// i = m / 4
+	je	.L1_20
+
+	ALIGN_4
+
+.L1_11:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,1), BO		// add number of values in B
+	leaq	(AO,%rax,4), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $4, %rax        // number of values in AO
+#else
+        addq    $1, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT4x1
+
+	sarq	$3, %rax			//  K / 8
+	je	.L1_16
+
+	ALIGN_5
+
+.L1_12:
+
+	KERNEL4x1
+
+	dec	%rax
+	jne	.L1_12
+
+
+.L1_16:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L1_19
+
+	ALIGN_4
+
+.L1_17:
+
+	KERNEL4x1_SUB
+
+	dec	%rax
+	jne	.L1_17
+	ALIGN_4
+
+
+.L1_19:
+
+	SAVE4x1
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax * SIZE
+        leaq    (BO, %rax, 1), BO		// number of values in B
+        leaq    (AO, %rax, 4), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $4, KK				// number of values in A
+#endif
+
+
+	decq	I			# i --
+	jg	.L1_11
+
+/**************************************************************************
+* Rest of M 
+***************************************************************************/
+.L1_20:
+	// Test rest of M
+
+	testq	$3, M
+	jz	.L1_100	
+
+
+.L1_30:
+	testq	$2, M		
+	jz	.L1_40
+
+	ALIGN_4
+
+.L1_31:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,1), BO		// add number of values in B
+	leaq	(AO,%rax,2), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $2, %rax        // number of values in AO
+#else
+        addq    $1, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT2x1
+
+	sarq	$3, %rax
+	je	.L1_36
+	ALIGN_4
+
+.L1_32:
+
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+	KERNEL2x1_SUB
+
+
+	dec %rax
+	jne	.L1_32
+
+.L1_36:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L1_39
+
+	ALIGN_4
+
+.L1_37:
+
+	KERNEL2x1_SUB
+
+	dec %rax
+	jne	.L1_37
+
+.L1_39:
+
+	SAVE2x1
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax * SIZE
+        leaq    (BO, %rax, 1), BO		// number of values in B
+        leaq    (AO, %rax, 2), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $2, KK				// number of values in A
+#endif
+
+
+.L1_40:
+	testq	$1, M		
+	jz	.L1_100		// to next 3 lines of N
+
+
+.L1_41:
+
+#if !defined(TRMMKERNEL) || \
+        (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+        (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+#else
+        movq    B, BO        
+        addq    $12 * SIZE, BO
+        movq    KK, %rax
+	salq	$3, %rax		// rax * SIZE
+	leaq	(BO,%rax,1), BO		// add number of values in B
+	leaq	(AO,%rax,1), AO		// add number of values in A
+#endif
+
+
+#ifndef TRMMKERNEL
+        movq    K, %rax
+#elif (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+        movq    K, %rax
+        subq    KK, %rax
+        movq    %rax, KKK
+#else
+        movq    KK, %rax
+#ifdef LEFT
+        addq    $1, %rax        // number of values in AO
+#else
+        addq    $1, %rax        // number of values in BO
+#endif
+        movq    %rax, KKK
+#endif
+
+	INIT1x1
+
+	sarq	$3,%rax
+	je	.L1_46
+
+	ALIGN_4
+
+.L1_42:
+
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+	KERNEL1x1_SUB
+
+	dec %rax
+	jne	.L1_42
+
+.L1_46:
+        movq    KKK, %rax
+
+	andq	$7, %rax		# if (k & 1)
+	je .L1_49
+
+	ALIGN_4
+
+.L1_47:
+
+	KERNEL1x1_SUB
+
+	dec	%rax
+	jne	.L1_47
+
+
+.L1_49:
+
+	SAVE1x1
+
+#if (defined(TRMMKERNEL) &&  defined(LEFT) &&  defined(TRANSA)) || \
+    (defined(TRMMKERNEL) && !defined(LEFT) && !defined(TRANSA))
+        movq    K, %rax
+        subq    KKK, %rax
+	salq	$3, %rax			// rax * SIZE
+        leaq    (BO, %rax, 1), BO		// number of values in B
+        leaq    (AO, %rax, 1), AO		// number of values in A
+#endif
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+        addq    $1, KK				// number of values in A
+#endif
+
+
+
+.L1_100:
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+        addq    $1, KK				// number of values in B
+#endif
+
+
+
+.L999:
+
+	vzeroupper
+
+	movq   		SP, %rsp
+	movq	   (%rsp), %rbx
+	movq	  8(%rsp), %rbp
+	movq	 16(%rsp), %r12
+	movq	 24(%rsp), %r13
+	movq	 32(%rsp), %r14
+	movq	 40(%rsp), %r15
+
+#ifdef WINDOWS_ABI
+	movq	 48(%rsp), %rdi
+	movq	 56(%rsp), %rsi
+	vmovups	 64(%rsp), %xmm6
+	vmovups	 80(%rsp), %xmm7
+	vmovups	 96(%rsp), %xmm8
+	vmovups	112(%rsp), %xmm9
+	vmovups	128(%rsp), %xmm10
+	vmovups	144(%rsp), %xmm11
+	vmovups	160(%rsp), %xmm12
+	vmovups	176(%rsp), %xmm13
+	vmovups	192(%rsp), %xmm14
+	vmovups	208(%rsp), %xmm15
+#endif
+
+	addq	$STACKSIZE, %rsp
+	ret
+
+	EPILOGUE
+
+
+
+
+
+#endif
diff --git a/kernel/x86_64/dgemv_n_4.c b/kernel/x86_64/dgemv_n_4.c
index 27df12bef..62016fc0b 100644
--- a/kernel/x86_64/dgemv_n_4.c
+++ b/kernel/x86_64/dgemv_n_4.c
@@ -37,48 +37,6 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 
 #define NBMAX 2048
-
-#ifndef HAVE_KERNEL_4x8
-
-static void dgemv_kernel_4x8(BLASLONG n, FLOAT **ap, FLOAT *xo, FLOAT *y, BLASLONG lda4, FLOAT *alpha)
-{
-	BLASLONG i;
-	FLOAT *a0,*a1,*a2,*a3;
-	FLOAT *b0,*b1,*b2,*b3;
-	FLOAT *x4;
-	FLOAT x[8];
-	a0 = ap[0];
-	a1 = ap[1];
-	a2 = ap[2];
-	a3 = ap[3];
-	b0 = a0 + lda4 ;
-	b1 = a1 + lda4 ;
-	b2 = a2 + lda4 ;
-	b3 = a3 + lda4 ;
-	x4 = x + 4;
-
-	for ( i=0; i<8; i++)
-		x[i] = xo[i] * *alpha;
-
-	for ( i=0; i< n; i+=4 )
-	{
-
-		y[i] += a0[i]*x[0] + a1[i]*x[1] + a2[i]*x[2] + a3[i]*x[3];		
-		y[i+1] += a0[i+1]*x[0] + a1[i+1]*x[1] + a2[i+1]*x[2] + a3[i+1]*x[3];		
-		y[i+2] += a0[i+2]*x[0] + a1[i+2]*x[1] + a2[i+2]*x[2] + a3[i+2]*x[3];		
-		y[i+3] += a0[i+3]*x[0] + a1[i+3]*x[1] + a2[i+3]*x[2] + a3[i+3]*x[3];		
-
-		y[i] += b0[i]*x4[0] + b1[i]*x4[1] + b2[i]*x4[2] + b3[i]*x4[3];		
-		y[i+1] += b0[i+1]*x4[0] + b1[i+1]*x4[1] + b2[i+1]*x4[2] + b3[i+1]*x4[3];		
-		y[i+2] += b0[i+2]*x4[0] + b1[i+2]*x4[1] + b2[i+2]*x4[2] + b3[i+2]*x4[3];		
-		y[i+3] += b0[i+3]*x4[0] + b1[i+3]*x4[1] + b2[i+3]*x4[2] + b3[i+3]*x4[3];		
-
-	}
-}
-	
-#endif
-
-
 #ifndef HAVE_KERNEL_4x4
 
 static void dgemv_kernel_4x4(BLASLONG n, FLOAT **ap, FLOAT *xo, FLOAT *y, FLOAT *alpha)
@@ -257,7 +215,6 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha, FLOAT *a, BLASLO
 	BLASLONG m3;
 	BLASLONG n2;
 	BLASLONG lda4 =  lda << 2;
-	BLASLONG lda8 =  lda << 3;
 	FLOAT xbuffer[8],*ybuffer;
 
         if ( m < 1 ) return(0);
@@ -265,23 +222,13 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha, FLOAT *a, BLASLO
 
 	ybuffer = buffer;
 	
-        if ( inc_x == 1 )
-	{
-		n1 = n >> 3 ;
-		n2 = n &  7 ;
-	}
-	else
-	{
-		n1 = n >> 2 ;
-		n2 = n &  3 ;
+	n1 = n >> 2 ;
+	n2 = n &  3 ;
 
-	}
-	
         m3 = m & 3  ;
         m1 = m & -4 ;
         m2 = (m & (NBMAX-1)) - m3 ;
 
-
 	y_ptr = y;
 
 	BLASLONG NB = NBMAX;
@@ -314,22 +261,12 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha, FLOAT *a, BLASLO
 
 
 			for( i = 0; i < n1 ; i++)
-			{
-				dgemv_kernel_4x8(NB,ap,x_ptr,ybuffer,lda4,&alpha);
-				ap[0] += lda8; 
-				ap[1] += lda8; 
-				ap[2] += lda8; 
-				ap[3] += lda8; 
-				a_ptr += lda8;
-				x_ptr += 8;	
-			}
-
-
-			if ( n2 & 4 )
 			{
 				dgemv_kernel_4x4(NB,ap,x_ptr,ybuffer,&alpha);
 				ap[0] += lda4; 
 				ap[1] += lda4; 
+				ap[2] += lda4; 
+				ap[3] += lda4; 
 				a_ptr += lda4;
 				x_ptr += 4;	
 			}
diff --git a/kernel/x86_64/dgemv_n_microk_haswell-4.c b/kernel/x86_64/dgemv_n_microk_haswell-4.c
index e1587b57c..b9f64407a 100644
--- a/kernel/x86_64/dgemv_n_microk_haswell-4.c
+++ b/kernel/x86_64/dgemv_n_microk_haswell-4.c
@@ -27,128 +27,6 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 
 
-#define HAVE_KERNEL_4x8 1
-static void dgemv_kernel_4x8( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, BLASLONG lda4, FLOAT *alpha) __attribute__ ((noinline));
-
-static void dgemv_kernel_4x8( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, BLASLONG lda4, FLOAT *alpha)
-{
-
-	BLASLONG register i = 0;
-
-	__asm__  __volatile__
-	(
-	"vzeroupper			 \n\t"
-	"vbroadcastsd    (%2), %%ymm12	 \n\t"	// x0 
-	"vbroadcastsd   8(%2), %%ymm13	 \n\t"	// x1 
-	"vbroadcastsd  16(%2), %%ymm14	 \n\t"	// x2 
-	"vbroadcastsd  24(%2), %%ymm15	 \n\t"	// x3 
-	"vbroadcastsd  32(%2), %%ymm0 	 \n\t"	// x4 
-	"vbroadcastsd  40(%2), %%ymm1 	 \n\t"	// x5 
-	"vbroadcastsd  48(%2), %%ymm2 	 \n\t"	// x6 
-	"vbroadcastsd  56(%2), %%ymm3 	 \n\t"	// x7 
-
-	"vbroadcastsd    (%9), %%ymm6 	 \n\t"	// alpha 
-
-        "testq          $0x04, %1                      \n\t"
-        "jz             2f                     \n\t"
-
-	"vmovupd	(%3,%0,8), %%ymm7	       \n\t"	// 4 * y
-	"vxorpd		%%ymm4 , %%ymm4, %%ymm4        \n\t"
-	"vxorpd		%%ymm5 , %%ymm5, %%ymm5        \n\t"
-
-	"vfmadd231pd   (%4,%0,8), %%ymm12, %%ymm4      \n\t" 
-	"vfmadd231pd   (%5,%0,8), %%ymm13, %%ymm5      \n\t" 
-	"vfmadd231pd   (%6,%0,8), %%ymm14, %%ymm4      \n\t" 
-	"vfmadd231pd   (%7,%0,8), %%ymm15, %%ymm5      \n\t" 
-
-	"vfmadd231pd   (%4,%8,8), %%ymm0 , %%ymm4      \n\t" 
-	"vfmadd231pd   (%5,%8,8), %%ymm1 , %%ymm5      \n\t" 
-	"vfmadd231pd   (%6,%8,8), %%ymm2 , %%ymm4      \n\t" 
-	"vfmadd231pd   (%7,%8,8), %%ymm3 , %%ymm5      \n\t" 
-
-	"vaddpd		%%ymm4 , %%ymm5 , %%ymm5       \n\t"
-	"vmulpd		%%ymm6 , %%ymm5 , %%ymm5       \n\t"
-	"vaddpd		%%ymm7 , %%ymm5 , %%ymm5       \n\t"
-
-
-	"vmovupd  %%ymm5,   (%3,%0,8)		       \n\t"	// 4 * y
-
-        "addq		$4 , %8	  	 	       \n\t"
-        "addq		$4 , %0	  	 	       \n\t"
-	"subq	        $4 , %1			       \n\t"		
-
-        "2:                                   \n\t"
-
-        "cmpq           $0, %1                         \n\t"
-        "je             3f                      \n\t"
-
-
-	".align 16				 \n\t"
-	"1:				 \n\t"
-
-	"vxorpd		%%ymm4 , %%ymm4, %%ymm4        \n\t"
-	"vxorpd		%%ymm5 , %%ymm5, %%ymm5        \n\t"
-	"vmovupd	(%3,%0,8), %%ymm8	       \n\t"	// 4 * y
-	"vmovupd      32(%3,%0,8), %%ymm9	       \n\t"	// 4 * y
-
-	"vfmadd231pd   (%4,%0,8), %%ymm12, %%ymm4      \n\t" 
-	"vfmadd231pd 32(%4,%0,8), %%ymm12, %%ymm5      \n\t" 
-	"vfmadd231pd   (%5,%0,8), %%ymm13, %%ymm4      \n\t" 
-	"vfmadd231pd 32(%5,%0,8), %%ymm13, %%ymm5      \n\t" 
-	"vfmadd231pd   (%6,%0,8), %%ymm14, %%ymm4      \n\t" 
-	"vfmadd231pd 32(%6,%0,8), %%ymm14, %%ymm5      \n\t" 
-	"vfmadd231pd   (%7,%0,8), %%ymm15, %%ymm4      \n\t" 
-	"vfmadd231pd 32(%7,%0,8), %%ymm15, %%ymm5      \n\t" 
-
-	"vfmadd231pd   (%4,%8,8), %%ymm0 , %%ymm4      \n\t" 
-        "addq		$8 , %0	  	 	       \n\t"
-	"vfmadd231pd 32(%4,%8,8), %%ymm0 , %%ymm5      \n\t" 
-	"vfmadd231pd   (%5,%8,8), %%ymm1 , %%ymm4      \n\t" 
-	"vfmadd231pd 32(%5,%8,8), %%ymm1 , %%ymm5      \n\t" 
-	"vfmadd231pd   (%6,%8,8), %%ymm2 , %%ymm4      \n\t" 
-	"vfmadd231pd 32(%6,%8,8), %%ymm2 , %%ymm5      \n\t" 
-	"vfmadd231pd   (%7,%8,8), %%ymm3 , %%ymm4      \n\t" 
-	"vfmadd231pd 32(%7,%8,8), %%ymm3 , %%ymm5      \n\t" 
-
-	"vfmadd231pd     %%ymm6 , %%ymm4 , %%ymm8      \n\t"
-	"vfmadd231pd     %%ymm6 , %%ymm5 , %%ymm9      \n\t"
-
-        "addq		$8 , %8	  	 	      \n\t"
-	"vmovupd  %%ymm8,-64(%3,%0,8)		      \n\t"	// 4 * y
-	"subq	        $8 , %1			      \n\t"		
-	"vmovupd  %%ymm9,-32(%3,%0,8)		      \n\t"	// 4 * y
-
-	"jnz		1b		      \n\t"
-
-        "3:                             \n\t"
-	"vzeroupper			        \n\t"
-
-	:
-        : 
-          "r" (i),	// 0	
-	  "r" (n),  	// 1
-          "r" (x),      // 2
-          "r" (y),      // 3
-          "r" (ap[0]),  // 4
-          "r" (ap[1]),  // 5
-          "r" (ap[2]),  // 6
-          "r" (ap[3]),  // 7
-          "r" (lda4),   // 8
-          "r" (alpha)   // 9
-	: "cc", 
-	  "%xmm0", "%xmm1", 
-	  "%xmm2", "%xmm3", 
-	  "%xmm4", "%xmm5", 
-	  "%xmm6", "%xmm7", 
-	  "%xmm8", "%xmm9", 
-	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
-	  "memory"
-	);
-
-} 
-
-
-
 #define HAVE_KERNEL_4x4 1
 static void dgemv_kernel_4x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT *alpha) __attribute__ ((noinline));
 
@@ -159,68 +37,59 @@ static void dgemv_kernel_4x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT
 
 	__asm__  __volatile__
 	(
-	"vzeroupper			 \n\t"
 	"vbroadcastsd    (%2), %%ymm12	 \n\t"	// x0 
 	"vbroadcastsd   8(%2), %%ymm13	 \n\t"	// x1 
 	"vbroadcastsd  16(%2), %%ymm14	 \n\t"	// x2 
 	"vbroadcastsd  24(%2), %%ymm15	 \n\t"	// x3 
 
+	"vmovups	(%4,%0,8), %%ymm0	 \n\t"
+	"vmovups	(%5,%0,8), %%ymm1	 \n\t"
+	"vmovups	(%6,%0,8), %%ymm2	 \n\t"
+	"vmovups	(%7,%0,8), %%ymm3	 \n\t"
 	"vbroadcastsd    (%8), %%ymm6 	 \n\t"	// alpha 
 
-        "testq          $0x04, %1                      \n\t"
-        "jz             2f                     \n\t"
-
-	"vxorpd		%%ymm4 , %%ymm4, %%ymm4        \n\t"
-	"vxorpd		%%ymm5 , %%ymm5, %%ymm5        \n\t"
-	"vmovupd	(%3,%0,8), %%ymm7	       \n\t"	// 4 * y
-
-	"vfmadd231pd   (%4,%0,8), %%ymm12, %%ymm4      \n\t" 
-	"vfmadd231pd   (%5,%0,8), %%ymm13, %%ymm5      \n\t" 
-	"vfmadd231pd   (%6,%0,8), %%ymm14, %%ymm4      \n\t" 
-	"vfmadd231pd   (%7,%0,8), %%ymm15, %%ymm5      \n\t" 
-
-	"vaddpd		%%ymm4 , %%ymm5 , %%ymm5       \n\t"
-	"vmulpd		%%ymm6 , %%ymm5 , %%ymm5       \n\t"
-	"vaddpd		%%ymm7 , %%ymm5 , %%ymm5       \n\t"
-
-	"vmovupd  %%ymm5,   (%3,%0,8)		       \n\t"	// 4 * y
-
-        "addq		$4 , %0	  	 	       \n\t"
-	"subq	        $4 , %1			       \n\t"		
-
-        "2:                                   \n\t"
-
-        "cmpq           $0, %1                         \n\t"
-        "je             3f                       \n\t"
-
+        "addq		$4 , %0	  	 	      \n\t"
+	"subq	        $4 , %1			      \n\t"		
+	"jz		2f		      \n\t"
 
 	".align 16				 \n\t"
 	"1:				 \n\t"
-	"vxorpd		%%ymm4 , %%ymm4, %%ymm4        \n\t"
-	"vxorpd		%%ymm5 , %%ymm5, %%ymm5        \n\t"
-	"vmovupd	(%3,%0,8), %%ymm8	       \n\t"	// 4 * y
-	"vmovupd      32(%3,%0,8), %%ymm9	       \n\t"	// 4 * y
 
-	"vfmadd231pd   (%4,%0,8), %%ymm12, %%ymm4      \n\t" 
-	"vfmadd231pd 32(%4,%0,8), %%ymm12, %%ymm5      \n\t" 
-	"vfmadd231pd   (%5,%0,8), %%ymm13, %%ymm4      \n\t" 
-	"vfmadd231pd 32(%5,%0,8), %%ymm13, %%ymm5      \n\t" 
-	"vfmadd231pd   (%6,%0,8), %%ymm14, %%ymm4      \n\t" 
-	"vfmadd231pd 32(%6,%0,8), %%ymm14, %%ymm5      \n\t" 
-	"vfmadd231pd   (%7,%0,8), %%ymm15, %%ymm4      \n\t" 
-	"vfmadd231pd 32(%7,%0,8), %%ymm15, %%ymm5      \n\t" 
+	"vmulpd        %%ymm0 , %%ymm12, %%ymm4      \n\t" 
+	"vmulpd        %%ymm1 , %%ymm13, %%ymm5      \n\t" 
+	"vmovups	(%4,%0,8), %%ymm0	 \n\t"
+	"vmovups	(%5,%0,8), %%ymm1	 \n\t"
+	"vfmadd231pd   %%ymm2 , %%ymm14, %%ymm4	     \n\t"
+	"vfmadd231pd   %%ymm3 , %%ymm15, %%ymm5	     \n\t"
+	"vmovups	(%6,%0,8), %%ymm2	 \n\t"
+	"vmovups	(%7,%0,8), %%ymm3	 \n\t"
 
+	"vmovups	-32(%3,%0,8), %%ymm8	       \n\t"	// 4 * y
+	"vaddpd		 %%ymm4 , %%ymm5 , %%ymm4      \n\t"
 	"vfmadd231pd     %%ymm6 , %%ymm4 , %%ymm8      \n\t"
-	"vfmadd231pd     %%ymm6 , %%ymm5 , %%ymm9      \n\t"
 
-	"vmovupd  %%ymm8,   (%3,%0,8)		      \n\t"	// 4 * y
-	"vmovupd  %%ymm9, 32(%3,%0,8)		      \n\t"	// 4 * y
+	"vmovups         %%ymm8,   -32(%3,%0,8)	      \n\t"	// 4 * y
 
-        "addq		$8 , %0	  	 	      \n\t"
-	"subq	        $8 , %1			      \n\t"		
+        "addq		$4 , %0	  	 	      \n\t"
+	"subq	        $4 , %1			      \n\t"		
 	"jnz		1b		      \n\t"
+	
+
+	"2:				 \n\t"
+
+	"vmulpd        %%ymm0 , %%ymm12, %%ymm4      \n\t" 
+	"vmulpd        %%ymm1 , %%ymm13, %%ymm5      \n\t" 
+	"vfmadd231pd   %%ymm2 , %%ymm14, %%ymm4	     \n\t"
+	"vfmadd231pd   %%ymm3 , %%ymm15, %%ymm5	     \n\t"
+
+
+	"vmovups	-32(%3,%0,8), %%ymm8	       \n\t"	// 4 * y
+	"vaddpd		 %%ymm4 , %%ymm5 , %%ymm4      \n\t"
+	"vfmadd231pd     %%ymm6 , %%ymm4 , %%ymm8      \n\t"
+
+	"vmovups  %%ymm8,   -32(%3,%0,8)	      \n\t"	// 4 * y
+
 
-        "3:                                    \n\t"
 	"vzeroupper			              \n\t"
 
 	:
diff --git a/kernel/x86_64/dgemv_n_microk_nehalem-4.c b/kernel/x86_64/dgemv_n_microk_nehalem-4.c
index 0d2c24d52..d8c29831a 100644
--- a/kernel/x86_64/dgemv_n_microk_nehalem-4.c
+++ b/kernel/x86_64/dgemv_n_microk_nehalem-4.c
@@ -27,150 +27,6 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 
 
-#define HAVE_KERNEL_4x8 1
-static void dgemv_kernel_4x8( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, BLASLONG lda4, FLOAT *alpha) __attribute__ ((noinline));
-
-static void dgemv_kernel_4x8( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, BLASLONG lda4, FLOAT *alpha)
-{
-
-	BLASLONG register i = 0;
-
-	__asm__  __volatile__
-	(
-	"movsd    (%2), %%xmm12	 \n\t"	// x0 
-	"movsd   8(%2), %%xmm13	 \n\t"	// x1 
-	"movsd  16(%2), %%xmm14	 \n\t"	// x2 
-	"movsd  24(%2), %%xmm15	 \n\t"	// x3 
-	"shufpd $0,  %%xmm12, %%xmm12\n\t"	
-	"shufpd $0,  %%xmm13, %%xmm13\n\t"	
-	"shufpd $0,  %%xmm14, %%xmm14\n\t"	
-	"shufpd $0,  %%xmm15, %%xmm15\n\t"	
-
-	"movsd  32(%2), %%xmm0	 \n\t"	// x4 
-	"movsd  40(%2), %%xmm1	 \n\t"	// x5 
-	"movsd  48(%2), %%xmm2	 \n\t"	// x6 
-	"movsd  56(%2), %%xmm3	 \n\t"	// x7 
-	"shufpd $0,  %%xmm0 , %%xmm0 \n\t"	
-	"shufpd $0,  %%xmm1 , %%xmm1 \n\t"	
-	"shufpd $0,  %%xmm2 , %%xmm2 \n\t"	
-	"shufpd $0,  %%xmm3 , %%xmm3 \n\t"	
-
-	"movsd    (%9), %%xmm6	     \n\t"	// alpha 
-	"shufpd $0,  %%xmm6 , %%xmm6 \n\t"	
-
-
-	".align 16				 \n\t"
-	"1:				 \n\t"
-	"xorpd           %%xmm4 , %%xmm4	 \n\t"
-	"xorpd           %%xmm5 , %%xmm5	 \n\t"
-	"movups             (%3,%0,8), %%xmm7          \n\t" // 2 * y
-
-	".align 2				       \n\t"
-	"movups             (%4,%0,8), %%xmm8          \n\t" 
-	"movups             (%5,%0,8), %%xmm9          \n\t" 
-	"movups             (%6,%0,8), %%xmm10         \n\t" 
-	"movups             (%7,%0,8), %%xmm11         \n\t" 
-	".align 2				       \n\t"
-	"mulpd		%%xmm12, %%xmm8		       \n\t"
-	"mulpd		%%xmm13, %%xmm9		       \n\t"
-	"mulpd		%%xmm14, %%xmm10	       \n\t"
-	"mulpd		%%xmm15, %%xmm11	       \n\t"
-	"addpd		%%xmm8 , %%xmm4		       \n\t"
-	"addpd		%%xmm9 , %%xmm5		       \n\t"
-	"addpd		%%xmm10, %%xmm4	               \n\t"
-	"addpd		%%xmm11, %%xmm5 	       \n\t"
-
-	"movups             (%4,%8,8), %%xmm8          \n\t" 
-	"movups             (%5,%8,8), %%xmm9          \n\t" 
-	"movups             (%6,%8,8), %%xmm10         \n\t" 
-	"movups             (%7,%8,8), %%xmm11         \n\t" 
-	".align 2				       \n\t"
-	"mulpd		%%xmm0 , %%xmm8		       \n\t"
-	"mulpd		%%xmm1 , %%xmm9		       \n\t"
-	"mulpd		%%xmm2 , %%xmm10	       \n\t"
-	"mulpd		%%xmm3 , %%xmm11	       \n\t"
-	"addpd		%%xmm8 , %%xmm4		       \n\t"
-	"addpd		%%xmm9 , %%xmm5		       \n\t"
-	"addpd		%%xmm10, %%xmm4	       	       \n\t"
-	"addpd		%%xmm11, %%xmm5 	       \n\t"
-
-	"addpd		%%xmm5 , %%xmm4 	       \n\t"
-	"mulpd		%%xmm6 , %%xmm4		       \n\t" 
-	"addpd		%%xmm4 , %%xmm7 	       \n\t"
-
-	"movups  %%xmm7 ,    (%3,%0,8)		       \n\t"	// 2 * y
-
-	"xorpd           %%xmm4 , %%xmm4	 \n\t"
-	"xorpd           %%xmm5 , %%xmm5	 \n\t"
-	"movups           16(%3,%0,8), %%xmm7          \n\t" // 2 * y
-
-	".align 2				       \n\t"
-	"movups           16(%4,%0,8), %%xmm8          \n\t" 
-	"movups           16(%5,%0,8), %%xmm9          \n\t" 
-	"movups           16(%6,%0,8), %%xmm10         \n\t" 
-	"movups           16(%7,%0,8), %%xmm11         \n\t" 
-	".align 2				       \n\t"
-	"mulpd		%%xmm12, %%xmm8		       \n\t"
-	"mulpd		%%xmm13, %%xmm9		       \n\t"
-	"mulpd		%%xmm14, %%xmm10	       \n\t"
-	"mulpd		%%xmm15, %%xmm11	       \n\t"
-	"addpd		%%xmm8 , %%xmm4		       \n\t"
-	"addpd		%%xmm9 , %%xmm5		       \n\t"
-	"addpd		%%xmm10, %%xmm4	               \n\t"
-	"addpd		%%xmm11, %%xmm5 	       \n\t"
-
-	"movups           16(%4,%8,8), %%xmm8          \n\t" 
-	"movups           16(%5,%8,8), %%xmm9          \n\t" 
-	"movups           16(%6,%8,8), %%xmm10         \n\t" 
-	"movups           16(%7,%8,8), %%xmm11         \n\t" 
-	".align 2				       \n\t"
-	"mulpd		%%xmm0 , %%xmm8		       \n\t"
-	"mulpd		%%xmm1 , %%xmm9		       \n\t"
-	"mulpd		%%xmm2 , %%xmm10	       \n\t"
-	"mulpd		%%xmm3 , %%xmm11	       \n\t"
-	"addpd		%%xmm8 , %%xmm4		       \n\t"
-	"addpd		%%xmm9 , %%xmm5		       \n\t"
-	"addpd		%%xmm10, %%xmm4	       	       \n\t"
-	"addpd		%%xmm11, %%xmm5 	       \n\t"
-
-        "addq		$4 , %8	  	 	       \n\t"
-	"addpd		%%xmm5 , %%xmm4 	       \n\t"
-	"mulpd		%%xmm6 , %%xmm4		       \n\t" 
-	"addpd		%%xmm4 , %%xmm7 	       \n\t"
-
-	"movups  %%xmm7 ,  16(%3,%0,8)		       \n\t"	// 2 * y
-
-        "addq		$4 , %0	  	 	       \n\t"
-	"subq	        $4 , %1			       \n\t"		
-	"jnz		1b		       \n\t"
-
-	:
-        : 
-          "r" (i),	// 0	
-	  "r" (n),  	// 1
-          "r" (x),      // 2
-          "r" (y),      // 3
-          "r" (ap[0]),  // 4
-          "r" (ap[1]),  // 5
-          "r" (ap[2]),  // 6
-          "r" (ap[3]),  // 7
-          "r" (lda4),   // 8
-          "r" (alpha)   // 9
-	: "cc", 
-	  "%xmm0", "%xmm1", 
-	  "%xmm2", "%xmm3", 
-	  "%xmm4", "%xmm5", 
-	  "%xmm6", "%xmm7", 
-	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
-	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
-	  "memory"
-	);
-
-} 
-
-
-
-
 #define HAVE_KERNEL_4x4 1
 static void dgemv_kernel_4x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT *alpha) __attribute__ ((noinline));
 
@@ -193,54 +49,105 @@ static void dgemv_kernel_4x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT
 	"movsd    (%8), %%xmm6	     \n\t"	// alpha 
 	"shufpd $0,  %%xmm6 , %%xmm6 \n\t"	
 
+	"movups             (%4,%0,8), %%xmm8          \n\t" 
+	"movups           16(%4,%0,8), %%xmm0          \n\t" 
+	"movups             (%5,%0,8), %%xmm9          \n\t" 
+	"movups           16(%5,%0,8), %%xmm1          \n\t" 
+	"movups             (%6,%0,8), %%xmm10         \n\t" 
+	"movups           16(%6,%0,8), %%xmm2          \n\t" 
+	"movups             (%7,%0,8), %%xmm11         \n\t" 
+	"movups           16(%7,%0,8), %%xmm3          \n\t" 
+
+        "addq		$4 , %0	  	 	       \n\t"
+	"subq	        $4 , %1			       \n\t"		
+	"jz		2f		       \n\t"
+
 	".align 16				 \n\t"
 	"1:				 \n\t"
+
 	"xorpd           %%xmm4 , %%xmm4	 \n\t"
 	"xorpd           %%xmm5 , %%xmm5	 \n\t"
-	"movups	       (%3,%0,8), %%xmm7	 \n\t"	// 2 * y
+	"movups	    -32(%3,%0,8), %%xmm7	 \n\t"	// 2 * y
+
+	"mulpd		%%xmm12, %%xmm8		       \n\t"
+	"mulpd		%%xmm12, %%xmm0		       \n\t"
+	"addpd		%%xmm8 , %%xmm4		       \n\t"
+	"addpd		%%xmm0 , %%xmm5		       \n\t"
 
 	"movups             (%4,%0,8), %%xmm8          \n\t" 
+	"movups           16(%4,%0,8), %%xmm0          \n\t" 
+
+	"mulpd		%%xmm13, %%xmm9		       \n\t"
+	"mulpd		%%xmm13, %%xmm1		       \n\t"
+	"addpd		%%xmm9 , %%xmm4		       \n\t"
+	"addpd		%%xmm1 , %%xmm5		       \n\t"
+
 	"movups             (%5,%0,8), %%xmm9          \n\t" 
+	"movups           16(%5,%0,8), %%xmm1          \n\t" 
+
+	"mulpd		%%xmm14, %%xmm10	       \n\t"
+	"mulpd		%%xmm14, %%xmm2 	       \n\t"
+	"addpd		%%xmm10 , %%xmm4	       \n\t"
+	"addpd		%%xmm2  , %%xmm5	       \n\t"
+
 	"movups             (%6,%0,8), %%xmm10         \n\t" 
+	"movups           16(%6,%0,8), %%xmm2          \n\t" 
+
+	"mulpd		%%xmm15, %%xmm11	       \n\t"
+	"mulpd		%%xmm15, %%xmm3 	       \n\t"
+	"addpd		%%xmm11 , %%xmm4	       \n\t"
+	"addpd		%%xmm3  , %%xmm5	       \n\t"
+
 	"movups             (%7,%0,8), %%xmm11         \n\t" 
-	"mulpd		%%xmm12, %%xmm8		       \n\t"
-	"mulpd		%%xmm13, %%xmm9		       \n\t"
-	"mulpd		%%xmm14, %%xmm10	       \n\t"
-	"mulpd		%%xmm15, %%xmm11	       \n\t"
-	"addpd		%%xmm8 , %%xmm4		       \n\t"
-	"addpd		%%xmm9 , %%xmm4		       \n\t"
-	"addpd		%%xmm10 , %%xmm4	       \n\t"
-	"addpd		%%xmm4 , %%xmm11	       \n\t"
+	"movups           16(%7,%0,8), %%xmm3          \n\t" 
 
-	"mulpd		%%xmm6 , %%xmm11	       \n\t" 
-	"addpd		%%xmm7 , %%xmm11 	       \n\t"
-	"movups  %%xmm11,    (%3,%0,8)		       \n\t"	// 2 * y
 
-	"xorpd           %%xmm4 , %%xmm4	 \n\t"
-	"xorpd           %%xmm5 , %%xmm5	 \n\t"
-	"movups	     16(%3,%0,8), %%xmm7	 \n\t"	// 2 * y
+	"mulpd		%%xmm6 , %%xmm4 	       \n\t" 
+	"addpd		%%xmm7 , %%xmm4  	       \n\t"
+	"movups	    -16(%3,%0,8), %%xmm7	 \n\t"	// 2 * y
+	"movups         %%xmm4 ,  -32(%3,%0,8)	       \n\t"	// 2 * y
 
-	"movups           16(%4,%0,8), %%xmm8          \n\t" 
-	"movups           16(%5,%0,8), %%xmm9          \n\t" 
-	"movups           16(%6,%0,8), %%xmm10         \n\t" 
-	"movups           16(%7,%0,8), %%xmm11         \n\t" 
-	"mulpd		%%xmm12, %%xmm8		       \n\t"
-	"mulpd		%%xmm13, %%xmm9		       \n\t"
-	"mulpd		%%xmm14, %%xmm10	       \n\t"
-	"mulpd		%%xmm15, %%xmm11	       \n\t"
-	"addpd		%%xmm8 , %%xmm4		       \n\t"
-	"addpd		%%xmm9 , %%xmm4		       \n\t"
-	"addpd		%%xmm10 , %%xmm4	       \n\t"
-	"addpd		%%xmm4 , %%xmm11	       \n\t"
-
-	"mulpd		%%xmm6 , %%xmm11	       \n\t" 
-	"addpd		%%xmm7 , %%xmm11 	       \n\t"
-	"movups  %%xmm11,  16(%3,%0,8)		       \n\t"	// 2 * y
+	"mulpd		%%xmm6 , %%xmm5 	       \n\t" 
+	"addpd		%%xmm7 , %%xmm5  	       \n\t"
+	"movups         %%xmm5 ,  -16(%3,%0,8)	       \n\t"	// 2 * y
 
         "addq		$4 , %0	  	 	       \n\t"
 	"subq	        $4 , %1			       \n\t"		
 	"jnz		1b		       \n\t"
 
+	"2:				 \n\t"
+
+	"xorpd           %%xmm4 , %%xmm4	 \n\t"
+	"xorpd           %%xmm5 , %%xmm5	 \n\t"
+
+	"mulpd		%%xmm12, %%xmm8		       \n\t"
+	"addpd		%%xmm8 , %%xmm4		       \n\t"
+	"mulpd		%%xmm13, %%xmm9		       \n\t"
+	"addpd		%%xmm9 , %%xmm4		       \n\t"
+	"mulpd		%%xmm14, %%xmm10	       \n\t"
+	"addpd		%%xmm10 , %%xmm4	       \n\t"
+	"mulpd		%%xmm15, %%xmm11	       \n\t"
+	"addpd		%%xmm11 , %%xmm4	       \n\t"
+
+	"mulpd		%%xmm12, %%xmm0		       \n\t"
+	"addpd		%%xmm0 , %%xmm5		       \n\t"
+	"mulpd		%%xmm13, %%xmm1		       \n\t"
+	"addpd		%%xmm1 , %%xmm5		       \n\t"
+	"mulpd		%%xmm14, %%xmm2 	       \n\t"
+	"addpd		%%xmm2 , %%xmm5	       \n\t"
+	"mulpd		%%xmm15, %%xmm3 	       \n\t"
+	"addpd		%%xmm3 , %%xmm5	       \n\t"
+
+	"movups	    -32(%3,%0,8), %%xmm7	 \n\t"	// 2 * y
+	"mulpd		%%xmm6 , %%xmm4 	       \n\t" 
+	"addpd		%%xmm7 , %%xmm4  	       \n\t"
+	"movups         %%xmm4 ,  -32(%3,%0,8)	       \n\t"	// 2 * y
+
+	"movups	    -16(%3,%0,8), %%xmm7	 \n\t"	// 2 * y
+	"mulpd		%%xmm6 , %%xmm5 	       \n\t" 
+	"addpd		%%xmm7 , %%xmm5  	       \n\t"
+	"movups         %%xmm5 ,  -16(%3,%0,8)	       \n\t"	// 2 * y
+
 	:
         : 
           "r" (i),	// 0	
@@ -253,8 +160,8 @@ static void dgemv_kernel_4x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT
           "r" (ap[3]),  // 7
           "r" (alpha)   // 8
 	: "cc", 
-	  "%xmm4", "%xmm5", 
-	  "%xmm6", "%xmm7", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
 	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
 	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
 	  "memory"
diff --git a/kernel/x86_64/dgemv_t_4.c b/kernel/x86_64/dgemv_t_4.c
index 5d85ecab7..7c550a759 100644
--- a/kernel/x86_64/dgemv_t_4.c
+++ b/kernel/x86_64/dgemv_t_4.c
@@ -293,7 +293,7 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha, FLOAT *a, BLASLO
         if ( n < 1 ) return(0);
 
 	xbuffer = buffer;
-	ytemp   = buffer + NBMAX;
+	ytemp   = buffer + (m < NBMAX ? m : NBMAX);
 	
 	n0 = n / NBMAX;
         n1 = (n % NBMAX)  >> 2 ;
diff --git a/kernel/x86_64/dger.c b/kernel/x86_64/dger.c
new file mode 100644
index 000000000..157a8ea7f
--- /dev/null
+++ b/kernel/x86_64/dger.c
@@ -0,0 +1,84 @@
+/*********************************************************************/
+/* Copyright 2009, 2010 The University of Texas at Austin.           */
+/* All rights reserved.                                              */
+/*                                                                   */
+/* Redistribution and use in source and binary forms, with or        */
+/* without modification, are permitted provided that the following   */
+/* conditions are met:                                               */
+/*                                                                   */
+/*   1. Redistributions of source code must retain the above         */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer.                                                  */
+/*                                                                   */
+/*   2. Redistributions in binary form must reproduce the above      */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer in the documentation and/or other materials       */
+/*      provided with the distribution.                              */
+/*                                                                   */
+/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
+/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
+/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
+/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
+/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
+/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
+/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
+/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
+/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
+/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
+/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
+/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
+/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
+/*    POSSIBILITY OF SUCH DAMAGE.                                    */
+/*                                                                   */
+/* The views and conclusions contained in the software and           */
+/* documentation are those of the authors and should not be          */
+/* interpreted as representing official policies, either expressed   */
+/* or implied, of The University of Texas at Austin.                 */
+/*********************************************************************/
+
+#include <stdio.h>
+#include "common.h"
+
+#if defined(SANDYBRIDGE)
+#include "dger_microk_sandy-2.c"
+#endif
+
+int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha,
+	 FLOAT *x, BLASLONG incx,
+	 FLOAT *y, BLASLONG incy,
+	 FLOAT *a, BLASLONG lda, FLOAT *buffer){
+
+	FLOAT *X    = x;
+
+	if (incx != 1) {
+    		X = buffer;
+    		COPY_K(m, x, incx, X, 1);
+	}
+
+	BLASLONG m1 = m & -16;
+
+	while (n > 0) 
+	{
+   		FLOAT y0 = alpha * *y;
+		if ( m1 > 0 )
+		{
+			#ifdef HAVE_KERNEL_16
+				dger_kernel_16(m1, X, a, &y0);
+			#else
+    				AXPYU_K(m1, 0, 0, y0, X, 1, a, 1, NULL, 0);
+			#endif
+		}
+	
+		if ( m > m1 )
+		{
+    			AXPYU_K(m-m1, 0, 0, y0, X+m1 , 1, a+m1, 1, NULL, 0);
+		}
+	
+    		a += lda;
+    		y += incy;
+    		n --;
+	}
+
+	return 0;
+}
+
diff --git a/kernel/x86_64/dger_microk_sandy-2.c b/kernel/x86_64/dger_microk_sandy-2.c
new file mode 100644
index 000000000..564f1356d
--- /dev/null
+++ b/kernel/x86_64/dger_microk_sandy-2.c
@@ -0,0 +1,124 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void dger_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void dger_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vmovddup		(%4), %%xmm0		    \n\t"  // alpha	
+	"prefetcht0	256(%3,%0,8)			    \n\t"
+	"vmovups	  (%3,%0,8), %%xmm8         \n\t"
+	"vmovups	16(%3,%0,8), %%xmm9         \n\t"
+	"vmovups	32(%3,%0,8), %%xmm10        \n\t"
+	"vmovups	48(%3,%0,8), %%xmm11        \n\t"
+
+	"prefetcht0	256(%2,%0,8)			    \n\t"
+	"vmovups	  (%2,%0,8), %%xmm4         \n\t"
+	"vmovups	16(%2,%0,8), %%xmm5         \n\t"
+	"vmovups	32(%2,%0,8), %%xmm6         \n\t"
+	"vmovups	48(%2,%0,8), %%xmm7        \n\t"
+
+	"addq		$8, %0	  	 	             \n\t"
+	"subq	        $8, %1			             \n\t"		
+	"jz		2f		             \n\t"
+
+	".align 8				            \n\t"
+	"1:				            \n\t"
+
+	"vmulpd		%%xmm4, %%xmm0, %%xmm4		\n\t"
+	"vaddpd		%%xmm8 , %%xmm4, %%xmm12	     \n\t"
+	"vmulpd		%%xmm5, %%xmm0, %%xmm5		\n\t"
+	"vaddpd		%%xmm9 , %%xmm5, %%xmm13	     \n\t"
+	"vmulpd		%%xmm6, %%xmm0, %%xmm6		\n\t"
+	"vaddpd		%%xmm10, %%xmm6, %%xmm14	     \n\t"
+	"vmulpd		%%xmm7, %%xmm0, %%xmm7		\n\t"
+	"vaddpd		%%xmm11, %%xmm7, %%xmm15	     \n\t"
+
+	"prefetcht0	256(%3,%0,8)			    \n\t"
+	"vmovups	  (%3,%0,8), %%xmm8         \n\t"
+	"vmovups	16(%3,%0,8), %%xmm9         \n\t"
+	"vmovups	32(%3,%0,8), %%xmm10        \n\t"
+	"vmovups	48(%3,%0,8), %%xmm11        \n\t"
+
+	"prefetcht0	256(%2,%0,8)			    \n\t"
+	"vmovups	  (%2,%0,8), %%xmm4         \n\t"
+	"vmovups	16(%2,%0,8), %%xmm5         \n\t"
+	"vmovups	32(%2,%0,8), %%xmm6         \n\t"
+	"vmovups	48(%2,%0,8), %%xmm7        \n\t"
+
+	"vmovups	%%xmm12,  -64(%3,%0,8)		     \n\t"
+	"vmovups	%%xmm13,  -48(%3,%0,8)		     \n\t"
+	"vmovups	%%xmm14,  -32(%3,%0,8)		     \n\t"
+	"vmovups	%%xmm15,  -16(%3,%0,8)		     \n\t"
+
+	"addq		$8, %0	  	 	             \n\t"
+	"subq	        $8, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+
+	"2:				            \n\t"
+	"vmulpd		%%xmm4, %%xmm0, %%xmm4		\n\t"
+	"vmulpd		%%xmm5, %%xmm0, %%xmm5		\n\t"
+	"vmulpd		%%xmm6, %%xmm0, %%xmm6		\n\t"
+	"vmulpd		%%xmm7, %%xmm0, %%xmm7		\n\t"
+
+	"vaddpd		%%xmm8 , %%xmm4, %%xmm12	     \n\t"
+	"vaddpd		%%xmm9 , %%xmm5, %%xmm13	     \n\t"
+	"vaddpd		%%xmm10, %%xmm6, %%xmm14	     \n\t"
+	"vaddpd		%%xmm11, %%xmm7, %%xmm15	     \n\t"
+
+	"vmovups	%%xmm12,  -64(%3,%0,8)		     \n\t"
+	"vmovups	%%xmm13,  -48(%3,%0,8)		     \n\t"
+	"vmovups	%%xmm14,  -32(%3,%0,8)		     \n\t"
+	"vmovups	%%xmm15,  -16(%3,%0,8)		     \n\t"
+
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dscal.c b/kernel/x86_64/dscal.c
new file mode 100644
index 000000000..e3e2b0d58
--- /dev/null
+++ b/kernel/x86_64/dscal.c
@@ -0,0 +1,238 @@
+/***************************************************************************
+Copyright (c) 2013 - 2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+#if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)
+#include "dscal_microk_bulldozer-2.c"
+#elif defined(SANDYBRIDGE)
+#include "dscal_microk_sandy-2.c"
+#elif defined(HASWELL)
+#include "dscal_microk_haswell-2.c"
+#endif
+
+
+#if !defined(HAVE_KERNEL_8)
+
+static void dscal_kernel_8( BLASLONG n, FLOAT *da , FLOAT *x )
+{
+
+	BLASLONG i;
+	FLOAT alpha = *da;
+
+	for( i=0; i<n; i+=8 )
+	{
+		x[0] *= alpha;	
+		x[1] *= alpha;	
+		x[2] *= alpha;	
+		x[3] *= alpha;	
+		x[4] *= alpha;	
+		x[5] *= alpha;	
+		x[6] *= alpha;	
+		x[7] *= alpha;	
+		x+=8;
+	}
+
+}
+
+
+static void dscal_kernel_8_zero( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	for( i=0; i<n; i+=8 )
+	{
+		x[0] = 0.0;	
+		x[1] = 0.0;	
+		x[2] = 0.0;	
+		x[3] = 0.0;	
+		x[4] = 0.0;	
+		x[5] = 0.0;	
+		x[6] = 0.0;	
+		x[7] = 0.0;	
+		x+=8;
+	}
+
+}
+
+#endif
+
+
+static void dscal_kernel_inc_8(BLASLONG n, FLOAT *alpha, FLOAT *x, BLASLONG inc_x)  __attribute__ ((noinline));
+
+static void dscal_kernel_inc_8(BLASLONG n, FLOAT *alpha, FLOAT *x, BLASLONG inc_x)
+{
+
+	FLOAT *x1;
+	BLASLONG inc_x3;
+
+	inc_x <<= 3;
+	inc_x3 = (inc_x << 1) + inc_x;
+
+        __asm__  __volatile__
+        (
+        "movddup               (%3), %%xmm0                 \n\t"  // alpha     
+
+	"leaq		(%1,%4,4), %2		            \n\t"
+
+        ".align 16                                          \n\t"
+
+        "1:                                                 \n\t"
+	"movsd	(%1)     , %%xmm4			    \n\t"
+	"movhpd (%1,%4,1), %%xmm4			    \n\t"
+	"movsd	(%1,%4,2), %%xmm5			    \n\t"
+	"movhpd (%1,%5,1), %%xmm5			    \n\t"
+
+	"movsd	(%2)     , %%xmm6			    \n\t"
+	"movhpd (%2,%4,1), %%xmm6			    \n\t"
+	"movsd	(%2,%4,2), %%xmm7			    \n\t"
+	"movhpd (%2,%5,1), %%xmm7			    \n\t"
+
+	"mulpd  %%xmm0, %%xmm4				    \n\t"
+	"mulpd  %%xmm0, %%xmm5				    \n\t"
+	"mulpd  %%xmm0, %%xmm6				    \n\t"
+	"mulpd  %%xmm0, %%xmm7				    \n\t"
+
+	"movsd  %%xmm4 , (%1)				    \n\t"
+	"movhpd %%xmm4 , (%1,%4,1)			    \n\t"
+	"movsd  %%xmm5 , (%1,%4,2)			    \n\t"
+	"movhpd %%xmm5 , (%1,%5,1)			    \n\t"
+
+	"movsd  %%xmm6 , (%2)				    \n\t"
+	"movhpd %%xmm6 , (%2,%4,1)			    \n\t"
+	"movsd  %%xmm7 , (%2,%4,2)			    \n\t"
+	"movhpd %%xmm7 , (%2,%5,1)			    \n\t"
+
+	"leaq   (%1,%4,8), %1				    \n\t"
+	"leaq   (%2,%4,8), %2				    \n\t"
+
+	"subq	$8, %0					    \n\t"
+	"jnz    1b					    \n\t"
+
+        :
+        :
+          "r" (n),      // 0
+          "r" (x),      // 1
+          "r" (x1),     // 2
+          "r" (alpha),  // 3
+          "r" (inc_x),  // 4
+          "r" (inc_x3)  // 5
+        : "cc", "%0", "%1", "%2",
+          "%xmm0", "%xmm1", "%xmm2", "%xmm3",
+          "%xmm4", "%xmm5", "%xmm6", "%xmm7",
+          "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+          "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+          "memory"
+        );
+
+
+}
+
+int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT *dummy, BLASLONG dummy2)
+{
+	BLASLONG i=0,j=0;
+
+	if ( inc_x != 1 )
+	{
+
+		if ( da == 0.0 )
+		{
+
+			BLASLONG n1 = n & -2;
+
+			while(j < n1)
+			{
+			
+				x[i]=0.0;
+				x[i+inc_x]=0.0;
+				i += 2*inc_x ;
+				j+=2;
+
+			}
+
+			while(j < n)
+			{
+			
+				x[i]=0.0;
+				i += inc_x ;
+				j++;
+
+			}
+		}
+		else
+		{
+
+			BLASLONG n1 = n & -8;
+			if ( n1 > 0 )
+			{
+				dscal_kernel_inc_8(n1, &da, x, inc_x);
+				i = n1 * inc_x;
+				j = n1;
+		        }			
+
+			while(j < n)
+			{
+			
+				x[i] *= da;
+				i += inc_x ;
+				j++;
+
+			}
+
+		}
+
+		return(0);
+	}
+
+	BLASLONG n1 = n & -8;
+	if ( n1 > 0 )
+	{
+		if ( da == 0.0 )
+			dscal_kernel_8_zero(n1 , &da , x);
+		else
+			dscal_kernel_8(n1 , &da , x);
+	}
+
+	if ( da == 0.0 )
+	{
+		for ( i=n1 ; i<n; i++ )
+		{
+			x[i] = 0.0;
+		}
+	}
+	else
+	{
+
+		for ( i=n1 ; i<n; i++ )
+		{
+			x[i] *= da;
+		}
+	}
+	return(0);
+}
+
+
diff --git a/kernel/x86_64/dscal_microk_bulldozer-2.c b/kernel/x86_64/dscal_microk_bulldozer-2.c
new file mode 100644
index 000000000..de53b0bc4
--- /dev/null
+++ b/kernel/x86_64/dscal_microk_bulldozer-2.c
@@ -0,0 +1,206 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+
+static void dscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void dscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	BLASLONG n1 = n >> 4 ;
+	BLASLONG n2 = n & 8   ;
+
+	__asm__  __volatile__
+	(
+	"vmovddup		(%2), %%xmm0		    \n\t"  // alpha	
+
+	"addq	$128, %1				    \n\t"
+
+	"cmpq 	$0, %0					    \n\t"
+	"je	4f					    \n\t" 
+
+	"vmulpd 	-128(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmulpd 	-112(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmulpd 	 -96(%1), %%xmm0, %%xmm6	    \n\t"
+	"vmulpd 	 -80(%1), %%xmm0, %%xmm7	    \n\t"
+
+	"vmulpd 	 -64(%1), %%xmm0, %%xmm8	    \n\t"
+	"vmulpd 	 -48(%1), %%xmm0, %%xmm9	    \n\t"
+	"vmulpd 	 -32(%1), %%xmm0, %%xmm10    	    \n\t"
+	"vmulpd 	 -16(%1), %%xmm0, %%xmm11           \n\t"
+
+	"subq	        $1 , %0			            \n\t"		
+	"jz		2f		             	    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+	"prefetcht0     256(%1)				    \n\t" 
+
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmulpd 	   0(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmulpd 	  16(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+	"vmulpd 	  32(%1), %%xmm0, %%xmm6	    \n\t"
+
+	"prefetcht0     320(%1)				    \n\t" 
+
+	"vmovups	%%xmm8  , -64(%1)		    \n\t"
+	"vmulpd 	  48(%1), %%xmm0, %%xmm7	    \n\t"
+	"vmovups	%%xmm9  , -48(%1)		    \n\t"
+	"vmulpd 	  64(%1), %%xmm0, %%xmm8	    \n\t"
+	"vmovups	%%xmm10 , -32(%1)		    \n\t"
+	"vmulpd 	  80(%1), %%xmm0, %%xmm9	    \n\t"
+	"vmovups	%%xmm11 , -16(%1)		    \n\t"
+
+	"vmulpd 	  96(%1), %%xmm0, %%xmm10    	    \n\t"
+	"vmulpd 	 112(%1), %%xmm0, %%xmm11           \n\t"
+
+
+	"addq		$128, %1	  	 	    \n\t"
+	"subq	        $1 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+ 
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+
+	"vmovups	%%xmm8  , -64(%1)		    \n\t"
+	"vmovups	%%xmm9  , -48(%1)		    \n\t"
+	"vmovups	%%xmm10 , -32(%1)		    \n\t"
+	"vmovups	%%xmm11 , -16(%1)		    \n\t"
+
+	"addq		$128, %1	  	 	    \n\t"
+
+	"4:				            	    \n\t"
+
+	"cmpq	$8  ,%3					    \n\t"
+	"jne	5f					    \n\t"
+
+	"vmulpd	    -128(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmulpd	    -112(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmulpd	     -96(%1), %%xmm0, %%xmm6     	    \n\t"
+	"vmulpd	     -80(%1), %%xmm0, %%xmm7     	    \n\t"
+
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+
+	"5:						    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n1),  	// 0
+          "r" (x),      // 1
+          "r" (alpha),  // 2
+	  "r" (n2)   	// 3
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void dscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void dscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	BLASLONG n1 = n >> 4 ;
+	BLASLONG n2 = n & 8 ;
+
+	__asm__  __volatile__
+	(
+	"vxorpd		%%xmm0, %%xmm0 , %%xmm0		    \n\t"  
+
+	"addq	$128, %1				    \n\t"
+
+	"cmpq 	$0, %0					    \n\t"
+	"je	2f					    \n\t" 
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	"vmovups	%%xmm0  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm0  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm0  , -96(%1)		    \n\t"
+	"vmovups	%%xmm0  , -80(%1)		    \n\t"
+
+	"vmovups	%%xmm0  , -64(%1)		    \n\t"
+	"vmovups	%%xmm0  , -48(%1)		    \n\t"
+	"vmovups	%%xmm0  , -32(%1)		    \n\t"
+	"vmovups	%%xmm0  , -16(%1)		    \n\t"
+
+	"addq		$128, %1	  	 	    \n\t"
+	"subq	        $1 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"cmpq	$8  ,%3					    \n\t"
+	"jne	4f					    \n\t"
+
+	"vmovups	%%xmm0  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm0  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm0  , -96(%1)		    \n\t"
+	"vmovups	%%xmm0  , -80(%1)		    \n\t"
+
+	"4:						    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n1),  	// 0
+          "r" (x),      // 1
+          "r" (alpha),  // 2
+	  "r" (n2)   	// 3
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dscal_microk_haswell-2.c b/kernel/x86_64/dscal_microk_haswell-2.c
new file mode 100644
index 000000000..07a9c804c
--- /dev/null
+++ b/kernel/x86_64/dscal_microk_haswell-2.c
@@ -0,0 +1,206 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+
+static void dscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void dscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	BLASLONG n1 = n >> 4 ;
+	BLASLONG n2 = n & 8   ;
+
+	__asm__  __volatile__
+	(
+	"vmovddup		(%2), %%xmm0		    \n\t"  // alpha	
+
+	"addq	$128, %1				    \n\t"
+
+	"cmpq 	$0, %0					    \n\t"
+	"je	4f					    \n\t" 
+
+	"vmulpd 	-128(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmulpd 	-112(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmulpd 	 -96(%1), %%xmm0, %%xmm6	    \n\t"
+	"vmulpd 	 -80(%1), %%xmm0, %%xmm7	    \n\t"
+
+	"vmulpd 	 -64(%1), %%xmm0, %%xmm8	    \n\t"
+	"vmulpd 	 -48(%1), %%xmm0, %%xmm9	    \n\t"
+	"vmulpd 	 -32(%1), %%xmm0, %%xmm10    	    \n\t"
+	"vmulpd 	 -16(%1), %%xmm0, %%xmm11           \n\t"
+
+	"subq	        $1 , %0			            \n\t"		
+	"jz		2f		             	    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+	// "prefetcht0     640(%1)				    \n\t" 
+
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmulpd 	   0(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmulpd 	  16(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+	"vmulpd 	  32(%1), %%xmm0, %%xmm6	    \n\t"
+
+	// "prefetcht0     704(%1)				    \n\t" 
+
+	"vmovups	%%xmm8  , -64(%1)		    \n\t"
+	"vmulpd 	  48(%1), %%xmm0, %%xmm7	    \n\t"
+	"vmovups	%%xmm9  , -48(%1)		    \n\t"
+	"vmulpd 	  64(%1), %%xmm0, %%xmm8	    \n\t"
+	"vmovups	%%xmm10 , -32(%1)		    \n\t"
+	"vmulpd 	  80(%1), %%xmm0, %%xmm9	    \n\t"
+	"vmovups	%%xmm11 , -16(%1)		    \n\t"
+
+	"vmulpd 	  96(%1), %%xmm0, %%xmm10    	    \n\t"
+	"vmulpd 	 112(%1), %%xmm0, %%xmm11           \n\t"
+
+
+	"addq		$128, %1	  	 	    \n\t"
+	"subq	        $1 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+ 
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+
+	"vmovups	%%xmm8  , -64(%1)		    \n\t"
+	"vmovups	%%xmm9  , -48(%1)		    \n\t"
+	"vmovups	%%xmm10 , -32(%1)		    \n\t"
+	"vmovups	%%xmm11 , -16(%1)		    \n\t"
+
+	"addq		$128, %1	  	 	    \n\t"
+
+	"4:				            	    \n\t"
+
+	"cmpq	$8  ,%3					    \n\t"
+	"jne	5f					    \n\t"
+
+	"vmulpd	    -128(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmulpd	    -112(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmulpd	     -96(%1), %%xmm0, %%xmm6     	    \n\t"
+	"vmulpd	     -80(%1), %%xmm0, %%xmm7     	    \n\t"
+
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+
+	"5:						    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n1),  	// 0
+          "r" (x),      // 1
+          "r" (alpha),  // 2
+	  "r" (n2)   	// 3
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void dscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void dscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	BLASLONG n1 = n >> 4 ;
+	BLASLONG n2 = n & 8 ;
+
+	__asm__  __volatile__
+	(
+	"vxorpd		%%xmm0, %%xmm0 , %%xmm0		    \n\t"  
+
+	"addq	$128, %1				    \n\t"
+
+	"cmpq 	$0, %0					    \n\t"
+	"je	2f					    \n\t" 
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	"vmovups	%%xmm0  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm0  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm0  , -96(%1)		    \n\t"
+	"vmovups	%%xmm0  , -80(%1)		    \n\t"
+
+	"vmovups	%%xmm0  , -64(%1)		    \n\t"
+	"vmovups	%%xmm0  , -48(%1)		    \n\t"
+	"vmovups	%%xmm0  , -32(%1)		    \n\t"
+	"vmovups	%%xmm0  , -16(%1)		    \n\t"
+
+	"addq		$128, %1	  	 	    \n\t"
+	"subq	        $1 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"cmpq	$8  ,%3					    \n\t"
+	"jne	4f					    \n\t"
+
+	"vmovups	%%xmm0  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm0  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm0  , -96(%1)		    \n\t"
+	"vmovups	%%xmm0  , -80(%1)		    \n\t"
+
+	"4:						    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n1),  	// 0
+          "r" (x),      // 1
+          "r" (alpha),  // 2
+	  "r" (n2)   	// 3
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dscal_microk_sandy-2.c b/kernel/x86_64/dscal_microk_sandy-2.c
new file mode 100644
index 000000000..f5bf5932f
--- /dev/null
+++ b/kernel/x86_64/dscal_microk_sandy-2.c
@@ -0,0 +1,206 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+
+static void dscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void dscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	BLASLONG n1 = n >> 4 ;
+	BLASLONG n2 = n & 8   ;
+
+	__asm__  __volatile__
+	(
+	"vmovddup		(%2), %%xmm0		    \n\t"  // alpha	
+
+	"addq	$128, %1				    \n\t"
+
+	"cmpq 	$0, %0					    \n\t"
+	"je	4f					    \n\t" 
+
+	"vmulpd 	-128(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmulpd 	-112(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmulpd 	 -96(%1), %%xmm0, %%xmm6	    \n\t"
+	"vmulpd 	 -80(%1), %%xmm0, %%xmm7	    \n\t"
+
+	"vmulpd 	 -64(%1), %%xmm0, %%xmm8	    \n\t"
+	"vmulpd 	 -48(%1), %%xmm0, %%xmm9	    \n\t"
+	"vmulpd 	 -32(%1), %%xmm0, %%xmm10    	    \n\t"
+	"vmulpd 	 -16(%1), %%xmm0, %%xmm11           \n\t"
+
+	"subq	        $1 , %0			            \n\t"		
+	"jz		2f		             	    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+	"prefetcht0     640(%1)				    \n\t" 
+
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmulpd 	   0(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmulpd 	  16(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+	"vmulpd 	  32(%1), %%xmm0, %%xmm6	    \n\t"
+
+	"prefetcht0     704(%1)				    \n\t" 
+
+	"vmovups	%%xmm8  , -64(%1)		    \n\t"
+	"vmulpd 	  48(%1), %%xmm0, %%xmm7	    \n\t"
+	"vmovups	%%xmm9  , -48(%1)		    \n\t"
+	"vmulpd 	  64(%1), %%xmm0, %%xmm8	    \n\t"
+	"vmovups	%%xmm10 , -32(%1)		    \n\t"
+	"vmulpd 	  80(%1), %%xmm0, %%xmm9	    \n\t"
+	"vmovups	%%xmm11 , -16(%1)		    \n\t"
+
+	"vmulpd 	  96(%1), %%xmm0, %%xmm10    	    \n\t"
+	"vmulpd 	 112(%1), %%xmm0, %%xmm11           \n\t"
+
+
+	"addq		$128, %1	  	 	    \n\t"
+	"subq	        $1 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+ 
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+
+	"vmovups	%%xmm8  , -64(%1)		    \n\t"
+	"vmovups	%%xmm9  , -48(%1)		    \n\t"
+	"vmovups	%%xmm10 , -32(%1)		    \n\t"
+	"vmovups	%%xmm11 , -16(%1)		    \n\t"
+
+	"addq		$128, %1	  	 	    \n\t"
+
+	"4:				            	    \n\t"
+
+	"cmpq	$8  ,%3					    \n\t"
+	"jne	5f					    \n\t"
+
+	"vmulpd	    -128(%1), %%xmm0, %%xmm4	    \n\t"
+	"vmulpd	    -112(%1), %%xmm0, %%xmm5	    \n\t"
+	"vmulpd	     -96(%1), %%xmm0, %%xmm6     	    \n\t"
+	"vmulpd	     -80(%1), %%xmm0, %%xmm7     	    \n\t"
+
+	"vmovups	%%xmm4  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm5  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm6  , -96(%1)		    \n\t"
+	"vmovups	%%xmm7  , -80(%1)		    \n\t"
+
+	"5:						    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n1),  	// 0
+          "r" (x),      // 1
+          "r" (alpha),  // 2
+	  "r" (n2)   	// 3
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void dscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void dscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	BLASLONG n1 = n >> 4 ;
+	BLASLONG n2 = n & 8 ;
+
+	__asm__  __volatile__
+	(
+	"vxorpd		%%xmm0, %%xmm0 , %%xmm0		    \n\t"  
+
+	"addq	$128, %1				    \n\t"
+
+	"cmpq 	$0, %0					    \n\t"
+	"je	2f					    \n\t" 
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	"vmovups	%%xmm0  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm0  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm0  , -96(%1)		    \n\t"
+	"vmovups	%%xmm0  , -80(%1)		    \n\t"
+
+	"vmovups	%%xmm0  , -64(%1)		    \n\t"
+	"vmovups	%%xmm0  , -48(%1)		    \n\t"
+	"vmovups	%%xmm0  , -32(%1)		    \n\t"
+	"vmovups	%%xmm0  , -16(%1)		    \n\t"
+
+	"addq		$128, %1	  	 	    \n\t"
+	"subq	        $1 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"cmpq	$8  ,%3					    \n\t"
+	"jne	4f					    \n\t"
+
+	"vmovups	%%xmm0  ,-128(%1)		    \n\t"
+	"vmovups	%%xmm0  ,-112(%1)		    \n\t"
+	"vmovups	%%xmm0  , -96(%1)		    \n\t"
+	"vmovups	%%xmm0  , -80(%1)		    \n\t"
+
+	"4:						    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n1),  	// 0
+          "r" (x),      // 1
+          "r" (alpha),  // 2
+	  "r" (n2)   	// 3
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dsymv_L.c b/kernel/x86_64/dsymv_L.c
index f6157f791..3f5e77e5f 100644
--- a/kernel/x86_64/dsymv_L.c
+++ b/kernel/x86_64/dsymv_L.c
@@ -30,6 +30,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)
 #include "dsymv_L_microk_bulldozer-2.c"
+#elif defined(HASWELL)
+#include "dsymv_L_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "dsymv_L_microk_sandy-2.c"
 #elif defined(NEHALEM)
 #include "dsymv_L_microk_nehalem-2.c"
 #endif
diff --git a/kernel/x86_64/dsymv_L_microk_haswell-2.c b/kernel/x86_64/dsymv_L_microk_haswell-2.c
new file mode 100644
index 000000000..bc5ec6b87
--- /dev/null
+++ b/kernel/x86_64/dsymv_L_microk_haswell-2.c
@@ -0,0 +1,129 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4x4 1
+static void dsymv_kernel_4x4( BLASLONG from, BLASLONG to, FLOAT **a, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2) __attribute__ ((noinline));
+
+static void dsymv_kernel_4x4(BLASLONG from, BLASLONG to, FLOAT **a, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorpd		%%ymm0 , %%ymm0 , %%ymm0     \n\t"	// temp2[0]
+	"vxorpd		%%ymm1 , %%ymm1 , %%ymm1     \n\t"	// temp2[1]
+	"vxorpd		%%ymm2 , %%ymm2 , %%ymm2     \n\t"	// temp2[2]
+	"vxorpd		%%ymm3 , %%ymm3 , %%ymm3     \n\t"	// temp2[3]
+	"vbroadcastsd   (%8),    %%ymm4	             \n\t"	// temp1[0]
+	"vbroadcastsd  8(%8),    %%ymm5	             \n\t"	// temp1[1]
+	"vbroadcastsd 16(%8),    %%ymm6	             \n\t"	// temp1[1]
+	"vbroadcastsd 24(%8),    %%ymm7	             \n\t"	// temp1[1]
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,8), %%ymm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,8), %%ymm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,8), %%ymm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,8), %%ymm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,8), %%ymm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,8), %%ymm15	           \n\t"  // 2 * a
+
+	"vfmadd231pd	%%ymm4, %%ymm12 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231pd	%%ymm8, %%ymm12 , %%ymm0  \n\t"  // temp2 += x * a
+
+	"vfmadd231pd	%%ymm5, %%ymm13 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231pd	%%ymm8, %%ymm13 , %%ymm1  \n\t"  // temp2 += x * a
+
+	"vfmadd231pd	%%ymm6, %%ymm14 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231pd	%%ymm8, %%ymm14 , %%ymm2  \n\t"  // temp2 += x * a
+
+	"vfmadd231pd	%%ymm7, %%ymm15 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231pd	%%ymm8, %%ymm15 , %%ymm3  \n\t"  // temp2 += x * a
+	"addq		$4 , %0	  	 	      \n\t"
+
+	"vmovups	%%ymm9 ,  -32(%3,%0,8)		   \n\t"
+
+	"cmpq		%0 , %1			      \n\t"
+	"jnz		1b		      \n\t"
+
+	"vmovsd		  (%9), %%xmm4		      \n\t"
+	"vmovsd		 8(%9), %%xmm5		      \n\t"
+	"vmovsd		16(%9), %%xmm6		      \n\t"
+	"vmovsd		24(%9), %%xmm7		      \n\t"
+
+	"vextractf128 $0x01, %%ymm0 , %%xmm12	      \n\t"
+	"vextractf128 $0x01, %%ymm1 , %%xmm13	      \n\t"
+	"vextractf128 $0x01, %%ymm2 , %%xmm14	      \n\t"
+	"vextractf128 $0x01, %%ymm3 , %%xmm15	      \n\t"
+
+	"vaddpd	        %%xmm0, %%xmm12, %%xmm0	      \n\t"
+	"vaddpd	        %%xmm1, %%xmm13, %%xmm1	      \n\t"
+	"vaddpd	        %%xmm2, %%xmm14, %%xmm2	      \n\t"
+	"vaddpd	        %%xmm3, %%xmm15, %%xmm3	      \n\t"
+
+	"vhaddpd        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddpd        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddpd        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddpd        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddsd		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddsd		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddsd		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddsd		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovsd         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm1 , 8(%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm2 ,16(%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm3 ,24(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (from),	// 0	
+	  "r" (to),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a[0]),	// 4
+          "r" (a[1]),	// 5
+          "r" (a[2]),	// 6
+          "r" (a[3]),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dsymv_L_microk_sandy-2.c b/kernel/x86_64/dsymv_L_microk_sandy-2.c
new file mode 100644
index 000000000..c87084915
--- /dev/null
+++ b/kernel/x86_64/dsymv_L_microk_sandy-2.c
@@ -0,0 +1,138 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4x4 1
+static void dsymv_kernel_4x4( BLASLONG from, BLASLONG to, FLOAT **a, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2) __attribute__ ((noinline));
+
+static void dsymv_kernel_4x4(BLASLONG from, BLASLONG to, FLOAT **a, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorpd		%%ymm0 , %%ymm0 , %%ymm0     \n\t"	// temp2[0]
+	"vxorpd		%%ymm1 , %%ymm1 , %%ymm1     \n\t"	// temp2[1]
+	"vxorpd		%%ymm2 , %%ymm2 , %%ymm2     \n\t"	// temp2[2]
+	"vxorpd		%%ymm3 , %%ymm3 , %%ymm3     \n\t"	// temp2[3]
+	"vbroadcastsd   (%8),    %%ymm4	             \n\t"	// temp1[0]
+	"vbroadcastsd  8(%8),    %%ymm5	             \n\t"	// temp1[1]
+	"vbroadcastsd 16(%8),    %%ymm6	             \n\t"	// temp1[1]
+	"vbroadcastsd 24(%8),    %%ymm7	             \n\t"	// temp1[1]
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,8), %%ymm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,8), %%ymm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,8), %%ymm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,8), %%ymm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,8), %%ymm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,8), %%ymm15	           \n\t"  // 2 * a
+
+	"vmulpd		%%ymm4, %%ymm12, %%ymm10	   \n\t"
+	"vaddpd		%%ymm9, %%ymm10, %%ymm9            \n\t"
+	"vmulpd		%%ymm8, %%ymm12, %%ymm11	   \n\t"
+	"vaddpd		%%ymm0, %%ymm11, %%ymm0            \n\t"
+
+	"vmulpd		%%ymm5, %%ymm13, %%ymm10	   \n\t"
+	"vaddpd		%%ymm9, %%ymm10, %%ymm9            \n\t"
+	"vmulpd		%%ymm8, %%ymm13, %%ymm11	   \n\t"
+	"vaddpd		%%ymm1, %%ymm11, %%ymm1            \n\t"
+
+	"vmulpd		%%ymm6, %%ymm14, %%ymm10	   \n\t"
+	"vaddpd		%%ymm9, %%ymm10, %%ymm9            \n\t"
+	"vmulpd		%%ymm8, %%ymm14, %%ymm11	   \n\t"
+	"vaddpd		%%ymm2, %%ymm11, %%ymm2            \n\t"
+
+	"vmulpd		%%ymm7, %%ymm15, %%ymm10	   \n\t"
+	"vaddpd		%%ymm9, %%ymm10, %%ymm9            \n\t"
+	"vmulpd		%%ymm8, %%ymm15, %%ymm11	   \n\t"
+	"vaddpd		%%ymm3, %%ymm11, %%ymm3            \n\t"
+
+	"addq		$4 , %0	  	 	      \n\t"
+
+	"vmovups	%%ymm9 ,  -32(%3,%0,8)		   \n\t"
+
+	"cmpq		%0 , %1			      \n\t"
+	"jnz		1b		      \n\t"
+
+	"vmovsd		  (%9), %%xmm4		      \n\t"
+	"vmovsd		 8(%9), %%xmm5		      \n\t"
+	"vmovsd		16(%9), %%xmm6		      \n\t"
+	"vmovsd		24(%9), %%xmm7		      \n\t"
+
+	"vextractf128 $0x01, %%ymm0 , %%xmm12	      \n\t"
+	"vextractf128 $0x01, %%ymm1 , %%xmm13	      \n\t"
+	"vextractf128 $0x01, %%ymm2 , %%xmm14	      \n\t"
+	"vextractf128 $0x01, %%ymm3 , %%xmm15	      \n\t"
+
+	"vaddpd	        %%xmm0, %%xmm12, %%xmm0	      \n\t"
+	"vaddpd	        %%xmm1, %%xmm13, %%xmm1	      \n\t"
+	"vaddpd	        %%xmm2, %%xmm14, %%xmm2	      \n\t"
+	"vaddpd	        %%xmm3, %%xmm15, %%xmm3	      \n\t"
+
+	"vhaddpd        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddpd        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddpd        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddpd        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddsd		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddsd		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddsd		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddsd		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovsd         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm1 , 8(%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm2 ,16(%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm3 ,24(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (from),	// 0	
+	  "r" (to),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a[0]),	// 4
+          "r" (a[1]),	// 5
+          "r" (a[2]),	// 6
+          "r" (a[3]),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dsymv_U.c b/kernel/x86_64/dsymv_U.c
index ecfaf5043..9f5ae3015 100644
--- a/kernel/x86_64/dsymv_U.c
+++ b/kernel/x86_64/dsymv_U.c
@@ -31,6 +31,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)
 #include "dsymv_U_microk_bulldozer-2.c"
+#elif defined(HASWELL)
+#include "dsymv_U_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "dsymv_U_microk_sandy-2.c"
 #elif defined(NEHALEM)
 #include "dsymv_U_microk_nehalem-2.c"
 #endif
diff --git a/kernel/x86_64/dsymv_U_microk_haswell-2.c b/kernel/x86_64/dsymv_U_microk_haswell-2.c
new file mode 100644
index 000000000..6ce384f93
--- /dev/null
+++ b/kernel/x86_64/dsymv_U_microk_haswell-2.c
@@ -0,0 +1,131 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4x4 1
+static void dsymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT *a3, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2) __attribute__ ((noinline));
+
+static void dsymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT *a3, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2)
+{
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorpd		%%ymm0 , %%ymm0 , %%ymm0     \n\t"	// temp2[0]
+	"vxorpd		%%ymm1 , %%ymm1 , %%ymm1     \n\t"	// temp2[1]
+	"vxorpd		%%ymm2 , %%ymm2 , %%ymm2     \n\t"	// temp2[2]
+	"vxorpd		%%ymm3 , %%ymm3 , %%ymm3     \n\t"	// temp2[3]
+	"vbroadcastsd   (%8),    %%ymm4	             \n\t"	// temp1[0]
+	"vbroadcastsd  8(%8),    %%ymm5	             \n\t"	// temp1[1]
+	"vbroadcastsd 16(%8),    %%ymm6	             \n\t"	// temp1[1]
+	"vbroadcastsd 24(%8),    %%ymm7	             \n\t"	// temp1[1]
+	"xorq           %0,%0                        \n\t"
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,8), %%ymm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,8), %%ymm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,8), %%ymm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,8), %%ymm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,8), %%ymm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,8), %%ymm15	           \n\t"  // 2 * a
+
+	"vfmadd231pd	%%ymm4, %%ymm12 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231pd	%%ymm8, %%ymm12 , %%ymm0  \n\t"  // temp2 += x * a
+
+	"vfmadd231pd	%%ymm5, %%ymm13 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231pd	%%ymm8, %%ymm13 , %%ymm1  \n\t"  // temp2 += x * a
+
+	"vfmadd231pd	%%ymm6, %%ymm14 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231pd	%%ymm8, %%ymm14 , %%ymm2  \n\t"  // temp2 += x * a
+
+	"vfmadd231pd	%%ymm7, %%ymm15 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231pd	%%ymm8, %%ymm15 , %%ymm3  \n\t"  // temp2 += x * a
+	"addq		$4 , %0	  	 	      \n\t"
+	"subq		$4 , %1	  	 	      \n\t"
+
+	"vmovups	%%ymm9 ,  -32(%3,%0,8)		   \n\t"
+
+	"jnz		1b		      \n\t"
+
+	"vmovsd		  (%9), %%xmm4		      \n\t"
+	"vmovsd		 8(%9), %%xmm5		      \n\t"
+	"vmovsd		16(%9), %%xmm6		      \n\t"
+	"vmovsd		24(%9), %%xmm7		      \n\t"
+
+	"vextractf128 $0x01, %%ymm0 , %%xmm12	      \n\t"
+	"vextractf128 $0x01, %%ymm1 , %%xmm13	      \n\t"
+	"vextractf128 $0x01, %%ymm2 , %%xmm14	      \n\t"
+	"vextractf128 $0x01, %%ymm3 , %%xmm15	      \n\t"
+
+	"vaddpd	        %%xmm0, %%xmm12, %%xmm0	      \n\t"
+	"vaddpd	        %%xmm1, %%xmm13, %%xmm1	      \n\t"
+	"vaddpd	        %%xmm2, %%xmm14, %%xmm2	      \n\t"
+	"vaddpd	        %%xmm3, %%xmm15, %%xmm3	      \n\t"
+
+	"vhaddpd        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddpd        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddpd        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddpd        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddsd		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddsd		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddsd		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddsd		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovsd         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm1 , 8(%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm2 ,16(%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm3 ,24(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a0),	// 4
+          "r" (a1),	// 5
+          "r" (a2),	// 6
+          "r" (a3),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dsymv_U_microk_sandy-2.c b/kernel/x86_64/dsymv_U_microk_sandy-2.c
new file mode 100644
index 000000000..212d4cf7b
--- /dev/null
+++ b/kernel/x86_64/dsymv_U_microk_sandy-2.c
@@ -0,0 +1,140 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4x4 1
+static void dsymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT *a3, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2) __attribute__ ((noinline));
+
+static void dsymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT *a3, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2)
+{
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorpd		%%ymm0 , %%ymm0 , %%ymm0     \n\t"	// temp2[0]
+	"vxorpd		%%ymm1 , %%ymm1 , %%ymm1     \n\t"	// temp2[1]
+	"vxorpd		%%ymm2 , %%ymm2 , %%ymm2     \n\t"	// temp2[2]
+	"vxorpd		%%ymm3 , %%ymm3 , %%ymm3     \n\t"	// temp2[3]
+	"vbroadcastsd   (%8),    %%ymm4	             \n\t"	// temp1[0]
+	"vbroadcastsd  8(%8),    %%ymm5	             \n\t"	// temp1[1]
+	"vbroadcastsd 16(%8),    %%ymm6	             \n\t"	// temp1[1]
+	"vbroadcastsd 24(%8),    %%ymm7	             \n\t"	// temp1[1]
+	"xorq           %0,%0                        \n\t"
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,8), %%ymm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,8), %%ymm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,8), %%ymm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,8), %%ymm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,8), %%ymm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,8), %%ymm15	           \n\t"  // 2 * a
+
+        "vmulpd         %%ymm4, %%ymm12, %%ymm10           \n\t"
+        "vaddpd         %%ymm9, %%ymm10, %%ymm9            \n\t"
+        "vmulpd         %%ymm8, %%ymm12, %%ymm11           \n\t"
+        "vaddpd         %%ymm0, %%ymm11, %%ymm0            \n\t"
+
+        "vmulpd         %%ymm5, %%ymm13, %%ymm10           \n\t"
+        "vaddpd         %%ymm9, %%ymm10, %%ymm9            \n\t"
+        "vmulpd         %%ymm8, %%ymm13, %%ymm11           \n\t"
+        "vaddpd         %%ymm1, %%ymm11, %%ymm1            \n\t"
+
+        "vmulpd         %%ymm6, %%ymm14, %%ymm10           \n\t"
+        "vaddpd         %%ymm9, %%ymm10, %%ymm9            \n\t"
+        "vmulpd         %%ymm8, %%ymm14, %%ymm11           \n\t"
+        "vaddpd         %%ymm2, %%ymm11, %%ymm2            \n\t"
+
+        "vmulpd         %%ymm7, %%ymm15, %%ymm10           \n\t"
+        "vaddpd         %%ymm9, %%ymm10, %%ymm9            \n\t"
+        "vmulpd         %%ymm8, %%ymm15, %%ymm11           \n\t"
+        "vaddpd         %%ymm3, %%ymm11, %%ymm3            \n\t"
+
+	"addq		$4 , %0	  	 	      \n\t"
+	"subq		$4 , %1	  	 	      \n\t"
+
+	"vmovups	%%ymm9 ,  -32(%3,%0,8)		   \n\t"
+
+	"jnz		1b		      \n\t"
+
+	"vmovsd		  (%9), %%xmm4		      \n\t"
+	"vmovsd		 8(%9), %%xmm5		      \n\t"
+	"vmovsd		16(%9), %%xmm6		      \n\t"
+	"vmovsd		24(%9), %%xmm7		      \n\t"
+
+	"vextractf128 $0x01, %%ymm0 , %%xmm12	      \n\t"
+	"vextractf128 $0x01, %%ymm1 , %%xmm13	      \n\t"
+	"vextractf128 $0x01, %%ymm2 , %%xmm14	      \n\t"
+	"vextractf128 $0x01, %%ymm3 , %%xmm15	      \n\t"
+
+	"vaddpd	        %%xmm0, %%xmm12, %%xmm0	      \n\t"
+	"vaddpd	        %%xmm1, %%xmm13, %%xmm1	      \n\t"
+	"vaddpd	        %%xmm2, %%xmm14, %%xmm2	      \n\t"
+	"vaddpd	        %%xmm3, %%xmm15, %%xmm3	      \n\t"
+
+	"vhaddpd        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddpd        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddpd        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddpd        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddsd		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddsd		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddsd		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddsd		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovsd         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm1 , 8(%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm2 ,16(%9)		\n\t"	// save temp2
+	"vmovsd         %%xmm3 ,24(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a0),	// 4
+          "r" (a1),	// 5
+          "r" (a2),	// 6
+          "r" (a3),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/dtrmm_kernel_4x8_haswell.c b/kernel/x86_64/dtrmm_kernel_4x8_haswell.c
new file mode 100644
index 000000000..ac8c97d03
--- /dev/null
+++ b/kernel/x86_64/dtrmm_kernel_4x8_haswell.c
@@ -0,0 +1,1546 @@
+#include "common.h"
+#include <stdbool.h>
+
+
+static void dtrmm_kernel_4x8( BLASLONG n, FLOAT *alpha ,FLOAT *a, FLOAT *b, FLOAT *C0, FLOAT *C1, FLOAT *C2,FLOAT *C3, FLOAT *C4, FLOAT *C5,FLOAT *C6, FLOAT *C7) __attribute__ ((noinline));
+
+static void dtrmm_kernel_4x8( BLASLONG n, FLOAT *alpha ,FLOAT *a, FLOAT *b, FLOAT *C0, FLOAT *C1, FLOAT *C2,FLOAT *C3, FLOAT *C4, FLOAT *C5,FLOAT *C6, FLOAT *C7)
+{
+
+		BLASLONG i = 0;
+		BLASLONG temp1 = n * 8;
+
+		 __asm__  __volatile__
+        	(
+		"	vxorpd	%%ymm4 , %%ymm4 , %%ymm4			\n\t"
+		"	vxorpd	%%ymm5 , %%ymm5 , %%ymm5			\n\t"
+		"	vxorpd	%%ymm6 , %%ymm6 , %%ymm6			\n\t"
+		"	vxorpd	%%ymm7 , %%ymm7 , %%ymm7			\n\t"
+		"	vxorpd	%%ymm8 , %%ymm8 , %%ymm8			\n\t"
+		"	vxorpd	%%ymm9 , %%ymm9 , %%ymm9			\n\t"
+		"	vxorpd	%%ymm10, %%ymm10, %%ymm10			\n\t"
+		"	vxorpd	%%ymm11, %%ymm11, %%ymm11			\n\t"
+
+		"	cmp $0, %1						\n\t"
+		"	jz 2f							\n\t"
+
+		"	.align 16						\n\t"
+		"1:								\n\t"
+		"	vmovups   	(%2,%0,4) , %%ymm0			\n\t"
+		"	vmovups   	(%3,%0,8) , %%ymm1			\n\t"
+		"	vmovups       32(%3,%0,8) , %%ymm2			\n\t"
+
+		"	vfmadd231pd	%%ymm0 , %%ymm1 , %%ymm4		\n\t"
+		"	vfmadd231pd	%%ymm0 , %%ymm2 , %%ymm8		\n\t"
+
+		"	vpermpd         $0xb1  , %%ymm0 , %%ymm0		\n\t"
+		"	vfmadd231pd	%%ymm0 , %%ymm1 , %%ymm5		\n\t"
+		"	vfmadd231pd	%%ymm0 , %%ymm2 , %%ymm9		\n\t"
+
+		"	vpermpd         $0x1b  , %%ymm0 , %%ymm0		\n\t"
+		"	vfmadd231pd	%%ymm0 , %%ymm1 , %%ymm6		\n\t"
+		"	vfmadd231pd	%%ymm0 , %%ymm2 , %%ymm10		\n\t"
+
+		"	vpermpd         $0xb1  , %%ymm0 , %%ymm0		\n\t"
+		"	vfmadd231pd	%%ymm0 , %%ymm1 , %%ymm7		\n\t"
+		"	vfmadd231pd	%%ymm0 , %%ymm2 , %%ymm11		\n\t"
+
+		"	addq	$8 , %0						\n\t"
+		"	cmp	%0 , %1						\n\t"
+		"	jne	1b						\n\t"
+
+		"2:								\n\t"
+
+		"	vbroadcastsd	(%4), %%ymm0				\n\t"
+
+		"	vmulpd		%%ymm0 , %%ymm4 , %%ymm4		\n\t"
+		"	vmulpd		%%ymm0 , %%ymm5 , %%ymm5		\n\t"
+		"	vmulpd		%%ymm0 , %%ymm6 , %%ymm6		\n\t"
+		"	vmulpd		%%ymm0 , %%ymm7 , %%ymm7		\n\t"
+		"	vmulpd		%%ymm0 , %%ymm8 , %%ymm8		\n\t"
+		"	vmulpd		%%ymm0 , %%ymm9 , %%ymm9		\n\t"
+		"	vmulpd		%%ymm0 , %%ymm10, %%ymm10		\n\t"
+		"	vmulpd		%%ymm0 , %%ymm11, %%ymm11		\n\t"
+
+		"	vpermpd 	$0xb1  , %%ymm5 , %%ymm5		\n\t"
+		"	vpermpd 	$0xb1  , %%ymm7 , %%ymm7		\n\t"
+
+		"	vblendpd 	$0x0a  , %%ymm5 , %%ymm4 , %%ymm0	\n\t"
+		"	vblendpd 	$0x05  , %%ymm5 , %%ymm4 , %%ymm1	\n\t"
+		"	vblendpd 	$0x0a  , %%ymm7 , %%ymm6 , %%ymm2	\n\t"
+		"	vblendpd 	$0x05  , %%ymm7 , %%ymm6 , %%ymm3	\n\t"
+
+		"	vpermpd 	$0x1b  , %%ymm2 , %%ymm2		\n\t"
+		"	vpermpd 	$0x1b  , %%ymm3 , %%ymm3		\n\t"
+		"	vpermpd 	$0xb1  , %%ymm2 , %%ymm2		\n\t"
+		"	vpermpd 	$0xb1  , %%ymm3 , %%ymm3		\n\t"
+
+		"	vblendpd 	$0x03  , %%ymm0 , %%ymm2 , %%ymm4	\n\t"
+		"	vblendpd 	$0x03  , %%ymm1 , %%ymm3 , %%ymm5	\n\t"
+		"	vblendpd 	$0x03  , %%ymm2 , %%ymm0 , %%ymm6	\n\t"
+		"	vblendpd 	$0x03  , %%ymm3 , %%ymm1 , %%ymm7	\n\t"
+
+		"	vmovups		%%ymm4 , (%5)				\n\t"
+		"	vmovups		%%ymm5 , (%6)				\n\t"
+		"	vmovups		%%ymm6 , (%7)				\n\t"
+		"	vmovups		%%ymm7 , (%8)				\n\t"
+
+		"	vpermpd 	$0xb1  , %%ymm9 , %%ymm9		\n\t"
+		"	vpermpd 	$0xb1  , %%ymm11, %%ymm11		\n\t"
+
+		"	vblendpd 	$0x0a  , %%ymm9 , %%ymm8 , %%ymm0	\n\t"
+		"	vblendpd 	$0x05  , %%ymm9 , %%ymm8 , %%ymm1	\n\t"
+		"	vblendpd 	$0x0a  , %%ymm11, %%ymm10, %%ymm2	\n\t"
+		"	vblendpd 	$0x05  , %%ymm11, %%ymm10, %%ymm3	\n\t"
+
+		"	vpermpd 	$0x1b  , %%ymm2 , %%ymm2		\n\t"
+		"	vpermpd 	$0x1b  , %%ymm3 , %%ymm3		\n\t"
+		"	vpermpd 	$0xb1  , %%ymm2 , %%ymm2		\n\t"
+		"	vpermpd 	$0xb1  , %%ymm3 , %%ymm3		\n\t"
+
+		"	vblendpd 	$0x03  , %%ymm0 , %%ymm2 , %%ymm4	\n\t"
+		"	vblendpd 	$0x03  , %%ymm1 , %%ymm3 , %%ymm5	\n\t"
+		"	vblendpd 	$0x03  , %%ymm2 , %%ymm0 , %%ymm6	\n\t"
+		"	vblendpd 	$0x03  , %%ymm3 , %%ymm1 , %%ymm7	\n\t"
+
+		"	vmovups		%%ymm4 , (%9)				\n\t"
+		"	vmovups		%%ymm5 , (%10)				\n\t"
+		"	vmovups		%%ymm6 , (%11)				\n\t"
+		"	vmovups		%%ymm7 , (%12)				\n\t"
+
+	        :
+        	:
+		"a" (i),	 // 0
+          	"r" (temp1),     // 1    
+          	"S" (a),         // 2
+          	"D" (b),         // 3
+          	"r" (alpha),     // 4
+		"r" (C0),	 // 5
+		"r" (C1),	 // 6
+		"r" (C2),	 // 7
+		"r" (C3),	 // 8
+		"r" (C4),	 // 9
+		"r" (C5),	 // 10
+		"r" (C6),	 // 11
+		"r" (C7) 	 // 12
+		: "cc",
+          	"%xmm0", "%xmm1", "%xmm2", "%xmm3",
+          	"%xmm4", "%xmm5", "%xmm6", "%xmm7",
+          	"%xmm8", "%xmm9", "%xmm10", "%xmm11",
+          	"%xmm12", "%xmm13", "%xmm14", "%xmm15",
+          	"memory"
+        	);
+
+
+
+
+}
+
+
+
+
+int CNAME(BLASLONG bm,BLASLONG bn,BLASLONG bk,FLOAT alpha,FLOAT* ba,FLOAT* bb,FLOAT* C,BLASLONG ldc ,BLASLONG offset)
+{
+
+   BLASLONG i,j,k;
+   FLOAT *C0,*C1,*C2,*C3,*C4,*C5,*C6,*C7,*ptrba,*ptrbb;
+
+   FLOAT res0_0;
+   FLOAT res0_1;
+   FLOAT res0_2;
+   FLOAT res0_3;
+
+   FLOAT res1_0;
+   FLOAT res1_1;
+   FLOAT res1_2;
+   FLOAT res1_3;
+
+   FLOAT res2_0;
+   FLOAT res2_1;
+   FLOAT res2_2;
+   FLOAT res2_3;
+
+   FLOAT res3_0;
+   FLOAT res3_1;
+   FLOAT res3_2;
+   FLOAT res3_3;
+
+   FLOAT res4_0;
+   FLOAT res4_1;
+   FLOAT res4_2;
+   FLOAT res4_3;
+
+   FLOAT res5_0;
+   FLOAT res5_1;
+   FLOAT res5_2;
+   FLOAT res5_3;
+
+   FLOAT res6_0;
+   FLOAT res6_1;
+   FLOAT res6_2;
+   FLOAT res6_3;
+
+   FLOAT res7_0;
+   FLOAT res7_1;
+   FLOAT res7_2;
+   FLOAT res7_3;
+
+   FLOAT a0;
+   FLOAT a1;
+
+   FLOAT b0;
+   FLOAT b1;
+   FLOAT b2;
+   FLOAT b3;
+   FLOAT b4;
+   FLOAT b5;
+   FLOAT b6;
+   FLOAT b7;
+
+   BLASLONG off, temp ;
+
+   bool left;
+   bool transposed;
+   bool backwards;
+
+#ifdef LEFT
+   left = true;
+#else
+   left = false;
+#endif
+
+#ifdef TRANSA
+   transposed = true;
+#else
+   transposed = false;
+#endif
+
+   backwards = left != transposed;
+
+   if (!left) {
+      off = -offset;
+   }
+
+
+   for (j=0; j<bn/8; j+=1) // do blocks of the Mx8 loops 
+   {
+        C0 = C;
+        C1 = C0+ldc;
+        C2 = C1+ldc;
+        C3 = C2+ldc;
+        C4 = C3+ldc;
+        C5 = C4+ldc;
+        C6 = C5+ldc;
+        C7 = C6+ldc;
+
+
+        if (left) {
+            off = offset;
+        }
+
+        ptrba = ba;
+
+        for (i=0; i<bm/4; i+=1) // do blocks of 4x4
+	{
+
+		ptrbb = bb;
+                if (backwards)
+                {
+		   ptrba += off*4; // number of values in A
+		   ptrbb += off*8; // number of values in B
+                }
+/*
+		res0_0 = 0;
+		res0_1 = 0;
+		res0_2 = 0;
+		res0_3 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+		res1_2 = 0;
+		res1_3 = 0;
+
+		res2_0 = 0;
+		res2_1 = 0;
+		res2_2 = 0;
+		res2_3 = 0;
+
+		res3_0 = 0;
+		res3_1 = 0;
+		res3_2 = 0;
+		res3_3 = 0;
+
+		res4_0 = 0;
+		res4_1 = 0;
+		res4_2 = 0;
+		res4_3 = 0;
+
+		res5_0 = 0;
+		res5_1 = 0;
+		res5_2 = 0;
+		res5_3 = 0;
+
+		res6_0 = 0;
+		res6_1 = 0;
+		res6_2 = 0;
+		res6_3 = 0;
+
+		res7_0 = 0;
+		res7_1 = 0;
+		res7_2 = 0;
+		res7_3 = 0;
+*/
+                temp = backwards ? bk-off :
+                             left ? off + 4 : // number of values in A
+                                    off + 8;  // number of values in B
+
+	        dtrmm_kernel_4x8( temp, &alpha , ptrba, ptrbb, C0, C1, C2, C3, C4, C5, C6, C7);
+
+		ptrba = ptrba + temp * 4;
+		ptrbb = ptrbb + temp * 8;
+
+/*
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+			b4 = ptrbb[4];
+			b5 = ptrbb[5];
+			b6 = ptrbb[6];
+			b7 = ptrbb[7];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+			res4_0 += a0*b4;
+			res5_0 += a0*b5;
+			res6_0 += a0*b6;
+			res7_0 += a0*b7;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+			res2_1 += a1*b2;
+			res3_1 += a1*b3;
+			res4_1 += a1*b4;
+			res5_1 += a1*b5;
+			res6_1 += a1*b6;
+			res7_1 += a1*b7;
+
+			a0 = ptrba[2];
+			res0_2 += a0*b0;
+			res1_2 += a0*b1;
+			res2_2 += a0*b2;
+			res3_2 += a0*b3;
+			res4_2 += a0*b4;
+			res5_2 += a0*b5;
+			res6_2 += a0*b6;
+			res7_2 += a0*b7;
+
+			a1 = ptrba[3];
+			res0_3 += a1*b0;
+			res1_3 += a1*b1;
+			res2_3 += a1*b2;
+			res3_3 += a1*b3;
+			res4_3 += a1*b4;
+			res5_3 += a1*b5;
+			res6_3 += a1*b6;
+			res7_3 += a1*b7;
+
+			ptrba = ptrba+4;
+			ptrbb = ptrbb+8;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+		res0_2 *= alpha;
+		res0_3 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+		res1_2 *= alpha;
+		res1_3 *= alpha;
+
+		res2_0 *= alpha;
+		res2_1 *= alpha;
+		res2_2 *= alpha;
+		res2_3 *= alpha;
+
+		res3_0 *= alpha;
+		res3_1 *= alpha;
+		res3_2 *= alpha;
+		res3_3 *= alpha;
+
+		res4_0 *= alpha;
+		res4_1 *= alpha;
+		res4_2 *= alpha;
+		res4_3 *= alpha;
+
+		res5_0 *= alpha;
+		res5_1 *= alpha;
+		res5_2 *= alpha;
+		res5_3 *= alpha;
+
+		res6_0 *= alpha;
+		res6_1 *= alpha;
+		res6_2 *= alpha;
+		res6_3 *= alpha;
+
+		res7_0 *= alpha;
+		res7_1 *= alpha;
+		res7_2 *= alpha;
+		res7_3 *= alpha;
+
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+		C0[2] = res0_2;
+		C0[3] = res0_3;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+		C1[2] = res1_2;
+		C1[3] = res1_3;
+
+		C2[0] = res2_0;
+		C2[1] = res2_1;
+		C2[2] = res2_2;
+		C2[3] = res2_3;
+
+		C3[0] = res3_0;
+		C3[1] = res3_1;
+		C3[2] = res3_2;
+		C3[3] = res3_3;
+
+		C4[0] = res4_0;
+		C4[1] = res4_1;
+		C4[2] = res4_2;
+		C4[3] = res4_3;
+
+		C5[0] = res5_0;
+		C5[1] = res5_1;
+		C5[2] = res5_2;
+		C5[3] = res5_3;
+
+		C6[0] = res6_0;
+		C6[1] = res6_1;
+		C6[2] = res6_2;
+		C6[3] = res6_3;
+
+		C7[0] = res7_0;
+		C7[1] = res7_1;
+		C7[2] = res7_2;
+		C7[3] = res7_3;
+*/
+		if (!backwards) {
+                    temp = bk-off;
+                    temp = left ? temp - 4 : // number of values in A
+                                  temp - 8;  // number of values in B
+
+                    ptrba += temp*4; // number of values in A
+		    ptrbb += temp*8; // number of values in B
+                }
+#ifdef LEFT
+		off += 4; // number of values in A
+#endif
+
+		C0 = C0+4;
+		C1 = C1+4;
+		C2 = C2+4;
+		C3 = C3+4;
+		C4 = C4+4;
+		C5 = C5+4;
+		C6 = C6+4;
+		C7 = C7+4;
+
+	}
+
+	if ( bm & 2 ) // do any 2x4 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*2;
+		ptrbb = bb + off*8;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+
+		res2_0 = 0;
+		res2_1 = 0;
+
+		res3_0 = 0;
+		res3_1 = 0;
+
+		res4_0 = 0;
+		res4_1 = 0;
+
+		res5_0 = 0;
+		res5_1 = 0;
+
+		res6_0 = 0;
+		res6_1 = 0;
+
+		res7_0 = 0;
+		res7_1 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+2;	// number of values in A
+#else
+		temp = off+8;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+			b4 = ptrbb[4];
+			b5 = ptrbb[5];
+			b6 = ptrbb[6];
+			b7 = ptrbb[7];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+			res4_0 += a0*b4;
+			res5_0 += a0*b5;
+			res6_0 += a0*b6;
+			res7_0 += a0*b7;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+			res2_1 += a1*b2;
+			res3_1 += a1*b3;
+			res4_1 += a1*b4;
+			res5_1 += a1*b5;
+			res6_1 += a1*b6;
+			res7_1 += a1*b7;
+
+			ptrba = ptrba+2;
+			ptrbb = ptrbb+8;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+
+		res2_0 *= alpha;
+		res2_1 *= alpha;
+
+		res3_0 *= alpha;
+		res3_1 *= alpha;
+
+		res4_0 *= alpha;
+		res4_1 *= alpha;
+
+		res5_0 *= alpha;
+		res5_1 *= alpha;
+
+		res6_0 *= alpha;
+		res6_1 *= alpha;
+
+		res7_0 *= alpha;
+		res7_1 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+
+		C2[0] = res2_0;
+		C2[1] = res2_1;
+
+		C3[0] = res3_0;
+		C3[1] = res3_1;
+
+		C4[0] = res4_0;
+		C4[1] = res4_1;
+
+		C5[0] = res5_0;
+		C5[1] = res5_1;
+
+		C6[0] = res6_0;
+		C6[1] = res6_1;
+
+		C7[0] = res7_0;
+		C7[1] = res7_1;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 2; // number of values in A
+#else
+		temp -= 8; // number of values in B
+#endif
+		ptrba += temp*2;
+		ptrbb += temp*8;
+#endif
+
+#ifdef LEFT
+		off += 2; // number of values in A
+#endif
+
+		C0 = C0+2;
+		C1 = C1+2;
+		C2 = C2+2;
+		C3 = C3+2;
+		C4 = C4+2;
+		C5 = C5+2;
+		C6 = C6+2;
+		C7 = C7+2;
+
+	}
+
+	if ( bm & 1 ) // do any 1x4 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*1;
+		ptrbb = bb + off*8;
+#endif
+
+		res0_0 = 0;
+		res1_0 = 0;
+		res2_0 = 0;
+		res3_0 = 0;
+		res4_0 = 0;
+		res5_0 = 0;
+		res6_0 = 0;
+		res7_0 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+1;	// number of values in A
+#else
+		temp = off+8;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+			b4 = ptrbb[4];
+			b5 = ptrbb[5];
+			b6 = ptrbb[6];
+			b7 = ptrbb[7];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+			res4_0 += a0*b4;
+			res5_0 += a0*b5;
+			res6_0 += a0*b6;
+			res7_0 += a0*b7;
+
+			ptrba = ptrba+1;
+			ptrbb = ptrbb+8;
+                }
+
+		res0_0 *= alpha;
+
+		res1_0 *= alpha;
+
+		res2_0 *= alpha;
+
+		res3_0 *= alpha;
+		res4_0 *= alpha;
+		res5_0 *= alpha;
+		res6_0 *= alpha;
+		res7_0 *= alpha;
+
+		C0[0] = res0_0;
+
+		C1[0] = res1_0;
+
+		C2[0] = res2_0;
+
+		C3[0] = res3_0;
+		C4[0] = res4_0;
+		C5[0] = res5_0;
+		C6[0] = res6_0;
+		C7[0] = res7_0;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 1; // number of values in A
+#else
+		temp -= 8; // number of values in B
+#endif
+		ptrba += temp*1;
+		ptrbb += temp*8;
+#endif
+
+#ifdef LEFT
+		off += 1; // number of values in A
+#endif
+
+		C0 = C0+1;
+		C1 = C1+1;
+		C2 = C2+1;
+		C3 = C3+1;
+		C4 = C4+1;
+		C5 = C5+1;
+		C6 = C6+1;
+		C7 = C7+1;
+
+	}
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+		off += 8;
+#endif
+
+        k = (bk<<3);
+        bb = bb+k;
+        i = (ldc<<3);
+        C = C+i;
+    }
+
+
+
+   for (j=0; j<(bn&4); j+=4) // do blocks of the Mx4 loops 
+   {
+        C0 = C;
+        C1 = C0+ldc;
+        C2 = C1+ldc;
+        C3 = C2+ldc;
+
+
+        if (left) {
+            off = offset;
+        }
+
+        ptrba = ba;
+
+        for (i=0; i<bm/4; i+=1) // do blocks of 4x4
+	{
+
+		ptrbb = bb;
+                if (backwards)
+                {
+		   ptrba += off*4; // number of values in A
+		   ptrbb += off*4; // number of values in B
+                }
+
+		res0_0 = 0;
+		res0_1 = 0;
+		res0_2 = 0;
+		res0_3 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+		res1_2 = 0;
+		res1_3 = 0;
+
+		res2_0 = 0;
+		res2_1 = 0;
+		res2_2 = 0;
+		res2_3 = 0;
+
+		res3_0 = 0;
+		res3_1 = 0;
+		res3_2 = 0;
+		res3_3 = 0;
+
+                temp = backwards ? bk-off :
+                             left ? off + 4 : // number of values in A
+                                    off + 4;  // number of values in B
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+			res2_1 += a1*b2;
+			res3_1 += a1*b3;
+
+			a0 = ptrba[2];
+			res0_2 += a0*b0;
+			res1_2 += a0*b1;
+			res2_2 += a0*b2;
+			res3_2 += a0*b3;
+
+			a1 = ptrba[3];
+			res0_3 += a1*b0;
+			res1_3 += a1*b1;
+			res2_3 += a1*b2;
+			res3_3 += a1*b3;
+
+			ptrba = ptrba+4;
+			ptrbb = ptrbb+4;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+		res0_2 *= alpha;
+		res0_3 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+		res1_2 *= alpha;
+		res1_3 *= alpha;
+
+		res2_0 *= alpha;
+		res2_1 *= alpha;
+		res2_2 *= alpha;
+		res2_3 *= alpha;
+
+		res3_0 *= alpha;
+		res3_1 *= alpha;
+		res3_2 *= alpha;
+		res3_3 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+		C0[2] = res0_2;
+		C0[3] = res0_3;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+		C1[2] = res1_2;
+		C1[3] = res1_3;
+
+		C2[0] = res2_0;
+		C2[1] = res2_1;
+		C2[2] = res2_2;
+		C2[3] = res2_3;
+
+		C3[0] = res3_0;
+		C3[1] = res3_1;
+		C3[2] = res3_2;
+		C3[3] = res3_3;
+
+		if (!backwards) {
+                    temp = bk-off;
+                    temp = left ? temp - 4 : // number of values in A
+                                  temp - 4;  // number of values in B
+
+                    ptrba += temp*4; // number of values in A
+		    ptrbb += temp*4; // number of values in B
+                }
+#ifdef LEFT
+		off += 4; // number of values in A
+#endif
+
+		C0 = C0+4;
+		C1 = C1+4;
+		C2 = C2+4;
+		C3 = C3+4;
+
+	}
+
+	if ( bm & 2 ) // do any 2x4 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*2;
+		ptrbb = bb + off*4;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+
+		res2_0 = 0;
+		res2_1 = 0;
+
+		res3_0 = 0;
+		res3_1 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+2;	// number of values in A
+#else
+		temp = off+4;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+			res2_1 += a1*b2;
+			res3_1 += a1*b3;
+
+			ptrba = ptrba+2;
+			ptrbb = ptrbb+4;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+
+		res2_0 *= alpha;
+		res2_1 *= alpha;
+
+		res3_0 *= alpha;
+		res3_1 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+
+		C2[0] = res2_0;
+		C2[1] = res2_1;
+
+		C3[0] = res3_0;
+		C3[1] = res3_1;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 2; // number of values in A
+#else
+		temp -= 4; // number of values in B
+#endif
+		ptrba += temp*2;
+		ptrbb += temp*4;
+#endif
+
+#ifdef LEFT
+		off += 2; // number of values in A
+#endif
+
+		C0 = C0+2;
+		C1 = C1+2;
+		C2 = C2+2;
+		C3 = C3+2;
+
+	}
+
+	if ( bm & 1 ) // do any 1x4 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*1;
+		ptrbb = bb + off*4;
+#endif
+
+		res0_0 = 0;
+		res1_0 = 0;
+		res2_0 = 0;
+		res3_0 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+1;	// number of values in A
+#else
+		temp = off+4;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+			b2 = ptrbb[2];
+			b3 = ptrbb[3];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+			res2_0 += a0*b2;
+			res3_0 += a0*b3;
+
+			ptrba = ptrba+1;
+			ptrbb = ptrbb+4;
+                }
+
+		res0_0 *= alpha;
+
+		res1_0 *= alpha;
+
+		res2_0 *= alpha;
+
+		res3_0 *= alpha;
+
+		C0[0] = res0_0;
+
+		C1[0] = res1_0;
+
+		C2[0] = res2_0;
+
+		C3[0] = res3_0;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 1; // number of values in A
+#else
+		temp -= 4; // number of values in B
+#endif
+		ptrba += temp*1;
+		ptrbb += temp*4;
+#endif
+
+#ifdef LEFT
+		off += 1; // number of values in A
+#endif
+
+		C0 = C0+1;
+		C1 = C1+1;
+		C2 = C2+1;
+		C3 = C3+1;
+
+	}
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+		off += 4;
+#endif
+
+        k = (bk<<2);
+        bb = bb+k;
+        i = (ldc<<2);
+        C = C+i;
+    }
+
+
+
+   for (j=0; j<(bn&2); j+=2) // do the Mx2 loops 
+   {
+        C0 = C;
+        C1 = C0+ldc;
+
+#if defined(TRMMKERNEL) && defined(LEFT)
+		off = offset;
+#endif
+
+
+        ptrba = ba;
+
+        for (i=0; i<bm/4; i+=1) // do blocks of 4x2
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*4;
+		ptrbb = bb + off*2;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+		res0_2 = 0;
+		res0_3 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+		res1_2 = 0;
+		res1_3 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+4;	// number of values in A
+#else
+		temp = off+2;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+
+			a0 = ptrba[2];
+			res0_2 += a0*b0;
+			res1_2 += a0*b1;
+
+			a1 = ptrba[3];
+			res0_3 += a1*b0;
+			res1_3 += a1*b1;
+
+			ptrba = ptrba+4;
+			ptrbb = ptrbb+2;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+		res0_2 *= alpha;
+		res0_3 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+		res1_2 *= alpha;
+		res1_3 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+		C0[2] = res0_2;
+		C0[3] = res0_3;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+		C1[2] = res1_2;
+		C1[3] = res1_3;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 4; // number of values in A
+#else
+		temp -= 2; // number of values in B
+#endif
+		ptrba += temp*4;
+		ptrbb += temp*2;
+#endif
+
+#ifdef LEFT
+		off += 4; // number of values in A
+#endif
+
+		C0 = C0+4;
+		C1 = C1+4;
+
+	}
+
+	if ( bm & 2 ) // do any 2x2 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*2;
+		ptrbb = bb + off*2;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+
+		res1_0 = 0;
+		res1_1 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+2;	// number of values in A
+#else
+		temp = off+2;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+			res1_1 += a1*b1;
+
+			ptrba = ptrba+2;
+			ptrbb = ptrbb+2;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+
+		res1_0 *= alpha;
+		res1_1 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+
+		C1[0] = res1_0;
+		C1[1] = res1_1;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 2; // number of values in A
+#else
+		temp -= 2; // number of values in B
+#endif
+		ptrba += temp*2;
+		ptrbb += temp*2;
+#endif
+
+#ifdef LEFT
+		off += 2; // number of values in A
+#endif
+
+		C0 = C0+2;
+		C1 = C1+2;
+
+	}
+
+	if ( bm & 1 ) // do any 1x2 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*1;
+		ptrbb = bb + off*2;
+#endif
+
+		res0_0 = 0;
+
+		res1_0 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+1;	// number of values in A
+#else
+		temp = off+2;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+			b1 = ptrbb[1];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+			res1_0 += a0*b1;
+
+			ptrba = ptrba+1;
+			ptrbb = ptrbb+2;
+                }
+
+		res0_0 *= alpha;
+
+		res1_0 *= alpha;
+
+		C0[0] = res0_0;
+
+		C1[0] = res1_0;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 1; // number of values in A
+#else
+		temp -= 2; // number of values in B
+#endif
+		ptrba += temp*1;
+		ptrbb += temp*2;
+#endif
+
+#ifdef LEFT
+		off += 1; // number of values in A
+#endif
+
+		C0 = C0+1;
+		C1 = C1+1;
+
+	}
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+		off += 2;
+#endif
+
+        k = (bk<<1);
+        bb = bb+k;
+        i = (ldc<<1);
+        C = C+i;
+    }
+
+
+
+
+
+
+
+   for (j=0; j<(bn&1); j+=1) // do the Mx1 loops
+   {
+        C0 = C;
+
+#if defined(TRMMKERNEL) &&  defined(LEFT)
+	off = offset;
+#endif
+
+        ptrba = ba;
+
+        for (i=0; i<bm/4; i+=1) // do blocks of 4x1 loops
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*4;
+		ptrbb = bb + off*1;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+		res0_2 = 0;
+		res0_3 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+4;	// number of values in A
+#else
+		temp = off+1;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+
+			a0 = ptrba[2];
+			res0_2 += a0*b0;
+
+			a1 = ptrba[3];
+			res0_3 += a1*b0;
+
+			ptrba = ptrba+4;
+			ptrbb = ptrbb+1;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+		res0_2 *= alpha;
+		res0_3 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+		C0[2] = res0_2;
+		C0[3] = res0_3;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 4; // number of values in A
+#else
+		temp -= 1; // number of values in B
+#endif
+		ptrba += temp*4;
+		ptrbb += temp*1;
+#endif
+
+#ifdef LEFT
+		off += 4; // number of values in A
+#endif
+
+		C0 = C0+4;
+
+	}
+
+	if ( bm & 2 ) // do any 2x1 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*2;
+		ptrbb = bb + off*1;
+#endif
+
+		res0_0 = 0;
+		res0_1 = 0;
+
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+2;	// number of values in A
+#else
+		temp = off+1;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+
+			a1 = ptrba[1];
+			res0_1 += a1*b0;
+
+			ptrba = ptrba+2;
+			ptrbb = ptrbb+1;
+                }
+
+		res0_0 *= alpha;
+		res0_1 *= alpha;
+
+		C0[0] = res0_0;
+		C0[1] = res0_1;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 2; // number of values in A
+#else
+		temp -= 1; // number of values in B
+#endif
+		ptrba += temp*2;
+		ptrbb += temp*1;
+#endif
+
+#ifdef LEFT
+		off += 2; // number of values in A
+#endif
+
+		C0 = C0+2;
+
+	}
+
+	if ( bm & 1 ) // do any 1x1 loop
+	{
+
+#if (defined(LEFT) &&  defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		ptrbb = bb;
+#else
+		ptrba += off*1;
+		ptrbb = bb + off*1;
+#endif
+
+		res0_0 = 0;
+
+
+#if (defined(LEFT) && !defined(TRANSA)) || (!defined(LEFT) && defined(TRANSA))
+		temp = bk-off;
+#elif defined(LEFT)
+		temp = off+1;	// number of values in A
+#else
+		temp = off+1;	// number of values in B
+#endif
+
+		for (k=0; k<temp; k++)
+                {
+			b0 = ptrbb[0];
+
+			a0 = ptrba[0];
+			res0_0 += a0*b0;
+
+			ptrba = ptrba+1;
+			ptrbb = ptrbb+1;
+                }
+
+		res0_0 *= alpha;
+
+		C0[0] = res0_0;
+
+
+#if ( defined(LEFT) && defined(TRANSA)) || (!defined(LEFT) && !defined(TRANSA))
+		temp = bk - off;
+#ifdef LEFT
+		temp -= 1; // number of values in A
+#else
+		temp -= 1; // number of values in B
+#endif
+		ptrba += temp*1;
+		ptrbb += temp*1;
+#endif
+
+#ifdef LEFT
+		off += 1; // number of values in A
+#endif
+
+		C0 = C0+1;
+
+	}
+
+
+
+#if defined(TRMMKERNEL) && !defined(LEFT)
+		off += 1;
+#endif
+
+        k = (bk<<0);
+        bb = bb+k;
+        C = C+ldc;
+   }
+   return 0;
+}
diff --git a/kernel/x86_64/saxpy.c b/kernel/x86_64/saxpy.c
index da81f1354..0b76c42f7 100644
--- a/kernel/x86_64/saxpy.c
+++ b/kernel/x86_64/saxpy.c
@@ -31,6 +31,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if defined(NEHALEM)
 #include "saxpy_microk_nehalem-2.c"
+#elif defined(HASWELL)
+#include "saxpy_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "saxpy_microk_sandy-2.c"
+#elif defined(PILEDRIVER) || defined(STEAMROLLER)
+#include "saxpy_microk_piledriver-2.c"
 #endif
 
 
@@ -69,7 +75,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-		int n1 = n & -16;
+		BLASLONG n1 = n & -32;
 
 		if ( n1 )
 			saxpy_kernel_16(n1, x, y , &da );
@@ -87,6 +93,29 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
 
 	}
 
+	BLASLONG n1 = n & -4;
+
+	while(i < n1)
+	{
+
+                FLOAT m1      = da * x[ix] ;
+                FLOAT m2      = da * x[ix+inc_x] ;
+                FLOAT m3      = da * x[ix+2*inc_x] ;
+                FLOAT m4      = da * x[ix+3*inc_x] ;
+
+                y[iy]         += m1 ;
+                y[iy+inc_y]   += m2 ;
+                y[iy+2*inc_y] += m3 ;
+                y[iy+3*inc_y] += m4 ;
+
+                ix  += inc_x*4 ;
+                iy  += inc_y*4 ;
+                i+=4 ;
+
+
+	}
+
+
 	while(i < n)
 	{
 
diff --git a/kernel/x86_64/saxpy_microk_haswell-2.c b/kernel/x86_64/saxpy_microk_haswell-2.c
new file mode 100644
index 000000000..496424207
--- /dev/null
+++ b/kernel/x86_64/saxpy_microk_haswell-2.c
@@ -0,0 +1,78 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void saxpy_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void saxpy_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss		(%4), %%ymm0		    \n\t"  // alpha	
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+        "vmovups                  (%3,%0,4), %%ymm12         \n\t"  // 8 * y
+        "vmovups                32(%3,%0,4), %%ymm13         \n\t"  // 8 * y
+        "vmovups                64(%3,%0,4), %%ymm14         \n\t"  // 8 * y
+        "vmovups                96(%3,%0,4), %%ymm15         \n\t"  // 8 * y
+	"vfmadd231ps       (%2,%0,4), %%ymm0  , %%ymm12  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     32(%2,%0,4), %%ymm0  , %%ymm13  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     64(%2,%0,4), %%ymm0  , %%ymm14  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     96(%2,%0,4), %%ymm0  , %%ymm15  	     \n\t"   // y += alpha * x
+	"vmovups	%%ymm12,   (%3,%0,4)		     \n\t"
+	"vmovups	%%ymm13, 32(%3,%0,4)		     \n\t"
+	"vmovups	%%ymm14, 64(%3,%0,4)		     \n\t"
+	"vmovups	%%ymm15, 96(%3,%0,4)		     \n\t"
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $32, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/saxpy_microk_piledriver-2.c b/kernel/x86_64/saxpy_microk_piledriver-2.c
new file mode 100644
index 000000000..204cf8bac
--- /dev/null
+++ b/kernel/x86_64/saxpy_microk_piledriver-2.c
@@ -0,0 +1,159 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void saxpy_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void saxpy_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n < 1024 )
+  {
+
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vbroadcastss		(%4), %%xmm0		    \n\t"  // alpha	
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+        "vmovups                  (%3,%0,4), %%xmm8          \n\t"  // 4 * y
+        "vmovups                16(%3,%0,4), %%xmm9          \n\t"  // 4 * y
+        "vmovups                32(%3,%0,4), %%xmm10         \n\t"  // 4 * y
+        "vmovups                48(%3,%0,4), %%xmm11         \n\t"  // 4 * y
+        "vmovups                64(%3,%0,4), %%xmm12         \n\t"  // 4 * y
+        "vmovups                80(%3,%0,4), %%xmm13         \n\t"  // 4 * y
+        "vmovups                96(%3,%0,4), %%xmm14         \n\t"  // 4 * y
+        "vmovups               112(%3,%0,4), %%xmm15         \n\t"  // 4 * y
+
+	"vfmadd231ps       (%2,%0,4), %%xmm0  , %%xmm8   	     \n\t"   // y += alpha * x
+	"vfmadd231ps     16(%2,%0,4), %%xmm0  , %%xmm9   	     \n\t"   // y += alpha * x
+	"vfmadd231ps     32(%2,%0,4), %%xmm0  , %%xmm10  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     48(%2,%0,4), %%xmm0  , %%xmm11  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     64(%2,%0,4), %%xmm0  , %%xmm12  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     80(%2,%0,4), %%xmm0  , %%xmm13  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     96(%2,%0,4), %%xmm0  , %%xmm14  	     \n\t"   // y += alpha * x
+	"vfmadd231ps    112(%2,%0,4), %%xmm0  , %%xmm15  	     \n\t"   // y += alpha * x
+
+	"vmovups	%%xmm8 ,   (%3,%0,4)		     \n\t"
+	"vmovups	%%xmm9 , 16(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm10, 32(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm12, 64(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm13, 80(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm14, 96(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm15,112(%3,%0,4)		     \n\t"
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $32, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vbroadcastss		(%4), %%xmm0		    \n\t"  // alpha	
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+	"prefetcht0		512(%3,%0,4)	             \n\t"
+	"prefetcht0		576(%3,%0,4)	             \n\t"
+
+        "vmovups                  (%3,%0,4), %%xmm8          \n\t"  // 4 * y
+        "vmovups                16(%3,%0,4), %%xmm9          \n\t"  // 4 * y
+        "vmovups                32(%3,%0,4), %%xmm10         \n\t"  // 4 * y
+        "vmovups                48(%3,%0,4), %%xmm11         \n\t"  // 4 * y
+        "vmovups                64(%3,%0,4), %%xmm12         \n\t"  // 4 * y
+        "vmovups                80(%3,%0,4), %%xmm13         \n\t"  // 4 * y
+        "vmovups                96(%3,%0,4), %%xmm14         \n\t"  // 4 * y
+        "vmovups               112(%3,%0,4), %%xmm15         \n\t"  // 4 * y
+
+	"prefetcht0		512(%2,%0,4)	             \n\t"
+	"prefetcht0		576(%2,%0,4)	             \n\t"
+	"vfmadd231ps       (%2,%0,4), %%xmm0  , %%xmm8   	     \n\t"   // y += alpha * x
+	"vfmadd231ps     16(%2,%0,4), %%xmm0  , %%xmm9   	     \n\t"   // y += alpha * x
+	"vfmadd231ps     32(%2,%0,4), %%xmm0  , %%xmm10  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     48(%2,%0,4), %%xmm0  , %%xmm11  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     64(%2,%0,4), %%xmm0  , %%xmm12  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     80(%2,%0,4), %%xmm0  , %%xmm13  	     \n\t"   // y += alpha * x
+	"vfmadd231ps     96(%2,%0,4), %%xmm0  , %%xmm14  	     \n\t"   // y += alpha * x
+	"vfmadd231ps    112(%2,%0,4), %%xmm0  , %%xmm15  	     \n\t"   // y += alpha * x
+
+	"vmovups	%%xmm8 ,   (%3,%0,4)		     \n\t"
+	"vmovups	%%xmm9 , 16(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm10, 32(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm12, 64(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm13, 80(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm14, 96(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm15,112(%3,%0,4)		     \n\t"
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $32, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/saxpy_microk_sandy-2.c b/kernel/x86_64/saxpy_microk_sandy-2.c
new file mode 100644
index 000000000..159a23175
--- /dev/null
+++ b/kernel/x86_64/saxpy_microk_sandy-2.c
@@ -0,0 +1,119 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void saxpy_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void saxpy_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss		(%4), %%ymm0		    \n\t"  // alpha	
+	"vmovups	  (%3,%0,4), %%ymm8         \n\t"
+	"vmovups	32(%3,%0,4), %%ymm9         \n\t"
+	"vmovups	64(%3,%0,4), %%ymm10        \n\t"
+	"vmovups	96(%3,%0,4), %%ymm11        \n\t"
+	"vmovups	  (%2,%0,4), %%ymm4         \n\t"
+	"vmovups	32(%2,%0,4), %%ymm5         \n\t"
+	"vmovups	64(%2,%0,4), %%ymm6         \n\t"
+	"vmovups	96(%2,%0,4), %%ymm7        \n\t"
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $32, %1			             \n\t"		
+	"jz		2f		             \n\t"
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmulps		%%ymm4, %%ymm0, %%ymm4		\n\t"
+	"vaddps		%%ymm8 , %%ymm4, %%ymm12	     \n\t"
+	"vmulps		%%ymm5, %%ymm0, %%ymm5		\n\t"
+	"vaddps		%%ymm9 , %%ymm5, %%ymm13	     \n\t"
+	"vmulps		%%ymm6, %%ymm0, %%ymm6		\n\t"
+	"vaddps		%%ymm10, %%ymm6, %%ymm14	     \n\t"
+	"vmulps		%%ymm7, %%ymm0, %%ymm7		\n\t"
+	"vaddps		%%ymm11, %%ymm7, %%ymm15	     \n\t"
+
+	"vmovups	  (%3,%0,4), %%ymm8         \n\t"
+	"vmovups	32(%3,%0,4), %%ymm9         \n\t"
+	"vmovups	64(%3,%0,4), %%ymm10        \n\t"
+	"vmovups	96(%3,%0,4), %%ymm11        \n\t"
+
+	"vmovups	  (%2,%0,4), %%ymm4         \n\t"
+	"vmovups	32(%2,%0,4), %%ymm5         \n\t"
+	"vmovups	64(%2,%0,4), %%ymm6         \n\t"
+	"vmovups	96(%2,%0,4), %%ymm7        \n\t"
+
+	"vmovups	%%ymm12, -128(%3,%0,4)		     \n\t"
+	"vmovups	%%ymm13,  -96(%3,%0,4)		     \n\t"
+	"vmovups	%%ymm14,  -64(%3,%0,4)		     \n\t"
+	"vmovups	%%ymm15,  -32(%3,%0,4)		     \n\t"
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $32, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+
+	"2:				            \n\t"
+	"vmulps		%%ymm4, %%ymm0, %%ymm4		\n\t"
+	"vmulps		%%ymm5, %%ymm0, %%ymm5		\n\t"
+	"vmulps		%%ymm6, %%ymm0, %%ymm6		\n\t"
+	"vmulps		%%ymm7, %%ymm0, %%ymm7		\n\t"
+
+	"vaddps		%%ymm8 , %%ymm4, %%ymm12	     \n\t"
+	"vaddps		%%ymm9 , %%ymm5, %%ymm13	     \n\t"
+	"vaddps		%%ymm10, %%ymm6, %%ymm14	     \n\t"
+	"vaddps		%%ymm11, %%ymm7, %%ymm15	     \n\t"
+
+	"vmovups	%%ymm12, -128(%3,%0,4)		     \n\t"
+	"vmovups	%%ymm13,  -96(%3,%0,4)		     \n\t"
+	"vmovups	%%ymm14,  -64(%3,%0,4)		     \n\t"
+	"vmovups	%%ymm15,  -32(%3,%0,4)		     \n\t"
+
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/sdot.c b/kernel/x86_64/sdot.c
index 6fec48175..a6da1fea7 100644
--- a/kernel/x86_64/sdot.c
+++ b/kernel/x86_64/sdot.c
@@ -28,10 +28,16 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #include "common.h"
 
-#if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)
+#if defined(BULLDOZER) 
 #include "sdot_microk_bulldozer-2.c"
+#elif defined(STEAMROLLER) || defined(PILEDRIVER)
+#include "sdot_microk_steamroller-2.c"
 #elif defined(NEHALEM)
 #include "sdot_microk_nehalem-2.c"
+#elif defined(HASWELL)
+#include "sdot_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "sdot_microk_sandy-2.c"
 #endif
 
 
@@ -74,7 +80,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-		int n1 = n & -16;
+		BLASLONG n1 = n & -32;
 
 		if ( n1 )
 			sdot_kernel_16(n1, x, y , &dot );
@@ -93,6 +99,18 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
 
 	}
 
+	BLASLONG n1 = n & -2;
+
+	while(i < n1)
+	{
+
+		dot += y[iy] * x[ix] + y[iy+inc_y] * x[ix+inc_x];
+		ix  += inc_x*2 ;
+		iy  += inc_y*2 ;
+		i+=2 ;
+
+	}
+
 	while(i < n)
 	{
 
diff --git a/kernel/x86_64/sdot_microk_haswell-2.c b/kernel/x86_64/sdot_microk_haswell-2.c
new file mode 100644
index 000000000..4051f9c1b
--- /dev/null
+++ b/kernel/x86_64/sdot_microk_haswell-2.c
@@ -0,0 +1,96 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void sdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void sdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vxorps		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorps		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorps		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorps		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,4), %%ymm12         \n\t"  // 2 * x
+        "vmovups                32(%2,%0,4), %%ymm13         \n\t"  // 2 * x
+        "vmovups                64(%2,%0,4), %%ymm14         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,4), %%ymm15         \n\t"  // 2 * x
+
+	"vfmadd231ps      (%3,%0,4), %%ymm12, %%ymm4 \n\t"  // 2 * y
+	"vfmadd231ps    32(%3,%0,4), %%ymm13, %%ymm5 \n\t"  // 2 * y
+	"vfmadd231ps    64(%3,%0,4), %%ymm14, %%ymm6 \n\t"  // 2 * y
+	"vfmadd231ps    96(%3,%0,4), %%ymm15, %%ymm7 \n\t"  // 2 * y
+
+	"addq		$32 , %0	  	     \n\t"
+	"subq	        $32 , %1		     \n\t"		
+	"jnz		1b		             \n\t"
+
+	"vextractf128	$1 , %%ymm4 , %%xmm12	     \n\t"
+	"vextractf128	$1 , %%ymm5 , %%xmm13	     \n\t"
+	"vextractf128	$1 , %%ymm6 , %%xmm14	     \n\t"
+	"vextractf128	$1 , %%ymm7 , %%xmm15	     \n\t"
+
+	"vaddps        %%xmm4, %%xmm12, %%xmm4	\n\t"
+	"vaddps        %%xmm5, %%xmm13, %%xmm5	\n\t"
+	"vaddps        %%xmm6, %%xmm14, %%xmm6	\n\t"
+	"vaddps        %%xmm7, %%xmm15, %%xmm7	\n\t"
+
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddps        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddps        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddps        %%xmm4, %%xmm4, %%xmm4	\n\t"
+	"vhaddps        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovss		%%xmm4,    (%4)		\n\t"
+	"vzeroupper				\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm4", "%xmm5", 
+	  "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/sdot_microk_sandy-2.c b/kernel/x86_64/sdot_microk_sandy-2.c
new file mode 100644
index 000000000..e265d16bd
--- /dev/null
+++ b/kernel/x86_64/sdot_microk_sandy-2.c
@@ -0,0 +1,101 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void sdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void sdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vxorps		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorps		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorps		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorps		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,4), %%ymm12         \n\t"  // 2 * x
+        "vmovups                32(%2,%0,4), %%ymm13         \n\t"  // 2 * x
+        "vmovups                64(%2,%0,4), %%ymm14         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,4), %%ymm15         \n\t"  // 2 * x
+
+	"vmulps      (%3,%0,4), %%ymm12, %%ymm12 \n\t"  // 2 * y
+	"vmulps    32(%3,%0,4), %%ymm13, %%ymm13 \n\t"  // 2 * y
+	"vmulps    64(%3,%0,4), %%ymm14, %%ymm14 \n\t"  // 2 * y
+	"vmulps    96(%3,%0,4), %%ymm15, %%ymm15 \n\t"  // 2 * y
+
+	"vaddps    %%ymm4 , %%ymm12, %%ymm4 \n\t"  // 2 * y
+	"vaddps    %%ymm5 , %%ymm13, %%ymm5 \n\t"  // 2 * y
+	"vaddps    %%ymm6 , %%ymm14, %%ymm6 \n\t"  // 2 * y
+	"vaddps    %%ymm7 , %%ymm15, %%ymm7 \n\t"  // 2 * y
+
+	"addq		$32 , %0	  	     \n\t"
+	"subq	        $32 , %1		     \n\t"		
+	"jnz		1b		             \n\t"
+
+	"vextractf128	$1 , %%ymm4 , %%xmm12	     \n\t"
+	"vextractf128	$1 , %%ymm5 , %%xmm13	     \n\t"
+	"vextractf128	$1 , %%ymm6 , %%xmm14	     \n\t"
+	"vextractf128	$1 , %%ymm7 , %%xmm15	     \n\t"
+
+	"vaddps        %%xmm4, %%xmm12, %%xmm4	\n\t"
+	"vaddps        %%xmm5, %%xmm13, %%xmm5	\n\t"
+	"vaddps        %%xmm6, %%xmm14, %%xmm6	\n\t"
+	"vaddps        %%xmm7, %%xmm15, %%xmm7	\n\t"
+
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddps        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddps        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddps        %%xmm4, %%xmm4, %%xmm4	\n\t"
+	"vhaddps        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovss		%%xmm4,    (%4)		\n\t"
+	"vzeroupper				\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm4", "%xmm5", 
+	  "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/sdot_microk_steamroller-2.c b/kernel/x86_64/sdot_microk_steamroller-2.c
new file mode 100644
index 000000000..6b8b2566b
--- /dev/null
+++ b/kernel/x86_64/sdot_microk_steamroller-2.c
@@ -0,0 +1,163 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void sdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void sdot_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+
+  if ( n < 4096 )
+  {
+
+	__asm__  __volatile__
+	(
+	"vxorps		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorps		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorps		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorps		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,4), %%xmm12         \n\t"  // 4 * x
+        "vmovups                16(%2,%0,4), %%xmm13         \n\t"  // 4 * x
+        "vmovups                32(%2,%0,4), %%xmm14         \n\t"  // 4 * x
+        "vmovups                48(%2,%0,4), %%xmm15         \n\t"  // 4 * x
+
+	"vfmadd231ps       (%3,%0,4), %%xmm12, %%xmm4 \n\t"  // 4 * y
+	"vfmadd231ps     16(%3,%0,4), %%xmm13, %%xmm5 \n\t"  // 4 * y
+        "vmovups                64(%2,%0,4), %%xmm0          \n\t"  // 4 * x
+        "vmovups                80(%2,%0,4), %%xmm1          \n\t"  // 4 * x
+	"vfmadd231ps     32(%3,%0,4), %%xmm14, %%xmm6 \n\t"  // 4 * y
+	"vfmadd231ps     48(%3,%0,4), %%xmm15, %%xmm7 \n\t"  // 4 * y
+
+        "vmovups                96(%2,%0,4), %%xmm2          \n\t"  // 4 * x
+        "vmovups               112(%2,%0,4), %%xmm3          \n\t"  // 4 * x
+
+	"vfmadd231ps     64(%3,%0,4), %%xmm0 , %%xmm4 \n\t"  // 4 * y
+	"vfmadd231ps     80(%3,%0,4), %%xmm1 , %%xmm5 \n\t"  // 4 * y
+	"vfmadd231ps     96(%3,%0,4), %%xmm2 , %%xmm6 \n\t"  // 4 * y
+	"vfmadd231ps    112(%3,%0,4), %%xmm3 , %%xmm7 \n\t"  // 4 * y
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $32, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddps        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddps        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddps        %%xmm4, %%xmm4, %%xmm4	\n\t"
+	"vhaddps        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovss		%%xmm4,    (%4)		\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+
+  }
+
+	__asm__  __volatile__
+	(
+	"vxorps		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorps		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorps		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorps		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+	"prefetcht0	512(%2,%0,4)		     \n\t"
+        "vmovups                  (%2,%0,4), %%xmm12         \n\t"  // 4 * x
+        "vmovups                16(%2,%0,4), %%xmm13         \n\t"  // 4 * x
+        "vmovups                32(%2,%0,4), %%xmm14         \n\t"  // 4 * x
+        "vmovups                48(%2,%0,4), %%xmm15         \n\t"  // 4 * x
+
+	"prefetcht0	512(%3,%0,4)		     \n\t"
+	"vfmadd231ps       (%3,%0,4), %%xmm12, %%xmm4 \n\t"  // 4 * y
+	"vfmadd231ps     16(%3,%0,4), %%xmm13, %%xmm5 \n\t"  // 4 * y
+	"prefetcht0	576(%2,%0,4)		     \n\t"
+        "vmovups                64(%2,%0,4), %%xmm0          \n\t"  // 4 * x
+        "vmovups                80(%2,%0,4), %%xmm1          \n\t"  // 4 * x
+	"prefetcht0	576(%3,%0,4)		     \n\t"
+	"vfmadd231ps     32(%3,%0,4), %%xmm14, %%xmm6 \n\t"  // 4 * y
+	"vfmadd231ps     48(%3,%0,4), %%xmm15, %%xmm7 \n\t"  // 4 * y
+
+        "vmovups                96(%2,%0,4), %%xmm2          \n\t"  // 4 * x
+        "vmovups               112(%2,%0,4), %%xmm3          \n\t"  // 4 * x
+
+	"vfmadd231ps     64(%3,%0,4), %%xmm0 , %%xmm4 \n\t"  // 4 * y
+	"vfmadd231ps     80(%3,%0,4), %%xmm1 , %%xmm5 \n\t"  // 4 * y
+	"vfmadd231ps     96(%3,%0,4), %%xmm2 , %%xmm6 \n\t"  // 4 * y
+	"vfmadd231ps    112(%3,%0,4), %%xmm3 , %%xmm7 \n\t"  // 4 * y
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $32, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+
+	"vaddps        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddps        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddps        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vhaddps        %%xmm4, %%xmm4, %%xmm4	\n\t"
+	"vhaddps        %%xmm4, %%xmm4, %%xmm4	\n\t"
+
+	"vmovss		%%xmm4,    (%4)		\n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/sgemm_kernel_16x4_sandy.S b/kernel/x86_64/sgemm_kernel_16x4_sandy.S
index 77c7e2f50..ea15cd87e 100644
--- a/kernel/x86_64/sgemm_kernel_16x4_sandy.S
+++ b/kernel/x86_64/sgemm_kernel_16x4_sandy.S
@@ -328,17 +328,17 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if !defined(TRMMKERNEL)
 
-	vaddps 	        (CO1), %xmm4,%xmm4
-	vaddps  1 * SIZE(CO1), %xmm5,%xmm5
+	vaddss 	        (CO1), %xmm4,%xmm4
+	vaddss  1 * SIZE(CO1), %xmm5,%xmm5
 
-	vaddps 	        (CO1, LDC), %xmm6,%xmm6
-	vaddps  1 * SIZE(CO1, LDC), %xmm7,%xmm7
+	vaddss 	        (CO1, LDC), %xmm6,%xmm6
+	vaddss  1 * SIZE(CO1, LDC), %xmm7,%xmm7
 
-	vaddps 	        (CO2), %xmm8,%xmm8
-	vaddps  1 * SIZE(CO2), %xmm9,%xmm9
+	vaddss 	        (CO2), %xmm8,%xmm8
+	vaddss  1 * SIZE(CO2), %xmm9,%xmm9
 
-	vaddps 	        (CO2, LDC), %xmm10,%xmm10
-	vaddps  1 * SIZE(CO2, LDC), %xmm11,%xmm11
+	vaddss 	        (CO2, LDC), %xmm10,%xmm10
+	vaddss  1 * SIZE(CO2, LDC), %xmm11,%xmm11
 
 #endif
 
@@ -389,10 +389,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if !defined(TRMMKERNEL)
 
-	vaddps 	        (CO1), %xmm4,%xmm4
-	vaddps 	        (CO1, LDC), %xmm6,%xmm6
-	vaddps 	        (CO2), %xmm8,%xmm8
-	vaddps 	        (CO2, LDC), %xmm10,%xmm10
+	vaddss 	        (CO1), %xmm4,%xmm4
+	vaddss 	        (CO1, LDC), %xmm6,%xmm6
+	vaddss 	        (CO2), %xmm8,%xmm8
+	vaddss 	        (CO2, LDC), %xmm10,%xmm10
 
 #endif
 
@@ -557,11 +557,11 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if !defined(TRMMKERNEL)
 
-	vaddps 	        (CO1), %xmm4,%xmm4
-	vaddps  1 * SIZE(CO1), %xmm5,%xmm5
+	vaddss 	        (CO1), %xmm4,%xmm4
+	vaddss  1 * SIZE(CO1), %xmm5,%xmm5
 
-	vaddps 	        (CO1, LDC), %xmm6,%xmm6
-	vaddps  1 * SIZE(CO1, LDC), %xmm7,%xmm7
+	vaddss 	        (CO1, LDC), %xmm6,%xmm6
+	vaddss  1 * SIZE(CO1, LDC), %xmm7,%xmm7
 
 #endif
 
@@ -597,8 +597,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #if !defined(TRMMKERNEL)
 
-	vaddps 	        (CO1), %xmm4,%xmm4
-	vaddps 	        (CO1, LDC), %xmm6,%xmm6
+	vaddss 	        (CO1), %xmm4,%xmm4
+	vaddss 	        (CO1, LDC), %xmm6,%xmm6
 
 #endif
 
diff --git a/kernel/x86_64/sgemv_n_4.c b/kernel/x86_64/sgemv_n_4.c
index 840ce9207..dc88ea098 100644
--- a/kernel/x86_64/sgemv_n_4.c
+++ b/kernel/x86_64/sgemv_n_4.c
@@ -376,6 +376,8 @@ int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha, FLOAT *a, BLASLO
 				sgemv_kernel_4x4(NB,ap,x_ptr,ybuffer,&alpha);
 				ap[0] += lda4; 
 				ap[1] += lda4; 
+				ap[2] += lda4; 
+				ap[3] += lda4; 
 				a_ptr += lda4;
 				x_ptr += 4;	
 			}
diff --git a/kernel/x86_64/sger.c b/kernel/x86_64/sger.c
new file mode 100644
index 000000000..84c056c0d
--- /dev/null
+++ b/kernel/x86_64/sger.c
@@ -0,0 +1,84 @@
+/*********************************************************************/
+/* Copyright 2009, 2010 The University of Texas at Austin.           */
+/* All rights reserved.                                              */
+/*                                                                   */
+/* Redistribution and use in source and binary forms, with or        */
+/* without modification, are permitted provided that the following   */
+/* conditions are met:                                               */
+/*                                                                   */
+/*   1. Redistributions of source code must retain the above         */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer.                                                  */
+/*                                                                   */
+/*   2. Redistributions in binary form must reproduce the above      */
+/*      copyright notice, this list of conditions and the following  */
+/*      disclaimer in the documentation and/or other materials       */
+/*      provided with the distribution.                              */
+/*                                                                   */
+/*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */
+/*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */
+/*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */
+/*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */
+/*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */
+/*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */
+/*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */
+/*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */
+/*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */
+/*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */
+/*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */
+/*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */
+/*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */
+/*    POSSIBILITY OF SUCH DAMAGE.                                    */
+/*                                                                   */
+/* The views and conclusions contained in the software and           */
+/* documentation are those of the authors and should not be          */
+/* interpreted as representing official policies, either expressed   */
+/* or implied, of The University of Texas at Austin.                 */
+/*********************************************************************/
+
+#include <stdio.h>
+#include "common.h"
+
+#if defined(SANDYBRIDGE)
+#include "sger_microk_sandy-2.c"
+#endif
+
+int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha,
+	 FLOAT *x, BLASLONG incx,
+	 FLOAT *y, BLASLONG incy,
+	 FLOAT *a, BLASLONG lda, FLOAT *buffer){
+
+	FLOAT *X    = x;
+
+	if (incx != 1) {
+    		X = buffer;
+    		COPY_K(m, x, incx, X, 1);
+	}
+
+	BLASLONG m1 = m & -16;
+
+	while (n > 0) 
+	{
+   		FLOAT y0 = alpha * *y;
+		if ( m1 > 0 )
+		{
+			#ifdef HAVE_KERNEL_16
+				sger_kernel_16(m1, X, a, &y0);
+			#else
+    				AXPYU_K(m1, 0, 0, y0, X, 1, a, 1, NULL, 0);
+			#endif
+		}
+	
+		if ( m > m1 )
+		{
+    			AXPYU_K(m-m1, 0, 0, y0, X+m1 , 1, a+m1, 1, NULL, 0);
+		}
+	
+    		a += lda;
+    		y += incy;
+    		n --;
+	}
+
+	return 0;
+}
+
diff --git a/kernel/x86_64/sger_microk_sandy-2.c b/kernel/x86_64/sger_microk_sandy-2.c
new file mode 100644
index 000000000..51c3bef3e
--- /dev/null
+++ b/kernel/x86_64/sger_microk_sandy-2.c
@@ -0,0 +1,124 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_16 1
+static void sger_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void sger_kernel_16( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vbroadcastss		(%4), %%xmm0		    \n\t"  // alpha	
+	"prefetcht0	256(%3,%0,4)			    \n\t"
+	"vmovups	  (%3,%0,4), %%xmm8         \n\t"
+	"vmovups	16(%3,%0,4), %%xmm9         \n\t"
+	"vmovups	32(%3,%0,4), %%xmm10        \n\t"
+	"vmovups	48(%3,%0,4), %%xmm11        \n\t"
+
+	"prefetcht0	256(%2,%0,4)			    \n\t"
+	"vmovups	  (%2,%0,4), %%xmm4         \n\t"
+	"vmovups	16(%2,%0,4), %%xmm5         \n\t"
+	"vmovups	32(%2,%0,4), %%xmm6         \n\t"
+	"vmovups	48(%2,%0,4), %%xmm7        \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
+	"jz		2f		             \n\t"
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmulps		%%xmm4, %%xmm0, %%xmm4		\n\t"
+	"vaddps		%%xmm8 , %%xmm4, %%xmm12	     \n\t"
+	"vmulps		%%xmm5, %%xmm0, %%xmm5		\n\t"
+	"vaddps		%%xmm9 , %%xmm5, %%xmm13	     \n\t"
+	"vmulps		%%xmm6, %%xmm0, %%xmm6		\n\t"
+	"vaddps		%%xmm10, %%xmm6, %%xmm14	     \n\t"
+	"vmulps		%%xmm7, %%xmm0, %%xmm7		\n\t"
+	"vaddps		%%xmm11, %%xmm7, %%xmm15	     \n\t"
+
+	"prefetcht0	256(%3,%0,4)			    \n\t"
+	"vmovups	  (%3,%0,4), %%xmm8         \n\t"
+	"vmovups	16(%3,%0,4), %%xmm9         \n\t"
+	"vmovups	32(%3,%0,4), %%xmm10        \n\t"
+	"vmovups	48(%3,%0,4), %%xmm11        \n\t"
+
+	"prefetcht0	256(%2,%0,4)			    \n\t"
+	"vmovups	  (%2,%0,4), %%xmm4         \n\t"
+	"vmovups	16(%2,%0,4), %%xmm5         \n\t"
+	"vmovups	32(%2,%0,4), %%xmm6         \n\t"
+	"vmovups	48(%2,%0,4), %%xmm7        \n\t"
+
+	"vmovups	%%xmm12,  -64(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm13,  -48(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm14,  -32(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm15,  -16(%3,%0,4)		     \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+
+	"2:				            \n\t"
+	"vmulps		%%xmm4, %%xmm0, %%xmm4		\n\t"
+	"vmulps		%%xmm5, %%xmm0, %%xmm5		\n\t"
+	"vmulps		%%xmm6, %%xmm0, %%xmm6		\n\t"
+	"vmulps		%%xmm7, %%xmm0, %%xmm7		\n\t"
+
+	"vaddps		%%xmm8 , %%xmm4, %%xmm12	     \n\t"
+	"vaddps		%%xmm9 , %%xmm5, %%xmm13	     \n\t"
+	"vaddps		%%xmm10, %%xmm6, %%xmm14	     \n\t"
+	"vaddps		%%xmm11, %%xmm7, %%xmm15	     \n\t"
+
+	"vmovups	%%xmm12,  -64(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm13,  -48(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm14,  -32(%3,%0,4)		     \n\t"
+	"vmovups	%%xmm15,  -16(%3,%0,4)		     \n\t"
+
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha)   // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/ssymv_L.c b/kernel/x86_64/ssymv_L.c
index a2b716b58..0997f108d 100644
--- a/kernel/x86_64/ssymv_L.c
+++ b/kernel/x86_64/ssymv_L.c
@@ -32,6 +32,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "ssymv_L_microk_bulldozer-2.c"
 #elif defined(NEHALEM)
 #include "ssymv_L_microk_nehalem-2.c"
+#elif defined(HASWELL)
+#include "ssymv_L_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "ssymv_L_microk_sandy-2.c"
 #endif
 
 
diff --git a/kernel/x86_64/ssymv_L_microk_haswell-2.c b/kernel/x86_64/ssymv_L_microk_haswell-2.c
new file mode 100644
index 000000000..516524528
--- /dev/null
+++ b/kernel/x86_64/ssymv_L_microk_haswell-2.c
@@ -0,0 +1,124 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4x4 1
+static void ssymv_kernel_4x4( BLASLONG from, BLASLONG to, FLOAT **a, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2) __attribute__ ((noinline));
+
+static void ssymv_kernel_4x4(BLASLONG from, BLASLONG to, FLOAT **a, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2)
+{
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorps		%%xmm0 , %%xmm0 , %%xmm0     \n\t"	// temp2[0]
+	"vxorps		%%xmm1 , %%xmm1 , %%xmm1     \n\t"	// temp2[1]
+	"vxorps		%%xmm2 , %%xmm2 , %%xmm2     \n\t"	// temp2[2]
+	"vxorps		%%xmm3 , %%xmm3 , %%xmm3     \n\t"	// temp2[3]
+	"vbroadcastss   (%8),    %%xmm4	             \n\t"	// temp1[0]
+	"vbroadcastss  4(%8),    %%xmm5	             \n\t"	// temp1[1]
+	"vbroadcastss  8(%8),    %%xmm6	             \n\t"	// temp1[1]
+	"vbroadcastss 12(%8),    %%xmm7	             \n\t"	// temp1[1]
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,4), %%xmm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,4), %%xmm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,4), %%xmm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,4), %%xmm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,4), %%xmm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,4), %%xmm15	           \n\t"  // 2 * a
+
+	"vfmadd231ps	%%xmm4, %%xmm12 , %%xmm9  \n\t"  // y     += temp1 * a
+	"vfmadd231ps	%%xmm8, %%xmm12 , %%xmm0  \n\t"  // temp2 += x * a
+
+	"vfmadd231ps	%%xmm5, %%xmm13 , %%xmm9  \n\t"  // y     += temp1 * a
+	"vfmadd231ps	%%xmm8, %%xmm13 , %%xmm1  \n\t"  // temp2 += x * a
+
+	"vfmadd231ps	%%xmm6, %%xmm14 , %%xmm9  \n\t"  // y     += temp1 * a
+	"vfmadd231ps	%%xmm8, %%xmm14 , %%xmm2  \n\t"  // temp2 += x * a
+
+	"vfmadd231ps	%%xmm7, %%xmm15 , %%xmm9  \n\t"  // y     += temp1 * a
+	"vfmadd231ps	%%xmm8, %%xmm15 , %%xmm3  \n\t"  // temp2 += x * a
+
+	"vmovups	%%xmm9 ,  (%3,%0,4)		   \n\t"
+
+	"addq		$4 , %0	  	 	      \n\t"
+	"cmpq		%0 , %1			      \n\t"
+	"jnz		1b		      \n\t"
+
+	"vmovss		  (%9), %%xmm4		      \n\t"
+	"vmovss		 4(%9), %%xmm5		      \n\t"
+	"vmovss		 8(%9), %%xmm6		      \n\t"
+	"vmovss		12(%9), %%xmm7		      \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddss		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddss		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddss		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddss		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovss         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovss         %%xmm1 , 4(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm2 , 8(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm3 ,12(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (from),	// 0	
+	  "r" (to),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a[0]),	// 4
+          "r" (a[1]),	// 5
+          "r" (a[2]),	// 6
+          "r" (a[3]),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/ssymv_L_microk_sandy-2.c b/kernel/x86_64/ssymv_L_microk_sandy-2.c
new file mode 100644
index 000000000..07293a964
--- /dev/null
+++ b/kernel/x86_64/ssymv_L_microk_sandy-2.c
@@ -0,0 +1,243 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4x4 1
+static void ssymv_kernel_4x4( BLASLONG from, BLASLONG to, FLOAT **a, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2) __attribute__ ((noinline));
+
+static void ssymv_kernel_4x4(BLASLONG from, BLASLONG to, FLOAT **a, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2)
+{
+
+  if ( ( to - from ) & 4 )
+  {
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorps		%%xmm0 , %%xmm0 , %%xmm0     \n\t"	// temp2[0]
+	"vxorps		%%xmm1 , %%xmm1 , %%xmm1     \n\t"	// temp2[1]
+	"vxorps		%%xmm2 , %%xmm2 , %%xmm2     \n\t"	// temp2[2]
+	"vxorps		%%xmm3 , %%xmm3 , %%xmm3     \n\t"	// temp2[3]
+	"vbroadcastss   (%8),    %%xmm4	             \n\t"	// temp1[0]
+	"vbroadcastss  4(%8),    %%xmm5	             \n\t"	// temp1[1]
+	"vbroadcastss  8(%8),    %%xmm6	             \n\t"	// temp1[1]
+	"vbroadcastss 12(%8),    %%xmm7	             \n\t"	// temp1[1]
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,4), %%xmm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,4), %%xmm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,4), %%xmm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,4), %%xmm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,4), %%xmm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,4), %%xmm15	           \n\t"  // 2 * a
+
+        "vmulps         %%xmm4, %%xmm12 , %%xmm10          \n\t"
+        "vmulps         %%xmm8, %%xmm12 , %%xmm11          \n\t"
+        "vaddps         %%xmm9, %%xmm10 , %%xmm9           \n\t"
+        "vaddps         %%xmm0, %%xmm11 , %%xmm0           \n\t"
+
+        "vmulps         %%xmm5, %%xmm13 , %%xmm10          \n\t"
+        "vmulps         %%xmm8, %%xmm13 , %%xmm11          \n\t"
+        "vaddps         %%xmm9, %%xmm10 , %%xmm9           \n\t"
+        "vaddps         %%xmm1, %%xmm11 , %%xmm1           \n\t"
+
+        "vmulps         %%xmm6, %%xmm14 , %%xmm10          \n\t"
+        "vmulps         %%xmm8, %%xmm14 , %%xmm11          \n\t"
+        "vaddps         %%xmm9, %%xmm10 , %%xmm9           \n\t"
+        "vaddps         %%xmm2, %%xmm11 , %%xmm2           \n\t"
+
+        "vmulps         %%xmm7, %%xmm15 , %%xmm10          \n\t"
+        "vmulps         %%xmm8, %%xmm15 , %%xmm11          \n\t"
+        "vaddps         %%xmm9, %%xmm10 , %%xmm9           \n\t"
+        "vaddps         %%xmm3, %%xmm11 , %%xmm3           \n\t"
+
+	"vmovups	%%xmm9 ,  (%3,%0,4)		   \n\t"
+
+	"addq		$4 , %0	  	 	      \n\t"
+	"cmpq		%0 , %1			      \n\t"
+	"jnz		1b		      \n\t"
+
+	"vmovss		  (%9), %%xmm4		      \n\t"
+	"vmovss		 4(%9), %%xmm5		      \n\t"
+	"vmovss		 8(%9), %%xmm6		      \n\t"
+	"vmovss		12(%9), %%xmm7		      \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddss		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddss		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddss		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddss		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovss         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovss         %%xmm1 , 4(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm2 , 8(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm3 ,12(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (from),	// 0	
+	  "r" (to),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a[0]),	// 4
+          "r" (a[1]),	// 5
+          "r" (a[2]),	// 6
+          "r" (a[3]),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+  }
+
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorps		%%ymm0 , %%ymm0 , %%ymm0     \n\t"	// temp2[0]
+	"vxorps		%%ymm1 , %%ymm1 , %%ymm1     \n\t"	// temp2[1]
+	"vxorps		%%ymm2 , %%ymm2 , %%ymm2     \n\t"	// temp2[2]
+	"vxorps		%%ymm3 , %%ymm3 , %%ymm3     \n\t"	// temp2[3]
+	"vbroadcastss   (%8),    %%ymm4	             \n\t"	// temp1[0]
+	"vbroadcastss  4(%8),    %%ymm5	             \n\t"	// temp1[1]
+	"vbroadcastss  8(%8),    %%ymm6	             \n\t"	// temp1[1]
+	"vbroadcastss 12(%8),    %%ymm7	             \n\t"	// temp1[1]
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,4), %%ymm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,4), %%ymm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,4), %%ymm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,4), %%ymm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,4), %%ymm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,4), %%ymm15	           \n\t"  // 2 * a
+
+        "vmulps         %%ymm4, %%ymm12 , %%ymm10          \n\t"
+        "vmulps         %%ymm8, %%ymm12 , %%ymm11          \n\t"
+        "vaddps         %%ymm9, %%ymm10 , %%ymm9           \n\t"
+        "vaddps         %%ymm0, %%ymm11 , %%ymm0           \n\t"
+
+        "vmulps         %%ymm5, %%ymm13 , %%ymm10          \n\t"
+        "vmulps         %%ymm8, %%ymm13 , %%ymm11          \n\t"
+        "vaddps         %%ymm9, %%ymm10 , %%ymm9           \n\t"
+        "vaddps         %%ymm1, %%ymm11 , %%ymm1           \n\t"
+
+        "vmulps         %%ymm6, %%ymm14 , %%ymm10          \n\t"
+        "vmulps         %%ymm8, %%ymm14 , %%ymm11          \n\t"
+        "vaddps         %%ymm9, %%ymm10 , %%ymm9           \n\t"
+        "vaddps         %%ymm2, %%ymm11 , %%ymm2           \n\t"
+
+        "vmulps         %%ymm7, %%ymm15 , %%ymm10          \n\t"
+        "vmulps         %%ymm8, %%ymm15 , %%ymm11          \n\t"
+        "vaddps         %%ymm9, %%ymm10 , %%ymm9           \n\t"
+        "vaddps         %%ymm3, %%ymm11 , %%ymm3           \n\t"
+
+	"vmovups	%%ymm9 ,  (%3,%0,4)		   \n\t"
+
+	"addq		$8 , %0	  	 	      \n\t"
+	"cmpq		%0 , %1			      \n\t"
+	"jnz		1b		      \n\t"
+
+	"vmovss		  (%9), %%xmm4		      \n\t"
+	"vmovss		 4(%9), %%xmm5		      \n\t"
+	"vmovss		 8(%9), %%xmm6		      \n\t"
+	"vmovss		12(%9), %%xmm7		      \n\t"
+
+        "vextractf128 $0x01, %%ymm0 , %%xmm12         \n\t"
+        "vextractf128 $0x01, %%ymm1 , %%xmm13         \n\t"
+        "vextractf128 $0x01, %%ymm2 , %%xmm14         \n\t"
+        "vextractf128 $0x01, %%ymm3 , %%xmm15         \n\t"
+
+        "vaddps         %%xmm0, %%xmm12, %%xmm0       \n\t"
+        "vaddps         %%xmm1, %%xmm13, %%xmm1       \n\t"
+        "vaddps         %%xmm2, %%xmm14, %%xmm2       \n\t"
+        "vaddps         %%xmm3, %%xmm15, %%xmm3       \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddss		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddss		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddss		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddss		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovss         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovss         %%xmm1 , 4(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm2 , 8(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm3 ,12(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (from),	// 0	
+	  "r" (to),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a[0]),	// 4
+          "r" (a[1]),	// 5
+          "r" (a[2]),	// 6
+          "r" (a[3]),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+
+} 
+
+
diff --git a/kernel/x86_64/ssymv_U.c b/kernel/x86_64/ssymv_U.c
index 0aadd3fd2..ed1e8236c 100644
--- a/kernel/x86_64/ssymv_U.c
+++ b/kernel/x86_64/ssymv_U.c
@@ -33,6 +33,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "ssymv_U_microk_bulldozer-2.c"
 #elif defined(NEHALEM)
 #include "ssymv_U_microk_nehalem-2.c"
+#elif defined(HASWELL)
+#include "ssymv_U_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "ssymv_U_microk_sandy-2.c"
 #endif
 
 #ifndef HAVE_KERNEL_4x4
diff --git a/kernel/x86_64/ssymv_U_microk_haswell-2.c b/kernel/x86_64/ssymv_U_microk_haswell-2.c
new file mode 100644
index 000000000..42f801c96
--- /dev/null
+++ b/kernel/x86_64/ssymv_U_microk_haswell-2.c
@@ -0,0 +1,136 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4x4 1
+static void ssymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT *a3, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2) __attribute__ ((noinline));
+
+static void ssymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT *a3, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2)
+{
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorps		%%ymm0 , %%ymm0 , %%ymm0     \n\t"	// temp2[0]
+	"vxorps		%%ymm1 , %%ymm1 , %%ymm1     \n\t"	// temp2[1]
+	"vxorps		%%ymm2 , %%ymm2 , %%ymm2     \n\t"	// temp2[2]
+	"vxorps		%%ymm3 , %%ymm3 , %%ymm3     \n\t"	// temp2[3]
+	"vbroadcastss   (%8),    %%ymm4	             \n\t"	// temp1[0]
+	"vbroadcastss  4(%8),    %%ymm5	             \n\t"	// temp1[1]
+	"vbroadcastss  8(%8),    %%ymm6	             \n\t"	// temp1[1]
+	"vbroadcastss 12(%8),    %%ymm7	             \n\t"	// temp1[1]
+	"xorq           %0,%0                        \n\t"
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,4), %%ymm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,4), %%ymm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,4), %%ymm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,4), %%ymm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,4), %%ymm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,4), %%ymm15	           \n\t"  // 2 * a
+
+	"vfmadd231ps	%%ymm4, %%ymm12 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231ps	%%ymm8, %%ymm12 , %%ymm0  \n\t"  // temp2 += x * a
+
+	"vfmadd231ps	%%ymm5, %%ymm13 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231ps	%%ymm8, %%ymm13 , %%ymm1  \n\t"  // temp2 += x * a
+
+	"vfmadd231ps	%%ymm6, %%ymm14 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231ps	%%ymm8, %%ymm14 , %%ymm2  \n\t"  // temp2 += x * a
+
+	"vfmadd231ps	%%ymm7, %%ymm15 , %%ymm9  \n\t"  // y     += temp1 * a
+	"vfmadd231ps	%%ymm8, %%ymm15 , %%ymm3  \n\t"  // temp2 += x * a
+
+	"vmovups	%%ymm9 ,  (%3,%0,4)	      \n\t"
+	"addq		$8 , %0	  	 	      \n\t"
+	"subq		$8 , %1	  	 	      \n\t"
+
+	"jnz		1b		      \n\t"
+
+	"vmovss		  (%9), %%xmm4		      \n\t"
+	"vmovss		 4(%9), %%xmm5		      \n\t"
+	"vmovss		 8(%9), %%xmm6		      \n\t"
+	"vmovss		12(%9), %%xmm7		      \n\t"
+
+	"vextractf128 $0x01, %%ymm0 , %%xmm12	      \n\t"
+	"vextractf128 $0x01, %%ymm1 , %%xmm13	      \n\t"
+	"vextractf128 $0x01, %%ymm2 , %%xmm14	      \n\t"
+	"vextractf128 $0x01, %%ymm3 , %%xmm15	      \n\t"
+
+	"vaddps	        %%xmm0, %%xmm12, %%xmm0	      \n\t"
+	"vaddps	        %%xmm1, %%xmm13, %%xmm1	      \n\t"
+	"vaddps	        %%xmm2, %%xmm14, %%xmm2	      \n\t"
+	"vaddps	        %%xmm3, %%xmm15, %%xmm3	      \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddsd		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddsd		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddsd		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddsd		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovss         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovss         %%xmm1 , 4(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm2 , 8(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm3 ,12(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a0),	// 4
+          "r" (a1),	// 5
+          "r" (a2),	// 6
+          "r" (a3),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/ssymv_U_microk_sandy-2.c b/kernel/x86_64/ssymv_U_microk_sandy-2.c
new file mode 100644
index 000000000..4b699af50
--- /dev/null
+++ b/kernel/x86_64/ssymv_U_microk_sandy-2.c
@@ -0,0 +1,144 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4x4 1
+static void ssymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT *a3, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2) __attribute__ ((noinline));
+
+static void ssymv_kernel_4x4(BLASLONG n, FLOAT *a0, FLOAT *a1, FLOAT *a2, FLOAT *a3, FLOAT *x, FLOAT *y, FLOAT *temp1, FLOAT *temp2)
+{
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper				     \n\t"
+	"vxorps		%%ymm0 , %%ymm0 , %%ymm0     \n\t"	// temp2[0]
+	"vxorps		%%ymm1 , %%ymm1 , %%ymm1     \n\t"	// temp2[1]
+	"vxorps		%%ymm2 , %%ymm2 , %%ymm2     \n\t"	// temp2[2]
+	"vxorps		%%ymm3 , %%ymm3 , %%ymm3     \n\t"	// temp2[3]
+	"vbroadcastss   (%8),    %%ymm4	             \n\t"	// temp1[0]
+	"vbroadcastss  4(%8),    %%ymm5	             \n\t"	// temp1[1]
+	"vbroadcastss  8(%8),    %%ymm6	             \n\t"	// temp1[1]
+	"vbroadcastss 12(%8),    %%ymm7	             \n\t"	// temp1[1]
+	"xorq           %0,%0                        \n\t"
+
+	".align 16				     \n\t"
+	"1:				     \n\t"
+
+	"vmovups	(%3,%0,4), %%ymm9	           \n\t"  // 2 * y
+	"vmovups	(%2,%0,4), %%ymm8	           \n\t"  // 2 * x
+
+	"vmovups	(%4,%0,4), %%ymm12	           \n\t"  // 2 * a
+	"vmovups	(%5,%0,4), %%ymm13	           \n\t"  // 2 * a
+	"vmovups	(%6,%0,4), %%ymm14	           \n\t"  // 2 * a
+	"vmovups	(%7,%0,4), %%ymm15	           \n\t"  // 2 * a
+
+	"vmulps		%%ymm4, %%ymm12 , %%ymm10	   \n\t"
+	"vaddps		%%ymm9, %%ymm10 , %%ymm9	   \n\t"
+	"vmulps		%%ymm8, %%ymm12 , %%ymm11	   \n\t"
+	"vaddps		%%ymm0, %%ymm11 , %%ymm0	   \n\t"
+
+	"vmulps		%%ymm5, %%ymm13 , %%ymm10	   \n\t"
+	"vaddps		%%ymm9, %%ymm10 , %%ymm9	   \n\t"
+	"vmulps		%%ymm8, %%ymm13 , %%ymm11	   \n\t"
+	"vaddps		%%ymm1, %%ymm11 , %%ymm1	   \n\t"
+
+	"vmulps		%%ymm6, %%ymm14 , %%ymm10	   \n\t"
+	"vaddps		%%ymm9, %%ymm10 , %%ymm9	   \n\t"
+	"vmulps		%%ymm8, %%ymm14 , %%ymm11	   \n\t"
+	"vaddps		%%ymm2, %%ymm11 , %%ymm2	   \n\t"
+
+	"vmulps		%%ymm7, %%ymm15 , %%ymm10	   \n\t"
+	"vaddps		%%ymm9, %%ymm10 , %%ymm9	   \n\t"
+	"vmulps		%%ymm8, %%ymm15 , %%ymm11	   \n\t"
+	"vaddps		%%ymm3, %%ymm11 , %%ymm3	   \n\t"
+
+	"vmovups	%%ymm9 ,  (%3,%0,4)	      \n\t"
+	"addq		$8 , %0	  	 	      \n\t"
+	"subq		$8 , %1	  	 	      \n\t"
+
+	"jnz		1b		      \n\t"
+
+	"vmovss		  (%9), %%xmm4		      \n\t"
+	"vmovss		 4(%9), %%xmm5		      \n\t"
+	"vmovss		 8(%9), %%xmm6		      \n\t"
+	"vmovss		12(%9), %%xmm7		      \n\t"
+
+	"vextractf128 $0x01, %%ymm0 , %%xmm12	      \n\t"
+	"vextractf128 $0x01, %%ymm1 , %%xmm13	      \n\t"
+	"vextractf128 $0x01, %%ymm2 , %%xmm14	      \n\t"
+	"vextractf128 $0x01, %%ymm3 , %%xmm15	      \n\t"
+
+	"vaddps	        %%xmm0, %%xmm12, %%xmm0	      \n\t"
+	"vaddps	        %%xmm1, %%xmm13, %%xmm1	      \n\t"
+	"vaddps	        %%xmm2, %%xmm14, %%xmm2	      \n\t"
+	"vaddps	        %%xmm3, %%xmm15, %%xmm3	      \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vhaddps        %%xmm0, %%xmm0, %%xmm0  \n\t"
+	"vhaddps        %%xmm1, %%xmm1, %%xmm1  \n\t"
+	"vhaddps        %%xmm2, %%xmm2, %%xmm2  \n\t"
+	"vhaddps        %%xmm3, %%xmm3, %%xmm3  \n\t"
+
+	"vaddsd		%%xmm4, %%xmm0, %%xmm0  \n\t"
+	"vaddsd		%%xmm5, %%xmm1, %%xmm1  \n\t"
+	"vaddsd		%%xmm6, %%xmm2, %%xmm2  \n\t"
+	"vaddsd		%%xmm7, %%xmm3, %%xmm3  \n\t"
+
+	"vmovss         %%xmm0 ,  (%9)		\n\t"	// save temp2
+	"vmovss         %%xmm1 , 4(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm2 , 8(%9)		\n\t"	// save temp2
+	"vmovss         %%xmm3 ,12(%9)		\n\t"	// save temp2
+	"vzeroupper				     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (a0),	// 4
+          "r" (a1),	// 5
+          "r" (a2),	// 6
+          "r" (a3),	// 8
+          "r" (temp1),  // 8
+          "r" (temp2)   // 9
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
diff --git a/kernel/x86_64/zaxpy.c b/kernel/x86_64/zaxpy.c
index 52a25c793..560acc7f9 100644
--- a/kernel/x86_64/zaxpy.c
+++ b/kernel/x86_64/zaxpy.c
@@ -29,8 +29,14 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 
 
-#if defined(BULLDOZER) || defined(PILEDRIVER) || defined(STEAMROLLER)
+#if defined(BULLDOZER)
 #include "zaxpy_microk_bulldozer-2.c"
+#elif defined(PILEDRIVER) || defined(STEAMROLLER)
+#include "zaxpy_microk_steamroller-2.c"
+#elif defined(HASWELL)
+#include "zaxpy_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "zaxpy_microk_sandy-2.c"
 #endif
 
 
@@ -78,13 +84,13 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da_r, FLOAT da_i,
 	if ( (inc_x == 1) && (inc_y == 1) )
 	{
 
-		int n1 = n & -4;
+		BLASLONG n1 = n & -16;
 
 		if ( n1 )
 		{
 			da[0] = da_r;
 			da[1] = da_i;
-			zaxpy_kernel_4(n1, x, y , &da );
+			zaxpy_kernel_4(n1, x, y , da );
 			ix = 2 * n1;
 		}
 		i = n1;
diff --git a/kernel/x86_64/zaxpy_microk_bulldozer-2.c b/kernel/x86_64/zaxpy_microk_bulldozer-2.c
index f9732cd4e..0e15761f7 100644
--- a/kernel/x86_64/zaxpy_microk_bulldozer-2.c
+++ b/kernel/x86_64/zaxpy_microk_bulldozer-2.c
@@ -31,89 +31,87 @@ static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __att
 static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
 {
 
+#if !defined(CONJ)
+	FLOAT mvec[2] = { -1.0, 1.0 };
+#else
+	FLOAT mvec[2] = { 1.0, -1.0 };
+#endif
 
 	BLASLONG register i = 0;
 
+  if ( n < 384 )
+  {
+
 	__asm__  __volatile__
 	(
+	"vzeroupper					    \n\t"
 	"vmovddup		(%4), %%xmm0		    \n\t"  // real part of alpha
 	"vmovddup	       8(%4), %%xmm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulpd		(%5), %%xmm1 , %%xmm1		    \n\t"
+#else
+	"vmulpd		(%5), %%xmm0 , %%xmm0		    \n\t"
+#endif
 
 	".align 16				            \n\t"
 	"1:				            \n\t"
 
-	"prefetcht0  768(%2,%0,8)                           \n\t"
 	"vmovups        (%2,%0,8), %%xmm5                   \n\t" // 1 complex values from x
+	".align 2					    \n\t"
 	"vmovups      16(%2,%0,8), %%xmm7                   \n\t" // 1 complex values from x
 	"vmovups      32(%2,%0,8), %%xmm9                   \n\t" // 1 complex values from x
 	"vmovups      48(%2,%0,8), %%xmm11                  \n\t" // 1 complex values from x
-	"prefetcht0  768(%3,%0,8)                           \n\t"
 
-#if !defined(CONJ)
-	"vfmaddpd    (%3,%0,8), %%xmm0 , %%xmm5, %%xmm12    \n\t"
+	"vmovups      64(%2,%0,8), %%xmm12                  \n\t" // 1 complex values from x
+	"vmovups      80(%2,%0,8), %%xmm13                  \n\t" // 1 complex values from x
+	"vmovups      96(%2,%0,8), %%xmm14                  \n\t" // 1 complex values from x
+	"vmovups     112(%2,%0,8), %%xmm15                  \n\t" // 1 complex values from x
+
 	"vpermilpd	$0x1 , %%xmm5 , %%xmm4 	    \n\t"  // exchange real and imag part
-	"vmulpd         %%xmm1, %%xmm4 , %%xmm4             \n\t"
-
-	"vfmaddpd  16(%3,%0,8), %%xmm0 , %%xmm7, %%xmm13    \n\t"
 	"vpermilpd	$0x1 , %%xmm7 , %%xmm6 	    \n\t"  // exchange real and imag part
-	"vmulpd         %%xmm1, %%xmm6 , %%xmm6             \n\t"
-
-	"vfmaddpd  32(%3,%0,8), %%xmm0 , %%xmm9, %%xmm14    \n\t"
 	"vpermilpd	$0x1 , %%xmm9 , %%xmm8 	    \n\t"  // exchange real and imag part
-	"vmulpd         %%xmm1, %%xmm8 , %%xmm8             \n\t"
-
-	"vfmaddpd  48(%3,%0,8), %%xmm0 , %%xmm11,%%xmm15    \n\t"
 	"vpermilpd	$0x1 , %%xmm11, %%xmm10 	    \n\t"  // exchange real and imag part
-	"vmulpd         %%xmm1, %%xmm10, %%xmm10            \n\t"
 
-        "vaddsubpd      %%xmm4, %%xmm12, %%xmm12            \n\t"
-        "vaddsubpd      %%xmm6, %%xmm13, %%xmm13            \n\t"
-        "vaddsubpd      %%xmm8, %%xmm14, %%xmm14            \n\t"
-        "vaddsubpd      %%xmm10,%%xmm15, %%xmm15            \n\t"
+	"vfmaddpd    (%3,%0,8), %%xmm0 , %%xmm5, %%xmm5     \n\t"
+	".align 2					    \n\t"
+	"vfmaddpd  16(%3,%0,8), %%xmm0 , %%xmm7, %%xmm7     \n\t"
+	"vfmaddpd  32(%3,%0,8), %%xmm0 , %%xmm9, %%xmm9     \n\t"
+	"vfmaddpd  48(%3,%0,8), %%xmm0 , %%xmm11,%%xmm11    \n\t"
 
-#else
+	"vfmaddpd	%%xmm5 , %%xmm1 , %%xmm4 , %%xmm5   \n\t"
+	"vfmaddpd	%%xmm7 , %%xmm1 , %%xmm6 , %%xmm7   \n\t"
+	"vfmaddpd	%%xmm9 , %%xmm1 , %%xmm8 , %%xmm9   \n\t"
+	"vfmaddpd	%%xmm11, %%xmm1 , %%xmm10, %%xmm11  \n\t"
 
-	"vmulpd		%%xmm0, %%xmm5, %%xmm4		    \n\t" // a_r*x_r, a_r*x_i
-	"vmulpd		%%xmm1, %%xmm5, %%xmm5 		    \n\t" // a_i*x_r, a_i*x_i
-	"vmulpd		%%xmm0, %%xmm7, %%xmm6		    \n\t" // a_r*x_r, a_r*x_i
-	"vmulpd		%%xmm1, %%xmm7, %%xmm7 		    \n\t" // a_i*x_r, a_i*x_i
-	"vmulpd		%%xmm0, %%xmm9, %%xmm8		    \n\t" // a_r*x_r, a_r*x_i
-	"vmulpd		%%xmm1, %%xmm9, %%xmm9 		    \n\t" // a_i*x_r, a_i*x_i
-	"vmulpd		%%xmm0, %%xmm11, %%xmm10	    \n\t" // a_r*x_r, a_r*x_i
-	"vmulpd		%%xmm1, %%xmm11, %%xmm11	    \n\t" // a_i*x_r, a_i*x_i
+	"vpermilpd	$0x1 , %%xmm12, %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm13, %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm14, %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm15, %%xmm10 	    \n\t"  // exchange real and imag part
 
-	"vpermilpd	$0x1 , %%xmm4 , %%xmm4 	    \n\t"  // exchange real and imag part
-	"vaddsubpd	%%xmm4 ,%%xmm5 , %%xmm4 	    \n\t"
-	"vpermilpd	$0x1 , %%xmm4 , %%xmm4 	    \n\t"  // exchange real and imag part
+	"vfmaddpd  64(%3,%0,8), %%xmm0 , %%xmm12, %%xmm12   \n\t"
+	"vfmaddpd  80(%3,%0,8), %%xmm0 , %%xmm13, %%xmm13   \n\t"
+	"vfmaddpd  96(%3,%0,8), %%xmm0 , %%xmm14, %%xmm14   \n\t"
+	"vfmaddpd 112(%3,%0,8), %%xmm0 , %%xmm15, %%xmm15   \n\t"
 
-	"vpermilpd	$0x1 , %%xmm6 , %%xmm6 	    \n\t"  // exchange real and imag part
-	"vaddsubpd	%%xmm6 ,%%xmm7 , %%xmm6 	    \n\t"
-	"vpermilpd	$0x1 , %%xmm6 , %%xmm6 	    \n\t"  // exchange real and imag part
+	"vfmaddpd	%%xmm12, %%xmm1 , %%xmm4 , %%xmm12  \n\t"
+	"vfmaddpd	%%xmm13, %%xmm1 , %%xmm6 , %%xmm13  \n\t"
+	"vfmaddpd	%%xmm14, %%xmm1 , %%xmm8 , %%xmm14  \n\t"
+	"vfmaddpd	%%xmm15, %%xmm1 , %%xmm10, %%xmm15  \n\t"
 
-	"vpermilpd	$0x1 , %%xmm8 , %%xmm8 	    \n\t"  // exchange real and imag part
-	"vaddsubpd	%%xmm8 ,%%xmm9 , %%xmm8 	    \n\t"
-	"vpermilpd	$0x1 , %%xmm8 , %%xmm8 	    \n\t"  // exchange real and imag part
+	"vmovups	%%xmm5 ,   (%3,%0,8)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%xmm7 , 16(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm9 , 32(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm12, 64(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm13, 80(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm14, 96(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm15,112(%3,%0,8)		    \n\t"
 
-	"vpermilpd	$0x1 , %%xmm10, %%xmm10	    \n\t"  // exchange real and imag part
-	"vaddsubpd	%%xmm10,%%xmm11, %%xmm10	    \n\t"
-	"vpermilpd	$0x1 , %%xmm10, %%xmm10	    \n\t"  // exchange real and imag part
-
-	"vaddpd	     (%3,%0,8) ,%%xmm4 , %%xmm12	    \n\t"
-	"vaddpd	   16(%3,%0,8) ,%%xmm6 , %%xmm13	    \n\t"
-	"vaddpd	   32(%3,%0,8) ,%%xmm8 , %%xmm14	    \n\t"
-	"vaddpd	   48(%3,%0,8) ,%%xmm10, %%xmm15	    \n\t"
-
-
-#endif
-
-	"vmovups	%%xmm12,   (%3,%0,8)		    \n\t"
-	"vmovups	%%xmm13, 16(%3,%0,8)		    \n\t"
-	"vmovups	%%xmm14, 32(%3,%0,8)		    \n\t"
-	"vmovups	%%xmm15, 48(%3,%0,8)		    \n\t"
-
-	"addq		$8 , %0	  	 	             \n\t"
-	"subq	        $4 , %1			             \n\t"		
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $8 , %1			             \n\t"		
 	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
 
 	:
         : 
@@ -121,7 +119,8 @@ static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
 	  "r" (n),  	// 1
           "r" (x),      // 2
           "r" (y),      // 3
-          "r" (alpha)   // 4
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
 	: "cc", 
 	  "%xmm0", "%xmm1",
 	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
@@ -129,7 +128,73 @@ static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
 	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
 	  "memory"
 	);
+	return;
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vmovddup		(%4), %%xmm0		    \n\t"  // real part of alpha
+	"vmovddup	       8(%4), %%xmm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulpd		(%5), %%xmm1 , %%xmm1		    \n\t"
+#else
+	"vmulpd		(%5), %%xmm0 , %%xmm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"prefetcht0	512(%2,%0,8)			    \n\t"
+	"vmovups        (%2,%0,8), %%xmm5                   \n\t" // 1 complex values from x
+	".align 2					    \n\t"
+	"vmovups      16(%2,%0,8), %%xmm7                   \n\t" // 1 complex values from x
+	"vmovups      32(%2,%0,8), %%xmm9                   \n\t" // 1 complex values from x
+	"vmovups      48(%2,%0,8), %%xmm11                  \n\t" // 1 complex values from x
+
+	"vpermilpd	$0x1 , %%xmm5 , %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm7 , %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm9 , %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm11, %%xmm10 	    \n\t"  // exchange real and imag part
+
+	"prefetcht0	512(%3,%0,8)			    \n\t"
+	"vfmaddpd    (%3,%0,8), %%xmm0 , %%xmm5, %%xmm5     \n\t"
+	".align 2					    \n\t"
+	"vfmaddpd  16(%3,%0,8), %%xmm0 , %%xmm7, %%xmm7     \n\t"
+	"vfmaddpd  32(%3,%0,8), %%xmm0 , %%xmm9, %%xmm9     \n\t"
+	"vfmaddpd  48(%3,%0,8), %%xmm0 , %%xmm11,%%xmm11    \n\t"
+
+	"vfmaddpd	%%xmm5 , %%xmm1 , %%xmm4 , %%xmm5   \n\t"
+	"vfmaddpd	%%xmm7 , %%xmm1 , %%xmm6 , %%xmm7   \n\t"
+	"vfmaddpd	%%xmm9 , %%xmm1 , %%xmm8 , %%xmm9   \n\t"
+	"vfmaddpd	%%xmm11, %%xmm1 , %%xmm10, %%xmm11  \n\t"
+
+	"vmovups	%%xmm5 ,   (%3,%0,8)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%xmm7 , 16(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm9 , 32(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,8)		    \n\t"
+
+	"addq		$8 , %0	  	 	             \n\t"
+	"subq	        $4, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "memory"
+	);
+
 
 } 
 
-
diff --git a/kernel/x86_64/zaxpy_microk_haswell-2.c b/kernel/x86_64/zaxpy_microk_haswell-2.c
new file mode 100644
index 000000000..e7e559502
--- /dev/null
+++ b/kernel/x86_64/zaxpy_microk_haswell-2.c
@@ -0,0 +1,132 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4 1
+static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+#if !defined(CONJ)
+	FLOAT mvec[4] = { -1.0, 1.0, -1.0, 1.0 };
+#else
+	FLOAT mvec[4] = { 1.0, -1.0, 1.0, -1.0 };
+#endif
+
+	BLASLONG register i = 0;
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vbroadcastsd		(%4), %%ymm0		    \n\t"  // real part of alpha
+	"vbroadcastsd	       8(%4), %%ymm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulpd		(%5), %%ymm1 , %%ymm1		    \n\t"
+#else
+	"vmulpd		(%5), %%ymm0 , %%ymm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmovups        (%2,%0,8), %%ymm5                   \n\t" // 2 complex values from x
+	".align 2					    \n\t"
+	"vmovups      32(%2,%0,8), %%ymm7                   \n\t" // 2 complex values from x
+	"vmovups      64(%2,%0,8), %%ymm9                   \n\t" // 2 complex values from x
+	"vmovups      96(%2,%0,8), %%ymm11                  \n\t" // 2 complex values from x
+
+	"vmovups     128(%2,%0,8), %%ymm12                  \n\t" // 2 complex values from x
+	"vmovups     160(%2,%0,8), %%ymm13                  \n\t" // 2 complex values from x
+	"vmovups     192(%2,%0,8), %%ymm14                  \n\t" // 2 complex values from x
+	"vmovups     224(%2,%0,8), %%ymm15                  \n\t" // 2 complex values from x
+
+	"vpermilpd	$0x5 , %%ymm5 , %%ymm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm7 , %%ymm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm9 , %%ymm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm11, %%ymm10 	    \n\t"  // exchange real and imag part
+
+	"vfmadd213pd    (%3,%0,8), %%ymm0 , %%ymm5          \n\t"
+	".align 2					    \n\t"
+	"vfmadd213pd  32(%3,%0,8), %%ymm0 , %%ymm7          \n\t"
+	"vfmadd213pd  64(%3,%0,8), %%ymm0 , %%ymm9          \n\t"
+	"vfmadd213pd  96(%3,%0,8), %%ymm0 , %%ymm11         \n\t"
+
+	"vfmadd231pd	%%ymm1 , %%ymm4 , %%ymm5   \n\t"
+	"vfmadd231pd	%%ymm1 , %%ymm6 , %%ymm7   \n\t"
+	"vfmadd231pd	%%ymm1 , %%ymm8 , %%ymm9   \n\t"
+	"vfmadd231pd	%%ymm1 , %%ymm10, %%ymm11  \n\t"
+
+	"vpermilpd	$0x5 , %%ymm12, %%ymm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm13, %%ymm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm14, %%ymm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm15, %%ymm10 	    \n\t"  // exchange real and imag part
+
+	"vfmadd213pd 128(%3,%0,8), %%ymm0 , %%ymm12         \n\t"
+	"vfmadd213pd 160(%3,%0,8), %%ymm0 , %%ymm13         \n\t"
+	"vfmadd213pd 192(%3,%0,8), %%ymm0 , %%ymm14         \n\t"
+	"vfmadd213pd 224(%3,%0,8), %%ymm0 , %%ymm15         \n\t"
+
+	"vfmadd231pd	%%ymm1 , %%ymm4 , %%ymm12  \n\t"
+	"vfmadd231pd	%%ymm1 , %%ymm6 , %%ymm13  \n\t"
+	"vfmadd231pd	%%ymm1 , %%ymm8 , %%ymm14  \n\t"
+	"vfmadd231pd	%%ymm1 , %%ymm10, %%ymm15  \n\t"
+
+	"vmovups	%%ymm5 ,   (%3,%0,8)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%ymm7 , 32(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm9 , 64(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm11, 96(%3,%0,8)		    \n\t"
+
+	"vmovups	%%ymm12,128(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm13,160(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm14,192(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm15,224(%3,%0,8)		    \n\t"
+
+	"addq		$32, %0	  	 	             \n\t"
+	"subq	        $16, %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15", 
+	  "memory"
+	);
+
+
+} 
+
diff --git a/kernel/x86_64/zaxpy_microk_sandy-2.c b/kernel/x86_64/zaxpy_microk_sandy-2.c
new file mode 100644
index 000000000..8b0a7ed05
--- /dev/null
+++ b/kernel/x86_64/zaxpy_microk_sandy-2.c
@@ -0,0 +1,198 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4 1
+static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+#if !defined(CONJ)
+	FLOAT mvec[4] = { -1.0, 1.0, -1.0, 1.0 };
+#else
+	FLOAT mvec[4] = { 1.0, -1.0, 1.0, -1.0 };
+#endif
+
+	BLASLONG register i = 0;
+
+  if ( n < 1280 )
+  {
+
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vbroadcastsd		(%4), %%ymm0		    \n\t"  // real part of alpha
+	"vbroadcastsd	       8(%4), %%ymm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulpd		(%5), %%ymm1 , %%ymm1		    \n\t"
+#else
+	"vmulpd		(%5), %%ymm0 , %%ymm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmovups        (%2,%0,8), %%ymm5                   \n\t" // 4 complex values from x
+	".align 2					    \n\t"
+	"vmovups      32(%2,%0,8), %%ymm7                   \n\t" // 4 complex values from x
+	"vmovups      64(%2,%0,8), %%ymm9                   \n\t" // 4 complex values from x
+	"vmovups      96(%2,%0,8), %%ymm11                  \n\t" // 4 complex values from x
+
+	"vpermilpd	$0x5 , %%ymm5 , %%ymm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm7 , %%ymm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm9 , %%ymm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm11, %%ymm10 	    \n\t"  // exchange real and imag part
+
+	"vmulpd	        %%ymm5 , %%ymm0 , %%ymm5	    \n\t"
+	"vmulpd	        %%ymm7 , %%ymm0 , %%ymm7	    \n\t"
+	"vmulpd	        %%ymm9 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulpd	        %%ymm11, %%ymm0 , %%ymm11           \n\t"
+
+	"vaddpd	        (%3,%0,8), %%ymm5 , %%ymm5          \n\t"
+	"vaddpd	      32(%3,%0,8), %%ymm7 , %%ymm7          \n\t"
+	"vaddpd	      64(%3,%0,8), %%ymm9 , %%ymm9          \n\t"
+	"vaddpd	      96(%3,%0,8), %%ymm11, %%ymm11         \n\t"
+
+	"vmulpd	        %%ymm4 , %%ymm1 , %%ymm4	    \n\t"
+	"vmulpd	        %%ymm6 , %%ymm1 , %%ymm6	    \n\t"
+	"vmulpd	        %%ymm8 , %%ymm1 , %%ymm8	    \n\t"
+	"vmulpd	        %%ymm10, %%ymm1 , %%ymm10           \n\t"
+
+	"vaddpd         %%ymm4 , %%ymm5 , %%ymm5            \n\t"
+	"vaddpd         %%ymm6 , %%ymm7 , %%ymm7            \n\t"
+	"vaddpd         %%ymm8 , %%ymm9 , %%ymm9            \n\t"
+	"vaddpd         %%ymm10, %%ymm11, %%ymm11           \n\t"
+
+	"vmovups	%%ymm5 ,   (%3,%0,8)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%ymm7 , 32(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm9 , 64(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm11, 96(%3,%0,8)		    \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $8 , %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "memory"
+	);
+	return;
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vbroadcastsd		(%4), %%ymm0		    \n\t"  // real part of alpha
+	"vbroadcastsd	       8(%4), %%ymm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulpd		(%5), %%ymm1 , %%ymm1		    \n\t"
+#else
+	"vmulpd		(%5), %%ymm0 , %%ymm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"prefetcht0	512(%2,%0,8)			    \n\t"
+	"prefetcht0	576(%2,%0,8)			    \n\t"
+	"vmovups        (%2,%0,8), %%ymm5                   \n\t" // 4 complex values from x
+	".align 2					    \n\t"
+	"vmovups      32(%2,%0,8), %%ymm7                   \n\t" // 4 complex values from x
+	"vmovups      64(%2,%0,8), %%ymm9                   \n\t" // 4 complex values from x
+	"vmovups      96(%2,%0,8), %%ymm11                  \n\t" // 4 complex values from x
+
+	"vpermilpd	$0x5 , %%ymm5 , %%ymm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm7 , %%ymm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm9 , %%ymm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x5 , %%ymm11, %%ymm10 	    \n\t"  // exchange real and imag part
+
+	"vmulpd	        %%ymm5 , %%ymm0 , %%ymm5	    \n\t"
+	"vmulpd	        %%ymm7 , %%ymm0 , %%ymm7	    \n\t"
+	"vmulpd	        %%ymm9 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulpd	        %%ymm11, %%ymm0 , %%ymm11           \n\t"
+
+	"prefetcht0	512(%3,%0,8)			    \n\t"
+	"prefetcht0	576(%3,%0,8)			    \n\t"
+	"vaddpd	        (%3,%0,8), %%ymm5 , %%ymm5          \n\t"
+	"vaddpd	      32(%3,%0,8), %%ymm7 , %%ymm7          \n\t"
+	"vaddpd	      64(%3,%0,8), %%ymm9 , %%ymm9          \n\t"
+	"vaddpd	      96(%3,%0,8), %%ymm11, %%ymm11         \n\t"
+
+	"vmulpd	        %%ymm4 , %%ymm1 , %%ymm4	    \n\t"
+	"vmulpd	        %%ymm6 , %%ymm1 , %%ymm6	    \n\t"
+	"vmulpd	        %%ymm8 , %%ymm1 , %%ymm8	    \n\t"
+	"vmulpd	        %%ymm10, %%ymm1 , %%ymm10           \n\t"
+
+	"vaddpd         %%ymm4 , %%ymm5 , %%ymm5            \n\t"
+	"vaddpd         %%ymm6 , %%ymm7 , %%ymm7            \n\t"
+	"vaddpd         %%ymm8 , %%ymm9 , %%ymm9            \n\t"
+	"vaddpd         %%ymm10, %%ymm11, %%ymm11           \n\t"
+
+	"vmovups	%%ymm5 ,   (%3,%0,8)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%ymm7 , 32(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm9 , 64(%3,%0,8)		    \n\t"
+	"vmovups	%%ymm11, 96(%3,%0,8)		    \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $8 , %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "memory"
+	);
+
+
+
+
+} 
+
diff --git a/kernel/x86_64/zaxpy_microk_steamroller-2.c b/kernel/x86_64/zaxpy_microk_steamroller-2.c
new file mode 100644
index 000000000..728d09213
--- /dev/null
+++ b/kernel/x86_64/zaxpy_microk_steamroller-2.c
@@ -0,0 +1,200 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_4 1
+static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *alpha) __attribute__ ((noinline));
+
+static void zaxpy_kernel_4( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *alpha)
+{
+
+#if !defined(CONJ)
+	FLOAT mvec[2] = { -1.0, 1.0 };
+#else
+	FLOAT mvec[2] = { 1.0, -1.0 };
+#endif
+
+	BLASLONG register i = 0;
+
+  if ( n < 640 )
+  {
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vmovddup		(%4), %%xmm0		    \n\t"  // real part of alpha
+	"vmovddup	       8(%4), %%xmm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulpd		(%5), %%xmm1 , %%xmm1		    \n\t"
+#else
+	"vmulpd		(%5), %%xmm0 , %%xmm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"vmovups        (%2,%0,8), %%xmm5                   \n\t" // 2 complex values from x
+	".align 2					    \n\t"
+	"vmovups      16(%2,%0,8), %%xmm7                   \n\t" // 2 complex values from x
+	"vmovups      32(%2,%0,8), %%xmm9                   \n\t" // 2 complex values from x
+	"vmovups      48(%2,%0,8), %%xmm11                  \n\t" // 2 complex values from x
+
+	"vmovups      64(%2,%0,8), %%xmm12                  \n\t" // 2 complex values from x
+	"vmovups      80(%2,%0,8), %%xmm13                  \n\t" // 2 complex values from x
+	"vmovups      96(%2,%0,8), %%xmm14                  \n\t" // 2 complex values from x
+	"vmovups     112(%2,%0,8), %%xmm15                  \n\t" // 2 complex values from x
+
+	"vpermilpd	$0x1 , %%xmm5 , %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm7 , %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm9 , %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm11, %%xmm10 	    \n\t"  // exchange real and imag part
+
+	"vfmadd213pd    (%3,%0,8), %%xmm0 , %%xmm5          \n\t"
+	".align 2					    \n\t"
+	"vfmadd213pd  16(%3,%0,8), %%xmm0 , %%xmm7          \n\t"
+	"vfmadd213pd  32(%3,%0,8), %%xmm0 , %%xmm9          \n\t"
+	"vfmadd213pd  48(%3,%0,8), %%xmm0 , %%xmm11         \n\t"
+
+	"vfmadd231pd	%%xmm1 , %%xmm4 , %%xmm5   \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm6 , %%xmm7   \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm8 , %%xmm9   \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm10, %%xmm11  \n\t"
+
+	"vpermilpd	$0x1 , %%xmm12, %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm13, %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm14, %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm15, %%xmm10 	    \n\t"  // exchange real and imag part
+
+	"vfmadd213pd  64(%3,%0,8), %%xmm0 , %%xmm12   \n\t"
+	"vfmadd213pd  80(%3,%0,8), %%xmm0 , %%xmm13   \n\t"
+	"vfmadd213pd  96(%3,%0,8), %%xmm0 , %%xmm14   \n\t"
+	"vfmadd213pd 112(%3,%0,8), %%xmm0 , %%xmm15   \n\t"
+
+	"vfmadd231pd	%%xmm1 , %%xmm4 , %%xmm12  \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm6 , %%xmm13  \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm8 , %%xmm14  \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm10, %%xmm15  \n\t"
+
+	"vmovups	%%xmm5 ,   (%3,%0,8)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%xmm7 , 16(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm9 , 32(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm12, 64(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm13, 80(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm14, 96(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm15,112(%3,%0,8)		    \n\t"
+
+	"addq		$16, %0	  	 	             \n\t"
+	"subq	        $8 , %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					    \n\t"
+	"vmovddup		(%4), %%xmm0		    \n\t"  // real part of alpha
+	"vmovddup	       8(%4), %%xmm1		    \n\t"  // imag part of alpha
+#if !defined(CONJ)
+	"vmulpd		(%5), %%xmm1 , %%xmm1		    \n\t"
+#else
+	"vmulpd		(%5), %%xmm0 , %%xmm0		    \n\t"
+#endif
+
+	".align 16				            \n\t"
+	"1:				            \n\t"
+
+	"prefetcht0	512(%2,%0,8)			    \n\t"
+	"vmovups        (%2,%0,8), %%xmm5                   \n\t" // 2 complex values from x
+	".align 2					    \n\t"
+	"vmovups      16(%2,%0,8), %%xmm7                   \n\t" // 2 complex values from x
+	"vmovups      32(%2,%0,8), %%xmm9                   \n\t" // 2 complex values from x
+	"vmovups      48(%2,%0,8), %%xmm11                  \n\t" // 2 complex values from x
+
+	"vpermilpd	$0x1 , %%xmm5 , %%xmm4 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm7 , %%xmm6 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm9 , %%xmm8 	    \n\t"  // exchange real and imag part
+	"vpermilpd	$0x1 , %%xmm11, %%xmm10 	    \n\t"  // exchange real and imag part
+
+	"prefetcht0	512(%3,%0,8)			    \n\t"
+	"vfmadd213pd    (%3,%0,8), %%xmm0 , %%xmm5          \n\t"
+	".align 2					    \n\t"
+	"vfmadd213pd  16(%3,%0,8), %%xmm0 , %%xmm7          \n\t"
+	"vfmadd213pd  32(%3,%0,8), %%xmm0 , %%xmm9          \n\t"
+	"vfmadd213pd  48(%3,%0,8), %%xmm0 , %%xmm11         \n\t"
+
+	"vfmadd231pd	%%xmm1 , %%xmm4 , %%xmm5   \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm6 , %%xmm7   \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm8 , %%xmm9   \n\t"
+	"vfmadd231pd	%%xmm1 , %%xmm10, %%xmm11  \n\t"
+
+	"vmovups	%%xmm5 ,   (%3,%0,8)		    \n\t"
+	".align 2					    \n\t"
+	"vmovups	%%xmm7 , 16(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm9 , 32(%3,%0,8)		    \n\t"
+	"vmovups	%%xmm11, 48(%3,%0,8)		    \n\t"
+
+	"addq		$8 , %0	  	 	             \n\t"
+	"subq	        $4 , %1			             \n\t"		
+	"jnz		1b		             \n\t"
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (alpha),  // 4
+          "r" (mvec)    // 5
+	: "cc", 
+	  "%xmm0", "%xmm1",
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "memory"
+	);
+
+
+} 
+
diff --git a/kernel/x86_64/zdot.c b/kernel/x86_64/zdot.c
new file mode 100644
index 000000000..eee00fd9f
--- /dev/null
+++ b/kernel/x86_64/zdot.c
@@ -0,0 +1,166 @@
+/***************************************************************************
+Copyright (c) 2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+
+#include "common.h"
+#include <complex.h>
+
+
+#if defined(BULLDOZER) 
+#include "zdot_microk_bulldozer-2.c"
+#elif defined(STEAMROLLER) || defined(PILEDRIVER)
+#include "zdot_microk_steamroller-2.c"
+#elif defined(HASWELL)
+#include "zdot_microk_haswell-2.c"
+#elif defined(SANDYBRIDGE)
+#include "zdot_microk_sandy-2.c"
+#endif
+
+
+#ifndef HAVE_KERNEL_8
+
+static void zdot_kernel_8(BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *d) __attribute__ ((noinline));
+
+static void zdot_kernel_8(BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *d)
+{
+	BLASLONG register i = 0;
+	FLOAT dot[4] = { 0.0, 0.0, 0.0, 0.0 };
+	BLASLONG j=0;
+
+	while( i < n )
+        {
+
+            dot[0] += x[j]   * y[j]   ;
+            dot[1] += x[j+1] * y[j+1] ;
+            dot[2] += x[j]   * y[j+1] ;
+            dot[3] += x[j+1] * y[j]   ;
+
+            dot[0] += x[j+2] * y[j+2] ;
+            dot[1] += x[j+3] * y[j+3] ;
+            dot[2] += x[j+2] * y[j+3] ;
+            dot[3] += x[j+3] * y[j+2] ;
+
+            dot[0] += x[j+4] * y[j+4] ;
+            dot[1] += x[j+5] * y[j+5] ;
+            dot[2] += x[j+4] * y[j+5] ;
+            dot[3] += x[j+5] * y[j+4] ;
+
+            dot[0] += x[j+6] * y[j+6] ;
+            dot[1] += x[j+7] * y[j+7] ;
+            dot[2] += x[j+6] * y[j+7] ;
+            dot[3] += x[j+7] * y[j+6] ;
+
+	    j+=8;
+            i+=4;
+
+        }
+	d[0] = dot[0];
+	d[1] = dot[1];
+	d[2] = dot[2];
+	d[3] = dot[3];
+
+}
+
+#endif
+
+FLOAT _Complex CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
+{
+	BLASLONG i;
+	BLASLONG ix,iy;
+	FLOAT _Complex result;
+	FLOAT  dot[4] = { 0.0, 0.0, 0.0 , 0.0 } ; 
+
+	if ( n <= 0 ) 
+	{
+	        __real__ result = 0.0 ;
+        	__imag__ result = 0.0 ;
+		return(result);
+
+	}
+
+	if ( (inc_x == 1) && (inc_y == 1) )
+	{
+
+		BLASLONG n1 = n & -8;
+
+		if ( n1 )
+			zdot_kernel_8(n1, x, y , dot );
+
+		i = n1;
+		BLASLONG j = i * 2;
+
+		while( i < n )
+		{
+
+			dot[0] += x[j]   * y[j]   ;
+			dot[1] += x[j+1] * y[j+1] ;
+			dot[2] += x[j]   * y[j+1] ;
+			dot[3] += x[j+1] * y[j]   ;
+
+			j+=2;
+			i++ ;
+
+		}
+
+
+	}
+	else
+	{
+		i=0;
+		ix=0;
+		iy=0;
+		inc_x <<= 1;
+		inc_y <<= 1;
+		while(i < n)
+		{
+
+			dot[0] += x[ix]   * y[iy]   ;
+			dot[1] += x[ix+1] * y[iy+1] ;
+			dot[2] += x[ix]   * y[iy+1] ;
+			dot[3] += x[ix+1] * y[iy]   ;
+
+			ix  += inc_x ;
+			iy  += inc_y ;
+			i++ ;
+
+		}
+	}
+
+#if !defined(CONJ)
+	__real__ result = dot[0] - dot[1];
+	__imag__ result = dot[2] + dot[3];
+#else
+	__real__ result = dot[0] + dot[1];
+	__imag__ result = dot[2] - dot[3];
+
+#endif
+
+	return(result);
+
+}
+
+
diff --git a/kernel/x86_64/zdot_microk_bulldozer-2.c b/kernel/x86_64/zdot_microk_bulldozer-2.c
new file mode 100644
index 000000000..30a9552d6
--- /dev/null
+++ b/kernel/x86_64/zdot_microk_bulldozer-2.c
@@ -0,0 +1,196 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void zdot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void zdot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n < 768  )
+  {
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%xmm0, %%xmm0, %%xmm0	             \n\t"
+	"vxorpd		%%xmm1, %%xmm1, %%xmm1	             \n\t"
+	"vxorpd		%%xmm2, %%xmm2, %%xmm2	             \n\t"
+	"vxorpd		%%xmm3, %%xmm3, %%xmm3	             \n\t"
+	"vxorpd		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorpd		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorpd		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorpd		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,8), %%xmm8          \n\t"  // 1 * x
+        "vmovups                16(%2,%0,8), %%xmm9          \n\t"  // 1 * x
+
+        "vmovups                  (%3,%0,8), %%xmm12         \n\t"  // 1 * y
+        "vmovups                16(%3,%0,8), %%xmm13         \n\t"  // 1 * y
+
+        "vmovups                32(%2,%0,8), %%xmm10         \n\t"  // 1 * x
+        "vmovups                48(%2,%0,8), %%xmm11         \n\t"  // 1 * x
+
+        "vmovups                32(%3,%0,8), %%xmm14         \n\t"  // 1 * y
+        "vmovups                48(%3,%0,8), %%xmm15         \n\t"  // 1 * y
+
+	"vfmaddpd       %%xmm0, %%xmm8 , %%xmm12, %%xmm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmaddpd       %%xmm1, %%xmm9 , %%xmm13, %%xmm1     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilpd      $0x1 , %%xmm12, %%xmm12               \n\t"
+	"vpermilpd      $0x1 , %%xmm13, %%xmm13               \n\t"
+
+	"vfmaddpd       %%xmm2, %%xmm10, %%xmm14, %%xmm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmaddpd       %%xmm3, %%xmm11, %%xmm15, %%xmm3     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilpd      $0x1 , %%xmm14, %%xmm14               \n\t"
+	"vpermilpd      $0x1 , %%xmm15, %%xmm15               \n\t"
+
+	"vfmaddpd       %%xmm4, %%xmm8 , %%xmm12, %%xmm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$8 , %0	  	 	             \n\t"
+	"vfmaddpd       %%xmm5, %%xmm9 , %%xmm13, %%xmm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmaddpd       %%xmm6, %%xmm10, %%xmm14, %%xmm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $4 , %1			             \n\t"		
+	"vfmaddpd       %%xmm7, %%xmm11, %%xmm15, %%xmm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddpd        %%xmm2, %%xmm3, %%xmm2	\n\t"
+	"vaddpd        %%xmm0, %%xmm2, %%xmm0	\n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+        return;
+
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%xmm0, %%xmm0, %%xmm0	             \n\t"
+	"vxorpd		%%xmm1, %%xmm1, %%xmm1	             \n\t"
+	"vxorpd		%%xmm2, %%xmm2, %%xmm2	             \n\t"
+	"vxorpd		%%xmm3, %%xmm3, %%xmm3	             \n\t"
+	"vxorpd		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorpd		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorpd		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorpd		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16				             \n\t"
+	"1:				             \n\t"
+	"prefetcht0            384(%2,%0,8)		     \n\t"
+        "vmovups                  (%2,%0,8), %%xmm8          \n\t"  // 1 * x
+        "vmovups                16(%2,%0,8), %%xmm9          \n\t"  // 1 * x
+
+	"prefetcht0            384(%3,%0,8)		     \n\t"
+        "vmovups                  (%3,%0,8), %%xmm12         \n\t"  // 1 * y
+        "vmovups                16(%3,%0,8), %%xmm13         \n\t"  // 1 * y
+
+        "vmovups                32(%2,%0,8), %%xmm10         \n\t"  // 1 * x
+        "vmovups                48(%2,%0,8), %%xmm11         \n\t"  // 1 * x
+
+        "vmovups                32(%3,%0,8), %%xmm14         \n\t"  // 1 * y
+        "vmovups                48(%3,%0,8), %%xmm15         \n\t"  // 1 * y
+
+	"vfmaddpd       %%xmm0, %%xmm8 , %%xmm12, %%xmm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmaddpd       %%xmm1, %%xmm9 , %%xmm13, %%xmm1     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilpd      $0x1 , %%xmm12, %%xmm12               \n\t"
+	"vpermilpd      $0x1 , %%xmm13, %%xmm13               \n\t"
+
+	"vfmaddpd       %%xmm2, %%xmm10, %%xmm14, %%xmm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmaddpd       %%xmm3, %%xmm11, %%xmm15, %%xmm3     \n\t"  // x_r * y_r, x_i * y_i
+
+	"vpermilpd      $0x1 , %%xmm14, %%xmm14               \n\t"
+	"vpermilpd      $0x1 , %%xmm15, %%xmm15               \n\t"
+
+	"vfmaddpd       %%xmm4, %%xmm8 , %%xmm12, %%xmm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$8 , %0	  	 	             \n\t"
+	"vfmaddpd       %%xmm5, %%xmm9 , %%xmm13, %%xmm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmaddpd       %%xmm6, %%xmm10, %%xmm14, %%xmm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $4 , %1			             \n\t"		
+	"vfmaddpd       %%xmm7, %%xmm11, %%xmm15, %%xmm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddpd        %%xmm2, %%xmm3, %%xmm2	\n\t"
+	"vaddpd        %%xmm0, %%xmm2, %%xmm0	\n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/zdot_microk_haswell-2.c b/kernel/x86_64/zdot_microk_haswell-2.c
new file mode 100644
index 000000000..810cb4439
--- /dev/null
+++ b/kernel/x86_64/zdot_microk_haswell-2.c
@@ -0,0 +1,210 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void zdot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void zdot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n <=1280 )
+  {
+
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%ymm0, %%ymm0, %%ymm0	             \n\t"
+	"vxorpd		%%ymm1, %%ymm1, %%ymm1	             \n\t"
+	"vxorpd		%%ymm2, %%ymm2, %%ymm2	             \n\t"
+	"vxorpd		%%ymm3, %%ymm3, %%ymm3	             \n\t"
+	"vxorpd		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorpd		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorpd		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorpd		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16			             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,8), %%ymm8          \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%ymm9          \n\t"  // 2 * x
+
+        "vmovups                  (%3,%0,8), %%ymm12         \n\t"  // 2 * y
+        "vmovups                32(%3,%0,8), %%ymm13         \n\t"  // 2 * y
+
+        "vmovups                64(%2,%0,8), %%ymm10         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,8), %%ymm11         \n\t"  // 2 * x
+
+        "vmovups                64(%3,%0,8), %%ymm14         \n\t"  // 2 * y
+        "vmovups                96(%3,%0,8), %%ymm15         \n\t"  // 2 * y
+
+	"vfmadd231pd       %%ymm8 , %%ymm12, %%ymm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231pd       %%ymm9 , %%ymm13, %%ymm1     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermpd      $0xb1 , %%ymm12, %%ymm12               \n\t"
+	"vpermpd      $0xb1 , %%ymm13, %%ymm13               \n\t"
+
+	"vfmadd231pd       %%ymm10, %%ymm14, %%ymm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231pd       %%ymm11, %%ymm15, %%ymm3     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermpd      $0xb1 , %%ymm14, %%ymm14               \n\t"
+	"vpermpd      $0xb1 , %%ymm15, %%ymm15               \n\t"
+
+	"vfmadd231pd       %%ymm8 , %%ymm12, %%ymm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$16 , %0	  	 	             \n\t"
+	"vfmadd231pd       %%ymm9 , %%ymm13, %%ymm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmadd231pd       %%ymm10, %%ymm14, %%ymm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $8 , %1			             \n\t"		
+	"vfmadd231pd       %%ymm11, %%ymm15, %%ymm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%ymm0, %%ymm1, %%ymm0	\n\t"
+	"vaddpd        %%ymm2, %%ymm3, %%ymm2	\n\t"
+	"vaddpd        %%ymm0, %%ymm2, %%ymm0	\n\t"
+
+	"vaddpd        %%ymm4, %%ymm5, %%ymm4	\n\t"
+	"vaddpd        %%ymm6, %%ymm7, %%ymm6	\n\t"
+	"vaddpd        %%ymm4, %%ymm6, %%ymm4	\n\t"
+
+	"vextractf128 $1 , %%ymm0 , %%xmm1	\n\t"
+	"vextractf128 $1 , %%ymm4 , %%xmm5	\n\t"
+
+	"vaddpd        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+  }
+
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%ymm0, %%ymm0, %%ymm0	             \n\t"
+	"vxorpd		%%ymm1, %%ymm1, %%ymm1	             \n\t"
+	"vxorpd		%%ymm2, %%ymm2, %%ymm2	             \n\t"
+	"vxorpd		%%ymm3, %%ymm3, %%ymm3	             \n\t"
+	"vxorpd		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorpd		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorpd		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorpd		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16			             \n\t"
+	"1:				             \n\t"
+	"prefetcht0	512(%2,%0,8)		     \n\t"
+        "vmovups                  (%2,%0,8), %%ymm8          \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%ymm9          \n\t"  // 2 * x
+
+	"prefetcht0	512(%3,%0,8)		     \n\t"
+        "vmovups                  (%3,%0,8), %%ymm12         \n\t"  // 2 * y
+        "vmovups                32(%3,%0,8), %%ymm13         \n\t"  // 2 * y
+
+	"prefetcht0	576(%2,%0,8)		     \n\t"
+        "vmovups                64(%2,%0,8), %%ymm10         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,8), %%ymm11         \n\t"  // 2 * x
+
+	"prefetcht0	576(%3,%0,8)		     \n\t"
+        "vmovups                64(%3,%0,8), %%ymm14         \n\t"  // 2 * y
+        "vmovups                96(%3,%0,8), %%ymm15         \n\t"  // 2 * y
+
+	"vfmadd231pd       %%ymm8 , %%ymm12, %%ymm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231pd       %%ymm9 , %%ymm13, %%ymm1     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermpd      $0xb1 , %%ymm12, %%ymm12               \n\t"
+	"vpermpd      $0xb1 , %%ymm13, %%ymm13               \n\t"
+
+	"vfmadd231pd       %%ymm10, %%ymm14, %%ymm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231pd       %%ymm11, %%ymm15, %%ymm3     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermpd      $0xb1 , %%ymm14, %%ymm14               \n\t"
+	"vpermpd      $0xb1 , %%ymm15, %%ymm15               \n\t"
+
+	"vfmadd231pd       %%ymm8 , %%ymm12, %%ymm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$16 , %0	  	 	             \n\t"
+	"vfmadd231pd       %%ymm9 , %%ymm13, %%ymm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmadd231pd       %%ymm10, %%ymm14, %%ymm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $8 , %1			             \n\t"		
+	"vfmadd231pd       %%ymm11, %%ymm15, %%ymm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%ymm0, %%ymm1, %%ymm0	\n\t"
+	"vaddpd        %%ymm2, %%ymm3, %%ymm2	\n\t"
+	"vaddpd        %%ymm0, %%ymm2, %%ymm0	\n\t"
+
+	"vaddpd        %%ymm4, %%ymm5, %%ymm4	\n\t"
+	"vaddpd        %%ymm6, %%ymm7, %%ymm6	\n\t"
+	"vaddpd        %%ymm4, %%ymm6, %%ymm4	\n\t"
+
+	"vextractf128 $1 , %%ymm0 , %%xmm1	\n\t"
+	"vextractf128 $1 , %%ymm4 , %%xmm5	\n\t"
+
+	"vaddpd        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+
+
+
+} 
+
+
diff --git a/kernel/x86_64/zdot_microk_sandy-2.c b/kernel/x86_64/zdot_microk_sandy-2.c
new file mode 100644
index 000000000..fd06612e6
--- /dev/null
+++ b/kernel/x86_64/zdot_microk_sandy-2.c
@@ -0,0 +1,222 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void zdot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void zdot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+if ( n < 1280 )
+{
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%ymm0, %%ymm0, %%ymm0	             \n\t"
+	"vxorpd		%%ymm1, %%ymm1, %%ymm1	             \n\t"
+	"vxorpd		%%ymm2, %%ymm2, %%ymm2	             \n\t"
+	"vxorpd		%%ymm3, %%ymm3, %%ymm3	             \n\t"
+	"vxorpd		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorpd		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorpd		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorpd		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16			             \n\t"
+	"1:				             \n\t"
+        "vmovups                  (%2,%0,8), %%ymm8          \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%ymm9          \n\t"  // 2 * x
+
+        "vmovups                  (%3,%0,8), %%ymm12         \n\t"  // 2 * y
+        "vmovups                32(%3,%0,8), %%ymm13         \n\t"  // 2 * y
+
+        "vmovups                64(%3,%0,8), %%ymm14         \n\t"  // 2 * y
+        "vmovups                96(%3,%0,8), %%ymm15         \n\t"  // 2 * y
+
+	"vmulpd		   %%ymm8 , %%ymm12, %%ymm10    \n\t"
+	"vmulpd		   %%ymm9 , %%ymm13, %%ymm11    \n\t"
+	"vpermilpd      $0x5 , %%ymm12, %%ymm12               \n\t"
+	"vpermilpd      $0x5 , %%ymm13, %%ymm13               \n\t"
+	"vaddpd		   %%ymm0 , %%ymm10, %%ymm0	\n\t"
+	"vaddpd		   %%ymm1 , %%ymm11, %%ymm1	\n\t"
+	"vmulpd		   %%ymm8 , %%ymm12, %%ymm10    \n\t"
+	"vmulpd		   %%ymm9 , %%ymm13, %%ymm11    \n\t"
+        "vmovups                64(%2,%0,8), %%ymm8         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,8), %%ymm9         \n\t"  // 2 * x
+	"vaddpd		   %%ymm4 , %%ymm10, %%ymm4	\n\t"
+	"vaddpd		   %%ymm5 , %%ymm11, %%ymm5	\n\t"
+
+
+	"vmulpd		   %%ymm8 , %%ymm14, %%ymm10    \n\t"
+	"vmulpd		   %%ymm9 , %%ymm15, %%ymm11    \n\t"
+	"vpermilpd      $0x5 , %%ymm14, %%ymm14               \n\t"
+	"vpermilpd      $0x5 , %%ymm15, %%ymm15               \n\t"
+	"vaddpd		   %%ymm2 , %%ymm10, %%ymm2	\n\t"
+	"vaddpd		   %%ymm3 , %%ymm11, %%ymm3	\n\t"
+	"vmulpd		   %%ymm8 , %%ymm14, %%ymm10    \n\t"
+	"addq		$16 , %0	  	 	\n\t"
+	"vmulpd		   %%ymm9 , %%ymm15, %%ymm11    \n\t"
+	"vaddpd		   %%ymm6 , %%ymm10, %%ymm6	\n\t"
+	"subq	        $8 , %1			        \n\t"		
+	"vaddpd		   %%ymm7 , %%ymm11, %%ymm7	\n\t"
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%ymm0, %%ymm1, %%ymm0	\n\t"
+	"vaddpd        %%ymm2, %%ymm3, %%ymm2	\n\t"
+	"vaddpd        %%ymm0, %%ymm2, %%ymm0	\n\t"
+
+	"vaddpd        %%ymm4, %%ymm5, %%ymm4	\n\t"
+	"vaddpd        %%ymm6, %%ymm7, %%ymm6	\n\t"
+	"vaddpd        %%ymm4, %%ymm6, %%ymm4	\n\t"
+
+	"vextractf128 $1 , %%ymm0 , %%xmm1	\n\t"
+	"vextractf128 $1 , %%ymm4 , %%xmm5	\n\t"
+
+	"vaddpd        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+	return;
+
+  }
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%ymm0, %%ymm0, %%ymm0	             \n\t"
+	"vxorpd		%%ymm1, %%ymm1, %%ymm1	             \n\t"
+	"vxorpd		%%ymm2, %%ymm2, %%ymm2	             \n\t"
+	"vxorpd		%%ymm3, %%ymm3, %%ymm3	             \n\t"
+	"vxorpd		%%ymm4, %%ymm4, %%ymm4	             \n\t"
+	"vxorpd		%%ymm5, %%ymm5, %%ymm5	             \n\t"
+	"vxorpd		%%ymm6, %%ymm6, %%ymm6	             \n\t"
+	"vxorpd		%%ymm7, %%ymm7, %%ymm7	             \n\t"
+
+	".align 16			             \n\t"
+	"1:				             \n\t"
+	"prefetcht0	512(%2,%0,8)		     \n\t"
+        "vmovups                  (%2,%0,8), %%ymm8          \n\t"  // 2 * x
+        "vmovups                32(%2,%0,8), %%ymm9          \n\t"  // 2 * x
+
+	"prefetcht0	512(%3,%0,8)		     \n\t"
+        "vmovups                  (%3,%0,8), %%ymm12         \n\t"  // 2 * y
+        "vmovups                32(%3,%0,8), %%ymm13         \n\t"  // 2 * y
+
+        "vmovups                64(%3,%0,8), %%ymm14         \n\t"  // 2 * y
+        "vmovups                96(%3,%0,8), %%ymm15         \n\t"  // 2 * y
+
+	"prefetcht0	576(%3,%0,8)		     \n\t"
+	"vmulpd		   %%ymm8 , %%ymm12, %%ymm10    \n\t"
+	"vmulpd		   %%ymm9 , %%ymm13, %%ymm11    \n\t"
+	"prefetcht0	576(%2,%0,8)		     \n\t"
+	"vpermilpd      $0x5 , %%ymm12, %%ymm12               \n\t"
+	"vpermilpd      $0x5 , %%ymm13, %%ymm13               \n\t"
+	"vaddpd		   %%ymm0 , %%ymm10, %%ymm0	\n\t"
+	"vaddpd		   %%ymm1 , %%ymm11, %%ymm1	\n\t"
+	"vmulpd		   %%ymm8 , %%ymm12, %%ymm10    \n\t"
+	"vmulpd		   %%ymm9 , %%ymm13, %%ymm11    \n\t"
+        "vmovups                64(%2,%0,8), %%ymm8         \n\t"  // 2 * x
+        "vmovups                96(%2,%0,8), %%ymm9         \n\t"  // 2 * x
+	"vaddpd		   %%ymm4 , %%ymm10, %%ymm4	\n\t"
+	"vaddpd		   %%ymm5 , %%ymm11, %%ymm5	\n\t"
+
+
+	"vmulpd		   %%ymm8 , %%ymm14, %%ymm10    \n\t"
+	"vmulpd		   %%ymm9 , %%ymm15, %%ymm11    \n\t"
+	"vpermilpd      $0x5 , %%ymm14, %%ymm14               \n\t"
+	"vpermilpd      $0x5 , %%ymm15, %%ymm15               \n\t"
+	"vaddpd		   %%ymm2 , %%ymm10, %%ymm2	\n\t"
+	"vaddpd		   %%ymm3 , %%ymm11, %%ymm3	\n\t"
+	"vmulpd		   %%ymm8 , %%ymm14, %%ymm10    \n\t"
+	"addq		$16 , %0	  	 	\n\t"
+	"vmulpd		   %%ymm9 , %%ymm15, %%ymm11    \n\t"
+	"vaddpd		   %%ymm6 , %%ymm10, %%ymm6	\n\t"
+	"subq	        $8 , %1			        \n\t"		
+	"vaddpd		   %%ymm7 , %%ymm11, %%ymm7	\n\t"
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%ymm0, %%ymm1, %%ymm0	\n\t"
+	"vaddpd        %%ymm2, %%ymm3, %%ymm2	\n\t"
+	"vaddpd        %%ymm0, %%ymm2, %%ymm0	\n\t"
+
+	"vaddpd        %%ymm4, %%ymm5, %%ymm4	\n\t"
+	"vaddpd        %%ymm6, %%ymm7, %%ymm6	\n\t"
+	"vaddpd        %%ymm4, %%ymm6, %%ymm4	\n\t"
+
+	"vextractf128 $1 , %%ymm0 , %%xmm1	\n\t"
+	"vextractf128 $1 , %%ymm4 , %%xmm5	\n\t"
+
+	"vaddpd        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+
+
+} 
+
+
diff --git a/kernel/x86_64/zdot_microk_steamroller-2.c b/kernel/x86_64/zdot_microk_steamroller-2.c
new file mode 100644
index 000000000..325f74ae3
--- /dev/null
+++ b/kernel/x86_64/zdot_microk_steamroller-2.c
@@ -0,0 +1,193 @@
+/***************************************************************************
+Copyright (c) 2014, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#define HAVE_KERNEL_8 1
+static void zdot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y , FLOAT *dot) __attribute__ ((noinline));
+
+static void zdot_kernel_8( BLASLONG n, FLOAT *x, FLOAT *y, FLOAT *dot)
+{
+
+
+	BLASLONG register i = 0;
+
+  if ( n < 640 )
+  {
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%xmm0, %%xmm0, %%xmm0	             \n\t"
+	"vxorpd		%%xmm1, %%xmm1, %%xmm1	             \n\t"
+	"vxorpd		%%xmm2, %%xmm2, %%xmm2	             \n\t"
+	"vxorpd		%%xmm3, %%xmm3, %%xmm3	             \n\t"
+	"vxorpd		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorpd		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorpd		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorpd		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16			             \n\t"
+	"1:				             \n\t"
+	//"prefetcht0            512(%2,%0,8)		     \n\t"
+        "vmovups                  (%2,%0,8), %%xmm8          \n\t"  // 1 * x
+        "vmovups                16(%2,%0,8), %%xmm9          \n\t"  // 1 * x
+
+	// "prefetcht0            512(%3,%0,8)		     \n\t"
+        "vmovups                  (%3,%0,8), %%xmm12         \n\t"  // 1 * y
+        "vmovups                16(%3,%0,8), %%xmm13         \n\t"  // 1 * y
+
+        "vmovups                32(%2,%0,8), %%xmm10         \n\t"  // 1 * x
+        "vmovups                48(%2,%0,8), %%xmm11         \n\t"  // 1 * x
+
+        "vmovups                32(%3,%0,8), %%xmm14         \n\t"  // 1 * y
+        "vmovups                48(%3,%0,8), %%xmm15         \n\t"  // 1 * y
+
+	"vfmadd231pd       %%xmm8 , %%xmm12, %%xmm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231pd       %%xmm9 , %%xmm13, %%xmm1     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermilpd      $0x1 , %%xmm13, %%xmm13               \n\t"
+	"vpermilpd      $0x1 , %%xmm12, %%xmm12               \n\t"
+	"vfmadd231pd       %%xmm10, %%xmm14, %%xmm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231pd       %%xmm11, %%xmm15, %%xmm3     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermilpd      $0x1 , %%xmm14, %%xmm14               \n\t"
+	"vpermilpd      $0x1 , %%xmm15, %%xmm15               \n\t"
+
+	"vfmadd231pd       %%xmm8 , %%xmm12, %%xmm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$8 , %0	  	 	             \n\t"
+	"vfmadd231pd       %%xmm9 , %%xmm13, %%xmm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmadd231pd       %%xmm10, %%xmm14, %%xmm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $4 , %1			             \n\t"		
+	"vfmadd231pd       %%xmm11, %%xmm15, %%xmm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddpd        %%xmm2, %%xmm3, %%xmm2	\n\t"
+	"vaddpd        %%xmm0, %%xmm2, %%xmm0	\n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+	return;
+  }
+
+
+	__asm__  __volatile__
+	(
+	"vzeroupper					     \n\t"
+	"vxorpd		%%xmm0, %%xmm0, %%xmm0	             \n\t"
+	"vxorpd		%%xmm1, %%xmm1, %%xmm1	             \n\t"
+	"vxorpd		%%xmm2, %%xmm2, %%xmm2	             \n\t"
+	"vxorpd		%%xmm3, %%xmm3, %%xmm3	             \n\t"
+	"vxorpd		%%xmm4, %%xmm4, %%xmm4	             \n\t"
+	"vxorpd		%%xmm5, %%xmm5, %%xmm5	             \n\t"
+	"vxorpd		%%xmm6, %%xmm6, %%xmm6	             \n\t"
+	"vxorpd		%%xmm7, %%xmm7, %%xmm7	             \n\t"
+
+	".align 16			             \n\t"
+	"1:				             \n\t"
+	"prefetcht0            512(%2,%0,8)		     \n\t"
+        "vmovups                  (%2,%0,8), %%xmm8          \n\t"  // 1 * x
+        "vmovups                16(%2,%0,8), %%xmm9          \n\t"  // 1 * x
+
+	"prefetcht0            512(%3,%0,8)		     \n\t"
+        "vmovups                  (%3,%0,8), %%xmm12         \n\t"  // 1 * y
+        "vmovups                16(%3,%0,8), %%xmm13         \n\t"  // 1 * y
+
+        "vmovups                32(%2,%0,8), %%xmm10         \n\t"  // 1 * x
+        "vmovups                48(%2,%0,8), %%xmm11         \n\t"  // 1 * x
+
+        "vmovups                32(%3,%0,8), %%xmm14         \n\t"  // 1 * y
+        "vmovups                48(%3,%0,8), %%xmm15         \n\t"  // 1 * y
+
+	"vfmadd231pd       %%xmm8 , %%xmm12, %%xmm0     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231pd       %%xmm9 , %%xmm13, %%xmm1     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermilpd      $0x1 , %%xmm13, %%xmm13               \n\t"
+	"vpermilpd      $0x1 , %%xmm12, %%xmm12               \n\t"
+	"vfmadd231pd       %%xmm10, %%xmm14, %%xmm2     \n\t"  // x_r * y_r, x_i * y_i
+	"vfmadd231pd       %%xmm11, %%xmm15, %%xmm3     \n\t"  // x_r * y_r, x_i * y_i
+	"vpermilpd      $0x1 , %%xmm14, %%xmm14               \n\t"
+	"vpermilpd      $0x1 , %%xmm15, %%xmm15               \n\t"
+
+	"vfmadd231pd       %%xmm8 , %%xmm12, %%xmm4     \n\t"  // x_r * y_i, x_i * y_r
+	"addq		$8 , %0	  	 	             \n\t"
+	"vfmadd231pd       %%xmm9 , %%xmm13, %%xmm5     \n\t"  // x_r * y_i, x_i * y_r
+	"vfmadd231pd       %%xmm10, %%xmm14, %%xmm6     \n\t"  // x_r * y_i, x_i * y_r
+	"subq	        $4 , %1			             \n\t"		
+	"vfmadd231pd       %%xmm11, %%xmm15, %%xmm7     \n\t"  // x_r * y_i, x_i * y_r
+
+	"jnz		1b		             \n\t"
+
+	"vaddpd        %%xmm0, %%xmm1, %%xmm0	\n\t"
+	"vaddpd        %%xmm2, %%xmm3, %%xmm2	\n\t"
+	"vaddpd        %%xmm0, %%xmm2, %%xmm0	\n\t"
+
+	"vaddpd        %%xmm4, %%xmm5, %%xmm4	\n\t"
+	"vaddpd        %%xmm6, %%xmm7, %%xmm6	\n\t"
+	"vaddpd        %%xmm4, %%xmm6, %%xmm4	\n\t"
+
+	"vmovups       %%xmm0,    (%4)		\n\t"
+	"vmovups       %%xmm4,  16(%4)		\n\t"
+	"vzeroupper					     \n\t"
+
+	:
+        : 
+          "r" (i),	// 0	
+	  "r" (n),  	// 1
+          "r" (x),      // 2
+          "r" (y),      // 3
+          "r" (dot)     // 4
+	: "cc", 
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11",
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+
+} 
+
+
diff --git a/kernel/x86_64/zscal.c b/kernel/x86_64/zscal.c
new file mode 100644
index 000000000..a96766032
--- /dev/null
+++ b/kernel/x86_64/zscal.c
@@ -0,0 +1,434 @@
+/***************************************************************************
+Copyright (c) 2013 - 2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+#include "common.h"
+
+
+#if defined(HASWELL)
+#include "zscal_microk_haswell-2.c"
+#elif defined(BULLDOZER)  || defined(PILEDRIVER)
+#include "zscal_microk_bulldozer-2.c"
+#elif defined(STEAMROLLER)
+#include "zscal_microk_steamroller-2.c"
+#endif
+
+
+#if !defined(HAVE_KERNEL_8)
+
+static void zscal_kernel_8( BLASLONG n, FLOAT *alpha , FLOAT *x )  __attribute__ ((noinline));
+static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha , FLOAT *x )  __attribute__ ((noinline));
+static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha , FLOAT *x )  __attribute__ ((noinline));
+static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha , FLOAT *x )  __attribute__ ((noinline));
+
+static void zscal_kernel_8( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	FLOAT da_r = alpha[0];
+	FLOAT da_i = alpha[1];
+	FLOAT t0,t1,t2,t3;
+
+	for( i=0; i<n; i+=4 )
+	{
+		t0 = da_r *x[0] - da_i *x[1];	
+		t1 = da_r *x[2] - da_i *x[3];	
+		t2 = da_r *x[4] - da_i *x[5];	
+		t3 = da_r *x[6] - da_i *x[7];	
+
+		x[1] = da_r * x[1] + da_i * x[0];
+		x[3] = da_r * x[3] + da_i * x[2];
+		x[5] = da_r * x[5] + da_i * x[4];
+		x[7] = da_r * x[7] + da_i * x[6];
+		
+		x[0] = t0;
+		x[2] = t1;
+		x[4] = t2;
+		x[6] = t3;
+
+		x+=8;
+	}
+
+}
+
+
+static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	FLOAT da_i = alpha[1];
+	FLOAT t0,t1,t2,t3;
+
+	for( i=0; i<n; i+=4 )
+	{
+		t0 =  - da_i *x[1];	
+		t1 =  - da_i *x[3];	
+		t2 =  - da_i *x[5];	
+		t3 =  - da_i *x[7];	
+
+		x[1] =  da_i * x[0];
+		x[3] =  da_i * x[2];
+		x[5] =  da_i * x[4];
+		x[7] =  da_i * x[6];
+		
+		x[0] = t0;
+		x[2] = t1;
+		x[4] = t2;
+		x[6] = t3;
+
+		x+=8;
+	}
+
+}
+
+
+static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	FLOAT da_r = alpha[0];
+	FLOAT t0,t1,t2,t3;
+
+	for( i=0; i<n; i+=4 )
+	{
+		t0 = da_r *x[0];	
+		t1 = da_r *x[2];	
+		t2 = da_r *x[4];	
+		t3 = da_r *x[6];	
+
+		x[1] = da_r * x[1];
+		x[3] = da_r * x[3];
+		x[5] = da_r * x[5];
+		x[7] = da_r * x[7];
+		
+		x[0] = t0;
+		x[2] = t1;
+		x[4] = t2;
+		x[6] = t3;
+
+		x+=8;
+	}
+
+}
+
+
+
+static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha , FLOAT *x )
+{
+
+	BLASLONG i;
+	for( i=0; i<n; i+=4 )
+	{
+		x[0] = 0.0;	
+		x[1] = 0.0;	
+		x[2] = 0.0;	
+		x[3] = 0.0;	
+		x[4] = 0.0;	
+		x[5] = 0.0;	
+		x[6] = 0.0;	
+		x[7] = 0.0;	
+		x+=8;
+	}
+
+}
+
+#endif
+
+static void zscal_kernel_inc_8(BLASLONG n, FLOAT *alpha, FLOAT *x, BLASLONG inc_x)  __attribute__ ((noinline));
+
+static void zscal_kernel_inc_8(BLASLONG n, FLOAT *alpha, FLOAT *x, BLASLONG inc_x)
+{
+
+	BLASLONG i;
+	BLASLONG inc_x2 = 2 * inc_x;
+	BLASLONG inc_x3 = inc_x2 + inc_x;
+	FLOAT t0,t1,t2,t3;
+	FLOAT da_r = alpha[0];
+	FLOAT da_i = alpha[1];
+
+	for ( i=0; i<n; i+=4 )
+	{
+		t0 = da_r * x[0]         - da_i *x[1];	
+		t1 = da_r * x[inc_x]     - da_i *x[inc_x  + 1];	
+		t2 = da_r * x[inc_x2]    - da_i *x[inc_x2 + 1];	
+		t3 = da_r * x[inc_x3]    - da_i *x[inc_x3 + 1];	
+
+		x[1]               = da_i * x[0]       + da_r * x[1];
+		x[inc_x  +1]       = da_i * x[inc_x]   + da_r * x[inc_x  +1];
+		x[inc_x2 +1]       = da_i * x[inc_x2]  + da_r * x[inc_x2 +1];
+		x[inc_x3 +1]       = da_i * x[inc_x3]  + da_r * x[inc_x3 +1];
+
+		x[0]        = t0;
+		x[inc_x]    = t1;
+		x[inc_x2]   = t2;
+		x[inc_x3]   = t3;
+
+		x+=4*inc_x;
+
+	}
+
+
+}
+
+
+int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da_r, FLOAT da_i, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT *dummy, BLASLONG dummy2)
+{
+	BLASLONG i=0,j=0;
+	FLOAT temp0;
+	FLOAT temp1;
+	FLOAT alpha[2];
+
+	if ( inc_x != 1 )
+	{
+		inc_x <<= 1;
+
+		if ( da_r == 0.0 )
+		{
+
+			BLASLONG n1 = n & -2;
+
+			if ( da_i == 0.0 )
+			{
+
+				while(j < n1)
+				{
+			
+					x[i]=0.0;
+					x[i+1]=0.0;
+					x[i+inc_x]=0.0;
+					x[i+1+inc_x]=0.0;
+					i += 2*inc_x ;
+					j+=2;
+
+				}
+
+				while(j < n)
+				{
+			
+					x[i]=0.0;
+					x[i+1]=0.0;
+					i += inc_x ;
+					j++;
+
+				}
+
+			}
+			else
+			{
+
+				while(j < n1)
+				{
+			
+					temp0        = -da_i * x[i+1];
+					x[i+1]       =  da_i * x[i];
+					x[i]         =  temp0;
+					temp1        = -da_i * x[i+1+inc_x];
+					x[i+1+inc_x] =  da_i * x[i+inc_x];
+					x[i+inc_x]   =  temp1;
+					i += 2*inc_x ;
+					j+=2;
+
+				}
+
+				while(j < n)
+				{
+			
+					temp0        = -da_i * x[i+1];
+					x[i+1]       =  da_i * x[i];
+					x[i]         =  temp0;
+					i += inc_x ;
+					j++;
+
+				}
+
+
+
+			}
+
+		}
+		else
+		{
+
+
+			if ( da_i == 0.0 )
+			{
+				BLASLONG n1 = n & -2;
+
+				while(j < n1)
+				{
+			
+					temp0        =  da_r * x[i];
+					x[i+1]       =  da_r * x[i+1];
+					x[i]         =  temp0;
+					temp1        =  da_r * x[i+inc_x];
+					x[i+1+inc_x] =  da_r * x[i+1+inc_x];
+					x[i+inc_x]   =  temp1;
+					i += 2*inc_x ;
+					j+=2;
+
+				}
+
+				while(j < n)
+				{
+			
+					temp0        =  da_r * x[i];
+					x[i+1]       =  da_r * x[i+1];
+					x[i]         =  temp0;
+					i += inc_x ;
+					j++;
+
+				}
+
+			}
+			else
+			{
+
+				BLASLONG n1 = n & -8;
+				if ( n1 > 0 )
+				{
+					alpha[0] = da_r;
+			                alpha[1] = da_i;
+					zscal_kernel_inc_8(n1, alpha, x, inc_x);
+					j = n1 ;
+					i = n1 * inc_x;
+				}
+
+				while(j < n)
+				{
+
+					temp0        =  da_r * x[i]   - da_i * x[i+1];
+					x[i+1]       =  da_r * x[i+1] + da_i * x[i];
+					x[i]         =  temp0;
+					i += inc_x ;
+					j++;
+
+				}
+
+			}
+
+		}
+
+		return(0);
+	}
+
+
+	BLASLONG n1 = n & -8;
+	if ( n1 > 0 )
+	{
+
+		alpha[0] = da_r;
+		alpha[1] = da_i;
+	
+		if ( da_r == 0.0 )
+			if ( da_i == 0 )
+				zscal_kernel_8_zero(n1 , alpha , x);
+			else
+				zscal_kernel_8_zero_r(n1 , alpha , x);
+		else
+			if ( da_i == 0 )
+				zscal_kernel_8_zero_i(n1 , alpha , x);
+			else
+				zscal_kernel_8(n1 , alpha , x);
+
+		i = n1 << 1;
+		j = n1;
+	}
+
+
+	if ( da_r == 0.0 )
+	{
+
+		if ( da_i == 0.0 )
+		{
+
+			while(j < n)
+			{
+		
+					x[i]=0.0;
+					x[i+1]=0.0;
+					i += 2 ;
+					j++;
+
+			}
+
+		}
+		else
+		{
+
+			while(j < n)
+			{
+			
+				temp0        = -da_i * x[i+1];
+				x[i+1]       =  da_i * x[i];
+				x[i]         =  temp0;
+				i += 2 ;
+				j++;
+
+			}
+
+		}
+
+	}
+	else
+	{
+
+		if ( da_i == 0.0 )
+		{
+
+			while(j < n)
+			{
+			
+					temp0        =  da_r * x[i];
+					x[i+1]       =  da_r * x[i+1];
+					x[i]         =  temp0;
+					i += 2 ;
+					j++;
+
+			}
+
+		}
+		else
+		{
+
+			while(j < n)
+			{
+
+				temp0        =  da_r * x[i]   - da_i * x[i+1];
+				x[i+1]       =  da_r * x[i+1] + da_i * x[i];
+				x[i]         =  temp0;
+				i += 2 ;
+				j++;
+
+			}
+
+		}		
+
+	}
+
+	return(0);
+}
+
+
diff --git a/kernel/x86_64/zscal_microk_bulldozer-2.c b/kernel/x86_64/zscal_microk_bulldozer-2.c
new file mode 100644
index 000000000..28fe73480
--- /dev/null
+++ b/kernel/x86_64/zscal_microk_bulldozer-2.c
@@ -0,0 +1,348 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+
+#define HAVE_KERNEL_8 1
+
+static void zscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vmovddup    		(%2), %%xmm0		    \n\t"  // da_r	
+	"vmovddup              8(%2), %%xmm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%xmm4		    \n\t"
+	"vmovups	-112(%1), %%xmm5		    \n\t"
+	"vmovups	 -96(%1), %%xmm6		    \n\t"
+	"vmovups	 -80(%1), %%xmm7		    \n\t"
+
+	"vpermilpd	$0x01 , %%xmm4, %%xmm12		    \n\t"
+	"vpermilpd	$0x01 , %%xmm5, %%xmm13		    \n\t"
+	"vpermilpd	$0x01 , %%xmm6, %%xmm14 	    \n\t"
+	"vpermilpd	$0x01 , %%xmm7, %%xmm15		    \n\t"
+
+	"subq	        $4 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	"prefetcht0     192(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulpd		%%xmm0, %%xmm4 , %%xmm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	 -64(%1), %%xmm4		    \n\t"
+	"vmulpd		%%xmm0, %%xmm5 , %%xmm9		    \n\t"
+	"vmovups	 -48(%1), %%xmm5		    \n\t"
+	"vmulpd		%%xmm0, %%xmm6 , %%xmm10	    \n\t" 
+	"vmovups	 -32(%1), %%xmm6		    \n\t"
+	"vmulpd		%%xmm0, %%xmm7 , %%xmm11	    \n\t" 
+	"vmovups	 -16(%1), %%xmm7		    \n\t"
+
+	"vmulpd		%%xmm1, %%xmm12, %%xmm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%xmm12 , %%xmm8 , %%xmm8	    \n\t"
+	"vmulpd		%%xmm1, %%xmm13, %%xmm13	    \n\t" 
+	"vaddsubpd	%%xmm13 , %%xmm9 , %%xmm9	    \n\t"
+	"vmulpd		%%xmm1, %%xmm14, %%xmm14	    \n\t" 
+	"vaddsubpd	%%xmm14 , %%xmm10, %%xmm10	    \n\t"
+	"vmulpd		%%xmm1, %%xmm15, %%xmm15	    \n\t" 
+	"vaddsubpd	%%xmm15 , %%xmm11, %%xmm11	    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vpermilpd	$0x01 , %%xmm4, %%xmm12		    \n\t"
+	"vpermilpd	$0x01 , %%xmm5, %%xmm13		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+	"vpermilpd	$0x01 , %%xmm6, %%xmm14 	    \n\t"
+	"vpermilpd	$0x01 , %%xmm7, %%xmm15		    \n\t"
+
+	"addq		$64  ,%1  	 	            \n\t"
+	"subq	        $4 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulpd		%%xmm0, %%xmm4 , %%xmm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulpd		%%xmm0, %%xmm5 , %%xmm9		    \n\t"
+	"vmulpd		%%xmm0, %%xmm6 , %%xmm10	    \n\t" 
+	"vmulpd		%%xmm0, %%xmm7 , %%xmm11	    \n\t" 
+
+	"vmulpd		%%xmm1, %%xmm12, %%xmm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%xmm12 , %%xmm8 , %%xmm8	    \n\t"
+	"vmulpd		%%xmm1, %%xmm13, %%xmm13	    \n\t" 
+	"vaddsubpd	%%xmm13 , %%xmm9 , %%xmm9	    \n\t"
+	"vmulpd		%%xmm1, %%xmm14, %%xmm14	    \n\t" 
+	"vaddsubpd	%%xmm14 , %%xmm10, %%xmm10	    \n\t"
+	"vmulpd		%%xmm1, %%xmm15, %%xmm15	    \n\t" 
+	"vaddsubpd	%%xmm15 , %%xmm11, %%xmm11	    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorpd	           %%xmm0, %%xmm0, %%xmm0	    \n\t"
+	"vmovddup              8(%2), %%xmm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%xmm4		    \n\t"
+	"vmovups	-112(%1), %%xmm5		    \n\t"
+	"vmovups	 -96(%1), %%xmm6		    \n\t"
+	"vmovups	 -80(%1), %%xmm7		    \n\t"
+
+	"vpermilpd	$0x01 , %%xmm4, %%xmm12		    \n\t"
+	"vpermilpd	$0x01 , %%xmm5, %%xmm13		    \n\t"
+	"vpermilpd	$0x01 , %%xmm6, %%xmm14 	    \n\t"
+	"vpermilpd	$0x01 , %%xmm7, %%xmm15		    \n\t"
+
+	"subq	        $4 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	 -64(%1), %%xmm4		    \n\t"
+	"vmovups	 -48(%1), %%xmm5		    \n\t"
+	"vmovups	 -32(%1), %%xmm6		    \n\t"
+	"vmovups	 -16(%1), %%xmm7		    \n\t"
+
+	"vmulpd		%%xmm1, %%xmm12, %%xmm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%xmm12 , %%xmm0 , %%xmm8	    \n\t"
+	"vmulpd		%%xmm1, %%xmm13, %%xmm13	    \n\t" 
+	"vaddsubpd	%%xmm13 , %%xmm0 , %%xmm9	    \n\t"
+	"vmulpd		%%xmm1, %%xmm14, %%xmm14	    \n\t" 
+	"vaddsubpd	%%xmm14 , %%xmm0 , %%xmm10	    \n\t"
+	"vmulpd		%%xmm1, %%xmm15, %%xmm15	    \n\t" 
+	"vaddsubpd	%%xmm15 , %%xmm0 , %%xmm11	    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vpermilpd	$0x01 , %%xmm4, %%xmm12		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vpermilpd	$0x01 , %%xmm5, %%xmm13		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vpermilpd	$0x01 , %%xmm6, %%xmm14 	    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+	"vpermilpd	$0x01 , %%xmm7, %%xmm15		    \n\t"
+
+	"addq		$64  ,%1  	 	            \n\t"
+	"subq	        $4 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"vmulpd		%%xmm1, %%xmm12, %%xmm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%xmm12 , %%xmm0 , %%xmm8	    \n\t"
+	"vmulpd		%%xmm1, %%xmm13, %%xmm13	    \n\t" 
+	"vaddsubpd	%%xmm13 , %%xmm0 , %%xmm9	    \n\t"
+	"vmulpd		%%xmm1, %%xmm14, %%xmm14	    \n\t" 
+	"vaddsubpd	%%xmm14 , %%xmm0 , %%xmm10	    \n\t"
+	"vmulpd		%%xmm1, %%xmm15, %%xmm15	    \n\t" 
+	"vaddsubpd	%%xmm15 , %%xmm0 , %%xmm11	    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
+static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vmovddup    		(%2), %%xmm0		    \n\t"  // da_r	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%xmm4		    \n\t"
+	"vmovups	-112(%1), %%xmm5		    \n\t"
+	"vmovups	 -96(%1), %%xmm6		    \n\t"
+	"vmovups	 -80(%1), %%xmm7		    \n\t"
+
+
+	"subq	        $4 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulpd		%%xmm0, %%xmm4 , %%xmm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	 -64(%1), %%xmm4		    \n\t"
+	"vmulpd		%%xmm0, %%xmm5 , %%xmm9		    \n\t"
+	"vmovups	 -48(%1), %%xmm5		    \n\t"
+	"vmulpd		%%xmm0, %%xmm6 , %%xmm10	    \n\t" 
+	"vmovups	 -32(%1), %%xmm6		    \n\t"
+	"vmulpd		%%xmm0, %%xmm7 , %%xmm11	    \n\t" 
+	"vmovups	 -16(%1), %%xmm7		    \n\t"
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+
+	"addq		$64  ,%1  	 	            \n\t"
+	"subq	        $4 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulpd		%%xmm0, %%xmm4 , %%xmm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulpd		%%xmm0, %%xmm5 , %%xmm9		    \n\t"
+	"vmulpd		%%xmm0, %%xmm6 , %%xmm10	    \n\t" 
+	"vmulpd		%%xmm0, %%xmm7 , %%xmm11	    \n\t" 
+
+	"vmovups	%%xmm8 , -128(%1)		    \n\t"
+	"vmovups	%%xmm9 , -112(%1)		    \n\t"
+	"vmovups	%%xmm10,  -96(%1)		    \n\t"
+	"vmovups	%%xmm11,  -80(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorpd	           %%xmm0, %%xmm0, %%xmm0	    \n\t"
+
+	"addq	$128, %1				    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	%%xmm0 , -128(%1)		    \n\t"
+	"vmovups	%%xmm0 , -112(%1)		    \n\t"
+	"vmovups	%%xmm0 ,  -96(%1)		    \n\t"
+	"vmovups	%%xmm0 ,  -80(%1)		    \n\t"
+
+	"addq		$64  ,%1  	 	            \n\t"
+	"subq	        $4 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
diff --git a/kernel/x86_64/zscal_microk_haswell-2.c b/kernel/x86_64/zscal_microk_haswell-2.c
new file mode 100644
index 000000000..a93308ec4
--- /dev/null
+++ b/kernel/x86_64/zscal_microk_haswell-2.c
@@ -0,0 +1,348 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+
+#define HAVE_KERNEL_8 1
+
+static void zscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastsd		(%2), %%ymm0		    \n\t"  // da_r	
+	"vbroadcastsd          8(%2), %%ymm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+	"vpermilpd	$0x05 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilpd	$0x05 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilpd	$0x05 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilpd	$0x05 , %%ymm7, %%ymm15		    \n\t"
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulpd		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmulpd		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmulpd		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmulpd		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmulpd		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%ymm12 , %%ymm8 , %%ymm8	    \n\t"
+	"vmulpd		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubpd	%%ymm13 , %%ymm9 , %%ymm9	    \n\t"
+	"vmulpd		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubpd	%%ymm14 , %%ymm10, %%ymm10	    \n\t"
+	"vmulpd		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubpd	%%ymm15 , %%ymm11, %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm4, %%ymm12		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm5, %%ymm13		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm6, %%ymm14 	    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm7, %%ymm15		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulpd		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulpd		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmulpd		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmulpd		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+
+	"vmulpd		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%ymm12 , %%ymm8 , %%ymm8	    \n\t"
+	"vmulpd		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubpd	%%ymm13 , %%ymm9 , %%ymm9	    \n\t"
+	"vmulpd		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubpd	%%ymm14 , %%ymm10, %%ymm10	    \n\t"
+	"vmulpd		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubpd	%%ymm15 , %%ymm11, %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorpd	           %%ymm0, %%ymm0, %%ymm0	    \n\t"
+	"vbroadcastsd          8(%2), %%ymm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+	"vpermilpd	$0x05 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilpd	$0x05 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilpd	$0x05 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilpd	$0x05 , %%ymm7, %%ymm15		    \n\t"
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmulpd		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%ymm12 , %%ymm0 , %%ymm8	    \n\t"
+	"vmulpd		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubpd	%%ymm13 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulpd		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubpd	%%ymm14 , %%ymm0 , %%ymm10	    \n\t"
+	"vmulpd		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubpd	%%ymm15 , %%ymm0 , %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm4, %%ymm12		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm5, %%ymm13		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm6, %%ymm14 	    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm7, %%ymm15		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"vmulpd		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%ymm12 , %%ymm0 , %%ymm8	    \n\t"
+	"vmulpd		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubpd	%%ymm13 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulpd		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubpd	%%ymm14 , %%ymm0 , %%ymm10	    \n\t"
+	"vmulpd		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubpd	%%ymm15 , %%ymm0 , %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
+static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastsd		(%2), %%ymm0		    \n\t"  // da_r	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulpd		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmulpd		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmulpd		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmulpd		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulpd		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulpd		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmulpd		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmulpd		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorpd	           %%ymm0, %%ymm0, %%ymm0	    \n\t"
+
+	"addq	$128, %1				    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	%%ymm0 , -128(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -64(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -32(%1)		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
diff --git a/kernel/x86_64/zscal_microk_steamroller-2.c b/kernel/x86_64/zscal_microk_steamroller-2.c
new file mode 100644
index 000000000..d611bf570
--- /dev/null
+++ b/kernel/x86_64/zscal_microk_steamroller-2.c
@@ -0,0 +1,349 @@
+/***************************************************************************
+Copyright (c) 2014-2015, The OpenBLAS Project
+All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+notice, this list of conditions and the following disclaimer in
+the documentation and/or other materials provided with the
+distribution.
+3. Neither the name of the OpenBLAS project nor the names of
+its contributors may be used to endorse or promote products
+derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
+USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+*****************************************************************************/
+
+
+#define HAVE_KERNEL_8 1
+
+static void zscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastsd		(%2), %%ymm0		    \n\t"  // da_r	
+	"vbroadcastsd          8(%2), %%ymm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+	"vpermilpd	$0x05 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilpd	$0x05 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilpd	$0x05 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilpd	$0x05 , %%ymm7, %%ymm15		    \n\t"
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	"prefetcht0     320(%1)				    \n\t"
+	"prefetcht0     384(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulpd		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmulpd		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmulpd		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmulpd		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmulpd		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%ymm12 , %%ymm8 , %%ymm8	    \n\t"
+	"vmulpd		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubpd	%%ymm13 , %%ymm9 , %%ymm9	    \n\t"
+	"vmulpd		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubpd	%%ymm14 , %%ymm10, %%ymm10	    \n\t"
+	"vmulpd		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubpd	%%ymm15 , %%ymm11, %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm4, %%ymm12		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm5, %%ymm13		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm6, %%ymm14 	    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm7, %%ymm15		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulpd		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulpd		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmulpd		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmulpd		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+
+	"vmulpd		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%ymm12 , %%ymm8 , %%ymm8	    \n\t"
+	"vmulpd		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubpd	%%ymm13 , %%ymm9 , %%ymm9	    \n\t"
+	"vmulpd		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubpd	%%ymm14 , %%ymm10, %%ymm10	    \n\t"
+	"vmulpd		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubpd	%%ymm15 , %%ymm11, %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorpd	           %%ymm0, %%ymm0, %%ymm0	    \n\t"
+	"vbroadcastsd          8(%2), %%ymm1		    \n\t"  // da_i 	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+	"vpermilpd	$0x05 , %%ymm4, %%ymm12		    \n\t"
+	"vpermilpd	$0x05 , %%ymm5, %%ymm13		    \n\t"
+	"vpermilpd	$0x05 , %%ymm6, %%ymm14 	    \n\t"
+	"vpermilpd	$0x05 , %%ymm7, %%ymm15		    \n\t"
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmulpd		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%ymm12 , %%ymm0 , %%ymm8	    \n\t"
+	"vmulpd		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubpd	%%ymm13 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulpd		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubpd	%%ymm14 , %%ymm0 , %%ymm10	    \n\t"
+	"vmulpd		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubpd	%%ymm15 , %%ymm0 , %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm4, %%ymm12		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm5, %%ymm13		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm6, %%ymm14 	    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+	"vpermilpd	$0x05 , %%ymm7, %%ymm15		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+	"vmulpd		%%ymm1, %%ymm12, %%ymm12	    \n\t" // da_i*x1 , da_i *x0
+	"vaddsubpd	%%ymm12 , %%ymm0 , %%ymm8	    \n\t"
+	"vmulpd		%%ymm1, %%ymm13, %%ymm13	    \n\t" 
+	"vaddsubpd	%%ymm13 , %%ymm0 , %%ymm9	    \n\t"
+	"vmulpd		%%ymm1, %%ymm14, %%ymm14	    \n\t" 
+	"vaddsubpd	%%ymm14 , %%ymm0 , %%ymm10	    \n\t"
+	"vmulpd		%%ymm1, %%ymm15, %%ymm15	    \n\t" 
+	"vaddsubpd	%%ymm15 , %%ymm0 , %%ymm11	    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
+static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vbroadcastsd		(%2), %%ymm0		    \n\t"  // da_r	
+
+	"addq	$128, %1				    \n\t"
+
+	"vmovups	-128(%1), %%ymm4		    \n\t"
+	"vmovups	 -96(%1), %%ymm5		    \n\t"
+	"vmovups	 -64(%1), %%ymm6		    \n\t"
+	"vmovups	 -32(%1), %%ymm7		    \n\t"
+
+
+	"subq	        $8 , %0			            \n\t"		
+	"jz	2f					    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmulpd		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmovups	   0(%1), %%ymm4		    \n\t"
+	"vmulpd		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmovups	  32(%1), %%ymm5		    \n\t"
+	"vmulpd		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmovups	  64(%1), %%ymm6		    \n\t"
+	"vmulpd		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+	"vmovups	  96(%1), %%ymm7		    \n\t"
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"2:				            	    \n\t"
+
+
+	"vmulpd		%%ymm0, %%ymm4 , %%ymm8		    \n\t" // da_r*x0 , da_r *x1
+	"vmulpd		%%ymm0, %%ymm5 , %%ymm9		    \n\t"
+	"vmulpd		%%ymm0, %%ymm6 , %%ymm10	    \n\t" 
+	"vmulpd		%%ymm0, %%ymm7 , %%ymm11	    \n\t" 
+
+	"vmovups	%%ymm8 , -128(%1)		    \n\t"
+	"vmovups	%%ymm9 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm10,  -64(%1)		    \n\t"
+	"vmovups	%%ymm11,  -32(%1)		    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));
+
+static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
+{
+
+
+	__asm__  __volatile__
+	(
+	"vxorpd	           %%ymm0, %%ymm0, %%ymm0	    \n\t"
+
+	"addq	$128, %1				    \n\t"
+
+	".align 16				            \n\t"
+	"1:				            	    \n\t"
+
+	//"prefetcht0     128(%1)				    \n\t"
+	// ".align 2				            \n\t"
+
+	"vmovups	%%ymm0 , -128(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -96(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -64(%1)		    \n\t"
+	"vmovups	%%ymm0 ,  -32(%1)		    \n\t"
+
+	"addq		$128 ,%1  	 	            \n\t"
+	"subq	        $8 , %0			            \n\t"		
+	"jnz		1b		             	    \n\t"
+
+	"vzeroupper					    \n\t"
+
+	:
+        : 
+	  "r" (n),  	// 0
+          "r" (x),      // 1
+          "r" (alpha)   // 2
+	: "cc", "%0", "%1",
+	  "%xmm0", "%xmm1", "%xmm2", "%xmm3", 
+	  "%xmm4", "%xmm5", "%xmm6", "%xmm7", 
+	  "%xmm8", "%xmm9", "%xmm10", "%xmm11", 
+	  "%xmm12", "%xmm13", "%xmm14", "%xmm15",
+	  "memory"
+	);
+
+} 
+
+
+
diff --git a/lapack-netlib/TESTING/EIG/Makefile b/lapack-netlib/TESTING/EIG/Makefile
index 128d59ae4..ef6ae0ec1 100644
--- a/lapack-netlib/TESTING/EIG/Makefile
+++ b/lapack-netlib/TESTING/EIG/Makefile
@@ -169,4 +169,4 @@ cchkee.o: cchkee.f
 zchkee.o: zchkee.f
 	$(FORTRAN) $(DRVOPTS) -c $< -o $@
 
-.f.o : ; $(FORTRAN) $(OPTS) -c $< -o $@
+.f.o : ; $(FORTRAN) $(DRVOPTS) -c $< -o $@
diff --git a/lapack-netlib/TESTING/LIN/Makefile b/lapack-netlib/TESTING/LIN/Makefile
index 2352da64c..44b05b794 100644
--- a/lapack-netlib/TESTING/LIN/Makefile
+++ b/lapack-netlib/TESTING/LIN/Makefile
@@ -338,4 +338,4 @@ zchkaa.o: zchkaa.f
 	$(FORTRAN) $(DRVOPTS) -c $< -o $@
  
 .f.o: 
-	$(FORTRAN) $(OPTS) -c $< -o $@
+	$(FORTRAN) $(DRVOPTS) -c $< -o $@
diff --git a/lapack-netlib/TESTING/sep.in b/lapack-netlib/TESTING/sep.in
index 19bd7c3da..6ef4329c9 100644
--- a/lapack-netlib/TESTING/sep.in
+++ b/lapack-netlib/TESTING/sep.in
@@ -1,11 +1,11 @@
 SEP:  Data file for testing Symmetric Eigenvalue Problem routines
 8                                 Number of values of N
-0 1 2 3 5 19 20 21                Values of N (dimension)
+0 1 2 3 5 18 19 21                Values of N (dimension)
 5                                 Number of values of NB
 1 3  3  3 10                      Values of NB (blocksize)
 2 2  2  2  2                      Values of NBMIN (minimum blocksize)
 1 0  5  9  1                      Values of NX (crossover point)
-160.0                             Threshold value
+300.0                             Threshold value
 T                                 Put T to test the LAPACK routines
 T                                 Put T to test the driver routines
 T                                 Put T to test the error exits
diff --git a/lapack-netlib/TESTING/zctest.in b/lapack-netlib/TESTING/zctest.in
index ef88cc0d9..48e88ec50 100644
--- a/lapack-netlib/TESTING/zctest.in
+++ b/lapack-netlib/TESTING/zctest.in
@@ -1,6 +1,6 @@
 Data file for testing ZCGESV/ZCPOSV LAPACK routines
 11                                      Number of values of M
-0 1 2 13 17 45 78 91 101 120 132        Values of M (row dimension)
+0 1 2 13 17 45 78 91 101 121 132        Values of M (row dimension)
 4                                       Number of values of NRHS
 1 2 15 16                               Values of NRHS (number of right hand sides)
 30.0                                    Threshold value of test ratio
diff --git a/lapack-netlib/lapacke/src/lapacke_clansy.c b/lapack-netlib/lapacke/src/lapacke_clansy.c
index 84a9d965a..eb9951145 100644
--- a/lapack-netlib/lapacke/src/lapacke_clansy.c
+++ b/lapack-netlib/lapacke/src/lapacke_clansy.c
@@ -51,7 +51,7 @@ float LAPACKE_clansy( int matrix_order, char norm, char uplo, lapack_int n,
 #endif
     /* Allocate memory for working array(s) */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
-        LAPACKE_lsame( norm, '0' ) ) {
+        LAPACKE_lsame( norm, 'O' ) ) {
         work = (float*)LAPACKE_malloc( sizeof(float) * MAX(1,n) );
         if( work == NULL ) {
             info = LAPACK_WORK_MEMORY_ERROR;
@@ -62,7 +62,7 @@ float LAPACKE_clansy( int matrix_order, char norm, char uplo, lapack_int n,
     res = LAPACKE_clansy_work( matrix_order, norm, uplo, n, a, lda, work );
     /* Release memory and exit */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
-        LAPACKE_lsame( norm, '0' ) ) {
+        LAPACKE_lsame( norm, 'O' ) ) {
         LAPACKE_free( work );
     }
 exit_level_0:
diff --git a/lapack-netlib/lapacke/src/lapacke_clantr.c b/lapack-netlib/lapacke/src/lapacke_clantr.c
index 77743f2d5..00ba34273 100644
--- a/lapack-netlib/lapacke/src/lapacke_clantr.c
+++ b/lapack-netlib/lapacke/src/lapacke_clantr.c
@@ -53,7 +53,7 @@ float LAPACKE_clantr( int matrix_order, char norm, char uplo, char diag,
     /* Allocate memory for working array(s) */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
         LAPACKE_lsame( norm, '0' ) ) {
-        work = (float*)LAPACKE_malloc( sizeof(float) * MAX(1,m) );
+        work = (float*)LAPACKE_malloc( sizeof(float) * MAX(1,MAX(m,n)) );
         if( work == NULL ) {
             info = LAPACK_WORK_MEMORY_ERROR;
             goto exit_level_0;
diff --git a/lapack-netlib/lapacke/src/lapacke_clantr_work.c b/lapack-netlib/lapacke/src/lapacke_clantr_work.c
index cb253a11e..1fa8cd923 100644
--- a/lapack-netlib/lapacke/src/lapacke_clantr_work.c
+++ b/lapack-netlib/lapacke/src/lapacke_clantr_work.c
@@ -47,7 +47,7 @@ float LAPACKE_clantr_work( int matrix_order, char norm, char uplo,
             info = info - 1;
         }
     } else if( matrix_order == LAPACK_ROW_MAJOR ) {
-        lapack_int lda_t = MAX(1,n);
+        lapack_int lda_t = MAX(1,m);
         lapack_complex_float* a_t = NULL;
         /* Check leading dimension(s) */
         if( lda < n ) {
@@ -57,13 +57,13 @@ float LAPACKE_clantr_work( int matrix_order, char norm, char uplo,
         }
         /* Allocate memory for temporary array(s) */
         a_t = (lapack_complex_float*)
-            LAPACKE_malloc( sizeof(lapack_complex_float) * lda_t * MAX(1,n) );
+            LAPACKE_malloc( sizeof(lapack_complex_float) * lda_t * MAX(1,MAX(m,n)) );
         if( a_t == NULL ) {
             info = LAPACK_TRANSPOSE_MEMORY_ERROR;
             goto exit_level_0;
         }
         /* Transpose input matrices */
-        LAPACKE_ctr_trans( matrix_order, uplo, diag, n, a, lda, a_t, lda_t );
+        LAPACKE_ctr_trans( matrix_order, uplo, diag, MAX(m,n), a, lda, a_t, lda_t );
         /* Call LAPACK function and adjust info */
         res = LAPACK_clantr( &norm, &uplo, &diag, &m, &n, a_t, &lda_t, work );
         info = 0;  /* LAPACK call is ok! */
diff --git a/lapack-netlib/lapacke/src/lapacke_cunmlq_work.c b/lapack-netlib/lapacke/src/lapacke_cunmlq_work.c
index 1cd20e1ca..5cf66424d 100644
--- a/lapack-netlib/lapacke/src/lapacke_cunmlq_work.c
+++ b/lapack-netlib/lapacke/src/lapacke_cunmlq_work.c
@@ -1,5 +1,5 @@
 /*****************************************************************************
-  Copyright (c) 2011, Intel Corp.
+  Copyright (c) 2014, Intel Corp.
   All rights reserved.
 
   Redistribution and use in source and binary forms, with or without
@@ -33,7 +33,7 @@
 
 #include "lapacke_utils.h"
 
-lapack_int LAPACKE_cunmlq_work( int matrix_order, char side, char trans,
+lapack_int LAPACKE_cunmlq_work( int matrix_layout, char side, char trans,
                                 lapack_int m, lapack_int n, lapack_int k,
                                 const lapack_complex_float* a, lapack_int lda,
                                 const lapack_complex_float* tau,
@@ -41,20 +41,22 @@ lapack_int LAPACKE_cunmlq_work( int matrix_order, char side, char trans,
                                 lapack_complex_float* work, lapack_int lwork )
 {
     lapack_int info = 0;
-    if( matrix_order == LAPACK_COL_MAJOR ) {
+    lapack_int r;
+    if( matrix_layout == LAPACK_COL_MAJOR ) {
         /* Call LAPACK function and adjust info */
         LAPACK_cunmlq( &side, &trans, &m, &n, &k, a, &lda, tau, c, &ldc, work,
                        &lwork, &info );
         if( info < 0 ) {
             info = info - 1;
         }
-    } else if( matrix_order == LAPACK_ROW_MAJOR ) {
+    } else if( matrix_layout == LAPACK_ROW_MAJOR ) {
+        r = LAPACKE_lsame( side, 'l' ) ? m : n;
         lapack_int lda_t = MAX(1,k);
         lapack_int ldc_t = MAX(1,m);
         lapack_complex_float* a_t = NULL;
         lapack_complex_float* c_t = NULL;
         /* Check leading dimension(s) */
-        if( lda < m ) {
+        if( lda < r ) {
             info = -8;
             LAPACKE_xerbla( "LAPACKE_cunmlq_work", info );
             return info;
@@ -84,8 +86,8 @@ lapack_int LAPACKE_cunmlq_work( int matrix_order, char side, char trans,
             goto exit_level_1;
         }
         /* Transpose input matrices */
-        LAPACKE_cge_trans( matrix_order, k, m, a, lda, a_t, lda_t );
-        LAPACKE_cge_trans( matrix_order, m, n, c, ldc, c_t, ldc_t );
+        LAPACKE_cge_trans( matrix_layout, k, m, a, lda, a_t, lda_t );
+        LAPACKE_cge_trans( matrix_layout, m, n, c, ldc, c_t, ldc_t );
         /* Call LAPACK function and adjust info */
         LAPACK_cunmlq( &side, &trans, &m, &n, &k, a_t, &lda_t, tau, c_t, &ldc_t,
                        work, &lwork, &info );
diff --git a/lapack-netlib/lapacke/src/lapacke_dlansy.c b/lapack-netlib/lapacke/src/lapacke_dlansy.c
index 5e6721ef8..3d9964202 100644
--- a/lapack-netlib/lapacke/src/lapacke_dlansy.c
+++ b/lapack-netlib/lapacke/src/lapacke_dlansy.c
@@ -51,7 +51,7 @@ double LAPACKE_dlansy( int matrix_order, char norm, char uplo, lapack_int n,
 #endif
     /* Allocate memory for working array(s) */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
-        LAPACKE_lsame( norm, '0' ) ) {
+        LAPACKE_lsame( norm, 'O' ) ) {
         work = (double*)LAPACKE_malloc( sizeof(double) * MAX(1,n) );
         if( work == NULL ) {
             info = LAPACK_WORK_MEMORY_ERROR;
@@ -62,7 +62,7 @@ double LAPACKE_dlansy( int matrix_order, char norm, char uplo, lapack_int n,
     res = LAPACKE_dlansy_work( matrix_order, norm, uplo, n, a, lda, work );
     /* Release memory and exit */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
-        LAPACKE_lsame( norm, '0' ) ) {
+        LAPACKE_lsame( norm, 'O' ) ) {
         LAPACKE_free( work );
     }
 exit_level_0:
diff --git a/lapack-netlib/lapacke/src/lapacke_dlantr.c b/lapack-netlib/lapacke/src/lapacke_dlantr.c
index 522122cb2..2cde1ebad 100644
--- a/lapack-netlib/lapacke/src/lapacke_dlantr.c
+++ b/lapack-netlib/lapacke/src/lapacke_dlantr.c
@@ -53,7 +53,7 @@ double LAPACKE_dlantr( int matrix_order, char norm, char uplo, char diag,
     /* Allocate memory for working array(s) */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
         LAPACKE_lsame( norm, '0' ) ) {
-        work = (double*)LAPACKE_malloc( sizeof(double) * MAX(1,m) );
+        work = (double*)LAPACKE_malloc( sizeof(double) * MAX(1,MAX(m,n)) );
         if( work == NULL ) {
             info = LAPACK_WORK_MEMORY_ERROR;
             goto exit_level_0;
diff --git a/lapack-netlib/lapacke/src/lapacke_dlantr_work.c b/lapack-netlib/lapacke/src/lapacke_dlantr_work.c
index 0a937bda9..44d638fa5 100644
--- a/lapack-netlib/lapacke/src/lapacke_dlantr_work.c
+++ b/lapack-netlib/lapacke/src/lapacke_dlantr_work.c
@@ -46,7 +46,7 @@ double LAPACKE_dlantr_work( int matrix_order, char norm, char uplo,
             info = info - 1;
         }
     } else if( matrix_order == LAPACK_ROW_MAJOR ) {
-        lapack_int lda_t = MAX(1,n);
+        lapack_int lda_t = MAX(1,m);
         double* a_t = NULL;
         /* Check leading dimension(s) */
         if( lda < n ) {
@@ -55,13 +55,13 @@ double LAPACKE_dlantr_work( int matrix_order, char norm, char uplo,
             return info;
         }
         /* Allocate memory for temporary array(s) */
-        a_t = (double*)LAPACKE_malloc( sizeof(double) * lda_t * MAX(1,n) );
+        a_t = (double*)LAPACKE_malloc( sizeof(double) * lda_t * MAX(1,MAX(m,n)) );
         if( a_t == NULL ) {
             info = LAPACK_TRANSPOSE_MEMORY_ERROR;
             goto exit_level_0;
         }
         /* Transpose input matrices */
-        LAPACKE_dtr_trans( matrix_order, uplo, diag, n, a, lda, a_t, lda_t );
+        LAPACKE_dtr_trans( matrix_order, uplo, diag, MAX(m,n), a, lda, a_t, lda_t );
         /* Call LAPACK function and adjust info */
         res = LAPACK_dlantr( &norm, &uplo, &diag, &m, &n, a_t, &lda_t, work );
         info = 0;  /* LAPACK call is ok! */
diff --git a/lapack-netlib/lapacke/src/lapacke_dormlq_work.c b/lapack-netlib/lapacke/src/lapacke_dormlq_work.c
index 9a7a997fe..99a7c3c71 100644
--- a/lapack-netlib/lapacke/src/lapacke_dormlq_work.c
+++ b/lapack-netlib/lapacke/src/lapacke_dormlq_work.c
@@ -1,5 +1,5 @@
 /*****************************************************************************
-  Copyright (c) 2011, Intel Corp.
+  Copyright (c) 2014, Intel Corp.
   All rights reserved.
 
   Redistribution and use in source and binary forms, with or without
@@ -33,27 +33,29 @@
 
 #include "lapacke_utils.h"
 
-lapack_int LAPACKE_dormlq_work( int matrix_order, char side, char trans,
+lapack_int LAPACKE_dormlq_work( int matrix_layout, char side, char trans,
                                 lapack_int m, lapack_int n, lapack_int k,
                                 const double* a, lapack_int lda,
                                 const double* tau, double* c, lapack_int ldc,
                                 double* work, lapack_int lwork )
 {
     lapack_int info = 0;
+    lapack_int r;
     lapack_int lda_t, ldc_t;
     double *a_t = NULL, *c_t = NULL;
-    if( matrix_order == LAPACK_COL_MAJOR ) {
+    if( matrix_layout == LAPACK_COL_MAJOR ) {
         /* Call LAPACK function and adjust info */
         LAPACK_dormlq( &side, &trans, &m, &n, &k, a, &lda, tau, c, &ldc, work,
                        &lwork, &info );
         if( info < 0 ) {
             info = info - 1;
         }
-    } else if( matrix_order == LAPACK_ROW_MAJOR ) {
+    } else if( matrix_layout == LAPACK_ROW_MAJOR ) {
+        r = LAPACKE_lsame( side, 'l' ) ? m : n;
         lda_t = MAX(1,k);
         ldc_t = MAX(1,m);
         /* Check leading dimension(s) */
-        if( lda < m ) {
+        if( lda < r ) {
             info = -8;
             LAPACKE_xerbla( "LAPACKE_dormlq_work", info );
             return info;
@@ -81,8 +83,8 @@ lapack_int LAPACKE_dormlq_work( int matrix_order, char side, char trans,
             goto exit_level_1;
         }
         /* Transpose input matrices */
-        LAPACKE_dge_trans( matrix_order, k, m, a, lda, a_t, lda_t );
-        LAPACKE_dge_trans( matrix_order, m, n, c, ldc, c_t, ldc_t );
+        LAPACKE_dge_trans( matrix_layout, k, m, a, lda, a_t, lda_t );
+        LAPACKE_dge_trans( matrix_layout, m, n, c, ldc, c_t, ldc_t );
         /* Call LAPACK function and adjust info */
         LAPACK_dormlq( &side, &trans, &m, &n, &k, a_t, &lda_t, tau, c_t, &ldc_t,
                        work, &lwork, &info );
diff --git a/lapack-netlib/lapacke/src/lapacke_slansy.c b/lapack-netlib/lapacke/src/lapacke_slansy.c
index 105ce4635..adad99b7d 100644
--- a/lapack-netlib/lapacke/src/lapacke_slansy.c
+++ b/lapack-netlib/lapacke/src/lapacke_slansy.c
@@ -51,7 +51,7 @@ float LAPACKE_slansy( int matrix_order, char norm, char uplo, lapack_int n,
 #endif
     /* Allocate memory for working array(s) */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
-        LAPACKE_lsame( norm, '0' ) ) {
+        LAPACKE_lsame( norm, 'O' ) ) {
         work = (float*)LAPACKE_malloc( sizeof(float) * MAX(1,n) );
         if( work == NULL ) {
             info = LAPACK_WORK_MEMORY_ERROR;
@@ -62,7 +62,7 @@ float LAPACKE_slansy( int matrix_order, char norm, char uplo, lapack_int n,
     res = LAPACKE_slansy_work( matrix_order, norm, uplo, n, a, lda, work );
     /* Release memory and exit */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
-        LAPACKE_lsame( norm, '0' ) ) {
+        LAPACKE_lsame( norm, 'O' ) ) {
         LAPACKE_free( work );
     }
 exit_level_0:
diff --git a/lapack-netlib/lapacke/src/lapacke_slantr.c b/lapack-netlib/lapacke/src/lapacke_slantr.c
index d6a512027..80313d118 100644
--- a/lapack-netlib/lapacke/src/lapacke_slantr.c
+++ b/lapack-netlib/lapacke/src/lapacke_slantr.c
@@ -53,7 +53,7 @@ float LAPACKE_slantr( int matrix_order, char norm, char uplo, char diag,
     /* Allocate memory for working array(s) */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
         LAPACKE_lsame( norm, '0' ) ) {
-        work = (float*)LAPACKE_malloc( sizeof(float) * MAX(1,m) );
+        work = (float*)LAPACKE_malloc( sizeof(float) * MAX(1,MAX(m,n)) );
         if( work == NULL ) {
             info = LAPACK_WORK_MEMORY_ERROR;
             goto exit_level_0;
diff --git a/lapack-netlib/lapacke/src/lapacke_slantr_work.c b/lapack-netlib/lapacke/src/lapacke_slantr_work.c
index 2389468d0..9032f7094 100644
--- a/lapack-netlib/lapacke/src/lapacke_slantr_work.c
+++ b/lapack-netlib/lapacke/src/lapacke_slantr_work.c
@@ -46,7 +46,7 @@ float LAPACKE_slantr_work( int matrix_order, char norm, char uplo,
             info = info - 1;
         }
     } else if( matrix_order == LAPACK_ROW_MAJOR ) {
-        lapack_int lda_t = MAX(1,n);
+        lapack_int lda_t = MAX(1,m);
         float* a_t = NULL;
         /* Check leading dimension(s) */
         if( lda < n ) {
@@ -55,13 +55,13 @@ float LAPACKE_slantr_work( int matrix_order, char norm, char uplo,
             return info;
         }
         /* Allocate memory for temporary array(s) */
-        a_t = (float*)LAPACKE_malloc( sizeof(float) * lda_t * MAX(1,n) );
+        a_t = (float*)LAPACKE_malloc( sizeof(float) * lda_t * MAX(1,MAX(m,n)) );
         if( a_t == NULL ) {
             info = LAPACK_TRANSPOSE_MEMORY_ERROR;
             goto exit_level_0;
         }
         /* Transpose input matrices */
-        LAPACKE_str_trans( matrix_order, uplo, diag, n, a, lda, a_t, lda_t );
+        LAPACKE_str_trans( matrix_order, uplo, diag, MAX(m,n), a, lda, a_t, lda_t );
         /* Call LAPACK function and adjust info */
         res = LAPACK_slantr( &norm, &uplo, &diag, &m, &n, a_t, &lda_t, work );
         info = 0;  /* LAPACK call is ok! */
diff --git a/lapack-netlib/lapacke/src/lapacke_sormlq_work.c b/lapack-netlib/lapacke/src/lapacke_sormlq_work.c
index 7a7464d18..bbf55bd84 100644
--- a/lapack-netlib/lapacke/src/lapacke_sormlq_work.c
+++ b/lapack-netlib/lapacke/src/lapacke_sormlq_work.c
@@ -1,5 +1,5 @@
 /*****************************************************************************
-  Copyright (c) 2011, Intel Corp.
+  Copyright (c) 2014, Intel Corp.
   All rights reserved.
 
   Redistribution and use in source and binary forms, with or without
@@ -33,27 +33,29 @@
 
 #include "lapacke_utils.h"
 
-lapack_int LAPACKE_sormlq_work( int matrix_order, char side, char trans,
+lapack_int LAPACKE_sormlq_work( int matrix_layout, char side, char trans,
                                 lapack_int m, lapack_int n, lapack_int k,
                                 const float* a, lapack_int lda,
                                 const float* tau, float* c, lapack_int ldc,
                                 float* work, lapack_int lwork )
 {
     lapack_int info = 0;
+    lapack_int r;
     lapack_int lda_t, ldc_t;
     float *a_t = NULL, *c_t = NULL;
-    if( matrix_order == LAPACK_COL_MAJOR ) {
+    if( matrix_layout == LAPACK_COL_MAJOR ) {
         /* Call LAPACK function and adjust info */
         LAPACK_sormlq( &side, &trans, &m, &n, &k, a, &lda, tau, c, &ldc, work,
                        &lwork, &info );
         if( info < 0 ) {
             info = info - 1;
         }
-    } else if( matrix_order == LAPACK_ROW_MAJOR ) {
+    } else if( matrix_layout == LAPACK_ROW_MAJOR ) {
+        r = LAPACKE_lsame( side, 'l' ) ? m : n;
         lda_t = MAX(1,k);
         ldc_t = MAX(1,m);
         /* Check leading dimension(s) */
-        if( lda < m ) {
+        if( lda < r ) {
             info = -8;
             LAPACKE_xerbla( "LAPACKE_sormlq_work", info );
             return info;
@@ -81,8 +83,8 @@ lapack_int LAPACKE_sormlq_work( int matrix_order, char side, char trans,
             goto exit_level_1;
         }
         /* Transpose input matrices */
-        LAPACKE_sge_trans( matrix_order, k, m, a, lda, a_t, lda_t );
-        LAPACKE_sge_trans( matrix_order, m, n, c, ldc, c_t, ldc_t );
+        LAPACKE_sge_trans( matrix_layout, k, m, a, lda, a_t, lda_t );
+        LAPACKE_sge_trans( matrix_layout, m, n, c, ldc, c_t, ldc_t );
         /* Call LAPACK function and adjust info */
         LAPACK_sormlq( &side, &trans, &m, &n, &k, a_t, &lda_t, tau, c_t, &ldc_t,
                        work, &lwork, &info );
diff --git a/lapack-netlib/lapacke/src/lapacke_zlansy.c b/lapack-netlib/lapacke/src/lapacke_zlansy.c
index 891437846..460a51a85 100644
--- a/lapack-netlib/lapacke/src/lapacke_zlansy.c
+++ b/lapack-netlib/lapacke/src/lapacke_zlansy.c
@@ -51,7 +51,7 @@ double LAPACKE_zlansy( int matrix_order, char norm, char uplo, lapack_int n,
 #endif
     /* Allocate memory for working array(s) */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
-        LAPACKE_lsame( norm, '0' ) ) {
+        LAPACKE_lsame( norm, 'O' ) ) {
         work = (double*)LAPACKE_malloc( sizeof(double) * MAX(1,n) );
         if( work == NULL ) {
             info = LAPACK_WORK_MEMORY_ERROR;
@@ -62,7 +62,7 @@ double LAPACKE_zlansy( int matrix_order, char norm, char uplo, lapack_int n,
     res = LAPACKE_zlansy_work( matrix_order, norm, uplo, n, a, lda, work );
     /* Release memory and exit */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
-        LAPACKE_lsame( norm, '0' ) ) {
+        LAPACKE_lsame( norm, 'O' ) ) {
         LAPACKE_free( work );
     }
 exit_level_0:
diff --git a/lapack-netlib/lapacke/src/lapacke_zlantr.c b/lapack-netlib/lapacke/src/lapacke_zlantr.c
index 887bc2eea..001ce68f6 100644
--- a/lapack-netlib/lapacke/src/lapacke_zlantr.c
+++ b/lapack-netlib/lapacke/src/lapacke_zlantr.c
@@ -53,7 +53,7 @@ double LAPACKE_zlantr( int matrix_order, char norm, char uplo, char diag,
     /* Allocate memory for working array(s) */
     if( LAPACKE_lsame( norm, 'i' ) || LAPACKE_lsame( norm, '1' ) ||
         LAPACKE_lsame( norm, '0' ) ) {
-        work = (double*)LAPACKE_malloc( sizeof(double) * MAX(1,m) );
+        work = (double*)LAPACKE_malloc( sizeof(double) * MAX(1,MAX(m,n)) );
         if( work == NULL ) {
             info = LAPACK_WORK_MEMORY_ERROR;
             goto exit_level_0;
diff --git a/lapack-netlib/lapacke/src/lapacke_zlantr_work.c b/lapack-netlib/lapacke/src/lapacke_zlantr_work.c
index 65e741428..8700a6ee2 100644
--- a/lapack-netlib/lapacke/src/lapacke_zlantr_work.c
+++ b/lapack-netlib/lapacke/src/lapacke_zlantr_work.c
@@ -47,7 +47,7 @@ double LAPACKE_zlantr_work( int matrix_order, char norm, char uplo,
             info = info - 1;
         }
     } else if( matrix_order == LAPACK_ROW_MAJOR ) {
-        lapack_int lda_t = MAX(1,n);
+        lapack_int lda_t = MAX(1,m);
         lapack_complex_double* a_t = NULL;
         /* Check leading dimension(s) */
         if( lda < n ) {
@@ -57,13 +57,13 @@ double LAPACKE_zlantr_work( int matrix_order, char norm, char uplo,
         }
         /* Allocate memory for temporary array(s) */
         a_t = (lapack_complex_double*)
-            LAPACKE_malloc( sizeof(lapack_complex_double) * lda_t * MAX(1,n) );
+            LAPACKE_malloc( sizeof(lapack_complex_double) * lda_t * MAX(1,MAX(m,n)) );
         if( a_t == NULL ) {
             info = LAPACK_TRANSPOSE_MEMORY_ERROR;
             goto exit_level_0;
         }
         /* Transpose input matrices */
-        LAPACKE_ztr_trans( matrix_order, uplo, diag, n, a, lda, a_t, lda_t );
+        LAPACKE_ztr_trans( matrix_order, uplo, diag, MAX(m,n), a, lda, a_t, lda_t );
         /* Call LAPACK function and adjust info */
         res = LAPACK_zlantr( &norm, &uplo, &diag, &m, &n, a_t, &lda_t, work );
         info = 0;  /* LAPACK call is ok! */
diff --git a/lapack-netlib/lapacke/src/lapacke_zunmlq_work.c b/lapack-netlib/lapacke/src/lapacke_zunmlq_work.c
index 8677ac0bc..38a2d947a 100644
--- a/lapack-netlib/lapacke/src/lapacke_zunmlq_work.c
+++ b/lapack-netlib/lapacke/src/lapacke_zunmlq_work.c
@@ -1,5 +1,5 @@
 /*****************************************************************************
-  Copyright (c) 2011, Intel Corp.
+  Copyright (c) 2014, Intel Corp.
   All rights reserved.
 
   Redistribution and use in source and binary forms, with or without
@@ -33,7 +33,7 @@
 
 #include "lapacke_utils.h"
 
-lapack_int LAPACKE_zunmlq_work( int matrix_order, char side, char trans,
+lapack_int LAPACKE_zunmlq_work( int matrix_layout, char side, char trans,
                                 lapack_int m, lapack_int n, lapack_int k,
                                 const lapack_complex_double* a, lapack_int lda,
                                 const lapack_complex_double* tau,
@@ -41,20 +41,22 @@ lapack_int LAPACKE_zunmlq_work( int matrix_order, char side, char trans,
                                 lapack_complex_double* work, lapack_int lwork )
 {
     lapack_int info = 0;
-    if( matrix_order == LAPACK_COL_MAJOR ) {
+    lapack_int r;
+    if( matrix_layout == LAPACK_COL_MAJOR ) {
         /* Call LAPACK function and adjust info */
         LAPACK_zunmlq( &side, &trans, &m, &n, &k, a, &lda, tau, c, &ldc, work,
                        &lwork, &info );
         if( info < 0 ) {
             info = info - 1;
         }
-    } else if( matrix_order == LAPACK_ROW_MAJOR ) {
+    } else if( matrix_layout == LAPACK_ROW_MAJOR ) {
+        r = LAPACKE_lsame( side, 'l' ) ? m : n;
         lapack_int lda_t = MAX(1,k);
         lapack_int ldc_t = MAX(1,m);
         lapack_complex_double* a_t = NULL;
         lapack_complex_double* c_t = NULL;
         /* Check leading dimension(s) */
-        if( lda < m ) {
+        if( lda < r ) {
             info = -8;
             LAPACKE_xerbla( "LAPACKE_zunmlq_work", info );
             return info;
@@ -84,8 +86,8 @@ lapack_int LAPACKE_zunmlq_work( int matrix_order, char side, char trans,
             goto exit_level_1;
         }
         /* Transpose input matrices */
-        LAPACKE_zge_trans( matrix_order, k, m, a, lda, a_t, lda_t );
-        LAPACKE_zge_trans( matrix_order, m, n, c, ldc, c_t, ldc_t );
+        LAPACKE_zge_trans( matrix_layout, k, m, a, lda, a_t, lda_t );
+        LAPACKE_zge_trans( matrix_layout, m, n, c, ldc, c_t, ldc_t );
         /* Call LAPACK function and adjust info */
         LAPACK_zunmlq( &side, &trans, &m, &n, &k, a_t, &lda_t, tau, c_t, &ldc_t,
                        work, &lwork, &info );
diff --git a/lapack/CMakeLists.txt b/lapack/CMakeLists.txt
new file mode 100644
index 000000000..de42e1ab6
--- /dev/null
+++ b/lapack/CMakeLists.txt
@@ -0,0 +1,98 @@
+
+include_directories(${CMAKE_SOURCE_DIR})
+
+
+set(LAPACK_SOURCES
+  getrf/getrf_single.c
+  potrf/potrf_U_single.c
+  potrf/potrf_L_single.c
+  lauum/lauum_U_single.c
+  lauum/lauum_L_single.c
+)
+
+# add a 'z' to filename for complex version
+set(LAPACK_MANGLED_SOURCES
+  getf2/getf2_k.c
+  lauu2/lauu2_U.c
+  lauu2/lauu2_L.c
+  potf2/potf2_U.c
+  potf2/potf2_L.c
+)
+
+# sources that need TRANS set
+# this has a 'z' version
+set(TRANS_SOURCES
+  getrs/getrs_single.c
+)
+
+# sources that need UNIT set
+# these do NOT have a z version
+set(UNIT_SOURCES
+  trtri/trtri_U_single.c
+  trtri/trtri_L_single.c
+)
+
+# these have a 'z' version
+set(UNIT_SOURCES2
+  trti2/trti2_U.c
+  trti2/trti2_L.c
+)
+
+GenerateNamedObjects("${LAPACK_SOURCES}")
+GenerateNamedObjects("${LAPACK_MANGLED_SOURCES}" "" "" false "" "" false 3)
+
+# TODO: laswp needs arch specific code
+GenerateNamedObjects("laswp/generic/laswp_k.c" "" "laswp_plus" false "" "" false 3)
+GenerateNamedObjects("laswp/generic/laswp_k.c" "MINUS" "laswp_minus" false "" "" false 3)
+
+if (SMP)
+
+  if (USE_OPENMP)
+    set(GETRF_SRC getrf/getrf_parallel_omp.c)
+  else ()
+    set(GETRF_SRC getrf/getrf_parallel.c)
+  endif ()
+
+  # these do not have 'z' versions
+  set(PARALLEL_SOURCES
+    ${GETRF_SRC}
+    lauum/lauum_U_parallel.c
+    lauum/lauum_L_parallel.c
+    potrf/potrf_U_parallel.c
+    potrf/potrf_L_parallel.c
+  )
+
+  # this has a z version
+  list(APPEND TRANS_SOURCES
+    getrs/getrs_parallel.c
+  )
+
+  # these do NOT have a z version
+  list(APPEND UNIT_SOURCES
+    trtri/trtri_U_parallel.c
+    trtri/trtri_L_parallel.c
+  )
+
+  GenerateNamedObjects("${PARALLEL_SOURCES}")
+endif ()
+
+foreach (float_type ${FLOAT_TYPES})
+  if (${float_type} STREQUAL "COMPLEX" OR ${float_type} STREQUAL "ZCOMPLEX")
+    foreach (trans_src ${TRANS_SOURCES})
+      string(REGEX MATCH "[a-z]/([a-z]+_)([a-z]+)" op_name ${trans_src})
+      string(REPLACE "/" "/z" ztrans_src ${trans_src})
+      GenerateNamedObjects("${ztrans_src}" "TRANS=1" "${CMAKE_MATCH_1}N_${CMAKE_MATCH_2}" false "" "" false ${float_type})
+      GenerateNamedObjects("${ztrans_src}" "TRANS=2" "${CMAKE_MATCH_1}T_${CMAKE_MATCH_2}" false "" "" false ${float_type})
+      GenerateNamedObjects("${ztrans_src}" "TRANS=3" "${CMAKE_MATCH_1}R_${CMAKE_MATCH_2}" false "" "" false ${float_type})
+      GenerateNamedObjects("${ztrans_src}" "TRANS=4" "${CMAKE_MATCH_1}C_${CMAKE_MATCH_2}" false "" "" false ${float_type})
+    endforeach ()
+  else ()
+    GenerateCombinationObjects("${TRANS_SOURCES}" "TRANS" "N" "" 4 "" false ${float_type})
+  endif ()
+endforeach ()
+
+GenerateCombinationObjects("${UNIT_SOURCES}" "UNIT" "N" "" 4)
+GenerateCombinationObjects("${UNIT_SOURCES2}" "UNIT" "N" "" 0 "" "" 3)
+
+add_library(lapack OBJECT ${OPENBLAS_SRC})
+
diff --git a/lapack/getrf/getrf_parallel.c b/lapack/getrf/getrf_parallel.c
index a76be3ba7..8fdf76987 100644
--- a/lapack/getrf/getrf_parallel.c
+++ b/lapack/getrf/getrf_parallel.c
@@ -67,7 +67,7 @@ double sqrt(double);
 #undef  GETRF_FACTOR
 #define GETRF_FACTOR 1.00
 
-static inline BLASLONG FORMULA1(BLASLONG M, BLASLONG N, BLASLONG IS, BLASLONG BK, BLASLONG T) {
+static __inline BLASLONG FORMULA1(BLASLONG M, BLASLONG N, BLASLONG IS, BLASLONG BK, BLASLONG T) {
 
   double m = (double)(M - IS - BK);
   double n = (double)(N - IS - BK);
@@ -373,7 +373,11 @@ blasint CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa,
 
   BLASLONG num_cpu;
 
+#ifdef _MSC_VER
+  BLASLONG flag[MAX_CPU_NUMBER * CACHE_LINE_SIZE];
+#else
   volatile BLASLONG flag[MAX_CPU_NUMBER * CACHE_LINE_SIZE] __attribute__((aligned(128)));
+#endif
 
 #ifndef COMPLEX
 #ifdef XDOUBLE
diff --git a/make.inc b/make.inc
index 485cb7d48..1fc95b0c6 100644
--- a/make.inc
+++ b/make.inc
@@ -1,6 +1,6 @@
 SHELL = /bin/sh
 PLAT = _LINUX
-DRVOPTS  = $(OPTS)
+DRVOPTS  = $(NOOPT)
 ARCHFLAGS= -ru
 #RANLIB   = ranlib
 
diff --git a/openblas_config_template.h b/openblas_config_template.h
index 3b3435b0e..942a8f547 100644
--- a/openblas_config_template.h
+++ b/openblas_config_template.h
@@ -59,7 +59,8 @@ typedef int blasint;
    extension since version 3.0.  If neither are available, use a compatible
    structure as fallback (see Clause 6.2.5.13 of the C99 standard). */
 #if (defined(__STDC_IEC_559_COMPLEX__) || __STDC_VERSION__ >= 199901L || \
-     (__GNUC__ >= 3 && !defined(__cplusplus)))
+     (__GNUC__ >= 3 && !defined(__cplusplus)) || \
+     _MSC_VER >= 1800) // Visual Studio 2013 supports complex
   #define OPENBLAS_COMPLEX_C99
 #ifndef __cplusplus
   #include <complex.h>
diff --git a/param.h b/param.h
index 18c711eb3..6c9ca83da 100644
--- a/param.h
+++ b/param.h
@@ -499,6 +499,98 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #endif
 
 
+#ifdef EXCAVATOR
+#define SNUMOPT         8
+#define DNUMOPT         4
+
+#define GEMM_DEFAULT_OFFSET_A  64
+#define GEMM_DEFAULT_OFFSET_B 832
+#define GEMM_DEFAULT_ALIGN 0x0fffUL
+
+
+
+#define QGEMM_DEFAULT_UNROLL_N 2
+#define CGEMM_DEFAULT_UNROLL_N 2
+#define ZGEMM_DEFAULT_UNROLL_N 2
+#define XGEMM_DEFAULT_UNROLL_N 1
+
+#ifdef ARCH_X86
+#define SGEMM_DEFAULT_UNROLL_N 4
+#define DGEMM_DEFAULT_UNROLL_N 4
+#define SGEMM_DEFAULT_UNROLL_M 4
+#define DGEMM_DEFAULT_UNROLL_M 2
+#define QGEMM_DEFAULT_UNROLL_M 2
+#define CGEMM_DEFAULT_UNROLL_M 2
+#define ZGEMM_DEFAULT_UNROLL_M 1
+#define XGEMM_DEFAULT_UNROLL_M 1
+#else
+#define SGEMM_DEFAULT_UNROLL_N 2
+#define DGEMM_DEFAULT_UNROLL_N 2
+#define SGEMM_DEFAULT_UNROLL_M 16
+#define DGEMM_DEFAULT_UNROLL_M 8
+#define QGEMM_DEFAULT_UNROLL_M 2
+#define CGEMM_DEFAULT_UNROLL_M 4
+#define ZGEMM_DEFAULT_UNROLL_M 2
+#define XGEMM_DEFAULT_UNROLL_M 1
+#define CGEMM3M_DEFAULT_UNROLL_N 4
+#define CGEMM3M_DEFAULT_UNROLL_M 8
+#define ZGEMM3M_DEFAULT_UNROLL_N 4
+#define ZGEMM3M_DEFAULT_UNROLL_M 4
+#define GEMV_UNROLL 8
+#endif
+
+#if defined(ARCH_X86_64)
+#define SGEMM_DEFAULT_P 768
+#define DGEMM_DEFAULT_P 576
+#define ZGEMM_DEFAULT_P 288
+#define CGEMM_DEFAULT_P 576
+#else
+#define SGEMM_DEFAULT_P 448
+#define DGEMM_DEFAULT_P 480
+#define ZGEMM_DEFAULT_P 112
+#define CGEMM_DEFAULT_P 224
+#endif
+#define QGEMM_DEFAULT_P 112
+#define XGEMM_DEFAULT_P  56
+
+#if defined(ARCH_X86_64)
+#define SGEMM_DEFAULT_Q 192
+#define DGEMM_DEFAULT_Q 160
+#define ZGEMM_DEFAULT_Q 160
+#define CGEMM_DEFAULT_Q 160
+#else
+#define SGEMM_DEFAULT_Q 224
+#define DGEMM_DEFAULT_Q 224
+#define ZGEMM_DEFAULT_Q 224
+#define CGEMM_DEFAULT_Q 224
+#endif
+#define QGEMM_DEFAULT_Q 224
+#define XGEMM_DEFAULT_Q 224
+
+#define CGEMM3M_DEFAULT_P 448
+#define ZGEMM3M_DEFAULT_P 224
+#define XGEMM3M_DEFAULT_P 112
+#define CGEMM3M_DEFAULT_Q 224
+#define ZGEMM3M_DEFAULT_Q 224
+#define XGEMM3M_DEFAULT_Q 224
+#define CGEMM3M_DEFAULT_R 12288
+#define ZGEMM3M_DEFAULT_R 12288
+#define XGEMM3M_DEFAULT_R 12288
+
+#define SGEMM_DEFAULT_R 12288
+#define QGEMM_DEFAULT_R qgemm_r
+#define DGEMM_DEFAULT_R 12288
+#define CGEMM_DEFAULT_R cgemm_r
+#define ZGEMM_DEFAULT_R zgemm_r
+#define XGEMM_DEFAULT_R xgemm_r
+
+#define SYMV_P  16
+#define HAVE_EXCLUSIVE_CACHE
+
+#define GEMM_THREAD gemm_thread_mn
+
+#endif
+
 #ifdef ATHLON
 
 #define SNUMOPT		4
@@ -1322,7 +1414,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define XGEMM_DEFAULT_UNROLL_M 1
 
 #define SGEMM_DEFAULT_UNROLL_N 4
-#define DGEMM_DEFAULT_UNROLL_N 4
+#define DGEMM_DEFAULT_UNROLL_N 8
 #define QGEMM_DEFAULT_UNROLL_N 2
 #define CGEMM_DEFAULT_UNROLL_N 2
 #define ZGEMM_DEFAULT_UNROLL_N 2
diff --git a/symcopy.h b/symcopy.h
index 48ccbd369..16172c046 100644
--- a/symcopy.h
+++ b/symcopy.h
@@ -43,7 +43,7 @@
 
 #if !defined(XDOUBLE) || !defined(QUAD_PRECISION)
 
-static inline void SYMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void SYMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -141,7 +141,7 @@ static inline void SYMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void SYMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void SYMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -232,7 +232,7 @@ static inline void SYMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
 }
 
 
-static inline void ZSYMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZSYMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -362,7 +362,7 @@ static inline void ZSYMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void ZSYMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZSYMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -486,7 +486,7 @@ static inline void ZSYMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void ZHEMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZHEMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -613,7 +613,7 @@ static inline void ZHEMCOPY_L(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void ZHEMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZHEMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -735,7 +735,7 @@ static inline void ZHEMCOPY_U(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
 }
 
 
-static inline void ZHEMCOPY_M(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZHEMCOPY_M(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -862,7 +862,7 @@ static inline void ZHEMCOPY_M(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void ZHEMCOPY_V(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZHEMCOPY_V(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -984,7 +984,7 @@ static inline void ZHEMCOPY_V(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
 }
 
 
-static inline void TRMCOPY_NL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void TRMCOPY_NL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -1082,7 +1082,7 @@ static inline void TRMCOPY_NL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void TRMCOPY_TL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void TRMCOPY_TL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -1180,7 +1180,7 @@ static inline void TRMCOPY_TL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void TRMCOPY_NU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void TRMCOPY_NU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -1270,7 +1270,7 @@ static inline void TRMCOPY_NU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void TRMCOPY_TU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void TRMCOPY_TU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -1360,7 +1360,7 @@ static inline void TRMCOPY_TU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void ZTRMCOPY_NL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZTRMCOPY_NL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -1490,7 +1490,7 @@ static inline void ZTRMCOPY_NL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void ZTRMCOPY_TL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZTRMCOPY_TL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -1620,7 +1620,7 @@ static inline void ZTRMCOPY_TL(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void ZTRMCOPY_NU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZTRMCOPY_NU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
@@ -1744,7 +1744,7 @@ static inline void ZTRMCOPY_NU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   }
 }
 
-static inline void ZTRMCOPY_TU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
+static __inline void ZTRMCOPY_TU(BLASLONG m, FLOAT *a, BLASLONG lda, FLOAT *b){
   BLASLONG is, js;
 
   FLOAT *aa1, *aa2;
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
new file mode 100644
index 000000000..cd4497117
--- /dev/null
+++ b/test/CMakeLists.txt
@@ -0,0 +1,38 @@
+include_directories(${CMAKE_SOURCE_DIR})
+
+enable_language(Fortran)
+
+set(OpenBLAS_Tests
+  sblat1 sblat2 sblat3
+  dblat1 dblat2 dblat3
+  cblat1 cblat2 cblat3
+  zblat1 zblat2 zblat3)
+
+foreach(test_bin ${OpenBLAS_Tests})
+add_executable(${test_bin} ${test_bin}.f)
+target_link_libraries(${test_bin} ${OpenBLAS_LIBNAME}_static)
+endforeach()
+
+# $1 exec, $2 input, $3 output_result
+FILE(WRITE ${CMAKE_CURRENT_BINARY_DIR}/test_helper.sh
+"rm -f $3\n"
+"$1 < $2\n"
+"grep -q FATAL $3\n"
+"if [ $? -eq 0 ]; then\n"
+"echo Error\n"
+"exit 1\n"
+"else\n"
+"exit 0\n"
+"fi\n"
+)
+
+set(float_types s d c z)
+foreach(float_type ${float_types})
+string(TOUPPER ${float_type} float_type_upper)
+add_test(NAME "${float_type}blas1" 
+  COMMAND "${CMAKE_CURRENT_BINARY_DIR}/${float_type}blat1")
+add_test(NAME "${float_type}blas2" 
+  COMMAND sh "${CMAKE_CURRENT_BINARY_DIR}/test_helper.sh" "${CMAKE_CURRENT_BINARY_DIR}/${float_type}blat2" "${PROJECT_SOURCE_DIR}/test/${float_type}blat2.dat" ${float_type_upper}BLAT2.SUMM)
+add_test(NAME "${float_type}blas3" 
+  COMMAND sh "${CMAKE_CURRENT_BINARY_DIR}/test_helper.sh" "${CMAKE_CURRENT_BINARY_DIR}/${float_type}blat3" "${PROJECT_SOURCE_DIR}/test/${float_type}blat3.dat" ${float_type_upper}BLAT3.SUMM)
+endforeach()
\ No newline at end of file