Merge branch 'develop'

.gitignore

@@ -8,6 +8,8 @@ lapack-3.1.1
 lapack-3.1.1.tgz
 lapack-3.4.1
 lapack-3.4.1.tgz
+lapack-3.4.2
+lapack-3.4.2.tgz
 *.so
 *.a
 .svn

Changelog.txt

@@ -1,4 +1,20 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.2.5
+26-Nov-2012
+common:
+	* Added NO_SHARED flag to disable generating the shared library.
+	* Compile LAPACKE with ILP64 modle when INTERFACE64=1 (#158)
+	* Export LAPACK 3.4.2 symbols in shared library. (#147)
+	* Only detect the number of physical CPU cores on Mac OSX. (#157)
+	* Fixed NetBSD build. (#155)
+	* Fixed compilation with TARGET=GENERIC. (#160)
+x86/x86-64:
+	* Restore the original CPU affinity when calling 
+	  openblas_set_num_threads(1) (#153)
+	* Fixed a SEGFAULT bug in dgemv_t when m is very large.(#154)
+MIPS64:
+
 ====================================================================
 Version 0.2.4
 8-Oct-2012

Makefile

@@ -80,6 +80,7 @@ endif
 	@echo
 
 shared :
+ifndef NO_SHARED
 ifeq ($(OSNAME), Linux)
 	$(MAKE) -C exports so
 	-ln -fs $(LIBSONAME) $(LIBPREFIX).so
@@ -103,6 +104,7 @@ endif
 ifeq ($(OSNAME), CYGWIN_NT)
 	$(MAKE) -C exports dll
 endif
+endif
 
 tests :
 ifndef NOFORTRAN
@@ -222,7 +224,11 @@ ifndef NOFORTRAN
 	-@echo "PNOOPT      = $(FPFLAGS) -O0" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "LOADOPTS    = $(FFLAGS) $(EXTRALIB)" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "CC          = $(CC)" >> $(NETLIB_LAPACK_DIR)/make.inc
+ifdef INTERFACE64
+	-@echo "CFLAGS      = $(CFLAGS) -DHAVE_LAPACK_CONFIG_H  -DLAPACK_ILP64" >> $(NETLIB_LAPACK_DIR)/make.inc
+else
 	-@echo "CFLAGS      = $(CFLAGS)" >> $(NETLIB_LAPACK_DIR)/make.inc
+endif
 	-@echo "ARCH        = $(AR)" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "ARCHFLAGS   = -ru" >> $(NETLIB_LAPACK_DIR)/make.inc
 	-@echo "RANLIB      = $(RANLIB)" >> $(NETLIB_LAPACK_DIR)/make.inc

Makefile.install

@@ -71,11 +71,9 @@ ifeq ($(OSNAME), Darwin)
 endif
 ifeq ($(OSNAME), WINNT)
 	-cp $(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)
-	-ln -fs $(OPENBLAS_LIBRARY_DIR)/$(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)/$(LIBPREFIX).dll
 endif
 ifeq ($(OSNAME), CYGWIN_NT)
 	-cp $(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)
-	-ln -fs $(OPENBLAS_LIBRARY_DIR)/$(LIBDLLNAME) $(OPENBLAS_LIBRARY_DIR)/$(LIBPREFIX).dll
 endif
 
 	@echo Install OK!

Makefile.rule

@@ -3,7 +3,7 @@
 #
 
 # This library's version
-VERSION = 0.2.4
+VERSION = 0.2.5
 
 # If you set the suffix, the library name will be libopenblas_$(LIBNAMESUFFIX).a
 # and libopenblas_$(LIBNAMESUFFIX).so. Meanwhile, the soname in shared library 
@@ -24,10 +24,13 @@ VERSION = 0.2.4
 # Fortran compiler. Default is g77.
 # FC = gfortran
 
-# Even you can specify cross compiler
+# Even you can specify cross compiler. Meanwhile, please set HOSTCC.
 # CC = x86_64-w64-mingw32-gcc
 # FC = x86_64-w64-mingw32-gfortran
 
+# If you use the cross compiler, please set this host compiler.
+# HOSTCC = gcc
+
 # If you need 32bit binary, define BINARY=32, otherwise define BINARY=64
 # BINARY=64
 
@@ -45,6 +48,9 @@ VERSION = 0.2.4
 # automatically detected by the the script.
 # NUM_THREADS = 24
 
+# if you don't need generate the shared library, please comment it in.
+# NO_SHARED = 1
+
 # If you don't need CBLAS interface, please comment it in.
 # NO_CBLAS = 1
 

Makefile.system

@@ -129,7 +129,7 @@ MD5SUM = md5 -r
 endif
 
 ifeq ($(OSNAME), NetBSD)
-MD5SUM = md5 -r
+MD5SUM = md5 -n
 endif
 
 ifeq ($(OSNAME), Linux)

common.h

@@ -351,7 +351,12 @@ typedef int blasint;
 #endif
 
 #define MMAP_ACCESS (PROT_READ | PROT_WRITE)
+
+#ifdef __NetBSD__
+#define MMAP_POLICY (MAP_PRIVATE | MAP_ANON)
+#else
 #define MMAP_POLICY (MAP_PRIVATE | MAP_ANONYMOUS)
+#endif
 
 #include "param.h"
 #include "common_param.h"

driver/others/blas_server.c

@@ -770,6 +770,19 @@ void goto_set_num_threads(int num_threads) {
 
   if (num_threads < 1) num_threads = blas_num_threads;
 
+#ifndef NO_AFFINITY
+  if (num_threads == 1) {
+    if (blas_cpu_number == 1){
+      //OpenBLAS is already single thread.
+      return; 
+    }else{
+      //From multi-threads to single thread
+      //Restore the original affinity mask
+      gotoblas_set_affinity(-1);
+    }
+  }
+#endif
+
   if (num_threads > MAX_CPU_NUMBER) num_threads = MAX_CPU_NUMBER;
 
   if (num_threads > blas_num_threads) {
@@ -800,6 +813,13 @@ void goto_set_num_threads(int num_threads) {
     UNLOCK_COMMAND(&server_lock);
   }
 
+#ifndef NO_AFFINITY
+  if(blas_cpu_number == 1 && num_threads > 1){
+    //Restore the thread 0 affinity.
+    gotoblas_set_affinity(0);
+  }
+#endif
+
   blas_cpu_number  = num_threads;
 
 #if defined(ARCH_MIPS64) 

driver/others/memory.c

@@ -185,7 +185,7 @@ int get_num_procs(void) {
 
 #endif
 
-#if defined(OS_FREEBSD) || defined(OS_DARWIN)
+#if defined(OS_FREEBSD) 
 
 int get_num_procs(void) {
   
@@ -206,6 +206,18 @@ int get_num_procs(void) {
 
 #endif
 
+#if defined(OS_DARWIN)
+int get_num_procs(void) {
+  static int nums = 0;
+  size_t len;
+  if (nums == 0){
+    len = sizeof(int);
+    sysctlbyname("hw.physicalcpu", &nums, &len, NULL, 0);
+  }
+  return nums;
+}
+#endif
+
 /*
 OpenBLAS uses the numbers of CPU cores in multithreading. 
 It can be set by openblas_set_num_threads(int num_threads);

exports/Makefile

@@ -119,7 +119,8 @@ so : ../$(LIBSONAME)
 
 endif
 
-ifeq ($(OSNAME), FreeBSD)
+#http://stackoverflow.com/questions/7656425/makefile-ifeq-logical-or
+ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD NetBSD))
 
 so : ../$(LIBSONAME)
 

exports/gensymbol

@@ -2669,7 +2669,8 @@
 if ($ARGV[5] == 1) {
 	#NO_LAPACK=1
 	@underscore_objs = (@blasobjs, @misc_underscore_objs);
-} elsif (-d "../lapack-3.1.1" || -d "../lapack-3.4.0" || -d "../lapack-3.4.1") {
+} elsif (-d "../lapack-3.1.1" || -d "../lapack-3.4.0" || -d "../lapack-3.4.1" ||
+         -d "../lapack-3.4.2") {
     @underscore_objs = (@blasobjs, @lapackobjs, @lapackobjs2, @misc_underscore_objs);
 } else {
     @underscore_objs = (@blasobjs, @lapackobjs, @misc_underscore_objs);

kernel/x86_64/dgemv_t.S

@@ -47,7 +47,7 @@
 
 #ifndef WINDOWS_ABI
 
-#define STACKSIZE	64
+#define STACKSIZE	128
 	
 #define OLD_M	  %rdi
 #define OLD_N	  %rsi
@@ -57,7 +57,10 @@
 #define STACK_Y		16 + STACKSIZE(%rsp)
 #define STACK_INCY	24 + STACKSIZE(%rsp)
 #define STACK_BUFFER	32 + STACKSIZE(%rsp)
-
+#define MMM	56(%rsp)
+#define NN	64(%rsp)
+#define AA	72(%rsp)
+#define LDAX	80(%rsp)
 #else
 
 #define STACKSIZE	256
@@ -132,27 +135,11 @@
 	movq	OLD_LDA,      LDA
 	movq	OLD_X,        X
 #else
-	movq	OLD_M,	      M
+	movq	OLD_M,	      MMM
-	movq	OLD_N,        N
+	movq	OLD_N,        NN
-	movq	OLD_A,        A
+	movq	OLD_A,        AA
-	movq	OLD_LDA,      LDA
+	movq	OLD_LDA,      LDAX
 #endif
-
-	movq	STACK_INCX,   INCX
-	movq	STACK_Y,      Y
-	movq	STACK_INCY,   INCY
-	movq	STACK_BUFFER, BUFFER
-
-	leaq	-1(INCX), %rax
-
-	leaq	(,LDA,  SIZE), LDA
-	leaq	(,INCX, SIZE), INCX
-	leaq	(,INCY, SIZE), INCY
-
-	leaq	(LDA, LDA, 2), LDA3
-
-	subq	$-16 * SIZE, A
-
 #ifdef HAVE_SSE3
 #ifndef WINDOWS_ABI
 	movddup	%xmm0, ALPHA
@@ -168,6 +155,39 @@
 	unpcklpd ALPHA, ALPHA
 #endif
 
+
+
+.L0x:
+	xorq	M,M
+	addq	$1,M
+	salq	$22,M
+	subq	M,MMM
+	jge .L00
+
+	movq	MMM,%rax
+	addq	M,%rax
+	jle	.L999x
+	movq	%rax,M
+
+.L00:	
+	movq	LDAX,LDA
+	movq	NN,N
+	movq	AA,A
+	movq	STACK_INCX,   INCX
+	movq	STACK_Y,      Y
+	movq	STACK_INCY,   INCY
+	movq	STACK_BUFFER, BUFFER
+
+	leaq	-1(INCX), %rax
+
+	leaq	(,LDA,  SIZE), LDA
+	leaq	(,INCX, SIZE), INCX
+	leaq	(,INCY, SIZE), INCY
+
+	leaq	(LDA, LDA, 2), LDA3
+
+	subq	$-16 * SIZE, A
+
 	testq	M, M
 	jle	.L999
 	testq	N, N
@@ -854,7 +874,6 @@
 
 .L21:
 #endif
-
 	subq	$4, N
 
 	leaq	16 * SIZE(BUFFER), X1
@@ -2461,6 +2480,12 @@
 	ALIGN_4
 
 .L999:
+	leaq	(, M, SIZE), %rax
+	addq %rax,AA
+	jmp .L0x;
+	ALIGN_4
+
+.L999x:
 	movq	  0(%rsp), %rbx
 	movq	  8(%rsp), %rbp
 	movq	 16(%rsp), %r12

param.h

@@ -1664,26 +1664,26 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #define XGEMM_DEFAULT_UNROLL_M 1
 #endif
 
-#define SGEMM_P sgemm_p
+#define SGEMM_DEFAULT_P sgemm_p
-#define DGEMM_P dgemm_p
+#define DGEMM_DEFAULT_P dgemm_p
-#define QGEMM_P qgemm_p
+#define QGEMM_DEFAULT_P qgemm_p
-#define CGEMM_P cgemm_p
+#define CGEMM_DEFAULT_P cgemm_p
-#define ZGEMM_P zgemm_p
+#define ZGEMM_DEFAULT_P zgemm_p
-#define XGEMM_P xgemm_p
+#define XGEMM_DEFAULT_P xgemm_p
 
-#define SGEMM_R sgemm_r
+#define SGEMM_DEFAULT_R sgemm_r
-#define DGEMM_R dgemm_r
+#define DGEMM_DEFAULT_R dgemm_r
-#define QGEMM_R qgemm_r
+#define QGEMM_DEFAULT_R qgemm_r
-#define CGEMM_R cgemm_r
+#define CGEMM_DEFAULT_R cgemm_r
-#define ZGEMM_R zgemm_r
+#define ZGEMM_DEFAULT_R zgemm_r
-#define XGEMM_R xgemm_r
+#define XGEMM_DEFAULT_R xgemm_r
 
-#define SGEMM_Q 128
+#define SGEMM_DEFAULT_Q 128
-#define DGEMM_Q 128
+#define DGEMM_DEFAULT_Q 128
-#define QGEMM_Q 128
+#define QGEMM_DEFAULT_Q 128
-#define CGEMM_Q 128
+#define CGEMM_DEFAULT_Q 128
-#define ZGEMM_Q 128
+#define ZGEMM_DEFAULT_Q 128
-#define XGEMM_Q 128
+#define XGEMM_DEFAULT_Q 128
 
 #define SYMV_P	16
 

patch.for_lapack-3.4.2

@@ -899,19 +899,64 @@ diff -ruN lapack-3.4.2.old/TESTING/LIN/Makefile lapack-3.4.2/TESTING/LIN/Makefil
  ../xlintsts: xlintsts
  	mv xlintsts $@
 diff -ruN lapack-3.4.2.old/lapacke/src/Makefile lapack-3.4.2/lapacke/src/Makefile
---- lapack-3.4.2.old/lapacke/src/Makefile	2012-04-02 22:16:32 +0200
+--- lapack-3.4.2.old/lapacke/src/Makefile	2012-09-21 04:21:29 +0200
-+++ lapack-3.4.2/lapacke/src/Makefile	2012-04-22 21:38:38 +0200
++++ lapack-3.4.2/lapacke/src/Makefile	2012-10-15 22:04:56 +0200
-@@ -2041,19 +2041,21 @@
+@@ -34,7 +34,7 @@
+ #
+ include ../../make.inc
+ 
+-SRC_OBJ = \
++CSRC_OBJ = \
+ lapacke_cbbcsd.o \
+ lapacke_cbbcsd_work.o \
+ lapacke_cbdsqr.o \
+@@ -526,7 +526,9 @@
+ lapacke_cupgtr.o \
+ lapacke_cupgtr_work.o \
+ lapacke_cupmtr.o \
+-lapacke_cupmtr_work.o \
++lapacke_cupmtr_work.o
++
++DSRC_OBJ = \
+ lapacke_dbbcsd.o \
+ lapacke_dbbcsd_work.o \
+ lapacke_dbdsdc.o \
+@@ -1012,7 +1014,9 @@
+ lapacke_dtrttp.o \
+ lapacke_dtrttp_work.o \
+ lapacke_dtzrzf.o \
+-lapacke_dtzrzf_work.o \
++lapacke_dtzrzf_work.o
++
++SSRC_OBJ = \
+ lapacke_sbbcsd.o \
+ lapacke_sbbcsd_work.o \
+ lapacke_sbdsdc.o \
+@@ -1492,7 +1496,9 @@
+ lapacke_strttp.o \
+ lapacke_strttp_work.o \
+ lapacke_stzrzf.o \
+-lapacke_stzrzf_work.o \
++lapacke_stzrzf_work.o
++
++ZSRC_OBJ = \
+ lapacke_zbbcsd.o \
+ lapacke_zbbcsd_work.o \
+ lapacke_zbdsqr.o \
+@@ -2041,19 +2047,29 @@
  lapacke_zlagsy.o \
  lapacke_zlagsy_work.o
  
 -ALLOBJ = $(SRC_OBJ) $(MATGEN_OBJ) 
-+OBJ_FILES := $(SRC_OBJ)
++COBJ_FILES := $(CSRC_OBJ)
++SOBJ_FILES := $(SSRC_OBJ)
++DOBJ_FILES := $(DSRC_OBJ)
++ZOBJ_FILES := $(ZSRC_OBJ)
  
 -ifdef USEXBLAS
 -ALLXOBJ = $(SXLASRC) $(DXLASRC) $(CXLASRC) $(ZXLASRC)
 +ifdef LAPACKE_EXTENDED
-+OBJ_FILES += $(SXLASRC) $(DXLASRC) $(CXLASRC) $(ZXLASRC)
++OBJ_FILES += $(SRCX_OBJ)
  endif
  
 -
@@ -924,9 +969,32 @@ diff -ruN lapack-3.4.2.old/lapacke/src/Makefile lapack-3.4.2/lapacke/src/Makefil
  
 -../../$(LAPACKELIB): $(ALLOBJ) $(ALLXOBJ)
 -	$(ARCH) $(ARCHFLAGS) ../../$(LAPACKELIB) $(ALLOBJ) $(ALLXOBJ)
-+../../$(LAPACKELIB): $(OBJ_FILES)
++../../$(LAPACKELIB): $(COBJ_FILES) $(DOBJ_FILES) $(SOBJ_FILES) $(ZOBJ_FILES) $(OBJ_FILES)
-+# http://hackage.haskell.org/trac/gtk2hs/ticket/1146
++	$(ARCH) $(ARCHFLAGS) ../../$(LAPACKELIB) $(COBJ_FILES)
-+	echo $(OBJ_FILES) | xargs -n 100 $(ARCH) $(ARCHFLAGS) ../../$(LAPACKELIB)
++	$(ARCH) $(ARCHFLAGS) ../../$(LAPACKELIB) $(DOBJ_FILES)
++	$(ARCH) $(ARCHFLAGS) ../../$(LAPACKELIB) $(SOBJ_FILES)
++	$(ARCH) $(ARCHFLAGS) ../../$(LAPACKELIB) $(ZOBJ_FILES)
++ifneq ($(strip $(OBJ_FILES)),)
++	$(ARCH) $(ARCHFLAGS) ../../$(LAPACKELIB) $(OBJ_FILES)
++endif
  	$(RANLIB) ../../$(LAPACKELIB)
  
  .c.o:
+diff -ruN lapack-3.4.2.old/lapacke/example/Makefile lapack-3.4.2/lapacke/example/Makefile
+--- lapack-3.4.2.old/lapacke/example/Makefile	2012-03-23 06:55:22.000000000 +0800
++++ lapack-3.4.2/lapacke/example/Makefile	2012-11-13 00:32:24.125449952 +0800
+@@ -4,12 +4,12 @@
+ 
+ xexample_DGESV_rowmajor: example_DGESV_rowmajor.o ../../$(LAPACKLIB) ../../$(LAPACKELIB)
+ 	$(LOADER) $(LOADOPTS)  example_DGESV_rowmajor.o \
+-         ../../$(LAPACKELIB) ../../$(LAPACKLIB) $(BLASLIB) -o $@
++         ../../$(LAPACKELIB) $(CEXTRALIB) -o $@
+ 	./$@
+ 
+ xexample_ZGESV_rowmajor: example_ZGESV_rowmajor.o ../../$(LAPACKLIB) ../../$(LAPACKELIB)
+ 	$(LOADER) $(LOADOPTS)  example_ZGESV_rowmajor.o \
+-         ../../$(LAPACKELIB) ../../$(LAPACKLIB) $(BLASLIB) -o $@
++         ../../$(LAPACKELIB) $(CEXTRALIB) -o $@
+ 	./$@
+ 
+ .c.o: