Compare commits

..

198 Commits

Author SHA1 Message Date
Zhang Xianyi
53e849f4fc Merge branch 'develop' 2015-10-27 15:44:50 -05:00
Zhang Xianyi
8447498b50 Update doc for OpenBLAS 0.2.15 version. [CI skipped] 2015-10-27 15:44:35 -05:00
Zhang Xianyi
b6519159f5 Merge pull request #674 from j-bo/develop
Fix #673
2015-10-27 10:49:22 -05:00
Zhang Xianyi
63c56d3da9 Only include complex.h since Android 5.0 2015-10-27 10:47:55 -05:00
j-bo
718d0f18e3 Merge pull request #1 from j-bo/j-bo-patch-673
Fix #673
2015-10-27 13:56:45 +01:00
j-bo
6040858b22 Fix #673
Add lacking headers declarations when compiling for Android ARM7
2015-10-27 13:55:24 +01:00
Zhang Xianyi
70642fe4ed Refs #668. Raise the signal when pthread_create fails.
Thank James K. Lowden for the patch.
2015-10-26 19:02:51 -05:00
Zhang Xianyi
79d4a62e10 Add AppVeyor badge. 2015-10-26 18:14:41 -05:00
Zhang Xianyi
1ac8c32f1d [ci skip] Build Visual Studio 12 Win64 on Appveyor 2015-10-26 18:08:54 -05:00
Zhang Xianyi
0b2ad98e48 Only test x64 Windows CI. 2015-10-27 05:11:07 +08:00
Zhang Xianyi
69363622a8 Fix DYNAMIC_ARCH=1 bug. 2015-10-27 05:10:40 +08:00
Zhang Xianyi
e6d754fddc Use AppVeyor Windows CI. 2015-10-26 15:08:17 -05:00
Zhang Xianyi
f74ff6da38 Merge branch 'cmake' into develop 2015-10-26 14:54:59 -05:00
Zhang Xianyi
2feef49fa8 Merge branch 'develop' into cmake
Conflicts:
	driver/others/memory.c
2015-10-26 14:54:34 -05:00
Zhang Xianyi
53b6023a6c Fix cmake bug on MSVC 32-bit. 2015-10-26 14:52:13 -05:00
Zhang Xianyi
309875de3c Fix cmake bug on x86 32-bit.
e.g. Build 32-bit on 64-bit Linux.
cmake -DBINARY=32
2015-10-27 02:54:53 +08:00
Zhang Xianyi
b809f99cee Add CBLAS test for CMAKE. 2015-10-26 23:42:21 +08:00
Zhang Xianyi
5a291606ad Refs #671. the return of i?max cannot larger than N. 2015-10-24 01:16:34 +08:00
Zhang Xianyi
1ce054fcb3 Refs #669. Fixed the build bug with gcc on Mac OS X. 2015-10-22 11:07:35 -05:00
Zhang Xianyi
8fade093aa Fixed cmake bug on Visual Studio. 2015-10-20 14:37:22 -05:00
Zhang Xianyi
96f0bbe067 Fixed cmake bug on haswell. 2015-10-21 02:24:54 +08:00
Zhang Xianyi
d8392c1245 Fixe cmake config bugs. 2015-10-20 04:30:55 +08:00
Zhang Xianyi
aca7d7e953 Detect cmake test result. 2015-10-20 03:35:25 +08:00
Zhang Xianyi
94b125255f Merge branch 'develop' into cmake
Conflicts:
	driver/others/memory.c
2015-10-13 04:46:08 +08:00
Zhang Xianyi
3684706a12 Include time.h. 2015-10-08 15:18:54 +00:00
Zhang Xianyi
90aa8e24b9 Refs #615. Import bug fixes for LAPACKE dormlq. 2015-10-07 02:31:51 +08:00
Zhang Xianyi
11ac4665c8 Fixed #654. Make sure the gotoblas_init function is run before all other static initializations. 2015-10-05 14:14:32 -05:00
Zhang Xianyi
c666158b79 Merge pull request #656 from stevengj/libname
default to lib$(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
2015-10-05 10:25:15 -05:00
Zhang Xianyi
ccf581f94d Merge pull request #659 from Keno/patch-2
Fix cross compilation suffix detection
2015-10-05 10:23:52 -05:00
Keno Fischer
e9493f69eb Fix cross compilation suffix detection
If the path involves `-`, this would have otherwise detected this as a cross compile suffix.
2015-10-05 00:58:07 -04:00
Steven G. Johnson
88bef3bffc default to lib$(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX), as discussed in #646: if you rename the symbols, it is best to rename the library 2015-10-01 15:07:04 -04:00
Zhang Xianyi
f27942a68a Fixed make TARGET=CORTEXA9 and CORTEXA15 bug. 2015-09-26 14:42:44 +00:00
Zhang Xianyi
0cc2b3de0b Merge pull request #652 from larsmans/fixes
Tiny fixes
2015-09-22 10:01:59 -05:00
Lars Buitinck
b9534bbd76 git ignore versioned .so files 2015-09-22 12:01:09 +02:00
Lars Buitinck
45c8b5e756 actually remove cblas_noconst.h
This file hasn't been used since 212463dce9.
2015-09-22 12:00:30 +02:00
Zhang Xianyi
a96a4cb012 Merge pull request #640 from kortschak/dlansy-fix
Fix LAPACK_*lansy routines
2015-09-10 10:36:57 -05:00
Zhang Xianyi
baec8f5cac Refs #638. Fixed compiling bug with clang on Mac OS X. 2015-09-10 10:32:07 -05:00
kortschak
d6e8459f20 Fix LAPACK_*lansy routines
Fixes #639.
2015-09-10 15:32:50 +09:30
Zhang Xianyi
dfe1eef33b Merge branch 'yuyichao-skylake-id' into develop 2015-09-09 10:48:15 -05:00
Zhang Xianyi
cc7cab8a45 Detect other Intel Skylake cores.
http://users.atw.hu/instlatx64/
2015-09-09 10:47:17 -05:00
Yichao Yu
61ae47eb99 Ref #632. Support Intel Skylake by Haswell kernels. 2015-09-09 11:07:33 -04:00
Zhang Xianyi
22353b1727 Merge pull request #634 from kortschak/lantr-trans-prep
Fix lantr preparation for row major matrices
2015-09-09 09:56:07 -05:00
kortschak
efffd28739 Fix lantr preparation for row major matrices 2015-09-09 09:25:48 +09:30
Zhang Xianyi
62cabef857 Merge pull request #633 from grisuthedragon/tune_imatcopy
Improved Ximatcopy when lda==ldb.
2015-09-08 13:59:08 -05:00
Martin Koehler
711ca33bc6 Improved Ximatcopy when lda==ldb.
The Ximatcopy functions create a copy of the input matrix
although they seem to work inplace. The new routines
XIMATCOPY_K_YY perform the operations inplace if the leading
dimension does not change.
2015-09-07 14:36:16 +02:00
Zhang Xianyi
40a3fed6b8 Merge pull request #630 from buffer51/develop
Fixed error in common.h for Android compilation introduced by e12cf11
2015-09-04 13:01:01 -05:00
buffer51
2297a2d989 Fixed error in common.h for Android compilation introduced by e12cf1123e 2015-09-03 20:54:21 -04:00
Zhang Xianyi
5408074941 Add notification. 2015-08-19 22:50:25 -05:00
Zhang Xianyi
bbcdf63bb4 Merge branch 'develop' of github.com:xianyi/OpenBLAS into develop 2015-08-19 22:48:55 -05:00
Zhang Xianyi
43eabab62f Merge pull request #619 from gitter-badger/gitter-badge
Add a Gitter chat badge to README.md
2015-08-19 22:26:20 -05:00
The Gitter Badger
50901943fd Added Gitter badge 2015-08-20 03:21:09 +00:00
Zhang Xianyi
7df0820160 Use C kernels for s/dgemv on x86. 2015-08-19 08:07:47 -05:00
Zhang Xianyi
17ee2237c3 Fixed cmake bug with NO_LAPACK=1 2015-08-18 22:43:42 -05:00
Zhang Xianyi
4b7381b7a4 Merge pull request #617 from notaz/arm_fixes
really fix ARM64 locking
2015-08-17 15:22:37 -05:00
Grazvydas Ignotas
abade3f896 really fix ARM64 locking 2015-08-17 01:27:45 +02:00
Zhang Xianyi
d1349e7a11 Merge pull request #616 from notaz/arm_fixes
ARM fixes
2015-08-16 17:16:18 -05:00
Grazvydas Ignotas
3efeaed0d8 correct a minor mistake 2015-08-16 20:12:04 +02:00
Grazvydas Ignotas
d38a1ddc7a use real armv5 support
there is no more requirement for ARMv6 instructions,
and VFP on ARMv5 is uncommon
2015-08-16 18:59:18 +02:00
Grazvydas Ignotas
6b92204a7c add fallback blas_lock implementation
to be used on armv5 and new platforms
2015-08-16 18:59:17 +02:00
Grazvydas Ignotas
f2ac1a5cee set ARMV7 for Cortex-A9 and Cortex-A15
otherwise some macros like YIELDING are not defined correctly
2015-08-16 18:59:17 +02:00
Grazvydas Ignotas
e12cf1123e add fallback rpcc implementation
- use on arm, arm64 and any new platform
- use faster integer math instead of double
- use similar scale as rdtsc so that timeouts work
2015-08-16 18:59:16 +02:00
Grazvydas Ignotas
d3e2f0a1af add missing barriers
should fix issue #597
2015-08-16 15:37:02 +02:00
Grazvydas Ignotas
c2323dd4d2 really fix ARM locking
- was writing 0 to lock variable, so was ineffective
- only exit loop if both lock was 0 and strex was successful
2015-08-16 15:18:42 +02:00
Zhang Xianyi
f8eba3d548 Fixed cmake build bugs on Linux. 2015-08-11 16:25:16 -05:00
Zhang Xianyi
40ab5cfc50 Merge branch 'hpanderson_cmake' into cmake 2015-08-11 03:31:55 +08:00
Zhang Xianyi
b7a8f9ad47 Merge branch 'cmake' of https://github.com/hpanderson/OpenBLAS into hpanderson_cmake 2015-08-11 03:31:07 +08:00
Zhang Xianyi
f874465bb8 Use cmake to build OpenBLAS GENERIC Target on MSVC x86 64-bit.
Disable CBLAS and LAPACK.
2015-08-10 14:10:44 -05:00
Zhang Xianyi
bb6e050509 Merge pull request #614 from xantares/cmake_version
install OpenBLASConfigVersion.cmake
2015-08-06 13:15:51 -05:00
xantares
87336b9acf install OpenBLASConfigVersion.cmake 2015-08-06 20:03:50 +02:00
Hank Anderson
19664f3ef4 Added missing lapacke.cmake file. 2015-08-06 07:40:06 -05:00
Zhang Xianyi
c50661e5b7 Merge pull request #613 from fabioperez/develop
Add POWER7/POWER8 as targets
2015-08-05 09:19:17 -05:00
Fábio Perez
b8d64a856a Add POWER7/POWER8 as targets 2015-08-05 11:02:39 -03:00
Zhang Xianyi
898fc7552a Merge pull request #612 from ibmsoe/ppc64le
ppc64le platform support (ELF ABI v2)
2015-08-04 16:58:24 -05:00
Zhang Xianyi
ab0a0a75fc Merge branch 'develop' into cmake 2015-08-03 23:59:01 -05:00
Zhang Xianyi
1cf2b10224 Use pure C generic target on x86 and x86_64.
make TARGET=GENERIC

?gemm3m is unimplemented on generic target.
2015-08-03 23:55:56 -05:00
Zhang Xianyi
7ac7e147d4 Fixed cmake building bugs on Linux. Disable LAPACK by default. 2015-08-04 04:37:05 +08:00
Matthew Brandyberry
7ba4fe5afb ppc64le platform support (ELF ABI v2) 2015-07-21 22:20:19 -05:00
Zhang Xianyi
a55377e9a4 Merge branch 'hpanderson_cmake' into cmake 2015-07-22 04:07:27 +08:00
Zhang Xianyi
dcd5ba4443 Merge branch 'cmake' of https://github.com/hpanderson/OpenBLAS into hpanderson_cmake 2015-07-22 04:06:39 +08:00
Zhang Xianyi
d0c51c4de9 Merge branch 'develop' 2015-03-24 15:07:07 -05:00
Hank Anderson
1d183dcda8 Added lapacke sources. 2015-02-25 16:51:08 -06:00
Hank Anderson
e19bf3a28b Removed MSVC cpuid func when using clang. 2015-02-25 14:44:49 -06:00
Hank Anderson
3649cfbd7b Fixed EPILOGUE for clang. 2015-02-25 12:23:26 -06:00
Hank Anderson
5ae8993752 Added intrinsics for MSVC. 2015-02-25 11:52:51 -06:00
Hank Anderson
84d90d6ed8 Fixed some compiler errors/warnings for clang. 2015-02-25 11:52:25 -06:00
Hank Anderson
518e2424a8 Fixed bad filename for cpuid.S compile. 2015-02-25 11:51:29 -06:00
Hank Anderson
00e373aea6 Added LAPACK sources directly to add_library call instead of OBJECT. 2015-02-25 10:18:18 -06:00
Hank Anderson
9eaea02f33 Added additional gemm defines for complex types. 2015-02-25 09:39:11 -06:00
Hank Anderson
ab7043373f Fixed bug generating trmv complex source names. 2015-02-24 15:18:41 -06:00
Hank Anderson
504cdb10ed Added check for MSVC before enabling fortran.
Currently forcing gfortran, instead of assuming ifort.
2015-02-24 14:31:45 -06:00
Hank Anderson
a8002b0c5f Separated getarch ASM file when using MSVC. 2015-02-24 14:31:18 -06:00
Hank Anderson
0553476fba Added TRANS defines for complex sources in lapack. 2015-02-24 14:30:35 -06:00
Hank Anderson
2416d9dbac Fixed TRANSA defines for complex sources in driver/level2. 2015-02-24 13:18:07 -06:00
Hank Anderson
0d8e227ea7 Changed strategy for setting preprocessor definitions.
Instead of generating separate object files for each permutation of
defines for a source file, GenerateNamedObjects now writes an entirely
new source file and inserts the defines as #define c statements.

This solves a problem I ran into with ar.exe where it was refusing to
link objects that had the same filename despite having different paths.
2015-02-24 12:26:33 -06:00
Hank Anderson
12d1fb2e40 Fixed incorrect object name in kernel CMakeLists.txt 2015-02-24 10:30:16 -06:00
Hank Anderson
1b7f427401 Added conj gemv objects for complex build. 2015-02-23 10:24:31 -06:00
Hank Anderson
b2284647a3 More complex objects. 2015-02-23 07:51:05 -06:00
Hank Anderson
a6116e5859 Added some more complex-only objects. 2015-02-22 17:49:28 -06:00
Hank Anderson
fb5d5bb971 Added defines for complex trmv. 2015-02-21 12:39:03 -06:00
Hank Anderson
371071d461 Added CONJ defines for trmm/trsm. 2015-02-21 10:59:02 -06:00
Hank Anderson
8a143516e3 Added alternate_name to a couple of the name mangling schemes.
Added zherk_k sources to driver/level3.
2015-02-20 17:03:33 -06:00
Hank Anderson
e5897ecb9b Added zherk_kernel.c objects to driver/level3. 2015-02-19 16:19:56 -06:00
Hank Anderson
714638c187 Added some TRMM objects for complex types. 2015-02-19 16:11:51 -06:00
Hank Anderson
e27c372e53 Fixed reuse of float_char from parent loop.
Fixed in/it/on/otcopy names.
2015-02-19 13:53:29 -06:00
Hank Anderson
f3f2b3d768 Added complex and single netlib-lapack fortran sources to lapack.cmake. 2015-02-19 12:26:11 -06:00
Hank Anderson
9492298048 Added other float types to Makefile.L3. 2015-02-18 13:01:05 -06:00
Hank Anderson
43725b82c5 ParseMakefileVars now replaces Makefile vars with CMake vars. 2015-02-18 12:23:17 -06:00
Hank Anderson
14fd3d35de Added checks for missing defines in kernel. 2015-02-18 10:25:01 -06:00
Hank Anderson
cebc07cebd ParseMakefileVars now recursively parses included makefiles. 2015-02-17 22:09:41 -06:00
Hank Anderson
33c5e8db7f Added a helper function for setting the L1 kernel defaults.
Added loop to build objects with different KERNEL defines.
2015-02-17 21:36:23 -06:00
Hank Anderson
67e39bd8fb Added mangled complex filenames to interface and lapack CMakeLists.txt. 2015-02-17 13:12:30 -06:00
Hank Anderson
9eb1499095 Added another param to GenerateNamedObjects to mangle complex source names.
There are a lot of sources for complex float types that are the same
names as the real sources, except with z prepended.
2015-02-17 10:30:28 -06:00
Hank Anderson
4662a0b13a Changed generate functions to iterate through a list of float types.
This will generate obj files for SINGLE/DOUBLE/COMPLEX/DOUBLE COMPLEX.
2015-02-15 17:44:37 -06:00
Hank Anderson
e74462a3f5 Moved declarations to start of functions to satisfy MSVC C89 implementation. 2015-02-11 11:16:57 -06:00
Hank Anderson
056ba26755 Changed a number of inline calls to use __inline.
MSVC doesn't inmplement C99, so can't use the inline keyword. __inline
appears to work in MSVC and GCC.
2015-02-11 11:13:17 -06:00
Hank Anderson
a0d9a7fd83 Changed _Complex types in common_level1.h to use the typedef. 2015-02-11 11:12:14 -06:00
Hank Anderson
5d3fc092e9 Added MSVC defines to common.h.
Don't have unistd.h in MSVC.

Chagned YIELDING to use the YeildProcessor macro.
2015-02-11 11:10:45 -06:00
Hank Anderson
c94fe71278 Removed incoming-stack-boundary for MSVC.
Made float type optional for GenerateNamedObjects.

Called GenerateNamedObjects for a couple of driver/others files that
needed NAME/CNAME set.
2015-02-11 10:54:14 -06:00
Hank Anderson
d60b49e5c5 Turned off uninizialized variable warning when compiling lapack-netlib. 2015-02-10 14:36:43 -06:00
Hank Anderson
64b5a0ef84 Added AUX files from lapack-netlib. 2015-02-10 14:29:05 -06:00
Hank Anderson
162791e30e Added common objects from kernel Makefile. 2015-02-10 12:42:05 -06:00
Hank Anderson
8743093bd7 Added aux files from lapack-netlib. 2015-02-10 11:47:46 -06:00
Hank Anderson
96cf6779ca Added DLA sources from lapack-netlib.
Can't use the lapack-netlib cmake files, since they are designed to
build a complete lapack/blas library. They have their own fortran
detection and flag setup and so on. Instead I'll just recreate the
makefiles I need.

Fixed a typo in the NAME defines.
2015-02-10 11:01:01 -06:00
Hank Anderson
3b20b62423 Fixed trti2 name. 2015-02-09 15:29:28 -06:00
Hank Anderson
6ddbfea700 Added generic laswp object. 2015-02-09 15:15:58 -06:00
Hank Anderson
c0624a26be Fixed some dgemm_copy function names. 2015-02-09 14:34:29 -06:00
Hank Anderson
4bfaf1ce66 Removed some list appends I missed. 2015-02-09 12:56:55 -06:00
Hank Anderson
e8c39138c6 Removed return value from GenerateNamedObjects.
It sets DBLAS_OBJS directly to save a bunch of list appending in the
CMakeLists.txt files.
2015-02-09 12:28:09 -06:00
Hank Anderson
f992799226 Added the rest of Makefile.L3. 2015-02-09 10:47:35 -06:00
Hank Anderson
4c65afcce1 Changed kernel filenames to vars. These will need to be read from KERNEL.
Added some kernel/L3 objects.
2015-02-09 09:52:14 -06:00
Hank Anderson
7fa5c4e2fd Fixed some case issues with ARCH.
Added some kernel and driver/others objects.
2015-02-08 15:29:18 -06:00
Hank Anderson
fa0e6a6c93 Added the rest of the L1 kernel makefile. 2015-02-07 21:37:46 -06:00
Hank Anderson
2f59135eb6 Added gemv to level2 CMakeLists.txt. 2015-02-07 21:15:21 -06:00
Hank Anderson
38681fb1c6 Added more kernel files. 2015-02-07 12:54:30 -06:00
Hank Anderson
6b5d26e07b Added SMP sources to level2 CMakeLists.txt. 2015-02-06 16:52:19 -06:00
Hank Anderson
13d2d48e67 Added yet another naming scheme for lapack functions. 2015-02-06 13:42:20 -06:00
Hank Anderson
189fadfde0 Started implementing kernel/Makefile in cmake. 2015-02-05 21:05:11 -06:00
Hank Anderson
627d5e7401 Added SMP objects to driver/level3. 2015-02-05 12:22:48 -06:00
Hank Anderson
943fa2fb58 Fixed object names in level2. 2015-02-05 10:49:11 -06:00
Hank Anderson
1b62a4f3c9 Changed some function parameters to optional. 2015-02-05 09:39:40 -06:00
Hank Anderson
461e691127 Codes when define is absent are now a parameter to AllCombinations.
The level3 object names should now be correct.
2015-02-05 09:23:47 -06:00
Hank Anderson
cfaf1c678f Added option to append define codes with an underscore.
Fixed the code array not getting reset on subsequent AllCombinations
calls.
2015-02-05 09:17:18 -06:00
Hank Anderson
0d7bad1f35 Changed GenerateObjects to append combination codes (e.g. dtrmm_TU). 2015-02-05 09:02:54 -06:00
Hank Anderson
373a1bdadb Converted lapack/Makefile to cmake. 2015-02-04 15:47:10 -06:00
Hank Anderson
2828f6630c Added SMP sources to COMMONOBJS. 2015-02-04 14:01:36 -06:00
Hank Anderson
58cff2fed8 Added CBLAS define/naming convention to GenerateNamedObjects. 2015-02-04 11:30:15 -06:00
Hank Anderson
5690cf3f0e Added override for function names in GenerateNamedObjects.
The BLAS interface folder should now be generated the correct objects
for the DOUBLE case.
2015-02-04 10:52:19 -06:00
Hank Anderson
a0aeda6187 Added function to set defines for the object names (e.g. -DNAME=dgemm). 2015-02-04 10:37:34 -06:00
Hank Anderson
84b3d760c4 Converted rest of Makefile.system to system.cmake. 2015-02-03 16:05:01 -06:00
Hank Anderson
0beea3a5a5 Converted LAPACK flags from Makefile.system. 2015-02-03 15:33:56 -06:00
Hank Anderson
560c96a9a7 Fixed newlines in some cmake files. 2015-02-03 15:11:15 -06:00
Hank Anderson
0ccfa60a53 Changed fortran compiler name to be uppercase and stripped of path/ext. 2015-02-03 15:09:37 -06:00
Hank Anderson
30be551502 Corrected fortran compiler name variables.
Fixed some typos.

Updated c_check to set ARCH and BINARY64/32.

Added version variables.
2015-02-03 14:21:22 -06:00
Hank Anderson
be1ce38f24 Fixed some missing parentheses. 2015-02-03 14:00:29 -06:00
Hank Anderson
e818ace11a Ported more of Makefile.system to CMake. 2015-02-03 13:34:41 -06:00
Hank Anderson
e4bfbd8258 Added fc.cmake (forgot it in last commit).
Moved a couple C compiler ifs from Makefile.system into cc.cmake.
2015-02-03 13:08:59 -06:00
Hank Anderson
2d5b442f5b Ported Fortran configuration code from Makefile.system to fc.cmake. 2015-02-03 12:32:23 -06:00
Hank Anderson
af11aff309 Ported C compiler settings from Makefile.system into new cmake file. 2015-02-03 12:00:49 -06:00
Hank Anderson
e66aa5f3b7 Ported arch dependent settings from Makefile.system to new cmake file. 2015-02-03 11:32:20 -06:00
Hank Anderson
31cf22cb4b Ported OS settings from Makefile.system into new cmake file. 2015-02-03 11:07:58 -06:00
Hank Anderson
20e593a44a Added cblas_ objects to interface CMakeLists.
Naming isn't right, though, not seeing cblas_xxxx exports in the
resulting library.
2015-02-02 16:25:30 -06:00
Hank Anderson
7194424fef Added missing common objects to the library. 2015-02-02 15:21:29 -06:00
Hank Anderson
d11bde60d0 DOUBLE define for DBLAS objects is now set in main CMakeLists.txt.
Since the objects are the same, could generate SINGLE/COMPLEX/etc here
without having to rewrite all the object enumeration code again.
2015-02-02 15:00:44 -06:00
Hank Anderson
9e154aba58 Added LAPACK object files to interface CMakeLists. 2015-02-02 12:31:15 -06:00
Hank Anderson
5057a4b4df Added openblas add_library call that uses DBLAS_OBJS ojbects. 2015-01-30 15:21:21 -06:00
Hank Anderson
3e8ea7a351 Added COMMONOBJS to driver/others CMakeLists.txt. 2015-01-30 14:06:14 -06:00
Hank Anderson
d3dcdddf75 Moved functions into util cmake file. 2015-01-30 13:47:40 -06:00
Hank Anderson
e5e7595bf9 Added paramater to GenerateObjects for defines that affect all sources. 2015-01-30 13:31:13 -06:00
Hank Anderson
7693887d61 Added empty set to the combinations generated by AllCombinations. 2015-01-30 13:01:11 -06:00
Hank Anderson
8d9b196e0d Moved loop over define combos into a function.
This function takes a set of sources and a set of preprocessor
definitions. It will iterate over the sources and build an object
file for each combination of preprocessor definitions for each
source file.
2015-01-30 12:14:44 -06:00
Hank Anderson
a6cf8aafc0 Updated level3/CMakeLists with correct defines using all combos. 2015-01-30 11:21:50 -06:00
Hank Anderson
dbdca7bf0c Added first pass at driver/level3 Makefile conversion.
Added a rather convoluted CMake function to find all combinations
of a given list. This will be useful for the object files that are
compiled multiple times with different combinations of preprocessor
definitions.
2015-01-29 22:53:11 -06:00
Hank Anderson
dabaecb2bc Moved getarch parsing code into a function. 2015-01-29 09:30:47 -06:00
Hank Anderson
8c23965da3 prebuild.cmake now reads the output from getarch into CMake vars. 2015-01-28 22:57:44 -06:00
Hank Anderson
61f21b5d03 getarch_2nd now appends its output to config.h/config_kernel.h 2015-01-28 22:20:15 -06:00
Hank Anderson
8ede4a8da4 getarch now compiles and sets config.h defines properly.
Still isn't parsed into CMake variables, and getarch_2 needs to
get the same treatment.
2015-01-28 17:18:26 -06:00
Hank Anderson
1c5b6bb4f7 Added CORE define to config.h in prebuild.cmake (temporarily). 2015-01-28 16:33:48 -06:00
Hank Anderson
c5f5c7a076 Updated c_check OS/compiler/bits detection. 2015-01-28 15:47:47 -06:00
Hank Anderson
9a508abdc7 Added first pass at driver/level2 makefile conversion. 2015-01-28 14:52:15 -06:00
Hank Anderson
5eefe18ae4 Added CMakeLists.txt for the first of the BLAS folders.
It only does the double precision compile currently.

I realized I didn't finish converting Makefile.system yet, so I made
a note of that.
2015-01-27 16:17:17 -06:00
Hank Anderson
1e8bb0e0e0 Fixed architecture detection when AMD64 in c_check. 2015-01-27 14:03:46 -06:00
Hank Anderson
864b8b31de Fixed incorrect case in OS_ definition in c_check. 2015-01-27 13:54:29 -06:00
Hank Anderson
d2d15e522f Started converting lib target to CMake.
The main part of this target is looping through the BLAS subfolders
and calling make on them. Need to add CMakeLists.txt for each of these
subfolders.
2015-01-27 12:23:35 -06:00
Hank Anderson
f4d1e7a265 Hardcoded NUM_CORES to get system.cmake working. 2015-01-27 11:37:39 -06:00
Hank Anderson
0f6bec0a32 cmake.prebuild now compiles getarch.
Doesn't actually run it yet.
2015-01-01 21:03:17 -06:00
Hank Anderson
92cdac5f87 Added MSVC functions to cpuid_x86.c to replace gcc-specific ASM. 2015-01-01 21:02:48 -06:00
Hank Anderson
1a41022e3e Added MSVC defines to cpuid.h and getarch.c. 2015-01-01 21:01:28 -06:00
Hank Anderson
e5c47e44f6 First pass at converting a few makefiles to CMake. 2014-12-30 21:53:00 -06:00
Zhang Xianyi
51ce5ef447 Merge branch 'develop' 2014-12-03 23:14:21 +08:00
Zhang Xianyi
37aee1f9b1 Merge branch 'develop' 2014-12-03 23:01:33 +08:00
Zhang Xianyi
7e4e195e82 Merge branch 'develop' 2014-10-13 17:10:41 +08:00
Zhang Xianyi
a7126c2ce4 Merge branch 'develop' 2014-08-18 11:16:14 +08:00
Zhang Xianyi
f20c0f9819 Merge branch 'develop' 2014-07-17 15:15:57 +08:00
Zhang Xianyi
21b5347fbe Merge branch 'develop' 2014-07-16 18:04:30 +08:00
Zhang Xianyi
f9991fd5f6 Merge branch 'develop' 2014-07-09 08:48:00 +08:00
Zhang Xianyi
da3d70420a Merge branch 'develop' 2014-06-29 10:46:22 +08:00
Zhang Xianyi
f773f492f3 Merge branch 'develop' 2014-06-10 21:55:47 +08:00
Zhang Xianyi
3e068e78e2 Merge branch 'release-0.2.9' 2014-03-06 17:45:31 +08:00
218 changed files with 8901 additions and 1504 deletions

3
.gitignore vendored
View File

@@ -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.*

View File

@@ -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

190
CMakeLists.txt Normal file
View File

@@ -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

View File

@@ -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]

View File

@@ -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

View File

@@ -26,8 +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

View File

@@ -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
@@ -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!

View File

@@ -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
@@ -114,9 +114,6 @@ NO_AFFINITY = 1
# Support for IEEE quad precision(it's *real* REAL*16)( under testing)
# QUAD_PRECISION = 1
# Support for integer matrix and vector (e.g. iaxpy)
# INTEGER_PRECISION = 1
# Theads are still working for a while after finishing BLAS operation
# to reduce thread activate/deactivate overhead. You can determine
# time out to improve performance. This number should be from 4 to 30
@@ -172,6 +169,9 @@ COMMON_PROF = -pg
# 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=

View File

@@ -309,10 +309,6 @@ CCOMMON_OPT += -DQUAD_PRECISION
NO_EXPRECISION = 1
endif
ifdef INTEGER_PRECISION
CCOMMON_OPT += -DINTEGER_PRECISION
endif
ifneq ($(ARCH), x86)
ifneq ($(ARCH), x86_64)
NO_EXPRECISION = 1
@@ -340,6 +336,11 @@ ifeq ($(ARCH), x86)
ifndef BINARY
NO_BINARY_MODE = 1
endif
ifeq ($(CORE), generic)
NO_EXPRECISION = 1
endif
ifndef NO_EXPRECISION
ifeq ($(F_COMPILER), GFORTRAN)
# ifeq logical or. GCC or LSB
@@ -358,6 +359,11 @@ endif
endif
ifeq ($(ARCH), x86_64)
ifeq ($(CORE), generic)
NO_EXPRECISION = 1
endif
ifndef NO_EXPRECISION
ifeq ($(F_COMPILER), GFORTRAN)
# ifeq logical or. GCC or LSB
@@ -874,12 +880,6 @@ ifdef USE_SIMPLE_THREADED_LEVEL3
CCOMMON_OPT += -DUSE_SIMPLE_THREADED_LEVEL3
endif
ifndef LIBNAMESUFFIX
LIBPREFIX = libopenblas
else
LIBPREFIX = libopenblas_$(LIBNAMESUFFIX)
endif
ifndef SYMBOLPREFIX
SYMBOLPREFIX =
endif
@@ -888,6 +888,12 @@ ifndef SYMBOLSUFFIX
SYMBOLSUFFIX =
endif
ifndef LIBNAMESUFFIX
LIBPREFIX = lib$(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)
else
LIBPREFIX = lib$(SYMBOLPREFIX)openblas$(SYMBOLSUFFIX)_$(LIBNAMESUFFIX)
endif
KERNELDIR = $(TOPDIR)/kernel/$(ARCH)
include $(TOPDIR)/Makefile.$(ARCH)

View File

@@ -4,7 +4,6 @@ QBLASOBJS_P = $(QBLASOBJS:.$(SUFFIX)=.$(PSUFFIX))
CBLASOBJS_P = $(CBLASOBJS:.$(SUFFIX)=.$(PSUFFIX))
ZBLASOBJS_P = $(ZBLASOBJS:.$(SUFFIX)=.$(PSUFFIX))
XBLASOBJS_P = $(XBLASOBJS:.$(SUFFIX)=.$(PSUFFIX))
IBLASOBJS_P = $(IBLASOBJS:.$(SUFFIX)=.$(PSUFFIX))
COMMONOBJS_P = $(COMMONOBJS:.$(SUFFIX)=.$(PSUFFIX))
@@ -23,18 +22,12 @@ BLASOBJS += $(QBLASOBJS) $(XBLASOBJS)
BLASOBJS_P += $(QBLASOBJS_P) $(XBLASOBJS_P)
endif
ifdef INTEGER_PRECISION
BLASOBJS += $(IBLASOBJS)
BLASOBJS_P += $(IBLASOBJS_P)
endif
$(SBLASOBJS) $(SBLASOBJS_P) : override CFLAGS += -UDOUBLE -UCOMPLEX
$(DBLASOBJS) $(DBLASOBJS_P) : override CFLAGS += -DDOUBLE -UCOMPLEX
$(QBLASOBJS) $(QBLASOBJS_P) : override CFLAGS += -DXDOUBLE -UCOMPLEX
$(CBLASOBJS) $(CBLASOBJS_P) : override CFLAGS += -UDOUBLE -DCOMPLEX
$(ZBLASOBJS) $(ZBLASOBJS_P) : override CFLAGS += -DDOUBLE -DCOMPLEX
$(XBLASOBJS) $(XBLASOBJS_P) : override CFLAGS += -DXDOUBLE -DCOMPLEX
$(IBLASOBJS) $(IBLASOBJS_P) : override CFLAGS += -DINTEGER -UCOMPLEX
$(SBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
$(DBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
@@ -42,7 +35,6 @@ $(QBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
$(CBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
$(ZBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
$(XBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
$(IBLASOBJS_P) : override CFLAGS += -DPROFILE $(COMMON_PROF)
libs :: $(BLASOBJS) $(COMMONOBJS)
$(AR) $(ARFLAGS) -ru $(TOPDIR)/$(LIBNAME) $^

View File

@@ -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.

View File

@@ -44,6 +44,8 @@ NANO
POWER4
POWER5
POWER6
POWER7
POWER8
PPCG4
PPC970
PPC970MP

42
appveyor.yml Normal file
View File

@@ -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!

View File

@@ -30,7 +30,7 @@ if ($ARGV[0] =~ /(.*)(-[.\d]+)/) {
$cross_suffix = $1;
}
} else {
if ($ARGV[0] =~ /(.*-)(.*)/) {
if ($ARGV[0] =~ /([^\/]*-)([^\/]*$)/) {
$cross_suffix = $1;
}
}

View File

@@ -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

115
cmake/arch.cmake Normal file
View File

@@ -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 ()

89
cmake/c_check.cmake Normal file
View File

@@ -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")

103
cmake/cc.cmake Normal file
View File

@@ -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 ()

60
cmake/export.cmake Normal file
View File

@@ -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()

66
cmake/f_check.cmake Normal file
View File

@@ -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)

200
cmake/fc.cmake Normal file
View File

@@ -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 ()

165
cmake/kernel.cmake Normal file
View File

@@ -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 ()

347
cmake/lapack.cmake Normal file
View File

@@ -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}")

2067
cmake/lapacke.cmake Normal file

File diff suppressed because it is too large Load Diff

104
cmake/os.cmake Normal file
View File

@@ -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 ()

113
cmake/prebuild.cmake Normal file
View File

@@ -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})

552
cmake/system.cmake Normal file
View File

@@ -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

346
cmake/utils.cmake Normal file
View File

@@ -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 ()

110
common.h
View File

@@ -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>
@@ -95,6 +98,10 @@ extern "C" {
#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
@@ -114,6 +121,7 @@ extern "C" {
#include <sys/shm.h>
#endif
#include <sys/time.h>
#include <time.h>
#include <unistd.h>
#include <math.h>
#ifdef SMP
@@ -276,11 +284,6 @@ typedef int blasint;
#define SIZE 8
#define BASE_SHIFT 3
#define ZBASE_SHIFT 4
#elif defined(INTEGER) //extend for integer matrix
#define FLOAT int
#define SIZE 4
#define BASE_SHIFT 2
#define ZBASE_SHIFT 3
#else
#define FLOAT float
#define SIZE 4
@@ -298,13 +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))
@@ -318,8 +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");
@@ -415,7 +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"
@@ -461,18 +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
@@ -489,6 +567,10 @@ typedef char* env_var_t;
#endif
#endif
#if defined(C_MSVC)
#define inline __inline
#endif
#ifndef ASSEMBLER
#ifndef MIN

View File

@@ -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)

View File

@@ -51,6 +51,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#ifndef ASSEMBLER
#if defined(ARMV6) || defined(ARMV7) || defined(ARMV8)
static void __inline blas_lock(volatile BLASULONG *address){
int register ret;
@@ -59,40 +61,29 @@ static void __inline blas_lock(volatile BLASULONG *address){
while (*address) {YIELDING;};
__asm__ __volatile__(
"1: \n\t"
"ldrex r2, [%1] \n\t"
"mov r2, #0 \n\t"
"strex r3, r2, [%1] \n\t"
"cmp r3, #0 \n\t"
"bne 1b \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")
@@ -140,4 +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

View File

@@ -45,42 +45,26 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
static void __inline blas_lock(volatile BLASULONG *address){
int register ret;
int register tmp;
long register ret;
do {
while (*address) {YIELDING;};
__asm__ __volatile__(
"1: \n\t"
"ldaxr %2, [%1] \n\t"
"mov %2, #0 \n\t"
"stlxr %w0, %2, [%1] \n\t"
"cbnz %w0, 1b \n\t"
"mov %0 , #0 \n\t"
: "=r"(ret), "=r"(address), "=r"(tmp)
: "1"(address)
: "memory", "%w0"
//, "%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;
}

View File

@@ -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

View File

@@ -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

View File

@@ -1,9 +0,0 @@
#ifndef COMMON_I_H
#define COMMON_I_H
#ifndef DYNAMIC_ARCH
#define IAXPYU_K iaxpy_k
#else
#error
#endif
#endif

View File

@@ -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);

View File

@@ -93,7 +93,6 @@ openblas_complex_xdouble BLASFUNC(xdotc) (blasint *, xdouble *, blasint *, xdo
void BLASFUNC(saxpy) (blasint *, float *, float *, blasint *, float *, blasint *);
void BLASFUNC(daxpy) (blasint *, double *, double *, blasint *, double *, blasint *);
void BLASFUNC(iaxpy) (blasint *, int *, int *, blasint *, int *, blasint *);
void BLASFUNC(qaxpy) (blasint *, xdouble *, xdouble *, blasint *, xdouble *, blasint *);
void BLASFUNC(caxpy) (blasint *, float *, float *, blasint *, float *, blasint *);
void BLASFUNC(zaxpy) (blasint *, double *, double *, blasint *, double *, blasint *);

View File

@@ -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);
@@ -60,8 +60,6 @@ int daxpy_k (BLASLONG, BLASLONG, BLASLONG, double,
double *, BLASLONG, double *, BLASLONG, double *, BLASLONG);
int qaxpy_k (BLASLONG, BLASLONG, BLASLONG, xdouble,
xdouble *, BLASLONG, xdouble *, BLASLONG, xdouble *, BLASLONG);
int iaxpy_k (BLASLONG, BLASLONG, BLASLONG, int,
int *, BLASLONG, int *, BLASLONG, int *, BLASLONG);
int caxpy_k (BLASLONG, BLASLONG, BLASLONG, float, float,
float *, BLASLONG, float *, BLASLONG, float *, BLASLONG);
int zaxpy_k (BLASLONG, BLASLONG, BLASLONG, double, double,

View File

@@ -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);

View File

@@ -47,10 +47,6 @@
#include "common_z.h"
#include "common_x.h"
#ifdef INTEGER_PRECISION
#include "common_i.h"
#endif
#ifndef COMPLEX
#ifdef XDOUBLE
@@ -638,10 +634,12 @@
#define OMATCOPY_K_RN DOMATCOPY_K_RN
#define OMATCOPY_K_CT DOMATCOPY_K_CT
#define OMATCOPY_K_RT DOMATCOPY_K_RT
#define GEADD_K DGEADD_K
#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
#elif defined(INTEGER)
#define AXPYU_K IAXPYU_K
#define GEADD_K DGEADD_K
#else
#define AMAX_K SAMAX_K
@@ -938,6 +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
@@ -1754,6 +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
@@ -2167,6 +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

View File

@@ -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

View File

@@ -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);

View File

@@ -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

View File

@@ -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

View File

@@ -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

View File

@@ -65,7 +65,6 @@ extern int blas_omp_linked;
#define BLAS_XDOUBLE 0x0002U
#define BLAS_REAL 0x0000U
#define BLAS_COMPLEX 0x0004U
#define BLAS_INTEGER 0x0008U
#define BLAS_TRANSA 0x0030U /* 2bit */
#define BLAS_TRANSA_N 0x0000U

View File

@@ -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
@@ -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 \

View File

@@ -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[];

View File

@@ -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

View File

@@ -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

View File

@@ -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");

View File

@@ -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;

View File

@@ -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
@@ -1133,6 +1159,16 @@ int get_cpuname(void){
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;
@@ -1147,6 +1183,17 @@ int get_cpuname(void){
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;
@@ -1608,6 +1655,16 @@ int get_coretype(void){
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;
@@ -1622,6 +1679,17 @@ int get_coretype(void){
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;

46
ctest/CMakeLists.txt Normal file
View File

@@ -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()

View File

@@ -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})

View File

@@ -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

View File

@@ -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;

View File

@@ -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;

View File

@@ -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

View File

@@ -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

View File

@@ -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

View File

@@ -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) {

View File

@@ -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);

View File

@@ -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);

View File

@@ -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);

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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})

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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;

View File

@@ -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})

View File

@@ -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
@@ -189,20 +191,6 @@ static void legacy_exec(void *func, int mode, blas_arg_t *args, void *sb){
args -> b, args -> ldb,
args -> c, args -> ldc, sb);
} else
#endif
#ifdef INTEGER_PRECISION
if (mode & BLAS_INTEGER){
/* REAL / Extended Double */
void (*afunc)(BLASLONG, BLASLONG, BLASLONG, int,
int *, BLASLONG, int *, BLASLONG,
int *, BLASLONG, void *) = func;
afunc(args -> m, args -> n, args -> k,
((int *)args -> alpha)[0],
args -> a, args -> lda,
args -> b, args -> ldb,
args -> c, args -> ldc, sb);
} else
#endif
if (mode & BLAS_DOUBLE){
/* REAL / Double */
@@ -279,7 +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;
@@ -439,6 +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;
@@ -468,7 +460,7 @@ static int blas_thread_server(void *arg){
//pthread_exit(NULL);
return 0;
return NULL;
}
#ifdef MONITOR
@@ -575,14 +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);
}
}
}
@@ -789,7 +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",
@@ -837,10 +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
}

View File

@@ -252,6 +252,15 @@ static gotoblas_t *get_coretype(void){
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
@@ -263,6 +272,15 @@ static gotoblas_t *get_coretype(void){
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:

View File

@@ -139,8 +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;
@@ -795,12 +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,
@@ -1153,6 +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
@@ -1399,6 +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) {

View File

@@ -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 */
@@ -546,7 +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"
@@ -748,7 +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 " \
@@ -763,7 +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 " \
@@ -796,8 +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
@@ -829,7 +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
@@ -924,7 +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
@@ -944,7 +949,7 @@ int main(int argc, char *argv[]){
#endif
#if defined(__i386__) || defined(__x86_64__)
#ifdef INTEL_AMD
#ifndef FORCE
get_sse();
#else
@@ -1024,7 +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

166
interface/CMakeLists.txt Normal file
View File

@@ -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})

View File

@@ -253,15 +253,6 @@ XBLAS3OBJS += xgemm3m.$(SUFFIX) xsymm3m.$(SUFFIX) xhemm3m.$(SUFFIX)
endif
endif
ifdef INTEGER_PRECISION
IBLAS1OBJS = \
iaxpy.$(SUFFIX)
IBLAS2OBJS =
IBLAS3OBJS =
endif
endif
HPLOBJS = dgemm.$(SUFFIX) dtrsm.$(SUFFIX) \
@@ -352,9 +343,6 @@ CZBLAS3OBJS = \
cblas_zomatcopy.$(SUFFIX) cblas_zimatcopy.$(SUFFIX) \
cblas_zgeadd.$(SUFFIX)
CIBLAS1OBJS = \
cblas_iaxpy.$(SUFFIX)
ifeq ($(SUPPORT_GEMM3M), 1)
@@ -384,10 +372,6 @@ ZBLAS1OBJS += $(CZBLAS1OBJS)
ZBLAS2OBJS += $(CZBLAS2OBJS)
ZBLAS3OBJS += $(CZBLAS3OBJS)
IBLAS1OBJS += $(CIBLAS1OBJS)
IBLAS2OBJS += $(CIBLAS2OBJS)
IBLAS3OBJS += $(CIBLAS3OBJS)
endif
SBLASOBJS = $(SBLAS1OBJS) $(SBLAS2OBJS) $(SBLAS3OBJS)
@@ -396,7 +380,6 @@ QBLASOBJS = $(QBLAS1OBJS) $(QBLAS2OBJS) $(QBLAS3OBJS)
CBLASOBJS = $(CBLAS1OBJS) $(CBLAS2OBJS) $(CBLAS3OBJS)
ZBLASOBJS = $(ZBLAS1OBJS) $(ZBLAS2OBJS) $(ZBLAS3OBJS)
XBLASOBJS = $(XBLAS1OBJS) $(XBLAS2OBJS) $(XBLAS3OBJS)
IBLASOBJS = $(IBLAS1OBJS) $(IBLAS2OBJS) $(IBLAS3OBJS)
#SLAPACKOBJS = \
# sgetrf.$(SUFFIX) sgetrs.$(SUFFIX) spotrf.$(SUFFIX) sgetf2.$(SUFFIX) \
@@ -475,10 +458,6 @@ ifdef QUAD_PRECISION
FUNCOBJS += $(QBLASOBJS) $(XBLASOBJS)
endif
ifdef INTEGER_PRECISION
FUNCOBJS += $(IBLASOBJS)
endif
FUNCALLFILES = $(FUNCOBJS:.$(SUFFIX)=)
include $(TOPDIR)/Makefile.tail
@@ -497,18 +476,17 @@ endif
clean ::
@rm -f functable.h
level1 : $(SBLAS1OBJS) $(DBLAS1OBJS) $(QBLAS1OBJS) $(CBLAS1OBJS) $(ZBLAS1OBJS) $(XBLAS1OBJS) $(IBLAS1OBJS)
level1 : $(SBLAS1OBJS) $(DBLAS1OBJS) $(QBLAS1OBJS) $(CBLAS1OBJS) $(ZBLAS1OBJS) $(XBLAS1OBJS)
$(AR) $(ARFLAGS) -ru $(TOPDIR)/$(LIBNAME) $^
level2 : $(SBLAS2OBJS) $(DBLAS2OBJS) $(QBLAS2OBJS) $(CBLAS2OBJS) $(ZBLAS2OBJS) $(XBLAS2OBJS) $(IBLAS2OBJS)
level2 : $(SBLAS2OBJS) $(DBLAS2OBJS) $(QBLAS2OBJS) $(CBLAS2OBJS) $(ZBLAS2OBJS) $(XBLAS2OBJS)
$(AR) $(ARFLAGS) -ru $(TOPDIR)/$(LIBNAME) $^
level3 : $(SBLAS3OBJS) $(DBLAS3OBJS) $(QBLAS3OBJS) $(CBLAS3OBJS) $(ZBLAS3OBJS) $(XBLAS3OBJS) $(IBLAS3OBJS)
level3 : $(SBLAS3OBJS) $(DBLAS3OBJS) $(QBLAS3OBJS) $(CBLAS3OBJS) $(ZBLAS3OBJS) $(XBLAS3OBJS)
$(AR) $(ARFLAGS) -ru $(TOPDIR)/$(LIBNAME) $^
$(CSBLASOBJS) $(CSBLASOBJS_P) $(CDBLASOBJS) $(CDBLASOBJS_P) $(CQBLASOBJS) $(CQBLASOBJS_P) \
$(CCBLASOBJS) $(CCBLASOBJS_P) $(CZBLASOBJS) $(CZBLASOBJS_P) $(CXBLASOBJS) $(CXBLASOBJS_P) \
$(CIBLASOBJS) $(CIBLASOBJS_P) : override CFLAGS += -DCBLAS
$(CCBLASOBJS) $(CCBLASOBJS_P) $(CZBLASOBJS) $(CZBLASOBJS_P) $(CXBLASOBJS) $(CXBLASOBJS_P) : override CFLAGS += -DCBLAS
srot.$(SUFFIX) srot.$(PSUFFIX) : rot.c
$(CC) $(CFLAGS) -c $< -o $(@F)
@@ -747,9 +725,6 @@ saxpy.$(SUFFIX) saxpy.$(PSUFFIX) : axpy.c
daxpy.$(SUFFIX) daxpy.$(PSUFFIX) : axpy.c
$(CC) $(CFLAGS) -c $< -o $(@F)
iaxpy.$(SUFFIX) iaxpy.$(PSUFFIX) : axpy.c
$(CC) $(CFLAGS) -c $< -o $(@F)
qaxpy.$(SUFFIX) qaxpy.$(PSUFFIX) : axpy.c
$(CC) $(CFLAGS) -c $< -o $(@F)
@@ -1462,9 +1437,6 @@ cblas_saxpy.$(SUFFIX) cblas_saxpy.$(PSUFFIX) : axpy.c
cblas_daxpy.$(SUFFIX) cblas_daxpy.$(PSUFFIX) : axpy.c
$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
cblas_iaxpy.$(SUFFIX) cblas_iaxpy.$(PSUFFIX) : axpy.c
$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)
cblas_caxpy.$(SUFFIX) cblas_caxpy.$(PSUFFIX) : zaxpy.c
$(CC) $(CFLAGS) -DCBLAS -c $< -o $(@F)

View File

@@ -103,8 +103,6 @@ void CNAME(blasint n, FLOAT alpha, FLOAT *x, blasint incx, FLOAT *y, blasint inc
mode = BLAS_XDOUBLE | BLAS_REAL;
#elif defined(DOUBLE)
mode = BLAS_DOUBLE | BLAS_REAL;
#elif defined(INTEGER)
mode = BLAS_INTEGER | BLAS_REAL;
#else
mode = BLAS_SINGLE | BLAS_REAL;
#endif

View File

@@ -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

View File

@@ -81,6 +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 *) = {
@@ -135,6 +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 *) = {
@@ -235,10 +241,10 @@ 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 <= (24.0 * 24.0 * (double) (GEMM_MULTITHREAD_THRESHOLD*GEMM_MULTITHREAD_THRESHOLD) ) )
nthreads_max = 1;

View File

@@ -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);

View File

@@ -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);

View File

@@ -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;

View File

@@ -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;

View File

@@ -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
}

View File

@@ -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;

Some files were not shown because too many files have changed in this diff Show More