platform/upstream/openblas.git
9 years agoAdded function to set defines for the object names (e.g. -DNAME=dgemm).
Hank Anderson [Wed, 4 Feb 2015 16:37:34 +0000 (10:37 -0600)]
Added function to set defines for the object names (e.g. -DNAME=dgemm).

9 years agoConverted rest of Makefile.system to system.cmake.
Hank Anderson [Tue, 3 Feb 2015 22:05:01 +0000 (16:05 -0600)]
Converted rest of Makefile.system to system.cmake.

9 years agoConverted LAPACK flags from Makefile.system.
Hank Anderson [Tue, 3 Feb 2015 21:33:56 +0000 (15:33 -0600)]
Converted LAPACK flags from Makefile.system.

9 years agoFixed newlines in some cmake files.
Hank Anderson [Tue, 3 Feb 2015 21:11:15 +0000 (15:11 -0600)]
Fixed newlines in some cmake files.

9 years agoChanged fortran compiler name to be uppercase and stripped of path/ext.
Hank Anderson [Tue, 3 Feb 2015 21:09:37 +0000 (15:09 -0600)]
Changed fortran compiler name to be uppercase and stripped of path/ext.

9 years agoCorrected fortran compiler name variables.
Hank Anderson [Tue, 3 Feb 2015 20:21:22 +0000 (14:21 -0600)]
Corrected fortran compiler name variables.

Fixed some typos.

Updated c_check to set ARCH and BINARY64/32.

Added version variables.

9 years agoFixed some missing parentheses.
Hank Anderson [Tue, 3 Feb 2015 20:00:29 +0000 (14:00 -0600)]
Fixed some missing parentheses.

9 years agoPorted more of Makefile.system to CMake.
Hank Anderson [Tue, 3 Feb 2015 19:34:41 +0000 (13:34 -0600)]
Ported more of Makefile.system to CMake.

9 years agoAdded fc.cmake (forgot it in last commit).
Hank Anderson [Tue, 3 Feb 2015 19:08:59 +0000 (13:08 -0600)]
Added fc.cmake (forgot it in last commit).

Moved a couple C compiler ifs from Makefile.system into cc.cmake.

9 years agoPorted Fortran configuration code from Makefile.system to fc.cmake.
Hank Anderson [Tue, 3 Feb 2015 18:32:23 +0000 (12:32 -0600)]
Ported Fortran configuration code from Makefile.system to fc.cmake.

9 years agoPorted C compiler settings from Makefile.system into new cmake file.
Hank Anderson [Tue, 3 Feb 2015 18:00:49 +0000 (12:00 -0600)]
Ported C compiler settings from Makefile.system into new cmake file.

9 years agoPorted arch dependent settings from Makefile.system to new cmake file.
Hank Anderson [Tue, 3 Feb 2015 17:32:20 +0000 (11:32 -0600)]
Ported arch dependent settings from Makefile.system to new cmake file.

9 years agoPorted OS settings from Makefile.system into new cmake file.
Hank Anderson [Tue, 3 Feb 2015 17:07:58 +0000 (11:07 -0600)]
Ported OS settings from Makefile.system into new cmake file.

9 years agoAdded cblas_ objects to interface CMakeLists.
Hank Anderson [Mon, 2 Feb 2015 22:25:30 +0000 (16:25 -0600)]
Added cblas_ objects to interface CMakeLists.

Naming isn't right, though, not seeing cblas_xxxx exports in the
resulting library.

9 years agoAdded missing common objects to the library.
Hank Anderson [Mon, 2 Feb 2015 21:21:29 +0000 (15:21 -0600)]
Added missing common objects to the library.

9 years agoDOUBLE define for DBLAS objects is now set in main CMakeLists.txt.
Hank Anderson [Mon, 2 Feb 2015 21:00:44 +0000 (15:00 -0600)]
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.

9 years agoAdded LAPACK object files to interface CMakeLists.
Hank Anderson [Mon, 2 Feb 2015 18:31:15 +0000 (12:31 -0600)]
Added LAPACK object files to interface CMakeLists.

9 years agoAdded openblas add_library call that uses DBLAS_OBJS ojbects.
Hank Anderson [Fri, 30 Jan 2015 21:21:21 +0000 (15:21 -0600)]
Added openblas add_library call that uses DBLAS_OBJS ojbects.

9 years agoAdded COMMONOBJS to driver/others CMakeLists.txt.
Hank Anderson [Fri, 30 Jan 2015 20:06:14 +0000 (14:06 -0600)]
Added COMMONOBJS to driver/others CMakeLists.txt.

9 years agoMoved functions into util cmake file.
Hank Anderson [Fri, 30 Jan 2015 19:47:40 +0000 (13:47 -0600)]
Moved functions into util cmake file.

9 years agoAdded paramater to GenerateObjects for defines that affect all sources.
Hank Anderson [Fri, 30 Jan 2015 19:31:13 +0000 (13:31 -0600)]
Added paramater to GenerateObjects for defines that affect all sources.

9 years agoAdded empty set to the combinations generated by AllCombinations.
Hank Anderson [Fri, 30 Jan 2015 19:01:11 +0000 (13:01 -0600)]
Added empty set to the combinations generated by AllCombinations.

9 years agoMoved loop over define combos into a function.
Hank Anderson [Fri, 30 Jan 2015 18:14:44 +0000 (12:14 -0600)]
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.

9 years agoUpdated level3/CMakeLists with correct defines using all combos.
Hank Anderson [Fri, 30 Jan 2015 17:21:50 +0000 (11:21 -0600)]
Updated level3/CMakeLists with correct defines using all combos.

9 years agoAdded first pass at driver/level3 Makefile conversion.
Hank Anderson [Fri, 30 Jan 2015 04:53:11 +0000 (22:53 -0600)]
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.

9 years agoMoved getarch parsing code into a function.
Hank Anderson [Thu, 29 Jan 2015 15:30:47 +0000 (09:30 -0600)]
Moved getarch parsing code into a function.

9 years agoprebuild.cmake now reads the output from getarch into CMake vars.
Hank Anderson [Thu, 29 Jan 2015 04:57:44 +0000 (22:57 -0600)]
prebuild.cmake now reads the output from getarch into CMake vars.

9 years agogetarch_2nd now appends its output to config.h/config_kernel.h
Hank Anderson [Thu, 29 Jan 2015 04:20:15 +0000 (22:20 -0600)]
getarch_2nd now appends its output to config.h/config_kernel.h

9 years agogetarch now compiles and sets config.h defines properly.
Hank Anderson [Wed, 28 Jan 2015 23:18:26 +0000 (17:18 -0600)]
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.

9 years agoAdded CORE define to config.h in prebuild.cmake (temporarily).
Hank Anderson [Wed, 28 Jan 2015 22:33:48 +0000 (16:33 -0600)]
Added CORE define to config.h in prebuild.cmake (temporarily).

9 years agoUpdated c_check OS/compiler/bits detection.
Hank Anderson [Wed, 28 Jan 2015 21:47:47 +0000 (15:47 -0600)]
Updated c_check OS/compiler/bits detection.

9 years agoAdded first pass at driver/level2 makefile conversion.
Hank Anderson [Wed, 28 Jan 2015 20:52:15 +0000 (14:52 -0600)]
Added first pass at driver/level2 makefile conversion.

9 years agoAdded CMakeLists.txt for the first of the BLAS folders.
Hank Anderson [Tue, 27 Jan 2015 22:17:17 +0000 (16:17 -0600)]
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.

9 years agoFixed architecture detection when AMD64 in c_check.
Hank Anderson [Tue, 27 Jan 2015 20:03:46 +0000 (14:03 -0600)]
Fixed architecture detection when AMD64 in c_check.

9 years agoFixed incorrect case in OS_ definition in c_check.
Hank Anderson [Tue, 27 Jan 2015 19:54:29 +0000 (13:54 -0600)]
Fixed incorrect case in OS_ definition in c_check.

9 years agoStarted converting lib target to CMake.
Hank Anderson [Tue, 27 Jan 2015 18:23:35 +0000 (12:23 -0600)]
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.

9 years agoHardcoded NUM_CORES to get system.cmake working.
Hank Anderson [Tue, 27 Jan 2015 17:37:39 +0000 (11:37 -0600)]
Hardcoded NUM_CORES to get system.cmake working.

9 years agocmake.prebuild now compiles getarch.
Hank Anderson [Fri, 2 Jan 2015 03:03:17 +0000 (21:03 -0600)]
cmake.prebuild now compiles getarch.

Doesn't actually run it yet.

9 years agoAdded MSVC functions to cpuid_x86.c to replace gcc-specific ASM.
Hank Anderson [Fri, 2 Jan 2015 03:02:48 +0000 (21:02 -0600)]
Added MSVC functions to cpuid_x86.c to replace gcc-specific ASM.

9 years agoAdded MSVC defines to cpuid.h and getarch.c.
Hank Anderson [Fri, 2 Jan 2015 03:01:28 +0000 (21:01 -0600)]
Added MSVC defines to cpuid.h and getarch.c.

9 years agoFirst pass at converting a few makefiles to CMake.
Hank Anderson [Wed, 31 Dec 2014 03:53:00 +0000 (21:53 -0600)]
First pass at converting a few makefiles to CMake.

9 years agoMerge pull request #486 from wernsaar/develop
Zhang Xianyi [Tue, 30 Dec 2014 18:36:23 +0000 (02:36 +0800)]
Merge pull request #486 from wernsaar/develop

Optimizations for steamroller

9 years agoMerge branch 'develop' of github.com:wernsaar/OpenBLAS into develop
Werner Saar [Tue, 30 Dec 2014 12:16:53 +0000 (20:16 +0800)]
Merge branch 'develop' of github.com:wernsaar/OpenBLAS into develop

9 years agoadded optimizations for steamroller
Werner Saar [Tue, 30 Dec 2014 12:14:45 +0000 (20:14 +0800)]
added optimizations for steamroller

9 years agoMerge pull request #483 from wernsaar/develop
Zhang Xianyi [Mon, 29 Dec 2014 04:00:16 +0000 (12:00 +0800)]
Merge pull request #483 from wernsaar/develop

added Steamroller as a  cpu target

9 years agobugfix in dynamic.c
Werner Saar [Sun, 28 Dec 2014 16:15:42 +0000 (17:15 +0100)]
bugfix in dynamic.c

9 years agoadded Steamroller as a target processor
Werner Saar [Sun, 28 Dec 2014 12:45:19 +0000 (13:45 +0100)]
added Steamroller as a target processor

9 years agoadded target processor STEAMROLLER
Werner Saar [Sun, 28 Dec 2014 12:16:46 +0000 (20:16 +0800)]
added target processor STEAMROLLER

9 years agoMerge pull request #481 from eschnett/develop
Zhang Xianyi [Fri, 26 Dec 2014 02:09:19 +0000 (10:09 +0800)]
Merge pull request #481 from eschnett/develop

Correct ilaver C declaration

9 years agoCorrect ilaver C declaration
Erik Schnetter [Thu, 25 Dec 2014 22:41:17 +0000 (17:41 -0500)]
Correct ilaver C declaration

9 years agoMerge pull request #479 from wernsaar/develop
Zhang Xianyi [Mon, 22 Dec 2014 16:59:41 +0000 (00:59 +0800)]
Merge pull request #479 from wernsaar/develop

workaround for sandybridge zgemm kernel

9 years agoRef #458: Backport, sandybrigde uses nehalem zgemm kernel
Werner Saar [Mon, 22 Dec 2014 16:01:18 +0000 (17:01 +0100)]
Ref #458: Backport, sandybrigde uses nehalem zgemm kernel

9 years agoincreased NMAX to 128
Werner Saar [Mon, 22 Dec 2014 13:04:27 +0000 (14:04 +0100)]
increased NMAX to 128

9 years agomodified sources for OS Darwin
Werner Saar [Fri, 19 Dec 2014 11:40:46 +0000 (12:40 +0100)]
modified sources for OS Darwin

9 years agosmall optimization on dgemm_kernel for N=1
Werner Saar [Thu, 18 Dec 2014 19:35:51 +0000 (20:35 +0100)]
small optimization on dgemm_kernel for N=1

9 years agoadded code for the size of n
Werner Saar [Wed, 17 Dec 2014 14:02:11 +0000 (15:02 +0100)]
added code for the size of n

9 years agomodified makefile for acml6.1
Werner Saar [Wed, 17 Dec 2014 13:12:21 +0000 (14:12 +0100)]
modified makefile for acml6.1

9 years agoFixed installation bug on Mac OSX.
Zhang Xianyi [Sat, 13 Dec 2014 05:05:06 +0000 (13:05 +0800)]
Fixed installation bug on Mac OSX.

9 years agoIncreased the Threshold value in sep.in
Werner Saar [Thu, 11 Dec 2014 13:57:41 +0000 (14:57 +0100)]
Increased the Threshold value in sep.in

9 years agoadded tests to sep.as as workaround for gfortran-4.8.x
Werner Saar [Thu, 11 Dec 2014 12:53:59 +0000 (13:53 +0100)]
added tests to sep.as as workaround for gfortran-4.8.x

9 years agoMerge pull request #475 from xantares/patch-2
Zhang Xianyi [Tue, 9 Dec 2014 09:57:43 +0000 (17:57 +0800)]
Merge pull request #475 from xantares/patch-2

add OpenBLAS_VERSION to cmake config file

9 years agoMerge pull request #474 from xantares/patch-1
Zhang Xianyi [Tue, 9 Dec 2014 09:57:16 +0000 (17:57 +0800)]
Merge pull request #474 from xantares/patch-1

set OPENBLAS_CMAKE_DIR to <prefix>/lib/cmake/<package_name>

9 years agoadd OpenBLAS_VERSION to cmake config file
xantares [Tue, 9 Dec 2014 09:34:41 +0000 (10:34 +0100)]
add OpenBLAS_VERSION to cmake config file

9 years agoset OPENBLAS_CMAKE_DIR to <prefix>/lib/cmake/<package_name>
xantares [Tue, 9 Dec 2014 09:18:18 +0000 (10:18 +0100)]
set OPENBLAS_CMAKE_DIR to <prefix>/lib/cmake/<package_name>

usually these files are more often located in this subdir

9 years agoMerge pull request #473 from wernsaar/develop
Zhang Xianyi [Mon, 8 Dec 2014 05:22:18 +0000 (13:22 +0800)]
Merge pull request #473 from wernsaar/develop

changed inline assembler labels to short form

9 years agochanged inline assembler labels to short form
Werner Saar [Sun, 7 Dec 2014 11:38:54 +0000 (12:38 +0100)]
changed inline assembler labels to short form

9 years agoMerge branch 'develop' of github.com:xianyi/OpenBLAS into develop
Zhang Xianyi [Wed, 3 Dec 2014 15:03:48 +0000 (23:03 +0800)]
Merge branch 'develop' of github.com:xianyi/OpenBLAS into develop

9 years agoUpdate the doc for 0.2.13 version.
Zhang Xianyi [Wed, 3 Dec 2014 15:00:29 +0000 (23:00 +0800)]
Update the doc for 0.2.13 version.

9 years agoFixed a bug of sgemm sandy bridge kernel.
Zhang Xianyi [Wed, 3 Dec 2014 09:38:41 +0000 (17:38 +0800)]
Fixed a bug of sgemm sandy bridge kernel.

Reported by Julia project. JuliaLang/julia#9084

9 years agoMerge pull request #471 from nolta/patch-4
Zhang Xianyi [Wed, 3 Dec 2014 04:53:20 +0000 (12:53 +0800)]
Merge pull request #471 from nolta/patch-4

c_check: set $hostarch to x86_64 instead of amd64

9 years agoMerge pull request #470 from nolta/patch-3
Zhang Xianyi [Wed, 3 Dec 2014 04:50:46 +0000 (12:50 +0800)]
Merge pull request #470 from nolta/patch-3

fix fortran compiler detection on FreeBSD

9 years agoc_check: set $hostarch to x86_64 instead of amd64
Mike Nolta [Wed, 3 Dec 2014 02:23:23 +0000 (21:23 -0500)]
c_check: set $hostarch to x86_64 instead of amd64

`uname -m` returns "amd64" on some systems.

9 years agofix fortran compiler detection on FreeBSD
Mike Nolta [Wed, 3 Dec 2014 01:47:40 +0000 (20:47 -0500)]
fix fortran compiler detection on FreeBSD

On FreeBSD, passing extra options to `which` causes it to report a non-zero status:

```
$ which gfortran48 -m64
/usr/local/bin/gfortran48
$ echo $?
1
```

```
$ which gfortran48
/usr/local/bin/gfortran48
$ echo $?
0
```

9 years agoRefs #461. Provide OpenBLASConfig.cmake to support CMake.
Zhang Xianyi [Fri, 28 Nov 2014 18:16:40 +0000 (02:16 +0800)]
Refs #461. Provide OpenBLASConfig.cmake to support CMake.

If you "make PREFIX=/path/to/OpenBLAS install" ,
The config file will be located in /path/to/OpenBLAS/cmake

Then, you can use "find_package(OpenBLAS)" at CMake.
cmake -DOpenBLAS_DIR=/path/to/OpenBLAS/cmake ..

9 years agoUpdate organization info.
Zhang Xianyi [Tue, 25 Nov 2014 07:28:58 +0000 (15:28 +0800)]
Update organization info.

9 years agoRefs #467. Added generic kernel file for x86_64.
Zhang Xianyi [Mon, 24 Nov 2014 07:34:48 +0000 (15:34 +0800)]
Refs #467. Added generic kernel file for x86_64.

9 years agoFixed #456. Merged the optimizations for APM's
Zhang Xianyi [Tue, 11 Nov 2014 14:21:04 +0000 (22:21 +0800)]
Fixed #456. Merged the optimizations for APM's
xgene-1 (aarch64).
Merge branch 'benedikt-huber-dave-patch' into develop

9 years ago # The first commit's message is:
Benedikt Huber [Thu, 9 Oct 2014 13:52:10 +0000 (06:52 -0700)]
 # The first commit's message is:
Optimizations for APM's xgene-1 (aarch64).

1) general system updates to support armv8 better.  Make all did not work, one needed to supply TARGET=ARMV8.
2) sgem 4x4 kernel in assembler using SIMD, and configuration changes to use it.
3) strmm 4x4 kernel in C.  Since the sgem kernel does 4x4, the trmm kernel must also do 4xN.

Added Dave Nuechterlein to the contributors list.

9 years agorefs #464. Fixed the bug of detecting L2 associative on x86.
Zhang Xianyi [Mon, 10 Nov 2014 09:15:34 +0000 (17:15 +0800)]
refs #464. Fixed the bug of detecting L2 associative on x86.

9 years ago#463 fixed a compiling bug on AIX.
Zhang Xianyi [Mon, 10 Nov 2014 06:39:56 +0000 (14:39 +0800)]
#463 fixed a compiling bug on AIX.

9 years agoMerge pull request #459 from tkelman/symbol-rename
Zhang Xianyi [Sat, 25 Oct 2014 11:49:03 +0000 (19:49 +0800)]
Merge pull request #459 from tkelman/symbol-rename

add SYMBOLPREFIX and SYMBOLSUFFIX makefile options

9 years agoadd SYMBOLPREFIX and SYMBOLSUFFIX makefile options
Tony Kelman [Sat, 25 Oct 2014 05:27:00 +0000 (22:27 -0700)]
add SYMBOLPREFIX and SYMBOLSUFFIX makefile options

for adding a prefix or suffix to all exported symbol names in the shared library
Useful to avoid conflicts with other BLAS libraries, especially when using
64 bit integer interfaces in OpenBLAS

Note that since OSX does not have the objcopy utility, setting these options
to non-empty values on Mac requires the objconv tool, available (GPL license)
from http://www.agner.org/optimize/#objconv

9 years agoUpdate dot to 0.2.12 version.
Zhang Xianyi [Mon, 13 Oct 2014 09:10:12 +0000 (17:10 +0800)]
Update dot to 0.2.12 version.

9 years agoRef #454: fixed bug in common_param.h
wernsaar [Tue, 23 Sep 2014 09:34:29 +0000 (11:34 +0200)]
Ref #454: fixed bug in common_param.h

9 years agoMerge pull request #453 from wernsaar/develop
Zhang Xianyi [Mon, 22 Sep 2014 08:47:54 +0000 (16:47 +0800)]
Merge pull request #453 from wernsaar/develop

Enabled GEMM3M functions

9 years agoupdated cblas.h and cblas_noconst.h
wernsaar [Sun, 21 Sep 2014 11:39:15 +0000 (13:39 +0200)]
updated cblas.h and cblas_noconst.h

9 years agoadded benchmark for gemm3m functions
wernsaar [Sun, 21 Sep 2014 10:00:41 +0000 (12:00 +0200)]
added benchmark for gemm3m functions

9 years agobugfix for GEMM3M functions
wernsaar [Sun, 21 Sep 2014 09:41:43 +0000 (11:41 +0200)]
bugfix for GEMM3M functions

9 years agoadded GEMM3M tests
wernsaar [Sun, 21 Sep 2014 08:55:08 +0000 (10:55 +0200)]
added GEMM3M tests

9 years agoenabled cblas gemm3m functions
wernsaar [Sat, 20 Sep 2014 15:20:02 +0000 (17:20 +0200)]
enabled cblas gemm3m functions

9 years agodisabled SYMM3M and HEMM3M functions because segment violations
wernsaar [Sat, 20 Sep 2014 13:27:40 +0000 (15:27 +0200)]
disabled SYMM3M and HEMM3M functions because segment violations

9 years agoadded test for CGEMM3M function
wernsaar [Sat, 20 Sep 2014 12:53:30 +0000 (14:53 +0200)]
added test for CGEMM3M function

9 years agoenabled use of GEMM3M functions
wernsaar [Sat, 20 Sep 2014 12:27:10 +0000 (14:27 +0200)]
enabled use of GEMM3M functions

9 years agoadded test for GEMM3M functions
wernsaar [Sat, 20 Sep 2014 12:21:42 +0000 (14:21 +0200)]
added test for GEMM3M functions

9 years agoupdated README.md
wernsaar [Wed, 17 Sep 2014 14:01:07 +0000 (16:01 +0200)]
updated README.md

9 years agoUpdate the doc for target list.
Zhang Xianyi [Wed, 17 Sep 2014 06:29:21 +0000 (14:29 +0800)]
Update the doc for target list.

9 years agoMerge pull request #451 from eshelman/patch-1
Zhang Xianyi [Wed, 17 Sep 2014 06:20:06 +0000 (14:20 +0800)]
Merge pull request #451 from eshelman/patch-1

Add HASWELL to TargetList.txt

9 years agoAdd HASWELL to TargetList.txt
Eliot Eshelman [Tue, 16 Sep 2014 22:26:45 +0000 (18:26 -0400)]
Add HASWELL to TargetList.txt

The Intel "Haswell" architecture is missing from the list of build targets.

9 years agoMerge pull request #449 from wernsaar/develop
Zhang Xianyi [Tue, 16 Sep 2014 06:33:48 +0000 (14:33 +0800)]
Merge pull request #449 from wernsaar/develop

optimized multithreading lower limits

9 years agooptimized multithreading lower limits
wernsaar [Mon, 15 Sep 2014 09:38:25 +0000 (11:38 +0200)]
optimized multithreading lower limits