208c7e7ca5
Use acq/rel semantics to pass flags/pointers in getrf_parallel.
...
The current implementation has locks, but the locks each only
have a critical section of one variable so atomic reads/writes
with barriers can be used to achieve the same behavior.
Like the previous patch, pthread_mutex_lock isn't fair, so in a
tight loop the previous thread that has the lock can keep it
starving another thread, even if that thread is about to write
the data that will stop the current thread from spinning.
On a 64c Arm system this improves performance by 20x on sgesv.goto.
2020-03-06 06:22:31 +00:00
4aa2d89217
Merge branch 'develop' into risc-v
2020-02-27 13:53:49 +08:00
c222b25b81
Correct generation of GETRF files by the CMAKE build
...
fixes #2396
2020-02-15 19:29:14 +01:00
9f7a9a32e3
Merge pull request #2252 from thrasibule/trtrs
...
Optimized ?trtrs
2019-09-12 21:45:47 +02:00
6cb47ea3f0
fix Makefile
2019-09-10 17:11:01 -04:00
f2becb777a
fix Makefile
2019-09-09 11:36:50 -04:00
9f6984fe4b
add missing files
2019-09-08 11:14:49 -04:00
42203dafdc
add logic
2019-09-08 11:14:49 -04:00
a4f17a9297
add missing objects
2019-09-08 11:14:49 -04:00
733d97b2df
add files
2019-09-08 11:14:49 -04:00
4de545aa7d
address minor warnings from gcc7
2019-09-07 10:21:08 +03:00
575a84398a
remove redundant code #2113
2019-05-07 23:46:54 +03:00
e882b239aa
Correct naming of getrf_parallel object
...
fixes #1984
2019-01-26 00:45:45 +01:00
3716267124
Change _STDC_VERSION__ to __STDC_VERSION__
...
Change-Id: Id3fa4e8d9eedd4ef7230df69b611e7f397301a42
2018-05-11 12:15:08 +08:00
c167a3d6f4
Added RISCV build
2018-04-16 14:08:31 -07:00
20c6c38e51
Merge branch 'develop' into atomic
2018-04-07 12:09:39 +02:00
8ec28ff461
Remove unguarded use of _Atomic and fix tabbing
2018-04-04 22:40:30 +02:00
bb9876db33
Fix thread races and infinite looping on systems with many cpus
...
On systems with more than 64 cpus, blas_quickdivide will sometimes return zero which creates bogus workloads when used for the stride calculation. This then leads to threads spinning incessantly waiting for a status change that never happens, as seen in #1497 .
This patch also fixes several data races that were found by helgrind and/or tsan while debugging the issue.
2018-04-04 18:16:52 +02:00
40160ff3c1
Use _Atomic instead of volatile for thread safety where C11 is supported
2018-03-10 00:15:44 +01:00
9fa986337d
add missing brackets to silence indentation warnings gcc721
2018-01-19 23:11:12 +01:00
d602b99386
LAPACK helpers in C that need care too
2018-01-02 14:38:50 +01:00
c7a8512d12
Cmake fixes for DYNAMIC_ARCH builds and whitespace in path names ( #1323 )
...
* prebuild.cmake: Put quotes around path names that may contain whitespace
(Copied from alexkaratakis' PR #1295 )
* kernel/CMakeLists.txt: Fix common_lapack header inclusion and DYNAMIC_ARCH generation of ?neg_tcopy and ?laswp_ncopy files
* lapack/CMakeLists.txt: Use correct template for ?laswp_(plus,minus) functions
2017-10-09 23:34:18 +02:00
37858d1146
Fix threading usage in CMake: s/SMP/USE_THREAD/
2017-08-19 15:07:42 +10:00
d245caa49a
Support out-of-source build
2017-08-01 15:16:14 +05:30
56762d5e4c
add lapack laswp for zarch
2017-04-13 15:38:59 +02:00
3918d17025
LAPACK: Fix lapack-test errors in ARM64 threaded version
2017-01-31 23:36:23 +05:30
209b63197e
prepared lapack/lauum for UNROLL values, that are not a power of two
2017-01-11 07:29:17 +01:00
c81dc6322f
prepared lapack/potrf functions for UNROLL values, that are not a power of two
2017-01-10 10:50:28 +01:00
3e1bbd6b5f
prepared lapack/getrf functions for UNROLL values, that are not a power of two
2017-01-09 12:57:26 +01:00
053044ae4d
Replace CMAKE_SOURCE_DIR/CMAKE_BINARY_DIR with PROJECT_SOURCE_DIR/PROJECT_BINARY_DIR
...
If OpenBLAS is built using add_subdirectory(OpenBlas) as part of another project
then the paths set by CMAKE_XXX_DIR are relative to the parent project
and not the OpenBLAS project.
2016-05-25 09:13:28 +02:00
956be69e1d
optimized getrf_single.c for POWER8
2016-05-17 16:19:53 +02:00
6a2bde7a2d
optimized dgemm and dgetrf for POWER8
2016-05-17 14:45:27 +02:00
2c3dfe2bf3
MIPS P5600(32 bit) and I6400(64 bit) cores support added.
...
Seperated mips and mips64 files.
Configurations support for mips 32 bit.
Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com>
2016-04-22 14:03:18 +05:30
4e1b521e27
Fix lapack complex implementation of lauu2 and potf2 for Android (use FLOAT instead of FLOAT[2] as imaginary part is not used).
2016-02-04 16:59:56 -05:00
13f0f8c10e
Refs #723 . Avoid out of boundary for getf2.
2016-01-26 09:14:57 -06:00
0553476fba
Added TRANS defines for complex sources in lapack.
2015-02-24 14:30:35 -06:00
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
f3f2b3d768
Added complex and single netlib-lapack fortran sources to lapack.cmake.
2015-02-19 12:26:11 -06:00
67e39bd8fb
Added mangled complex filenames to interface and lapack CMakeLists.txt.
2015-02-17 13:12:30 -06:00
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
e74462a3f5
Moved declarations to start of functions to satisfy MSVC C89 implementation.
2015-02-11 11:16:57 -06:00
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
3b20b62423
Fixed trti2 name.
2015-02-09 15:29:28 -06:00
6ddbfea700
Added generic laswp object.
2015-02-09 15:15:58 -06:00
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
13d2d48e67
Added yet another naming scheme for lapack functions.
2015-02-06 13:42:20 -06:00
373a1bdadb
Converted lapack/Makefile to cmake.
2015-02-04 15:47:10 -06:00
6c2ead30f0
Remove all trailing whitespace except lapack-netlib
...
Signed-off-by: Timothy Gu <timothygu99@gmail.com>
2014-06-27 12:05:18 -07:00
c26bbee489
enabled abd tested optimized trtri lapack functions
2014-05-23 10:55:39 +02:00
c4ccb3fbb2
removed lapack/getri because it was never used
2014-05-21 14:21:19 +02:00
a748d3a75d
enabled optimized trti2 lapack functions again
2014-05-21 11:02:07 +02:00
dbaeea7b59
enabled lauu2 and lauum lapack functions again
2014-05-21 09:49:18 +02:00
4f98f8c9b3
enabled and tested optimized potrf lapack functions
2014-05-18 21:42:37 +02:00
536875d463
enabled and tested optimized getrs lapack functions
2014-05-18 21:13:56 +02:00
ac029f81b3
enabled and tested optimized dgetrf function
2014-05-18 19:07:51 +02:00
a35a1a9ae7
changed makefiles for lapack development
2014-05-07 11:33:02 +02:00
4be4db590c
Merge remote branch 'origin/develop' into armv7
2013-12-01 13:16:41 +01:00
fe5f46c330
added experimental support for ARMV8
2013-11-24 15:47:00 +01:00
5048a80032
Refs #283 . Fixed the incorrect usage of long data type for Windows 64.
2013-11-14 13:46:42 +08:00
73770e60b8
Refs #309 . Fixed trtri_U single thread computational bug.
2013-11-07 01:08:39 +08:00
95aedfa0ff
added missing file arm/Makefile in lapack/laswp
2013-11-03 11:19:32 +01:00
a07cc39571
Refs #266 . Fixed the compiling bug with Open64 5.0.
2013-07-31 14:41:39 +08:00
fd0c388681
Refs #191 . A walk around for dtrtri_U single thread bug.
...
This function caused the failure of ERKALE serial test.
I replaced it with LAPACK source code.
2013-07-14 22:16:30 +08:00
32d2ca3035
Refs #214 , #221 , #246 . Fixed the getrf overflow bug on Windows.
...
I used a smaller threshold since the stack size is 1MB on windows.
2013-07-11 03:20:02 +08:00
5d3312142a
Refs #221 #246 . Fixed the overflowing stack bug in mutlithreading BLAS3.
...
When NUM_THREADS(MAX_CPU_NUNBERS) is very large ,e.g. 256.
typedef struct {
volatile BLASLONG working[MAX_CPU_NUMBER][CACHE_LINE_SIZE * DIVIDE_RATE];
} job_t;
job_t job[MAX_CPU_NUMBER];
The job array is equal 8MB.
Thus, We use malloc instead of stack allocation.
2013-07-08 01:07:05 +08:00
4c2123c334
Fixed the overflowing bug in single thread cholesky factorization.
2013-02-23 13:00:52 +08:00
7bd1834d59
Refs #130 Fixed laswp building bug with DYNAMIC_ARCH=1.
2012-08-09 20:36:29 +08:00
1b056c5328
Refs #130 Prevent reading ipiv array beyond the bound in ?laswp. Use laswp instead of laswp_oncopy in getrf.
2012-08-09 20:06:51 +08:00
342bbc3871
Import GotoBLAS2 1.13 BSD version codes.
2011-01-24 14:54:24 +00:00
Ali Saidi
Xianyi Zhang
Martin Kroeker
Guillaume Horel
Andrew
Zhiyong Dang
Jerry Zhao
Andrew
Sacha Refshauge
Isuru Fernando
Dan Horák
Ashwin Sekhar T K
Werner Saar
John Biddiscombe
Shivraj Patil
buffer51
Zhang Xianyi
Hank Anderson
Timothy Gu
Zhang Xianyi