88928650c4
Merge pull request #2883 from martin-frbg/issue2872
...
Minor CMAKE fixes
2020-10-11 10:30:33 +02:00
82a497ec5d
restore PRESCOTT default for DYNAMIC_LIST
2020-10-11 00:43:09 +02:00
de27e4f5fb
Stop DYNAMIC_ARCH build if the toplevel source contains a stray config_kernel.h from a gmake build
...
This is unlikely to happen in practice, but if it does, the rogue file would get included instead of the dynamically generated version for each target_core, leading to very confusing errors like "invalid operands (undefined UND and ABS sections)" in compilation of the assembly kernels as macros like PREFETCH would remain undefined
2020-10-11 00:40:22 +02:00
e1b7123bbe
Merge pull request #2867 from Qiyu8/usimd-floatdot
...
Optimize the performance of dot by using universal intrinsics in X86/ARM
2020-10-10 12:10:25 +02:00
f32d34a015
add sse3 compiler flag
2020-10-10 10:36:15 +08:00
599777ecb7
Merge pull request #2879 from martin-frbg/issue2839
...
Default BLAS3_MEM_ALLOC_THRESHOLD on all platforms to 32
2020-10-06 23:26:52 +02:00
a5feea6611
make BLAS3_MEM_ALLOC_THRESHOLD configurable on non-Windows
2020-10-04 23:01:06 +02:00
dc8e4e1959
Reduce the BLAS3 heap allocation threshold to 32 and mark it as configurable
2020-10-04 22:59:24 +02:00
cccd1438da
Merge pull request #93 from xianyi/develop
...
rebase
2020-10-04 22:57:11 +02:00
f032d8966e
Merge pull request #2874 from Flamefire/memory_fixes
...
Avoid out of bounds access on invalid memory free
2020-10-04 15:16:51 +02:00
f6e4cf2f9d
Merge pull request #2876 from Flamefire/omp_fork_fix
...
Lazyly reinit threads after a fork in OMP mode
2020-10-03 22:52:17 +02:00
9828343e12
Merge pull request #2878 from brada4/asms
...
fix clang std=c18 compilation on aarch64
2020-10-03 22:51:49 +02:00
d2333e7842
aarch64 fix std=c18 compilation
2020-10-03 18:00:34 +03:00
3094fc6c83
Lazyly reinit threads after a fork in OMP mode
...
This initializes the per-thread memory buffers which get
cleared/released on a fork via pthread_at_fork. Not doing so leads to
each thread calling blas_memory_alloc on almost every execution which
slows down the code significantly as the threads race for the memory
allocation using locks to serialize that.
2020-10-01 15:41:42 +02:00
3c05f54df8
Avoid out of bounds access on invalid memory free
2020-10-01 10:48:45 +02:00
dee7c49938
Fix TABs and trailing space
2020-10-01 10:43:16 +02:00
d3c0d6811b
Merge pull request #2873 from martin-frbg/issue2871
...
Check for __linux rather than linux in cpuid code and benchmarks
2020-10-01 06:38:22 +02:00
9637cd1fd1
Merge pull request #2865 from thisch/backticks
...
Consolidate usage of backticks for build options
2020-10-01 06:38:06 +02:00
2367726578
Remove redundant status message
2020-09-30 23:28:49 +02:00
5464eb13ea
Change ifdef linux to __linux for C11 compatibility
2020-09-30 22:59:41 +02:00
e1574cbc83
Change ifdef linux to __linux for C11 compatibility
...
and add a fallback for unsupported operating systems in detect()
2020-09-30 22:50:21 +02:00
0b2bb5696a
Change ifdef linux to __linux for C11 compatibility
2020-09-30 22:47:25 +02:00
a7d5d0078d
Change ifdef linux to __linux for C11 compatibility
2020-09-30 22:46:25 +02:00
be40440ec5
Change ifdef linux to __linux for C11 compatibility
2020-09-30 22:45:18 +02:00
2bf70c8e3b
Change ifdef linux to __linux for C11 compatibility
2020-09-30 22:43:25 +02:00
60e6c68e38
Adapt ARM architect
2020-09-29 16:36:14 +08:00
64629cb5c7
Merge pull request #91 from xianyi/develop
...
rebase
2020-09-28 22:48:53 +02:00
1b1a757f5f
Optimize the performance of dot by using universal intrinsics in X86/ARM
2020-09-28 20:36:53 +08:00
0d98ce202c
Merge pull request #2866 from RajalakshmiSR/p10_dcopy
...
Optimize dcopy/zcopy for POWER10
2020-09-28 07:22:54 +02:00
2df4235e00
Optimize dcopy/zcopy for POWER10
...
This patch makes use of new POWER10 vector pair instructions for
loads and stores. Tested in simulator and no new failures.
2020-09-27 21:42:32 -05:00
fe8cd5ae7e
Consolidate usage of backticks for build options
...
There were some build options in the README that were not
highlighted. Now all are highlighted.
2020-09-28 00:42:17 +02:00
ba31c8f5f9
Merge pull request #2853 from Qiyu8/usimd-daxpy
...
Optimize the performance of daxpy by using universal intrinsics
2020-09-27 23:19:59 +02:00
e961d4d609
Merge pull request #2864 from martin-frbg/lapack445
...
FIx underflow/rounding errors in LAPACK (S,D)LANV2
2020-09-27 23:11:17 +02:00
7ed25e9e10
FIx underflow/rounding errors in LAPACK (S,D)LANV2
...
Reference-LAPACK PR 445, fixing their issue 263
2020-09-27 22:59:20 +02:00
7b169379e0
Merge pull request #2863 from martin-frbg/readmefixes
...
Readmefixes
2020-09-27 22:50:25 +02:00
7f539fb850
Update cpu list, outline cmake build, clarify scope of set_num_threads extension
2020-09-27 22:48:41 +02:00
caf7a12295
Merge pull request #90 from xianyi/develop
...
rebase
2020-09-27 22:35:45 +02:00
72b5b73647
Merge pull request #2850 from xiaojiayuan111/develop
...
fix a bug of trmm
2020-09-27 12:12:35 +02:00
881c15179f
remove default support for FMA4 on zen architect
2020-09-27 09:35:50 +08:00
dfaafd3b55
Merge pull request #2854 from martin-frbg/travis-graviton
...
Add an AWS-Graviton2 build to Travis CI
2020-09-23 21:59:18 +02:00
f2e9a24e1a
Add AWS Graviton2 build
2020-09-23 19:02:20 +02:00
61fae59298
Merge pull request #88 from xianyi/develop
...
rebase
2020-09-22 23:15:33 +02:00
33d22f99f1
Merge pull request #2851 from martin-frbg/travis-xcode12
...
Add an OSX build with xcode12
2020-09-22 21:44:55 +02:00
5ba01dd1a8
Add an OSX build with xcode12
2020-09-22 17:26:19 +02:00
14f7dad3b7
performance improved
2020-09-22 16:52:15 +08:00
06cf73a239
fix a bug of trmm
2020-09-22 16:47:10 +08:00
325b539c26
Optimize the performance of daxpy by using universal intrinsics
2020-09-22 10:38:35 +08:00
0f112077e6
Merge pull request #2847 from mhillenibm/fixup_cscal
...
s390x: fix cscal and zscal implementations
2020-09-21 22:22:43 +02:00
22aa81f3e5
s390x: fix cscal and zscal implementations
...
The implementation of complex scalar * vector multiplication for Z14
makes some LAPACK tests fail because the numerical differences to the
reference implementation exceed the threshold (as can be seen by running
make lapack-test and replacing kernel/zarch/cscal.c with a generic
implementation for comparison).
The complex multiplication uses terms of the form a * b + c * d for both
real and imaginary parts. The assembly code (and compiler-emitted code
as well) uses fused multiply add operations for the second product and
sum. The results can be "surprising", for example when both terms in the
imaginary part nearly cancel each other out. In that case, the second
product contributes more digits to the sum than the first product that
has been rounded before.
One option is to use separate multiplications (which then round the same
way) and a distinct add. Change the code to pursue that path, by (1)
requesting the compiler not to contract the operations into FMAs and (2)
replacing the assembly kernel with corresponding vectorized C code
(where change 1 also applies).
Signed-off-by: Marius Hillenbrand <mhillen@linux.ibm.com>
2020-09-21 13:10:05 +02:00
77ea73f5e5
s390x: for clang use fp-contract=on instead of fast
...
Make clang slightly more cautious when contracting floating-point
operations (e.g., when applying fused multiply add) by setting
-ffp-contract=on (instead of fast).
Signed-off-by: Marius Hillenbrand <mhillen@linux.ibm.com>
2020-09-21 11:32:08 +02:00
Martin Kroeker
Qiyu8
User User-User
Alexander Grund
Rajalakshmi Srinivasaraghavan
Thomas Hisch
y00512012
Marius Hillenbrand