a83a59b038
Use generic kernels for ishama,shasum,shdot,shrot
2020-04-18 15:53:51 +02:00
0a19bd813c
Use generic codes for shamax and shcopy
2020-04-18 12:52:51 +02:00
e7afe8a969
Define AXPBY_K fallback for float16
2020-04-18 11:10:15 +02:00
f361de30a3
Use generic axpy.c for SHAXPY as x86 lacks saxpy.c
2020-04-18 11:07:16 +02:00
9f6d6f6cb6
use saxpy.c instead of axpy.S for SHAXPY
2020-04-17 22:27:58 +02:00
22bb50fb81
cmake fixes
2020-04-17 13:35:17 -05:00
236a3d8ce6
Merge pull request #2563 from zelong-1024/develop
...
[OpenBLAS]: benchmark error of potrf
2020-04-16 11:45:32 +02:00
6b7ef6543a
[OpenBLAS]: benchmark error of potrf
...
[description]: when the matrix size goes higher than 5800 during the cpotrf test, error info, such as "Potrf info = 5679", will be returned on ARM64 and x86 machines. Uplo = L & F.
[solution]: changed the func for building the matrix so that the complex Hermitian matrix can stay positive definite during the computation.
[dts]:
2020-04-16 10:55:10 +08:00
67cc4b9e16
Fix warnings in clang and export symbol
2020-04-15 19:15:23 -05:00
250e6f8039
Merge pull request #2557 from martin-frbg/dronebadge
...
Update and reformat README
2020-04-15 20:23:43 +02:00
7a6d0016b0
Merge pull request #2556 from martin-frbg/epicdrone
...
Add a drone.io multithread test for x86_64
2020-04-15 20:23:17 +02:00
e8e8a6e608
Restore USE_OPENMP in the x86 thread test
2020-04-15 19:26:12 +02:00
579811fb6a
Move all 19.04-based jobs back to ubuntu 18.04
2020-04-15 17:38:33 +02:00
a87793e03c
Fix DYNAMIC_ARCH compilation errors
2020-04-15 09:09:50 -05:00
ac6a22ae78
Update header
2020-04-14 22:58:39 -05:00
ff010f496e
Build shgemm for all architecture
2020-04-14 20:38:53 -05:00
7eb55504b1
RFC : Add half precision gemm for bfloat16 in OpenBLAS
...
This patch adds support for bfloat16 data type matrix multiplication kernel.
For architectures that don't support bfloat16, it is defined as unsigned short
(2 bytes). Default unroll sizes can be changed as per architecture as done for
SGEMM and for now 8 and 4 are used for M and N. Size of ncopy/tcopy can be
changed as per architecture requirement and for now, size 2 is used.
Added shgemm in kernel/power/KERNEL.POWER9 and tested in powerpc64le and
powerpc64. For reference, added a small test compare_sgemm_shgemm.c to compare
sgemm and shgemm output.
This patch does not cover OpenBLAS test, benchmark and lapack tests for shgemm.
Complex type implementation can be discussed and added once this is approved.
2020-04-14 14:55:08 -05:00
84a9614345
try x86_64 test without openmp
2020-04-14 19:18:35 +02:00
b969533703
Add drone.io badge, mention EMAG8180 support, reformat the DYNAMIC_ARCH paragraph
2020-04-14 10:53:28 +02:00
0f08f3efa6
Add a multithread test for x86_64
2020-04-13 22:46:12 +02:00
c861b2a7bd
Merge pull request #2553 from martin-frbg/issue2444
...
Add a read memory barrier to the traversal of the buffer slot list
2020-04-13 21:28:59 +02:00
cf62adffbb
Merge pull request #2555 from martin-frbg/issue1137
...
Handle unaligned data in the SSE2 copy kernel
2020-04-13 18:29:56 +02:00
3eec7d382c
ARMV7 does not support DMB ISHLD, use DMB ISH
2020-04-13 15:56:31 +02:00
5b0093b5fe
Convert aligned moves to unaligned
...
should have no performance impact on reasonably modern cpus and fixes occasional crashes in actual user code.
2020-04-13 14:58:52 +02:00
f41600e66f
Add a read barrier in the traversing of the buffer list
...
Needed on systems with weak memory ordering - the inferior, partially working fix from #2544 was already removed in #2551
2020-04-13 12:34:02 +02:00
f5efecb7ca
Add (empty) read barrier definition
2020-04-13 12:24:10 +02:00
a52bdd9d7b
Add (empty) read barrier definition
2020-04-13 12:22:35 +02:00
db3226a646
Add (empty) read barrier definition
2020-04-13 12:18:48 +02:00
69b6e258d8
Add (empty) read barrier definition
2020-04-13 12:17:41 +02:00
3d4db4d002
Add read barrier definition
2020-04-13 12:16:44 +02:00
99dde1d2c9
Add read barrier definition
2020-04-13 12:14:58 +02:00
ee6b3df02c
Add read barrier definition
2020-04-13 12:14:06 +02:00
25e879fe92
Add (empty) read barrier definition
2020-04-13 12:12:54 +02:00
d237dc1360
Add read barrier definition
2020-04-13 12:11:58 +02:00
8692456226
Add read barrier definition
2020-04-13 12:10:37 +02:00
d1d69e1b9a
Add read barrier definition
2020-04-13 12:09:24 +02:00
20d0cb2f65
Merge pull request #46 from xianyi/develop
...
rebase
2020-04-13 12:06:40 +02:00
e7f0da9295
Merge pull request #2551 from martin-frbg/issue2538-2
...
Increase BUFFER_SIZEs and add a safeguard; supply GEMM_R for POWER8/9
2020-04-12 22:34:41 +02:00
e9bfa2291a
Fix parameter overflow
2020-04-12 19:47:02 +02:00
2a28448a96
Add safeguards for sufficient BUFFER_SIZE
2020-04-12 19:45:36 +02:00
a33d177430
Increase default BUFFER_SIZE on ARM, ZARCH and newer x86_64, add GEMM_R for POWER8/9
...
As shown in #2538 , default buffersizes on some platforms were smaller than required in memory.c
and the requirement could never be fulfilled for a calculated GEMM_R on PPC given the fomula used
2020-04-12 19:44:48 +02:00
f73391c9c9
Merge pull request #45 from xianyi/develop
...
rebase
2020-04-12 19:39:05 +02:00
7905383cb5
Merge pull request #2547 from sharvil/develop
...
Add API to set thread affinity on Linux.
2020-04-11 00:35:38 +02:00
a8cbd451bf
Merge pull request #2541 from bapt/develop
...
libname: treat FreeBSD and DragonFly like linux and sunos
2020-04-11 00:35:07 +02:00
eecd8c3204
Merge pull request #2548 from gxw-loongson/develop
...
Add a GENERIC target for 64bit MIPS
2020-04-11 00:34:04 +02:00
ea85eb2e02
Merge pull request #2549 from martin-frbg/fixthreadtest
...
Match thread count in cpp_thread_test to host capability
2020-04-10 23:54:40 +02:00
66f89c0aaf
Match thread count to machine capability
2020-04-10 22:06:44 +02:00
8d07cf9b67
Fix compilation problem on loongson platform
...
Using "make TARGET=GENERIC" on loongson platform will get the following
error messages:
"make[1]: *** No rule to make target 'sgemm_incopy.o', needed by 'libs'"
Add kernel/mips64/KERNEL.generic to slove the problem.
2020-04-09 19:28:15 +08:00
7b4773b24d
Add API to set thread affinity on Linux.
...
Issue: #2545
2020-04-08 12:49:35 -07:00
69f277f8ee
Add another memory barrier for ARM and a multicore test run on ThunderX to help detect such issues ( #2544 )
...
* Add another memory barrier in memory.c to prevent races in memory slot allocation
* Add an all-core test on Drone.io's ThunderX platform and modify dgemm_tester to use all 96 cores
2020-04-08 11:04:51 +02:00
Martin Kroeker
Rajalakshmi Srinivasaraghavan
l00536773
Rajalakshmi Srinivasaraghavan
gxw
Sharvil Nanavati