c8f53b85ce
Merge remote-tracking branch 'origin/develop' into vectorizeBF16GEMV
2024-10-11 11:10:20 -05:00
e52d9b4cf1
Merge pull request #4928 from austinpagan/czgemm_in_c
...
CGEMM & ZGEMM using C code, Power only, P10 only.
2024-10-09 20:26:21 +02:00
0b7fb5c791
CGEMM & ZGEMM using C code.
2024-10-09 09:42:23 -05:00
d6bb8dcfd1
Common code.
2024-10-06 14:13:43 -05:00
c9e92348a6
Handle inf/nan if dummy2 flag is set
2024-10-06 19:57:17 +02:00
9ac0fb0111
Merge branch 'develop' into vectorizeBF16GEMV
2024-10-04 06:49:53 -05:00
d714013ab9
change sgemm kernel to 4x4 as the 16x4 altivec goes out of bounds
2024-10-03 22:04:20 +02:00
915a6d6e44
Add casting.
2024-10-03 14:08:21 -05:00
7ec3c16d82
Remove beta from optimized functions.
2024-10-03 13:27:33 -05:00
7cc00f68c9
Remove more duplicate.
2024-10-01 11:23:32 -05:00
e238a68c03
Remove duplicate.
2024-10-01 11:06:23 -05:00
32095b0cbb
Remove parameter.
2024-10-01 09:32:42 -05:00
c8788208c8
Fixing block issue with transpose version.
2024-09-27 13:27:03 -05:00
d7c0d87cd1
Small changes.
2024-09-26 15:21:29 -05:00
eb6f3a05ef
Common MMA code.
2024-09-26 09:28:56 -05:00
fb287d17fc
Common code.
2024-09-25 16:31:36 -05:00
8ab6245771
Small change.
2024-09-24 16:50:21 -05:00
df19375560
Almost final code for MMA.
2024-09-24 16:30:01 -05:00
05aa63e738
More MMA BF16 GEMV code.
2024-09-24 12:54:02 -05:00
c9ce37d527
Force vector pairs in clang.
2024-09-23 08:43:58 -05:00
89a12fa083
MMA BF16 GEMV code.
2024-09-23 06:32:14 -05:00
7947970f9d
Move common code.
2024-09-13 06:22:13 -05:00
72216d28c2
Fix bug with inc_y adding results twice.
2024-09-11 08:47:32 -05:00
2f142ee857
More common code.
2024-09-09 14:41:55 -05:00
39fd29f1de
Minor improvement and turn off BF16 GEMV forwarding by default.
2024-09-08 18:28:31 -05:00
8541b25e1d
Special case beta is one.
2024-09-06 14:48:48 -05:00
76227e2948
Initial commit for vectorized BF16 GEMV. Added GEMM_GEMV_FORWARD_BF16 to enable using BF16 GEMV for one dimension matrices. Updated unit test to support inc_x != 1 or inc_y for GEMV.
2024-09-06 14:03:31 -05:00
1a7b8c650d
Merge branch 'develop' into betterPowerGEMVTail
2024-08-01 14:59:12 -05:00
f5d04318e3
Merge branch 'OpenMathLib:develop' into scalfixes
2024-07-21 13:43:43 +02:00
73f8866ffb
make NAN handling depend on DUMMY2 parameter
2024-07-21 13:42:47 +02:00
db98f8753f
Try to fix LAPACK testing failures on P7.
...
1. Remove the FADD insn from the GEMV Transpose code.
2. Remove the FADD insn from GEMM and ZGEMM code.
3. Reorder the compution of the Imaginary part in ZGEMM code.
2024-07-19 02:08:19 -04:00
b9bfc8ce09
make NAN handling depend on dummy2 parameter
2024-07-17 23:29:50 +02:00
ba47c7f4f3
Vectorize reduction stage of sgemv_t.
2024-07-16 15:57:24 -05:00
cb154832f8
Vectorize SBGEMM incopy - 4x faster.
2024-07-09 13:10:03 -05:00
2a5fe97e3b
temporarily(?) disable the alpha=0 branch as it does not handle INF,NAN
2024-06-27 16:21:57 +02:00
7f8f037a36
handle INF and NAN in input
2024-06-22 16:03:30 +02:00
f1248b849d
handle INF and NAN in input
2024-06-22 15:55:29 +02:00
e112191b54
POWER: Fix issues in zscal to address lapack failures
...
This patch fixes following lapack failures with clang compiler on POWER.
zed.out: ZVX: 18 out of 5190 tests failed to pass the threshold
zgd.out: ZGV drivers: 25 out of 1092 tests failed to pass the threshold
zgd.out: ZGV drivers: 6 out of 1092 tests failed to pass the threshold
2024-05-22 08:00:06 -05:00
aa259b141d
Merge pull request #4704 from amritahs-ibm/saxpy_perf_fix
...
Fix regression SAXPY when compiler with OpenXL compiler.
2024-05-18 19:11:25 +02:00
3a1417671a
POWER: Fixing endianness issue in cswap/zswap kernel for AIX
2024-05-15 19:36:46 -05:00
87b3d9054f
Fix regression SAXPY when compiler with OpenXL compiler.
...
SAXPY built with OpenXL regresses when compared to SAXPY
built with gcc. OpenXL compiler doesn't know that the
SAXPY inner kernel assembly is a 64 element loop and
to it the remainder loop is the main loop. It vectorizes
and interleaves the remainder to be a 48 elements per
iteration loop. With a max of 63 iterations, a 48 element
loop is mostly not going to get executed, so the 1 element
scalar loop that is the remainder after that is probably
mostly what gets executed.
This can be fixed by adding a pragma, loop interleave_count(2)
which will result in 8 element loop.
Signed-off-by: Amrita H S <amritahs@linux.vnet.ibm.com>
2024-05-12 23:27:55 -05:00
99384933ff
Revert "Merge pull request #4532 from austinpagan/cgemm_zgemm_c_code"
...
This reverts commit accea15551b925353006
2024-03-01 07:57:39 -06:00
accea15551
Merge pull request #4532 from austinpagan/cgemm_zgemm_c_code
...
Cgemm zgemm c code
2024-02-27 22:07:07 +01:00
87ba528d8b
Changed C files to straighten out indentation. Removed commented lines from other file.
2024-02-01 18:46:07 -06:00
ddac75e0ef
Adding .C versions of CGEMM and ZGEMM
2024-02-01 12:24:25 -06:00
2bb7ea64a1
Only vectorize 64-bit version for Power8.
2024-02-01 08:11:43 -06:00
09bb48d1b9
Vectorize in-copy packing/copying for SGEMM - 4X faster.
2024-01-30 09:13:16 -06:00
058dd2a4cb
Replace two vector loads with one vector pair load and fix endianess of stores - DGEMM versions.
2024-01-08 14:16:09 -06:00
d9653af018
KERNEL.PPC970, KERNEL.PPCG4: unbreak CMake parsing
...
Fixes: https://github.com/OpenMathLib/OpenBLAS/issues/4366
2023-12-14 12:00:11 +08:00
4e738e561a
Replace two vector loads with one vector pair load and fix endianess of stores.
2023-12-08 12:36:08 -06:00
Chip Kerchner
Martin Kroeker
Gordon Fossum
Hong Bo Peng
Rajalakshmi Srinivasaraghavan
Amrita H S
austinpagan
barracuda156