bae45d94d1
scal benchmark: eliminate y, move init/timing out of loop
...
Removing y avoids cache effects (if y is the size of the L1 cache, the
main array x is removed from it).
Moving init and timing out of the loop makes the scal benchmark behave like
the gemm benchmark, and allows higher accuracy for smaller test cases since
the loop overhead is much smaller than the timing overhead.
Example:
OPENBLAS_LOOPS=10000 ./dscal.goto 1024 8192 1024
on AMD Zen2 (7532) with 32k (4k doubles) L1 cache per core.
Before
From : 1024 To : 8192 Step = 1024 Inc_x = 1 Inc_y = 1 Loops = 10000
SIZE Flops
1024 : 5627.08 MFlops 0.000000 sec
2048 : 5907.34 MFlops 0.000000 sec
3072 : 5553.30 MFlops 0.000001 sec
4096 : 5446.38 MFlops 0.000001 sec
5120 : 5504.61 MFlops 0.000001 sec
6144 : 5501.80 MFlops 0.000001 sec
7168 : 5547.43 MFlops 0.000001 sec
8192 : 5548.46 MFlops 0.000001 sec
After
From : 1024 To : 8192 Step = 1024 Inc_x = 1 Inc_y = 1 Loops = 10000
SIZE Flops
1024 : 6310.28 MFlops 0.000000 sec
2048 : 6396.29 MFlops 0.000000 sec
3072 : 6439.14 MFlops 0.000000 sec
4096 : 6327.14 MFlops 0.000001 sec
5120 : 5628.24 MFlops 0.000001 sec
6144 : 5616.41 MFlops 0.000001 sec
7168 : 5553.13 MFlops 0.000001 sec
8192 : 5600.88 MFlops 0.000001 sec
We can see the L1->L2 switchover point is now where it should be, and the
number of flops for L1 is more accurate.
2022-11-29 08:02:45 -05:00
f92dd6e303
change line endings from CRLF to LF
2022-11-17 10:18:36 +01:00
9e6b060bf3
Fix comment.
...
It stores the pointer, not an offset (that would be an alternative approach).
2022-10-20 20:11:09 -04:00
9959a60873
Benchmarks: align malloc'ed buffers.
...
Benchmarks should allocate with cacheline (often 64 bytes) alignment
to avoid unreliable timings. This technique, storing the offset in the
byte before the pointer, doesn't require C11's aligned_alloc for
compatibility with older compilers.
For example, Glibc's x86_64 malloc returns 16-byte aligned buffers, which is
not sufficient for AVX/AVX2 (32-byte preferred) or AVX512 (64-byte).
2022-10-20 13:28:20 -04:00
f119e26354
Fix flipped indices in benchmark for gemv
...
Fixes #3439
2021-11-03 12:45:09 +01:00
14e33e0f7e
Handle OPENBLAS_LOOPS in SYR2 benchmark
2021-07-10 21:27:53 +02:00
4ed99c2ce3
Merge pull request #3292 from martin-frbg/syrk_limit
...
Add lower limit for multithreading in xSYRK
2021-07-07 20:46:28 +02:00
a4543e4918
Handle OPENBLAS_LOOP
2021-07-04 16:59:43 +02:00
dcfc5cf714
Handle OPENBLAS_LOOPS for more stable results
2021-07-01 17:39:37 +02:00
06e3b07ecb
Handle OPENBLAS_LOOPS and OPENBLAS_TEST options
2021-07-01 17:38:45 +02:00
1f8bda71b9
Add OPENBLAS_LOOPS support to potrf/potrs/potri benchmark
2021-06-26 23:46:00 +02:00
d57c681a6d
Fix compilation on older OSX versions
2021-03-26 22:29:29 +01:00
38dcf3454b
Support timing Apple M1
2021-03-02 17:50:55 +01:00
f917c26e83
Refractoring remaining benchmark cases.
2020-10-26 10:25:05 +08:00
dd6ebdfdab
Refactor the performance measurement system
2020-10-23 10:32:03 +08:00
7ae9e8960e
Change "HALF" and "sh" to "BFLOAT16" and "sb"
2020-10-12 00:08:29 +02:00
5464eb13ea
Change ifdef linux to __linux for C11 compatibility
2020-09-30 22:59:41 +02:00
6f8fad87c5
Use POSIX2001 clock.gettime for higher resolution
2020-09-05 19:44:01 +02:00
ced49466f0
Use the fortran compiler to link LAPACK-related benchmarks
...
to fix linking problems with (at least) the AMD version of flang that creates dependencies on more than just the fortran runtime.
2020-05-29 13:35:51 +02:00
6e270f91ec
add support for RETURN_BY_STACK semantics, e.g. clang
2020-05-29 13:29:10 +02:00
ce90e2bd3f
Include shgemm in benchtest
...
This patch is to enable benchtest for half precision gemm
when BUILD_HALF is set during make.
2020-05-11 09:57:46 -05:00
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
717c604aeb
Merge pull request #2515 from zelong-1024/develop
...
[OpenBLAS]: benchmark for her/her2 LEVEL2 functions
2020-03-16 21:59:55 +01:00
ce33da4cab
Merge pull request #2513 from aaawuanjun/develop
...
[OpenBlas]: Add benchmark tpsv file and modify benchmark/Makefile
2020-03-16 21:58:55 +01:00
d45c53ecf1
[OpenBLAS]: benchmark for her/her2 LEVEL2 functions
...
[description]: benchmark for her/her2
[solution]: added benchmark for her/her2, modified makefile in benchmark
[dts]:
2020-03-16 11:19:05 +08:00
c2840997db
Merge pull request #2508 from liujingjue/develop
...
[OpenBLAS]:fix the iamax benchmark error
2020-03-14 14:21:30 +01:00
c0649aa694
Merge pull request #2506 from xiaofengF/develop
...
Add benchmark for SPMV and fix segmentation fault when data size >= 50000
2020-03-14 13:08:36 +01:00
2428dc9fd3
[OpenBlas]: Add benchmark tpsv file and modify benchmark/Makefile
...
[Description]: Solve lack of tpsv benchmark.
2020-03-14 09:11:08 +08:00
a0a3bf7c81
[OpenBLAS]:fix the iamax benchmark error
...
[Description]:the result for i?amax is not MFlops, it is MBytes
2020-03-13 10:58:39 +08:00
ae3f2c2e49
Remove cspmv and zspmv to remove the error occured in travis CI
2020-03-11 17:02:34 +08:00
649733ff15
Only keep spmv.goto and spmv.atlas
2020-03-11 15:48:58 +08:00
3e8f1c6cc5
[OpenBlas]:Add benchmark tpmv.c and modify Makefile
...
[Description]:Solve the problem of missing tpmv.c benchmark file
2020-03-11 12:31:48 +08:00
2f4c5bb3a9
Update spmv.c: solve segmentation fault when m and n are larger than 50000
2020-03-11 10:30:09 +08:00
047dfb216d
Merge pull request #2501 from jijiwawa/Fix_mistakes
...
Fix pr #2487 error
2020-03-10 16:44:40 +01:00
cd8871f1a1
Use the correct unit of measure
2020-03-10 19:26:06 -04:00
08e1d8cbae
Modify Makefile in Benchmark
2020-03-10 14:32:18 +08:00
ff40a4e726
Add benchmark for SPMV
2020-03-10 14:22:18 +08:00
c5bdd21352
Add benchmark for ?amax, ?max, ?amin, ?min, i?max, i?amin and i?min.
2020-03-09 14:59:03 +08:00
b6a6ccbbea
Merge pull request #2495 from ZuoQ3/develop
...
add benchmark for axpby test
2020-03-08 08:09:58 +01:00
8b720f7365
Merge pull request #2494 from shengyang-3390/develop
...
add benchmark for csrot and zdrot
2020-03-07 23:04:21 +01:00
14df234edb
Merge pull request #2489 from jijiwawa/brightness
...
Remove redundant code
2020-03-07 22:26:00 +01:00
bbeda55b7b
add trmm.c
2020-03-07 13:09:19 -05:00
efcf89aec7
Remove redundant code
2020-03-07 12:03:05 -05:00
0c8162eba6
Add benchmark file axpby.c and modify benchmark/Makefile to test s/d/c/zaxpby
2020-03-07 17:48:55 +08:00
09c7a191bd
add benchmark for csrot and zdrot
...
modified: benchmark/Makefile
modified: benchmark/rot.c
2020-03-07 15:17:49 +08:00
dca3e0cf20
Merge pull request #2491 from chenxuqiang/hbmv_benchmark
...
benchmark/hpmv&hbmv: add benchmark/hpmv.c and benchmark/hbmv.c
2020-03-06 15:06:42 +01:00
c9f8db979b
Merge pull request #2490 from shengyang-3390/develop
...
Add benchmark file rotm.c and modify benchmark/Makefile to test s/drotm
2020-03-06 15:05:55 +01:00
97c36ca58c
Merge branch 'develop' into develop
2020-03-06 14:41:40 +01:00
9f5a74f3c7
Merge pull request #2486 from qqqil/develop
...
add benchmark for trsv
2020-03-06 14:30:09 +01:00
2afb10975d
Merge pull request #2485 from Darkness303/develop
...
Add syr2 benchmark
2020-03-06 14:29:27 +01:00
Bart Oldeman
Martin Kroeker
Marius Hillenbrand
Qiyu8
Rajalakshmi Srinivasaraghavan
l00536773
wuanjun 00447568
l00546269
jayfely@qq.com
s00527847
s00548429
zq
shengyang