fe93153404
Merge pull request #413 from CNugteren/CLBlast-412-python-runtime-libs
...
Add library dir on Linux for pyclblast
2021-02-04 20:45:40 +01:00
d57f8065ea
Added second Windows library path
2021-02-04 20:13:02 +01:00
c78c649844
Add library path for Windows as well
2021-01-30 14:28:11 +01:00
bbcb357a71
Add library dir on Linux for pyclblast
2021-01-29 20:48:05 +01:00
07837a5c2d
Update pyclblast package version number
2021-01-21 20:49:31 +01:00
a5ef06ec57
Merge pull request #410 from jamesjer/master
...
Use reference types to prevent unnecessary copying
2021-01-21 19:56:18 +01:00
dc82a1fbc8
Use reference types to prevent unnecessary copying
2021-01-20 10:21:36 -07:00
70016e8698
Updated to version 1.5.2
2021-01-19 21:19:12 +01:00
0ee39af5ed
Add tuning results for TITAN RTX
2020-10-10 13:01:12 +02:00
481d86665f
Add tuning results for Radeon RX Vega
2020-10-10 12:56:28 +02:00
e6e2519eaa
Merge pull request #400 from baryluk/patch-6
...
Allow single graph / subplot on plot
2020-10-05 21:19:34 +02:00
ea199c3469
Allow single graph / subplot on plot
...
`plt.subplots` tries to be special, and return array or not-array depending on a number of subplots.
It is not actually helpful, and IMHO bad design.
Make it always `ndarray`.
The `and not type(axes) is np.ndarray`, is just in case matplotlib decides to make their behavior more uniform. For now work around it.
Also, no need for `ndarray.flat` really.
Confirmed to work with existing benchmarks (i.e. rows=2, cols=3), and with single graphs (rows=1, cols=1).
2020-10-05 12:11:17 +00:00
3462d7fa85
Merge pull request #399 from baryluk/patch-3
...
Fix a typo in benchmark when running fp 16 vs 32
2020-10-04 16:43:21 +02:00
eb967a0943
Fix a typo in benchmark when running fp 16 vs 32
...
The intention here was to limit the iteration range to common indexes only.
Fix that.
2020-10-04 10:22:00 +00:00
615e5f0ff2
Merge pull request #397 from baryluk/patch-1
...
Fix Python SyntaxWarning
2020-10-04 11:07:56 +02:00
cdcfbbc8bc
Merge pull request #398 from baryluk/patch-2
...
Fix --load_from_disk argument help message
2020-10-04 11:07:09 +02:00
2dfe7c5c23
Fix --load_from_disk argument help message
2020-10-04 08:17:16 +00:00
45fd085395
Fix Python SyntaxWarning
...
There is no guarantee that all empty strings objects are the same or share object with `""` literal.
2020-10-04 08:12:50 +00:00
46fb748a96
Merge pull request #396 from CNugteren/CLBlast-395-fix-benchmark-script
...
Fix a Python 3 bug in the benchmark script
2020-10-03 10:50:43 +02:00
0abd62a0e7
Fix a Python 3 bug in the benchmark script
2020-10-02 20:32:58 +02:00
b4cd2b04e9
Added FUNDING.yml file
2020-08-16 10:33:47 +02:00
41f344d1a6
Merge pull request #392 from 9prady9/fix_Program_getIR
...
Fix Program::GetIR to handle programs with multiple devices
2020-06-07 19:52:49 +02:00
dff65e9217
Add a cautionary note in Program::GetIR and mention the fix in CHANGELOG
2020-06-07 21:13:33 +05:30
aec71699f8
Fix Program::GetIR to handle programs with multiple devices
2020-06-05 12:00:45 +05:30
da0e657d39
Merge pull request #389 from CNugteren/CLBlast-385-version-defines
...
Added version number defines
2020-05-13 20:28:58 +02:00
396ac0278a
Added CLBLAST_VERSION_MAJOR/MINOR/PATCH defines in headers to store version numbering
2020-05-12 14:43:25 +02:00
0826bfe683
Merge pull request #388 from CNugteren/CLBlast-381-gemm-direct-tuner-failure
...
Fixed tuners global workgroup size
2020-05-11 22:39:48 +02:00
c369cf1a16
Increase display width of the local/global sizes
2020-05-11 20:26:33 +02:00
4a6c7c37a3
Made sure that the global workgroup size is a multiple of the local size in the tuners
2020-05-10 20:28:23 +02:00
69a4b4d4b0
Added logging of local/global workgroup sizes when run the tuners
2020-05-10 20:08:28 +02:00
9abc416785
Merge pull request #386 from CNugteren/CLBlast-384-pyclblast-missing-routines
...
PyCLBlast: add missing batched routines
2020-05-10 18:23:41 +02:00
0870e76fba
Updated PyCLBlast version number
2020-05-10 14:55:03 +02:00
0b7ce8033c
Added a sample to demonstrate a batched routine
2020-05-10 14:54:50 +02:00
b94e81af10
Added pyclblast bindings for the 3 batched routines
2020-05-10 12:26:25 +02:00
5f4b3ffcf7
Merge pull request #383 from CNugteren/CLBlast-382-improve-tuner
...
Move queue creation out of the tuner loop
2020-05-04 20:26:42 +02:00
bbb2031bf3
Move queue creation out of the tuner loop
2020-05-03 20:30:55 +02:00
78300ccbea
Merge pull request #378 from CNugteren/CLBlast-377-fix-amax-amin
...
Change amax/amin behaviour
2020-03-15 11:34:31 +01:00
5f97d64505
Update API documentation
2020-03-08 11:29:47 +01:00
b46853660e
Made it more likely (but no guarantees) for amax/amin to return the first index
2020-03-08 11:26:49 +01:00
7fab29304c
Added sample to play around with XAMAX routine
2020-03-08 11:26:18 +01:00
e3ce88154a
Silenced a new OpenCL warning message
2020-03-08 10:14:59 +01:00
8433985051
Updated to version 1.5.1
2020-02-18 10:29:40 +01:00
bf4e4198b7
Merge pull request #376 from CNugteren/fix_tuner_exception_catching
...
Catches all exceptions of the tuners
2020-02-18 10:23:43 +01:00
49eb490ee1
Catches all exceptions of the tuners
2020-02-17 22:07:51 +01:00
8a19667e75
Merge pull request #372 from trantila/master
...
Reduced number of TestMatrix calls for the batched xgemm routines.
2019-12-15 09:33:53 +01:00
21b66ca761
Reduce TestMatrix calls for xgemmstridedbatched.
...
Replace the looped test by a single one with the offset of the last batch.
2019-12-09 22:17:24 +02:00
bf50c4e53e
Reduce TestMatrix calls for xgemmbatched.
...
Replace the looped test by a single one with the maximal found offset.
2019-12-09 22:13:52 +02:00
6ac74008b6
Added notion of fixes in XhadFaster
2019-09-06 19:33:30 +02:00
701ac9bf76
Merge pull request #368 from etomzak/master
...
Fix out-of-bounds read/write in XhadFaster
2019-09-06 19:30:52 +02:00
9560193a9e
Fix out-of-bounds read/write in XhadFaster
...
Fix an error in XhadFaster where data would be written beyond the end of zgm.
The kernel loop assumed that there was always enough work for each thread to
process WPT items, but this was not enforced. It's possible to detect the
overflow with the "canary" buffer regions, but for SHAD, kCanarySize must be
~500 (much larger than the normal 127).
This commit may improve the performance of XhadFaster, since the kernel was
performing 2x work in some cases (once over real data, once over garbage).
Courtesy of Codeplay Software Ltd.
2019-09-04 12:55:25 +01:00
Cedric Nugteren
Jerry James
Witold Baryluk
Pradeep Garigipati
Tarmo Räntilä
etomzak