ROCm / hipBLASLt

hipBLASLt is a library that provides general matrix-matrix operations with a flexible API and extends functionalities beyond a traditional BLAS library
https://rocm.docs.amd.com/projects/hipBLASLt/en/latest/index.html
MIT License
63 stars 88 forks source link

Tune Aquavanjaram942 Grid sizes for BBS NN #1120

Closed pkamd closed 1 month ago

pkamd commented 2 months ago

Updated 8 grid points to address perf tuning request. Ensure changes doesn't degrade existing grid point performance.

Note - Same changes as old PR.

pkamd commented 2 months ago

All hipBLASLt tests passed -

[----------] 1 test from ExtOpTest/ExtOpAMaxWithScaleUnsupportedDatatypeTest
[ RUN      ] ExtOpTest/ExtOpAMaxWithScaleUnsupportedDatatypeTest.amaxWithScaleFailureUnsupportedDatatype/0
[       OK ] ExtOpTest/ExtOpAMaxWithScaleUnsupportedDatatypeTest.amaxWithScaleFailureUnsupportedDatatype/0 (0 ms)
[----------] 1 test from ExtOpTest/ExtOpAMaxWithScaleUnsupportedDatatypeTest (0 ms total)

[----------] Global test environment tear-down
[==========] 48206 tests from 13 test suites ran. (2234494 ms total)
[  PASSED  ] 48206 tests.
hipBLASLt version: 1000

command line: ./hipblaslt-test
StaceyLai commented 1 month ago

If your commit involves changes to a YAML file, please take note.

Due to PR-1117 (https://github.com/ROCm/hipBLASLt/pull/1117) adding ActivationType: hipblaslt_all, all YAML files now need to include the ActivationType field (for hipblaslt logic YAML, use ActivationType: hipblaslt_all (not all); for tensile test YAML, you can use either all or hipblaslt_all).

Correspondingly, in the Logic YAML filename, the _A that originally represented ActivationType: all should also be changed to _HA. These changes have been implemented in PR-1117.

nakajee commented 1 month ago

I removed noCI label.

pkamd commented 1 month ago
[       OK ] ExtOpTest/ExtOpAMaxWithScaleUnsupportedDatatypeTest.amaxWithScaleFailureUnsupportedDatatype/0 (0 ms)
[----------] 1 test from ExtOpTest/ExtOpAMaxWithScaleUnsupportedDatatypeTest (0 ms total)

[----------] Global test environment tear-down
[==========] 48206 tests from 13 test suites ran. (2371439 ms total)
[  PASSED  ] 48206 tests.
hipBLASLt version: 1000

command line: ./hipblaslt-test