Wmma support for gemm_multiply_multiply_wp (#3278)

* Initial implementation with splitK support

* Add gfx11 support

* Fix compilation error

* Add instances

* Add irregular instances

* Fix GetBuffer arguments

* Minor changes

* Address review comments

* Fix compilation errors

* Fix copyright header
This commit is contained in:
Enrico Degregori
2025-12-03 16:38:23 +01:00
committed by GitHub
parent f29b67cf9b
commit 161835533b
30 changed files with 2482 additions and 86 deletions

View File

@@ -2,8 +2,8 @@
# SPDX-License-Identifier: MIT
if(GPU_TARGETS MATCHES "gfx9[45]|gfx12")
add_gtest_executable(test_gemm_multiply_multiply_wp_xdl_fp8 test_gemm_multiply_multiply_wp_xdl_fp8.cpp)
add_gtest_executable(test_gemm_multiply_multiply_wp_fp8 test_gemm_multiply_multiply_wp_fp8.cpp)
if(result EQUAL 0)
target_link_libraries(test_gemm_multiply_multiply_wp_xdl_fp8 PRIVATE utility device_gemm_multiply_multiply_wp_instance)
target_link_libraries(test_gemm_multiply_multiply_wp_fp8 PRIVATE utility device_gemm_multiply_multiply_wp_instance)
endif()
endif()