leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 29 Packages Projects Releases Wiki Activity

Updates to fused epilogue (#383)

Browse Source 
* Enhancements and fixes to fused GEMM and Convolution epilogue.
* Need to explicitly list cudart as unit test library dependency.
This commit is contained in:
Andrew Kerr
2021-12-17 16:04:43 -05:00
committed by GitHub
parent 4e666e1dfd
commit ec4f7e5194
24 changed files with 372 additions and 193 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CMakeLists.txt
+11 -1
View File
@@ -197,9 +197,19 @@ list(APPEND CUTLASS_CUDA_NVCC_FLAGS -DCUTLASS_TEST_LEVEL=${CUTLASS_TEST_LEVEL})
list(APPEND CUTLASS_CUDA_CLANG_FLAGS -DCUTLASS_TEST_LEVEL=${CUTLASS_TEST_LEVEL})
if (CUTLASS_TEST_ENABLE_CACHED_RESULTS)
list(APPEND CUTLASS_CUDA_NVCC_FLAGS -DCUTLASS_TEST_ENABLE_CACHED_RESULTS=1)
message(STATUS "Enable caching of reference results in conv unit tests")
list(APPEND CUTLASS_CUDA_NVCC_FLAGS -DCUTLASS_TEST_ENABLE_CACHED_RESULTS=1)
endif()
set(CUTLASS_CONV_UNIT_TEST_RIGOROUS_SIZE_ENABLED ON CACHE BOOL "Enable/Disable rigorous conv problem sizes in conv unit tests")
if (CUTLASS_CONV_UNIT_TEST_RIGOROUS_SIZE_ENABLED)
message(STATUS "Enable rigorous conv problem sizes in conv unit tests")
list(APPEND CUTLASS_CUDA_NVCC_FLAGS -DCUTLASS_CONV_UNIT_TEST_RIGOROUS_SIZE_ENABLED=1)
endif()
#
# CUDA 10.1 introduces "mma" in PTX performing collective matrix multiply operations.
#