v4.0 update. (#2371)

This commit is contained in:
Junkai-Wu
2025-06-06 02:39:20 -04:00
committed by GitHub
parent 2e2af190bd
commit 8bdbfca682
254 changed files with 29751 additions and 1980 deletions
+8 -2
View File
@@ -175,11 +175,13 @@ if (CUDA_VERSION VERSION_GREATER_EQUAL 12.0)
endif()
if (CUDA_VERSION VERSION_GREATER_EQUAL 12.8)
list(APPEND CUTLASS_NVCC_ARCHS_SUPPORTED 100 100a 101 101a 120 120a)
list(APPEND CUTLASS_NVCC_ARCHS_SUPPORTED 100 100a 120 120a)
list(APPEND CUTLASS_NVCC_ARCHS_SUPPORTED 101 101a)
endif()
if (CUDA_VERSION VERSION_GREATER_EQUAL 12.9)
list(APPEND CUTLASS_NVCC_ARCHS_SUPPORTED 100f 101f 120f)
list(APPEND CUTLASS_NVCC_ARCHS_SUPPORTED 100f 120f)
list(APPEND CUTLASS_NVCC_ARCHS_SUPPORTED 101f)
endif()
set(CUTLASS_NVCC_ARCHS ${CUTLASS_NVCC_ARCHS_SUPPORTED} CACHE STRING "The SM architectures requested.")
@@ -344,6 +346,10 @@ if(CUTLASS_ENABLE_SM90_EXTENDED_MMA_SHAPES)
list(APPEND CUTLASS_CUDA_NVCC_FLAGS -DCUTE_SM90_EXTENDED_MMA_SHAPES_ENABLED)
endif()
if (CUTLASS_NVCC_ARCHS MATCHES 100f OR CUTLASS_NVCC_ARCHS MATCHES 101f)
list(APPEND CUTLASS_CUDA_NVCC_FLAGS -DCUTLASS_SM100_FAMILY_ARCHS_ENABLED)
endif()
set(CUTLASS_SKIP_REDUCTION_INIT OFF CACHE BOOL "Disable init reduction workspace")
#