Skip to content

Commit 662c27e

Browse files
committed
udpate CMakeLists for seperate Blackwell fp8 entry
1 parent bbd291e commit 662c27e

File tree

1 file changed

+14
-3
lines changed

1 file changed

+14
-3
lines changed

CMakeLists.txt

Lines changed: 14 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -419,9 +419,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
419419
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND FP4_ARCHS)
420420
set(SRCS
421421
"csrc/quantization/fp4/nvfp4_quant_kernels.cu"
422-
"csrc/quantization/fp4/nvfp4_scaled_mm_kernels.cu"
423-
"csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm100_fp8.cu"
424-
)
422+
"csrc/quantization/fp4/nvfp4_scaled_mm_kernels.cu")
425423
set_gencode_flags_for_srcs(
426424
SRCS "${SRCS}"
427425
CUDA_ARCHS "${FP4_ARCHS}")
@@ -434,6 +432,19 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
434432
set(FP4_ARCHS)
435433
endif()
436434

435+
# FP8 Blackwell Archs
436+
cuda_archs_loose_intersection(BLACKWELL_ARCHS "10.0;10.1;12.0" "${CUDA_ARCHS}")
437+
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND BLACKWELL_ARCHS)
438+
set(SRCS
439+
"csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm100_fp8.cu"
440+
)
441+
list(APPEND VLLM_EXT_SRC "${SRCS}")
442+
message(STATUS "Building FP8 for archs: ${BLACKWELL_ARCHS}")
443+
else()
444+
# clear BLACKWELL_ARCHS
445+
set(BLACKWELL_ARCHS)
446+
endif()
447+
437448
#
438449
# Machete kernels
439450

0 commit comments

Comments
 (0)