A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 13, 04:30 UTC
[Bugfix] Build hadacore kernels on >SM90 (#28748)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Nov 14, 2025 at 22:51 UTC
edfe49818959b1a1a0b7e8ef7ffcdc39d9903ec6
1 file changed
+1
-1
CMakeLists.txt
+1
-1
@@ -861,7 +861,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
861
endif()
862
863
# Hadacore kernels
864
- cuda_archs_loose_intersection(HADACORE_ARCHS "8.0;8.9;9.0" "${CUDA_ARCHS}")
864
+ cuda_archs_loose_intersection(HADACORE_ARCHS "8.0+PTX;9.0+PTX" "${CUDA_ARCHS}")
865
if(HADACORE_ARCHS)
866
set(SRCS "csrc/quantization/hadamard/hadacore/hadamard_transform_cuda.cu")
867
set_gencode_flags_for_srcs(