Skip to content

Commit 4a7e973

Browse files
amadeuszszMax-Bin
andauthored
feat(autoware_tensorrt_vad): update nvcc flags (#12044)
Signed-off-by: Amadeusz Szymko <amadeusz.szymko.2@tier4.jp> Co-authored-by: Max-Bin <vborisw@gmail.com>
1 parent 6b14d56 commit 4a7e973

File tree

1 file changed

+8
-5
lines changed

1 file changed

+8
-5
lines changed

e2e/autoware_tensorrt_vad/CMakeLists.txt

Lines changed: 8 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -64,13 +64,16 @@ if(TRT_AVAIL AND CUDA_AVAIL)
6464
endif()
6565

6666
# CUDA architecture settings
67-
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_75,code=sm_75")
68-
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_80,code=sm_80")
6967
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_86,code=sm_86")
68+
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_87,code=sm_87")
7069
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_89,code=sm_89")
71-
# Generate both binary and PTX for sm_90 for forward compatibility with newer architectures (e.g., Blackwell)
72-
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_90,code=sm_90")
73-
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_90,code=compute_90")
70+
if(CUDA_VERSION VERSION_LESS "13.0")
71+
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_101,code=sm_101")
72+
else() # CUDA 13.0 renamed SM101 to SM110
73+
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_110,code=sm_110")
74+
endif()
75+
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_120,code=sm_120")
76+
list(APPEND CUDA_NVCC_FLAGS "-gencode arch=compute_120,code=compute_120")
7477

7578
cuda_add_library(${PROJECT_NAME}_cuda_lib SHARED
7679
lib/networks/preprocess/multi_camera_preprocess_kernel.cu

0 commit comments

Comments
 (0)