mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-17 03:19:48 +00:00
add cuda extract_asm script
[ROCm/composable_kernel commit: e6c86f81b5]
This commit is contained in:
@@ -15,11 +15,12 @@ cmake
|
||||
-D DEVICE_BACKEND=CUDA \
|
||||
-D BOOST_ROOT="/package/install/boost_1.67.0" \
|
||||
-D CUDA_COMMON_INCLUDE_DIR="/home/chao/code/test_feature/cuda_common/cuda_10.0_common/inc" \
|
||||
-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -arch=sm_61" \
|
||||
-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -gencode=arch=compute_61,code=sm_61 -Xptxas -v -Xptxas -v -maxrregcount=128" \
|
||||
${MY_PROJECT_SOURCE}
|
||||
|
||||
|
||||
#-D CMAKE_CUDA_COMPILER="/package/install/cuda_10.0/bin/nvcc" \
|
||||
#-D CMAKE_CUDA_FLAGS="-G -lineinfo --source-in-ptx -keep -Xptxas -v -arch=sm_61" \
|
||||
#-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -arch=sm_61" \
|
||||
#-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -arch=sm_61 -Xptxas -v -maxrregcount=128" \
|
||||
#-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -gencode=arch=compute_61,code=sm_61" \
|
||||
#-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -gencode=arch=compute_61,code=sm_61 -Xptxas -v -Xptxas -v -maxrregcount=128" \
|
||||
#-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -gencode=arch=compute_61,code=sm_61 -Xptxas -v -gencode=arch=compute_70,code=sm_70" \
|
||||
#-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -gencode=arch=compute_61,code=sm_61 -Xptxas -v -gencode=arch=compute_70,code=sm_70 -Xptxas -v -maxrregcount=128" \
|
||||
|
||||
1
script/extract_asm-cuda.sh
Executable file
1
script/extract_asm-cuda.sh
Executable file
@@ -0,0 +1 @@
|
||||
cuobjdump -xelf all ./driver/driver && nvdisasm --print-code -g driver.sm_61.cubin > driver.sm_61.asm && nvdisasm --print-code -g driver.sm_70.cubin > driver.sm_70.asm
|
||||
Reference in New Issue
Block a user