remove wrong cuda arch that cause big performance gap

fix #226
This commit is contained in:
Francesco Gatti
2021-05-16 22:16:21 +02:00
parent a638592fc7
commit 6b8ae1e27c
+1 -1
View File
@@ -25,7 +25,7 @@ add_definitions(-DTKDNN_PATH="${CMAKE_CURRENT_SOURCE_DIR}")
find_package(CUDA 9.0 REQUIRED)
SET(CUDA_SEPARABLE_COMPILATION ON)
#set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -arch=sm_30 --compiler-options '-fPIC'")
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --maxrregcount=32 -arch=sm_61 )
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --maxrregcount=32)
find_package(CUDNN REQUIRED)
include_directories(${CUDNN_INCLUDE_DIR})