Skip to content

GitLab

"...composable_kernel.git" did not exist on "d32a67a9b6f58de5e05f65d2bb8b834a253c94bd"
Commit 9b4fdeee authored by Jing Zhang's avatar Jing Zhang
Browse files

fix

parent df6dd915
Hide whitespace changes
Inline Side-by-side
Showing with 12 additions and 11 deletions
driver/src/conv_driver.cpp
View file @ 9b4fdeee
...@@ -20,6 +20,7 @@ ...@@ -20,6 +20,7 @@
//#include "device_convolution_implicit_gemm_v3_nchw_cyxk_nkhw.hpp" //#include "device_convolution_implicit_gemm_v3_nchw_cyxk_nkhw.hpp"
#include "device_convolution_implicit_gemm_v4r1_nchw_kcyx_nkhw.hpp" #include "device_convolution_implicit_gemm_v4r1_nchw_kcyx_nkhw.hpp"
#include "device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw.hpp" #include "device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw.hpp"
#include "device_convolution_implicit_gemm_v4r4_xdlops_nchw_kcyx_nkhw.hpp"
int main(int argc, char* argv[]) int main(int argc, char* argv[])
{ {
...@@ -133,14 +134,14 @@ int main(int argc, char* argv[]) ...@@ -133,14 +134,14 @@ int main(int argc, char* argv[])
#elif 1 #elif 1
// 3x3, 299x299 stride=2 // 3x3, 299x299 stride=2
constexpr index_t N = 128; constexpr index_t N = 128;
constexpr index_t C = 3; constexpr index_t C = 1024;
constexpr index_t HI = 299; constexpr index_t HI = 14;
constexpr index_t WI = 299; constexpr index_t WI = 14;
constexpr index_t K = 32; constexpr index_t K = 1024;
constexpr index_t Y = 3; constexpr index_t Y = 1;
constexpr index_t X = 3; constexpr index_t X = 1;
using ConvStrides = Sequence<2, 2>; using ConvStrides = Sequence<1, 1>;
using ConvDilations = Sequence<1, 1>; using ConvDilations = Sequence<1, 1>;
using LeftPads = Sequence<0, 0>; using LeftPads = Sequence<0, 0>;
......
script/cmake-cuda_docker.sh
View file @ 9b4fdeee
#!/bin/bash #!/bin/bash
MY_PROJECT_SOURCE=../../../ MY_PROJECT_SOURCE=../
MY_PROJECT_INSTALL=../install.dir MY_PROJECT_INSTALL=../install.dir
export CUDA_ROOT=/usr/local/cuda export CUDA_ROOT=/usr/local/cuda
...@@ -15,9 +15,9 @@ cmake ...@@ -15,9 +15,9 @@ cmake
-D CMAKE_VERBOSE_MAKEFILE:BOOL=ON \ -D CMAKE_VERBOSE_MAKEFILE:BOOL=ON \
-D DEVICE_BACKEND=NVIDIA \ -D DEVICE_BACKEND=NVIDIA \
-D CUDA_COMMON_INCLUDE_DIR="/root/NVIDIA_CUDA-10.1_Samples/common/inc" \ -D CUDA_COMMON_INCLUDE_DIR="/root/NVIDIA_CUDA-10.1_Samples/common/inc" \
-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 -maxrregcount=128" \ -D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -gencode=arch=compute_70,code=sm_70 -Xptxas -v -maxrregcount=128" \
${MY_PROJECT_SOURCE} ${MY_PROJECT_SOURCE}
#-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_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 -maxrregcount=128" \ #-D CMAKE_CUDA_FLAGS="-ccbin clang++ -m64 -Xcompiler -fopenmp -lineinfo --source-in-ptx -keep -Xptxas -v -gencode=arch=compute_70,code=sm_70 -Xptxas -v -maxrregcount=128" \
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment