Skip to content

GitLab

"src/include/ConstantTensorDescriptor.hpp" did not exist on "b2888adfbe103ae3d9006af87d5871b69cbf00ba"
Commit c39c573e authored by Chao Liu's avatar Chao Liu
Browse files

refactor

parent c9af4dec
Hide whitespace changes
Inline Side-by-side
Showing with 8 additions and 8 deletions
driver/device_implicit_gemm_convolution_2_cnhw_srck_knhw.cuh
View file @ c39c573e
...@@ -102,8 +102,8 @@ void device_implicit_gemm_convolution_2_cnhw_srck_knhw(InDesc, ...@@ -102,8 +102,8 @@ void device_implicit_gemm_convolution_2_cnhw_srck_knhw(InDesc,
constexpr unsigned KPerThread = 8; constexpr unsigned KPerThread = 8;
constexpr unsigned CPerThread = 1; constexpr unsigned CPerThread = 1;
constexpr unsigned GemmThreadPerClusterRow = 4; constexpr unsigned GemmRowThreadPerCluster = 4;
constexpr unsigned GemmThreadPerClusterColumn = 4; constexpr unsigned GemmColumnThreadPerCluster = 4;
constexpr unsigned InBlockCopyThreadPerDim0 = 2; constexpr unsigned InBlockCopyThreadPerDim0 = 2;
constexpr unsigned InBlockCopyThreadPerDim1 = 64; constexpr unsigned InBlockCopyThreadPerDim1 = 64;
...@@ -149,8 +149,8 @@ void device_implicit_gemm_convolution_2_cnhw_srck_knhw(InDesc, ...@@ -149,8 +149,8 @@ void device_implicit_gemm_convolution_2_cnhw_srck_knhw(InDesc,
BPerThread, BPerThread,
KPerThread, KPerThread,
CPerThread, CPerThread,
GemmThreadPerClusterRow, GemmRowThreadPerCluster,
GemmThreadPerClusterColumn, GemmColumnThreadPerCluster,
InBlockCopyThreadPerDim0, InBlockCopyThreadPerDim0,
InBlockCopyThreadPerDim1> InBlockCopyThreadPerDim1>
<<<grid_dim, block_dim>>>(in_cnhw_desc, <<<grid_dim, block_dim>>>(in_cnhw_desc,
......
src/include/gridwise_implicit_gemm_convolution_3_cnhw_srck_knhw.cuh
View file @ c39c573e
...@@ -20,8 +20,8 @@ template <unsigned GridSize, ...@@ -20,8 +20,8 @@ template <unsigned GridSize,
unsigned BPerThread, unsigned BPerThread,
unsigned KPerThread, unsigned KPerThread,
unsigned CPerThread, unsigned CPerThread,
unsigned GemmThreadPerClusterRow, unsigned GemmRowThreadPerCluster,
unsigned GemmThreadPerClusterColumn, unsigned GemmColumnThreadPerCluster,
unsigned InBlockCopyThreadPerDim0, unsigned InBlockCopyThreadPerDim0,
unsigned InBlockCopyThreadPerDim1> unsigned InBlockCopyThreadPerDim1>
__global__ void __global__ void
...@@ -177,8 +177,8 @@ gridwise_implicit_gemm_convolution_3_cnhw_srck_knhw(InGlobalDesc, ...@@ -177,8 +177,8 @@ gridwise_implicit_gemm_convolution_3_cnhw_srck_knhw(InGlobalDesc,
false, false,
false, false,
CPerThread, CPerThread,
GemmThreadPerClusterRow, GemmRowThreadPerCluster,
GemmThreadPerClusterColumn, GemmColumnThreadPerCluster,
true>{}; true>{};
// LDS // LDS
......
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