ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-11 17:00:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

debugging

Browse Source
This commit is contained in:
Chao Liu
2019-04-05 18:54:26 -05:00
parent bae2333791
commit 0983d205ad
5 changed files with 62 additions and 109 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
driver/device_implicit_gemm_convolution_2_chwn_cyxk_khwn.hpp
View File

@@ -191,7 +191,7 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
constexpr index_t WeiBlockCopyDataPerRead = 4;
constexpr index_t BlockSize = 256;
#elif 0
#elif 1
// 1x1, 14x14, Vega 20, disable lds_double_buffer, enable register double buffer
constexpr index_t BPerBlock = 64;
constexpr index_t KPerBlock = 128;
@@ -208,9 +208,6 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
constexpr index_t GemmNLevel1Cluster = 4;
constexpr index_t GemmKPerThreadLoop = 1;
constexpr index_t GemmThreadPerColumnPerCluster = 8;
constexpr index_t GemmThreadPerRowPerCluster = 8;
constexpr index_t InBlockCopyThreadPerDim0 = 4;
constexpr index_t InBlockCopyThreadPerDim1 = 16;

4
driver/driver.hip.cpp
View File

@@ -580,7 +580,7 @@ int main(int argc, char* argv[])
constexpr index_t HPad = 0;
constexpr index_t WPad = 0;
#elif 1
#elif 0
// 1x1 filter, 14x14 image, C = 2048
constexpr index_t N = 128;
constexpr index_t C = 2048;
@@ -592,7 +592,7 @@ int main(int argc, char* argv[])
constexpr index_t HPad = 0;
constexpr index_t WPad = 0;
#elif 0
#elif 1
// 1x1 filter, 14x14 image, C = 512
constexpr index_t N = 128;
constexpr index_t C = 512;