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

Jing's ds_read inline asm

Browse Source
This commit is contained in:
Chao Liu
2019-03-28 19:46:29 -05:00
parent 766b0a9eaf
commit d6d9a8e4ce
6 changed files with 126 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
driver/device_implicit_gemm_convolution_2_chwn_cyxk_khwn.hpp
View File

@@ -190,8 +190,8 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
constexpr index_t WeiBlockCopyDataPerRead = 4;
constexpr index_t BlockSize = 256;
#elif 1
// 1x1, 14x14, Vega 10
#elif 0
// 1x1, 14x14, Vega 20
constexpr index_t BPerBlock = 64;
constexpr index_t KPerBlock = 128;
constexpr index_t CPerBlock = 8;
@@ -219,6 +219,36 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
constexpr index_t InBlockCopyDataPerRead = 4;
constexpr index_t WeiBlockCopyDataPerRead = 4;
constexpr index_t BlockSize = 128;
#elif 1
// 1x1, 14x14, Vega 20, hack CPerBlock = 1
constexpr index_t BPerBlock = 64;
constexpr index_t KPerBlock = 128;
constexpr index_t CPerBlock = 1;
constexpr index_t BPerThread = 8;
constexpr index_t KPerThread = 8;
constexpr index_t GemmMPerThreadSubC = 4;
constexpr index_t GemmNPerThreadSubC = 4;
constexpr index_t GemmMLevel0Cluster = 4;
constexpr index_t GemmNLevel0Cluster = 2;
constexpr index_t GemmMLevel1Cluster = 4;
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;
constexpr index_t WeiBlockCopyThreadPerDim0 = 4;
constexpr index_t WeiBlockCopyThreadPerDim1 = 16;
constexpr index_t InBlockCopyDataPerRead = 4;
constexpr index_t WeiBlockCopyDataPerRead = 4;
constexpr index_t BlockSize = 128;
#endif