ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 01:10:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

added implicit gemm v4r4 and double buffer

Browse Source
This commit is contained in:
Chao Liu
2019-08-03 00:19:19 -05:00
parent c01af89928
commit c2d246696f
3 changed files with 420 additions and 188 deletions
Download Patch File Download Diff File Expand all files Collapse all files

73
driver/include/device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw.hpp
View File

@@ -4,7 +4,7 @@
#include "tensor.hpp"
#include "gridwise_convolution_kernel_wrapper.hpp"
#include "gridwise_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw.hpp"
//#include "gridwise_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw_lds_double_buffer.hpp"
#include "gridwise_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw_lds_double_buffer.hpp"
using namespace ck;
@@ -132,39 +132,44 @@ void device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw(InDesc,
printf("%s: BlockSize %u, GridSize %u \n", __func__, BlockSize, GridSize);
constexpr auto gridwise_conv =
GridwiseConvolutionImplicitGemm_v4r4_nchw_kcyx_nkhw<GridSize,
BlockSize,
T,
decltype(in_nchw_desc),
decltype(wei_kcyx_desc),
decltype(out_nkhw_desc),
ConvStrides,
ConvDilations,
BPerBlock,
KPerBlock,
EPerBlock,
GemmMPerThreadSubC,
GemmNPerThreadSubC,
GemmMLevel0Cluster,
GemmNLevel0Cluster,
GemmMLevel1Cluster,
GemmNLevel1Cluster,
GemmKPerThreadLoop,
GemmDataPerReadA,
GemmDataPerReadB,
InBlockCopySubLengths_E_B,
InBlockCopyClusterLengths_E_B,
InBlockCopyThreadClusterArrangeOrder,
InBlockCopySrcAccessOrder,
InBlockCopyDstAccessOrder,
InBlockCopyDataPerAccess_B,
WeiBlockCopySubLengths_E_K,
WeiBlockCopyClusterLengths_E_K,
WeiBlockCopyThreadClusterArrangeOrder,
WeiBlockCopySrcAccessOrder,
WeiBlockCopyDstAccessOrder,
WeiBlockCopySrcDataPerRead_E,
WeiBlockCopyDstDataPerWrite_K>{};
#if 0
GridwiseConvolutionImplicitGemm_v4r4_nchw_kcyx_nkhw
#else
GridwiseConvolutionImplicitGemm_v4r4_nchw_kcyx_nkhw_lds_double_buffer
#endif
<GridSize,
BlockSize,
T,
decltype(in_nchw_desc),
decltype(wei_kcyx_desc),
decltype(out_nkhw_desc),
ConvStrides,
ConvDilations,
BPerBlock,
KPerBlock,
EPerBlock,
GemmMPerThreadSubC,
GemmNPerThreadSubC,
GemmMLevel0Cluster,
GemmNLevel0Cluster,
GemmMLevel1Cluster,
GemmNLevel1Cluster,
GemmKPerThreadLoop,
GemmDataPerReadA,
GemmDataPerReadB,
InBlockCopySubLengths_E_B,
InBlockCopyClusterLengths_E_B,
InBlockCopyThreadClusterArrangeOrder,
InBlockCopySrcAccessOrder,
InBlockCopyDstAccessOrder,
InBlockCopyDataPerAccess_B,
WeiBlockCopySubLengths_E_K,
WeiBlockCopyClusterLengths_E_K,
WeiBlockCopyThreadClusterArrangeOrder,
WeiBlockCopySrcAccessOrder,
WeiBlockCopyDstAccessOrder,
WeiBlockCopySrcDataPerRead_E,
WeiBlockCopyDstDataPerWrite_K>{};
for(index_t i = 0; i < nrepeat; ++i)
{