ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 09:16:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

removing dependency on old tensor descriptor

Browse Source
This commit is contained in:
Chao Liu
2019-09-26 11:49:05 -05:00
parent 0f52c4c0e4
commit 51a9fa1dbd
10 changed files with 162 additions and 151 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
driver/src/driver.cpp
View File

@@ -14,7 +14,7 @@
//#include "device_convolution_implicit_gemm_v1_nchw_cyxk_nkhw.hpp"
//#include "device_convolution_implicit_gemm_v2_chwn_cyxk_khwn.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_v4r1_nchw_kcyx_nkhw_padded.hpp"
//#include "device_convolution_implicit_gemm_v4r2_nchw_kcyx_nkhw.hpp"
//#include "device_convolution_implicit_gemm_v4r3_nchw_kcyx_nkhw.hpp"
@@ -103,7 +103,7 @@ int main(int argc, char* argv[])
using LeftPads = Sequence<0, 0>;
using RightPads = Sequence<0, 0>;
#elif 0
#elif 1
// 1x1 filter, 8x8 image
// cudnn@V100 68%, ck@V100 72%, ck@P100 52%, ck@VII 42%
constexpr index_t N = 64;
@@ -366,6 +366,10 @@ int main(int argc, char* argv[])
ostream_ConstantTensorDescriptor(in_nchw_desc, std::cout << "in_nchw_desc: ");
ostream_ConstantTensorDescriptor(wei_kcyx_desc, std::cout << "wei_kcyx_desc: ");
ostream_ConstantTensorDescriptor(out_nkhw_desc, std::cout << "out_nkhw_desc: ");
print_sequence("LeftPads", LeftPads{});
print_sequence("RightPads", RightPads{});
print_sequence("ConvStrides", ConvStrides{});
print_sequence("ConvDilations", ConvDilations{});
using in_data_t = float;
using out_data_t = float;
@@ -444,7 +448,7 @@ int main(int argc, char* argv[])
ConvStrides{},
ConvDilations{},
nrepeat);
#elif 1
#elif 0
device_convolution_implicit_gemm_v4r1_nchw_kcyx_nkhw_padded(in_nchw_desc,
in_nchw,
wei_kcyx_desc,
@@ -486,7 +490,7 @@ int main(int argc, char* argv[])
ConvStrides{},
ConvDilations{},
nrepeat);
#elif 0
#elif 1
device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw_padded(in_nchw_desc,
in_nchw,
wei_kcyx_desc,