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

refactor

Browse Source
This commit is contained in:
Chao Liu
2019-04-01 15:17:22 -05:00
parent d058d16407
commit e43d7bc63c
13 changed files with 873 additions and 928 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
driver/device_implicit_gemm_convolution_2_chwn_cyxk_khwn.hpp
View File

@@ -270,7 +270,7 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
for(index_t i = 0; i < nrepeat; ++i)
{
float time = launch_kernel(
constexpr auto gridwise_conv =
#if 1
gridwise_implicit_gemm_convolution_2_chwn_cyxk_khwn
#else
@@ -301,12 +301,14 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
WeiBlockCopyThreadPerDim0,
WeiBlockCopyThreadPerDim1,
InBlockCopyDataPerRead,
WeiBlockCopyDataPerRead>,
dim3(GridSize),
dim3(BlockSize),
static_cast<T*>(in_chwn_device_buf.GetDeviceBuffer()),
static_cast<T*>(wei_cyxk_device_buf.GetDeviceBuffer()),
static_cast<T*>(out_khwn_device_buf.GetDeviceBuffer()));
WeiBlockCopyDataPerRead>();
float time = launch_kernel(gridwise_conv.Run,
dim3(GridSize),
dim3(BlockSize),
static_cast<T*>(in_chwn_device_buf.GetDeviceBuffer()),
static_cast<T*>(wei_cyxk_device_buf.GetDeviceBuffer()),
static_cast<T*>(out_khwn_device_buf.GetDeviceBuffer()));
printf("Elapsed time : %f ms\n", time);
usleep(std::min(time * 1000, float(10000)));

2
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 0
#elif 1
// 1x1 filter, 14x14 image, C = 2048
constexpr index_t N = 128;
constexpr index_t C = 2048;