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
2019-03-06 12:34:31 -06:00
parent 04c5527d07
commit 8edbc659b8
2 changed files with 11 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
driver/driver.hip.cpp
View File

@@ -577,10 +577,10 @@ int main(int argc, char* argv[])
ostream_ConstantTensorDescriptor(wei_kcsr_desc, std::cout << "wei_kcsr_desc: ");
ostream_ConstantTensorDescriptor(out_nkhw_desc, std::cout << "out_nkhw_desc: ");
Tensor<float> in_nchw(make_TensorDescriptor(in_nchw_desc));
Tensor<float> wei_kcsr(make_TensorDescriptor(wei_kcsr_desc));
Tensor<float> out_nkhw_host(make_TensorDescriptor(out_nkhw_desc));
Tensor<float> out_nkhw_device(make_TensorDescriptor(out_nkhw_desc));
Tensor<half> in_nchw(make_TensorDescriptor(in_nchw_desc));
Tensor<half> wei_kcsr(make_TensorDescriptor(wei_kcsr_desc));
Tensor<half> out_nkhw_host(make_TensorDescriptor(out_nkhw_desc));
Tensor<half> out_nkhw_device(make_TensorDescriptor(out_nkhw_desc));
std::size_t num_thread = std::thread::hardware_concurrency();
@@ -633,7 +633,7 @@ int main(int argc, char* argv[])
if(do_verification)
{
#if 1
#if 0
if(Y == 3 && X == 3)
{
host_winograd_3x3_convolution(in_nchw, wei_kcsr, out_nkhw_host, lower_pads, upper_pads);