ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 17:26:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

tidy up

Browse Source
This commit is contained in:
Chao Liu
2019-04-08 10:48:29 -05:00
parent c9fa46af0b
commit 268d1c717c
14 changed files with 120 additions and 102 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
src/include/blockwise_direct_convolution.hip.hpp
View File

@@ -93,10 +93,11 @@ __device__ void blockwise_direct_convolution(InBlockDesc,
Float p_out_thread[out_thread_desc.GetElementSpace()];
threadwise_4d_tensor_copy(out_block_desc,
p_out_block + out_block_desc.Get1dIndex(n_thread_data_begin,
k_thread_data_begin,
ho_thread_data_begin,
wo_thread_data_begin),
p_out_block +
out_block_desc.Get1dIndex(n_thread_data_begin,
k_thread_data_begin,
ho_thread_data_begin,
wo_thread_data_begin),
out_thread_desc,
p_out_thread,
out_thread_desc.GetLengths());
@@ -107,10 +108,11 @@ __device__ void blockwise_direct_convolution(InBlockDesc,
// threadwise convolution
threadwise_direct_convolution_2(
in_thread_block_desc,
p_in_block + in_block_desc.Get1dIndex(n_thread_data_begin,
c_thread_data_begin,
hi_thread_data_begin,
wi_thread_data_begin),
p_in_block +
in_block_desc.Get1dIndex(n_thread_data_begin,
c_thread_data_begin,
hi_thread_data_begin,
wi_thread_data_begin),
wei_thread_block_desc,
p_wei_block +
wei_block_desc.Get1dIndex(k_thread_data_begin, c_thread_data_begin, 0, 0),
@@ -122,10 +124,11 @@ __device__ void blockwise_direct_convolution(InBlockDesc,
threadwise_4d_tensor_copy(out_thread_desc,
p_out_thread,
out_block_desc,
p_out_block + out_block_desc.Get1dIndex(n_thread_data_begin,
k_thread_data_begin,
ho_thread_data_begin,
wo_thread_data_begin),
p_out_block +
out_block_desc.Get1dIndex(n_thread_data_begin,
k_thread_data_begin,
ho_thread_data_begin,
wo_thread_data_begin),
out_thread_desc.GetLengths());
}
}