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

enabled ds_read_b128 and ds_write_b128 on hip c++

Browse Source
This commit is contained in:
Chao Liu
2019-04-09 19:05:44 -05:00
parent 471830a052
commit e624df922d
5 changed files with 48 additions and 85 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
driver/device_implicit_gemm_convolution_2_chwn_cyxk_khwn.hpp
View File

@@ -189,7 +189,7 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
constexpr index_t WeiBlockCopyDataPerRead = 4;
constexpr index_t BlockSize = 256;
#elif 1
#elif 0
// 1x1, 14x14, Pascal, enable lds_double_buffer, disable register double buffer
constexpr index_t BPerBlock = 64;
constexpr index_t KPerBlock = 128;
@@ -219,7 +219,7 @@ void device_implicit_gemm_convolution_2_chwn_cyxk_khwn(InDesc,
constexpr index_t OutThreadCopyDataPerWrite = 4;
constexpr index_t BlockSize = 128;
#elif 0
#elif 1
// 1x1, 14x14, Vega 20, enable lds_double_buffer, disable register_double_buffer
constexpr index_t BPerBlock = 128;
constexpr index_t KPerBlock = 128;

8
driver/driver.hip.cpp
View File

@@ -409,7 +409,7 @@ int main(int argc, char* argv[])
constexpr index_t HPad = 0;
constexpr index_t WPad = 0;
#elif 1
#elif 0
// 3x3, 34x34
constexpr index_t N = 64;
constexpr index_t C = 256;
@@ -583,7 +583,7 @@ 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;
@@ -667,9 +667,9 @@ int main(int argc, char* argv[])
device_direct_convolution_2_nchw_kcyx_nkhw
#elif 0
device_direct_convolution_2_vectorized_nchw_kcyx_nkhw
#elif 1
device_implicit_gemm_convolution_1_chwn_cyxk_khwn
#elif 0
device_implicit_gemm_convolution_1_chwn_cyxk_khwn
#elif 1
device_implicit_gemm_convolution_2_chwn_cyxk_khwn
#endif
(in_nchw_desc, in_nchw, wei_kcyx_desc, wei_kcyx, out_nkhw_desc, out_nkhw_device, nrepeat);