Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
gaoqiong
composable_kernel
Commits
cfff66cd
Commit
cfff66cd
authored
Nov 30, 2019
by
Chao Liu
Browse files
refactor
parent
ecd3240b
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
33 additions
and
34 deletions
+33
-34
driver/include/device_convolution_backward_data_implicit_gemm_v1r1_nchw_kcyx_nkhw.hpp
...ution_backward_data_implicit_gemm_v1r1_nchw_kcyx_nkhw.hpp
+32
-33
driver/src/conv_bwd_data_driver.cpp
driver/src/conv_bwd_data_driver.cpp
+1
-1
No files found.
driver/include/device_convolution_backward_data_implicit_gemm_v1r1_nchw_kcyx_nkhw.hpp
View file @
cfff66cd
...
...
@@ -3,7 +3,7 @@
#include "device.hpp"
#include "tensor.hpp"
#include "gridwise_operation_wrapper.hpp"
#include "gridwise_convolution_backward_data_implicit_gemm_v1r1_nchw_kcyx_nkhw
_lds_double_buffer
.hpp"
#include "gridwise_convolution_backward_data_implicit_gemm_v1r1_nchw_kcyx_nkhw.hpp"
template
<
typename
T
,
typename
InDesc
,
...
...
@@ -84,38 +84,37 @@ void device_convolution_backward_data_implicit_gemm_v1r1_nchw_kcyx_nkhw(InDesc i
printf
(
"%s: BlockSize %u, GridSize %u
\n
"
,
__func__
,
BlockSize
,
GridSize
);
constexpr
auto
gridwise_conv
=
GridwiseConvolutionBackwardDataImplicitGemm_v1r1_nchw_kcyx_nkhw_lds_double_buffer
<
GridSize
,
BlockSize
,
T
,
T
,
decltype
(
in_nchw_desc
),
decltype
(
wei_kcyx_desc
),
decltype
(
out_nkhw_desc
),
ConvStrides
,
ConvDilations
,
LeftPads
,
RightPads
,
BPerBlock
,
EPerBlock
,
KPerBlock
,
GemmMPerThreadSubC
,
GemmNPerThreadSubC
,
GemmMLevel0Cluster
,
GemmNLevel0Cluster
,
GemmMLevel1Cluster
,
GemmNLevel1Cluster
,
GemmKPerThreadLoop
,
GemmDataPerReadA
,
GemmDataPerReadB
,
OutBlockCopySubLengths_K_B
,
OutBlockCopyClusterLengths_K_B
,
OutBlockCopyDataPerAccess_B
,
WeiBlockCopySubLengths_K_E
,
WeiBlockCopyClusterLengths_K_E
,
WeiBlockCopyDataPerAccess_E
,
InThreadCopyDataPerAccess_B
>
{};
constexpr
auto
gridwise_conv
=
GridwiseConvolutionBackwardDataImplicitGemm_v1r1_nchw_kcyx_nkhw
<
GridSize
,
BlockSize
,
T
,
T
,
decltype
(
in_nchw_desc
),
decltype
(
wei_kcyx_desc
),
decltype
(
out_nkhw_desc
),
ConvStrides
,
ConvDilations
,
LeftPads
,
RightPads
,
BPerBlock
,
EPerBlock
,
KPerBlock
,
GemmMPerThreadSubC
,
GemmNPerThreadSubC
,
GemmMLevel0Cluster
,
GemmNLevel0Cluster
,
GemmMLevel1Cluster
,
GemmNLevel1Cluster
,
GemmKPerThreadLoop
,
GemmDataPerReadA
,
GemmDataPerReadB
,
OutBlockCopySubLengths_K_B
,
OutBlockCopyClusterLengths_K_B
,
OutBlockCopyDataPerAccess_B
,
WeiBlockCopySubLengths_K_E
,
WeiBlockCopyClusterLengths_K_E
,
WeiBlockCopyDataPerAccess_E
,
InThreadCopyDataPerAccess_B
>
{};
for
(
index_t
i
=
0
;
i
<
nrepeat
;
++
i
)
{
...
...
driver/src/conv_bwd_data_driver.cpp
View file @
cfff66cd
...
...
@@ -345,7 +345,7 @@ int main(int argc, char* argv[])
#endif
}
#if
0
#if
1
device_convolution_backward_data_implicit_gemm_v1r1_nchw_kcyx_nkhw
#else
device_convolution_backward_data_implicit_gemm_v1r2_nchw_kcyx_nkhw
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment