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
1762eb16
Commit
1762eb16
authored
Mar 02, 2021
by
Chao Liu
Browse files
fix bug
parent
ac87ed7c
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
7 additions
and
7 deletions
+7
-7
composable_kernel/include/utility/config.amd.hpp.in
composable_kernel/include/utility/config.amd.hpp.in
+3
-3
driver/include/device_dynamic_convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk.hpp
...convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk.hpp
+1
-1
driver/src/conv_driver.cpp
driver/src/conv_driver.cpp
+3
-3
No files found.
composable_kernel/include/utility/config.amd.hpp.in
View file @
1762eb16
...
@@ -7,17 +7,17 @@
...
@@ -7,17 +7,17 @@
#endif
#endif
#include "bfloat16_dev.hpp"
#include "bfloat16_dev.hpp"
#if
1
#if
0
#define CK_AMD_GPU_GFX906 1
#define CK_AMD_GPU_GFX906 1
#elif 0
#elif 0
#define CK_AMD_GPU_GFX908 1
#define CK_AMD_GPU_GFX908 1
#else
#else
#define CK_AMD_GPU_GFX
906
1
#define CK_AMD_GPU_GFX
1030
1
#endif
#endif
#if defined(CK_AMD_GPU_GFX906) || defined(CK_AMD_GPU_GFX908)
#if defined(CK_AMD_GPU_GFX906) || defined(CK_AMD_GPU_GFX908)
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x00020000
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x00020000
#elif defined(CK_AMD_GPU_GFX
_
1030)
#elif defined(CK_AMD_GPU_GFX1030)
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x31014000
#define CK_BUFFER_RESOURCE_3RD_DWORD 0x31014000
#endif
#endif
...
...
driver/include/device_dynamic_convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk.hpp
View file @
1762eb16
...
@@ -175,7 +175,7 @@ void device_dynamic_convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk(InDesc
...
@@ -175,7 +175,7 @@ void device_dynamic_convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk(InDesc
constexpr
index_t
GemmBBlockTransferDstScalarPerVector_GemmN
=
1
;
constexpr
index_t
GemmBBlockTransferDstScalarPerVector_GemmN
=
1
;
constexpr
index_t
GemmCThreadTransferDstScalarPerVector_GemmM1
=
2
;
constexpr
index_t
GemmCThreadTransferDstScalarPerVector_GemmM1
=
2
;
#elif
1
#elif
0
// cdata = 64, BlockSize = 64, 16x256x2
// cdata = 64, BlockSize = 64, 16x256x2
constexpr
index_t
BlockSize
=
64
;
constexpr
index_t
BlockSize
=
64
;
...
...
driver/src/conv_driver.cpp
View file @
1762eb16
...
@@ -77,7 +77,7 @@ int main(int argc, char* argv[])
...
@@ -77,7 +77,7 @@ int main(int argc, char* argv[])
using
LeftPads
=
Sequence
<
1
,
1
>
;
using
LeftPads
=
Sequence
<
1
,
1
>
;
using
RightPads
=
Sequence
<
1
,
1
>
;
using
RightPads
=
Sequence
<
1
,
1
>
;
#elif
0
#elif
1
constexpr
index_t
N
=
1
;
constexpr
index_t
N
=
1
;
constexpr
index_t
C
=
4
;
constexpr
index_t
C
=
4
;
constexpr
index_t
HI
=
1080
;
constexpr
index_t
HI
=
1080
;
...
@@ -701,7 +701,7 @@ int main(int argc, char* argv[])
...
@@ -701,7 +701,7 @@ int main(int argc, char* argv[])
LeftPads
{},
LeftPads
{},
RightPads
{},
RightPads
{},
nrepeat
);
nrepeat
);
#elif
0
#elif
1
device_convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk
(
in_nchw_desc
,
device_convolution_forward_implicit_gemm_v4r4_nhwc_kyxc_nhwk
(
in_nchw_desc
,
in_nchw
,
in_nchw
,
wei_kcyx_desc
,
wei_kcyx_desc
,
...
@@ -713,7 +713,7 @@ int main(int argc, char* argv[])
...
@@ -713,7 +713,7 @@ int main(int argc, char* argv[])
LeftPads
{},
LeftPads
{},
RightPads
{},
RightPads
{},
nrepeat
);
nrepeat
);
#elif
1
#elif
0
device_dynamic_convolution_forward_implicit_gemm_v4r4_nchw_kcyx_nkhw
(
in_nchw_desc
,
device_dynamic_convolution_forward_implicit_gemm_v4r4_nchw_kcyx_nkhw
(
in_nchw_desc
,
in_nchw
,
in_nchw
,
wei_kcyx_desc
,
wei_kcyx_desc
,
...
...
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