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_ROCM
Commits
5db68230
Commit
5db68230
authored
Feb 28, 2024
by
Jing Zhang
Browse files
Merge branch 'navi4x_wmma' of github.com:ROCm/composable_kernel-internal into navi4x_wmma
parents
14b422d7
2052dfc9
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
4 additions
and
4 deletions
+4
-4
include/ck/tensor_operation/gpu/block/blockwise_gemm_wmma.hpp
...ude/ck/tensor_operation/gpu/block/blockwise_gemm_wmma.hpp
+2
-2
include/ck/tensor_operation/gpu/grid/gridwise_gemm_wmma.hpp
include/ck/tensor_operation/gpu/grid/gridwise_gemm_wmma.hpp
+1
-1
script/cmake-ck-dev.sh
script/cmake-ck-dev.sh
+1
-1
No files found.
include/ck/tensor_operation/gpu/block/blockwise_gemm_wmma.hpp
View file @
5db68230
...
@@ -416,7 +416,7 @@ struct BlockwiseGemmWMMA
...
@@ -416,7 +416,7 @@ struct BlockwiseGemmWMMA
static
constexpr
auto
a_thread_desc_
=
make_naive_tensor_descriptor
(
static
constexpr
auto
a_thread_desc_
=
make_naive_tensor_descriptor
(
make_tuple
(
Number
<
KPack
/
A_K1
/
A_KRow
>
{},
Number
<
MRepeat
>
{},
I1
,
I1
,
I1
,
Number
<
A_K1
>
{}),
make_tuple
(
Number
<
KPack
/
A_K1
/
A_KRow
>
{},
Number
<
MRepeat
>
{},
I1
,
I1
,
I1
,
Number
<
A_K1
>
{}),
make_tuple
(
Number
<
A_K1
>
{},
make_tuple
(
Number
<
A_K1
>
{},
Number
<
KPack
/
A_K
Row
>
{},
Number
<
A_KRow
*
A_K
1
>
{},
Number
<
A_K1
>
{},
Number
<
A_K1
>
{},
Number
<
A_K1
>
{},
Number
<
A_K1
>
{},
Number
<
A_K1
>
{},
Number
<
A_K1
>
{},
...
@@ -425,7 +425,7 @@ struct BlockwiseGemmWMMA
...
@@ -425,7 +425,7 @@ struct BlockwiseGemmWMMA
static
constexpr
auto
b_thread_desc_
=
make_naive_tensor_descriptor
(
static
constexpr
auto
b_thread_desc_
=
make_naive_tensor_descriptor
(
make_tuple
(
Number
<
KPack
/
B_K1
/
B_KRow
>
{},
Number
<
NRepeat
>
{},
I1
,
I1
,
I1
,
Number
<
B_K1
>
{}),
make_tuple
(
Number
<
KPack
/
B_K1
/
B_KRow
>
{},
Number
<
NRepeat
>
{},
I1
,
I1
,
I1
,
Number
<
B_K1
>
{}),
make_tuple
(
Number
<
B_K1
>
{},
make_tuple
(
Number
<
B_K1
>
{},
Number
<
KPack
/
B_K
Row
>
{},
Number
<
B_KRow
*
B_K
1
>
{},
Number
<
B_K1
>
{},
Number
<
B_K1
>
{},
Number
<
B_K1
>
{},
Number
<
B_K1
>
{},
Number
<
B_K1
>
{},
Number
<
B_K1
>
{},
...
...
include/ck/tensor_operation/gpu/grid/gridwise_gemm_wmma.hpp
View file @
5db68230
...
@@ -135,7 +135,7 @@ struct GridwiseGemm_Wmma
...
@@ -135,7 +135,7 @@ struct GridwiseGemm_Wmma
static
constexpr
auto
MWaves
=
MPerBlock
/
(
MRepeat
*
MPerWmma
);
static
constexpr
auto
MWaves
=
MPerBlock
/
(
MRepeat
*
MPerWmma
);
static
constexpr
auto
NWaves
=
NPerBlock
/
(
NRepeat
*
NPerWmma
);
static
constexpr
auto
NWaves
=
NPerBlock
/
(
NRepeat
*
NPerWmma
);
static
constexpr
auto
WmmaK
=
K1
==
16
?
32
:
16
;
static
constexpr
auto
WmmaK
=
(
K1
==
16
)
?
32
:
16
;
using
ThisThreadBlock
=
ThisThreadBlock
<
BlockSize
>
;
using
ThisThreadBlock
=
ThisThreadBlock
<
BlockSize
>
;
...
...
script/cmake-ck-dev.sh
View file @
5db68230
...
@@ -11,7 +11,7 @@ cmake
...
@@ -11,7 +11,7 @@ cmake
-D
CMAKE_CXX_FLAGS
=
"-std=c++17 -O3 -ftemplate-backtrace-limit=0 -fPIE -Wno-gnu-line-marker"
\
-D
CMAKE_CXX_FLAGS
=
"-std=c++17 -O3 -ftemplate-backtrace-limit=0 -fPIE -Wno-gnu-line-marker"
\
-D
CMAKE_BUILD_TYPE
=
Release
\
-D
CMAKE_BUILD_TYPE
=
Release
\
-D
BUILD_DEV
=
ON
\
-D
BUILD_DEV
=
ON
\
-D
GPU_TARGETS
=
"gfx
908;gfx90a;gfx94
0"
\
-D
GPU_TARGETS
=
"gfx
120
0"
\
-D
CMAKE_VERBOSE_MAKEFILE:BOOL
=
ON
\
-D
CMAKE_VERBOSE_MAKEFILE:BOOL
=
ON
\
-D
USE_BITINT_EXTENSION_INT4
=
OFF
\
-D
USE_BITINT_EXTENSION_INT4
=
OFF
\
${
MY_PROJECT_SOURCE
}
${
MY_PROJECT_SOURCE
}
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