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
d5c5d2a3
Commit
d5c5d2a3
authored
Jan 21, 2025
by
Jakub Piasecki
Browse files
add reviewers comments
parent
c2945b96
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
65 additions
and
305 deletions
+65
-305
example/ck_tile/03_gemm/instances/gemm_api.cpp
example/ck_tile/03_gemm/instances/gemm_api.cpp
+64
-304
example/ck_tile/03_gemm/run_gemm_example.inc
example/ck_tile/03_gemm/run_gemm_example.inc
+1
-1
No files found.
example/ck_tile/03_gemm/instances/gemm_api.cpp
View file @
d5c5d2a3
...
...
@@ -18,188 +18,68 @@ float gemm(const gemm_traits& t, const ck_tile::GemmHostArgs& a, const ck_tile::
{
if
(
a
.
M
>
512
)
{
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Row
,
Row
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Row
,
Row
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
else
{
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Row
,
Row
,
Row
,
128
,
32
,
64
,
4
,
1
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Row
,
Row
,
Row
,
128
,
32
,
64
,
4
,
1
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
}
else
if
(
t
.
is_a_rowmajor
&&
!
t
.
is_b_rowmajor
&&
t
.
is_c_rowmajor
)
{
if
(
a
.
M
>
512
)
{
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Row
,
Col
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Row
,
Col
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
else
{
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Row
,
Col
,
Row
,
128
,
32
,
64
,
4
,
1
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Row
,
Col
,
Row
,
128
,
32
,
64
,
4
,
1
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
}
else
if
(
!
t
.
is_a_rowmajor
&&
t
.
is_b_rowmajor
&&
t
.
is_c_rowmajor
)
{
if
(
a
.
M
>
512
)
{
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Col
,
Row
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Col
,
Row
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
else
{
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Col
,
Row
,
Row
,
128
,
128
,
32
,
2
,
2
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Col
,
Row
,
Row
,
128
,
128
,
32
,
2
,
2
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
}
else
if
(
!
t
.
is_a_rowmajor
&&
!
t
.
is_b_rowmajor
&&
t
.
is_c_rowmajor
)
{
if
(
a
.
M
>
512
)
{
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Col
,
Col
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Col
,
Col
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
else
{
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Col
,
Col
,
Row
,
128
,
128
,
32
,
2
,
2
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
FP16
,
FP16
,
FP32
,
FP16
,
Col
,
Col
,
Row
,
128
,
128
,
32
,
2
,
2
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
}
else
...
...
@@ -213,188 +93,68 @@ float gemm(const gemm_traits& t, const ck_tile::GemmHostArgs& a, const ck_tile::
{
if
(
a
.
M
>
512
)
{
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Row
,
Row
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Row
,
Row
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
else
{
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Row
,
Row
,
Row
,
128
,
32
,
64
,
4
,
1
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Row
,
Row
,
Row
,
128
,
32
,
64
,
4
,
1
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
}
else
if
(
t
.
is_a_rowmajor
&&
!
t
.
is_b_rowmajor
&&
t
.
is_c_rowmajor
)
{
if
(
a
.
M
>
512
)
{
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Row
,
Col
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Row
,
Col
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
else
{
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Row
,
Col
,
Row
,
128
,
32
,
64
,
4
,
1
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Row
,
Col
,
Row
,
128
,
32
,
64
,
4
,
1
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
}
else
if
(
!
t
.
is_a_rowmajor
&&
t
.
is_b_rowmajor
&&
t
.
is_c_rowmajor
)
{
if
(
a
.
M
>
512
)
{
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Col
,
Row
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Col
,
Row
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
else
{
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Col
,
Row
,
Row
,
128
,
128
,
32
,
2
,
2
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Col
,
Row
,
Row
,
128
,
128
,
32
,
2
,
2
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
}
else
if
(
!
t
.
is_a_rowmajor
&&
!
t
.
is_b_rowmajor
&&
t
.
is_c_rowmajor
)
{
if
(
a
.
M
>
512
)
{
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Col
,
Col
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Col
,
Col
,
Row
,
256
,
256
,
32
,
2
,
2
,
1
,
32
,
32
,
16
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
else
{
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Col
,
Col
,
Row
,
128
,
128
,
32
,
2
,
2
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format off
// ADataType, BDataType, AccDataType, CDataType, ALayout, BLayout, CLayout, M_Tile, N_Tile, K_Tile, M_Warp, N_Warp, K_Warp, M_Warp_Tile, N_Warp_Tile, K_Warp_Tile, PadM, PadN, PadK
return
gemm_
<
gemm_traits_
<
BF16
,
BF16
,
FP32
,
BF16
,
Col
,
Col
,
Row
,
128
,
128
,
32
,
2
,
2
,
1
,
32
,
32
,
8
,
false
,
false
,
false
>>
(
a
,
s
);
// clang-format on
}
}
else
...
...
example/ck_tile/03_gemm/run_gemm_example.inc
View file @
d5c5d2a3
// SPDX-License-Identifier: MIT
// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2024
-2025
, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
auto
create_args
(
int
argc
,
char
*
argv
[])
...
...
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