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
256c4947
Commit
256c4947
authored
May 10, 2024
by
illsilin
Browse files
fix clang format and test/wmma_op
parent
7f015d49
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
18 additions
and
2 deletions
+18
-2
include/ck_tile/core/config.hpp
include/ck_tile/core/config.hpp
+2
-2
test/wmma_op/wmma_op_util.hpp
test/wmma_op/wmma_op_util.hpp
+16
-0
No files found.
include/ck_tile/core/config.hpp
View file @
256c4947
...
@@ -143,8 +143,8 @@
...
@@ -143,8 +143,8 @@
#define CK_TILE_BUFFER_RESOURCE_3RD_DWORD 0x00020000
#define CK_TILE_BUFFER_RESOURCE_3RD_DWORD 0x00020000
#elif defined(__gfx1030__) // for GPU code
#elif defined(__gfx1030__) // for GPU code
#define CK_TILE_BUFFER_RESOURCE_3RD_DWORD 0x31014000
#define CK_TILE_BUFFER_RESOURCE_3RD_DWORD 0x31014000
#elif defined(__gfx1100__) || defined(__gfx1101__) || defined(__gfx1102__) ||
defined(__gfx1103__) ||
\
#elif defined(__gfx1100__) || defined(__gfx1101__) || defined(__gfx1102__) || \
defined(__gfx1200__) || defined(__gfx1201__) // for GPU code
defined(__gfx1103__) ||
defined(__gfx1200__) || defined(__gfx1201__) // for GPU code
#define CK_TILE_BUFFER_RESOURCE_3RD_DWORD 0x31004000
#define CK_TILE_BUFFER_RESOURCE_3RD_DWORD 0x31004000
#endif
#endif
...
...
test/wmma_op/wmma_op_util.hpp
View file @
256c4947
...
@@ -140,10 +140,18 @@ __global__ void matmul(const src_t* a, const src_t* b, dst_t* c)
...
@@ -140,10 +140,18 @@ __global__ void matmul(const src_t* a, const src_t* b, dst_t* c)
p_shared
[
8
*
16
*
lane_hi
+
8
*
lane_lo
+
ele
+
16
*
16
]
=
b_temp
[
ele
];
p_shared
[
8
*
16
*
lane_hi
+
8
*
lane_lo
+
ele
+
16
*
16
]
=
b_temp
[
ele
];
}
}
#ifdef __gfx12__
asm
volatile
(
"\
s_wait_dscnt 0x0
\n
\
s_barrier_signal -1
\n
\
s_barrier_wait -1 \
"
::
);
#else
asm
volatile
(
"\
asm
volatile
(
"\
s_waitcnt lgkmcnt(0)
\n
\
s_waitcnt lgkmcnt(0)
\n
\
s_barrier \
s_barrier \
"
::
);
"
::
);
#endif
for
(
int
ele
=
0
;
ele
<
16
;
++
ele
)
for
(
int
ele
=
0
;
ele
<
16
;
++
ele
)
{
{
...
@@ -155,10 +163,18 @@ __global__ void matmul(const src_t* a, const src_t* b, dst_t* c)
...
@@ -155,10 +163,18 @@ __global__ void matmul(const src_t* a, const src_t* b, dst_t* c)
a_frag
[
ele
]
=
p_shared
[(
ele
/
8
)
*
16
*
8
+
8
*
lane
+
ele
%
8
];
a_frag
[
ele
]
=
p_shared
[(
ele
/
8
)
*
16
*
8
+
8
*
lane
+
ele
%
8
];
}
}
#ifdef __gfx12__
asm
volatile
(
"\
s_wait_dscnt 0x0
\n
\
s_barrier_signal -1
\n
\
s_barrier_wait -1 \
"
::
);
#else
asm
volatile
(
"\
asm
volatile
(
"\
s_waitcnt lgkmcnt(0)
\n
\
s_waitcnt lgkmcnt(0)
\n
\
s_barrier \
s_barrier \
"
::
);
"
::
);
#endif
// sync threads, similar to mma_sync
// sync threads, similar to mma_sync
// __syncthreads();
// __syncthreads();
...
...
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