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
fengzch-das
nunchaku
Commits
b1fec976
Commit
b1fec976
authored
Nov 10, 2024
by
sxtyzhangzk
Committed by
Zhekai Zhang
Nov 10, 2024
Browse files
[major] fix build on windows
parent
3969b561
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
30 additions
and
25 deletions
+30
-25
src/kernels/gemm_utils.cuh
src/kernels/gemm_utils.cuh
+1
-1
src/kernels/gemm_w4a4.cu
src/kernels/gemm_w4a4.cu
+29
-24
No files found.
src/kernels/gemm_utils.cuh
View file @
b1fec976
...
...
@@ -60,7 +60,7 @@ static void store(T *addr, T val) {
}
if
constexpr
(
sizeof
(
T
)
==
4
)
{
__stcg
(
reinterpret_cast
<
uint
*>
(
addr
),
*
reinterpret_cast
<
uint
*>
(
&
val
));
__stcg
(
reinterpret_cast
<
u
nsigned
int
*>
(
addr
),
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
val
));
return
;
}
if
constexpr
(
sizeof
(
T
)
==
8
)
{
...
...
src/kernels/gemm_w4a4.cu
View file @
b1fec976
...
...
@@ -7,7 +7,12 @@
#include "dispatch_utils.h"
#pragma nv_diag_suppress 177
#ifdef _MSC_VER
#define ALWAYSINLINE [[msvc::forceinline]]
#else
#define ALWAYSINLINE __attribute__((always_inline))
#endif
// #define ENABLE_NAN_CHECK 1
#if ENABLE_NAN_CHECK
...
...
@@ -196,12 +201,12 @@ public:
:
"=f"
(
psum
.
data
[
0
]),
"=f"
(
psum
.
data
[
1
]),
"=f"
(
psum
.
data
[
2
]),
"=f"
(
psum
.
data
[
3
])
:
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
b
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
b
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
b
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
b
.
data
[
1
])),
// "r"(0), "r"(0), "r"(0), "r"(0)
"f"
(
psum
.
data
[
0
]),
"f"
(
psum
.
data
[
1
]),
"f"
(
psum
.
data
[
2
]),
"f"
(
psum
.
data
[
3
])
...
...
@@ -215,12 +220,12 @@ public:
:
"=f"
(
psum
.
data
[
4
]),
"=f"
(
psum
.
data
[
5
]),
"=f"
(
psum
.
data
[
6
]),
"=f"
(
psum
.
data
[
7
])
:
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
b
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
b
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
b
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
b
.
data
[
3
])),
// "r"(0), "r"(0), "r"(0), "r"(0)
"f"
(
psum
.
data
[
4
]),
"f"
(
psum
.
data
[
5
]),
"f"
(
psum
.
data
[
6
]),
"f"
(
psum
.
data
[
7
])
);
...
...
@@ -236,12 +241,12 @@ public:
:
"=f"
(
psum
.
data
[
0
]),
"=f"
(
psum
.
data
[
1
]),
"=f"
(
psum
.
data
[
2
]),
"=f"
(
psum
.
data
[
3
])
:
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
b
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
b
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
b
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
b
.
data
[
1
])),
// "r"(0), "r"(0), "r"(0), "r"(0)
"f"
(
psum
.
data
[
0
]),
"f"
(
psum
.
data
[
1
]),
"f"
(
psum
.
data
[
2
]),
"f"
(
psum
.
data
[
3
])
...
...
@@ -255,12 +260,12 @@ public:
:
"=f"
(
psum
.
data
[
4
]),
"=f"
(
psum
.
data
[
5
]),
"=f"
(
psum
.
data
[
6
]),
"=f"
(
psum
.
data
[
7
])
:
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
a
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
b
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
uint
*>
(
&
b
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
0
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
1
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
a
.
data
[
3
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
b
.
data
[
2
])),
"r"
(
*
reinterpret_cast
<
u
nsigned
int
*>
(
&
b
.
data
[
3
])),
// "r"(0), "r"(0), "r"(0), "r"(0)
"f"
(
psum
.
data
[
4
]),
"f"
(
psum
.
data
[
5
]),
"f"
(
psum
.
data
[
6
]),
"f"
(
psum
.
data
[
7
])
);
...
...
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