Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
TonyTonyFun
Paddle
提交
4478389c
P
Paddle
项目概览
TonyTonyFun
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
0
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
列表
看板
标记
里程碑
合并请求
0
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
0
Issue
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
未验证
提交
4478389c
编写于
8月 31, 2022
作者:
R
ronnywang
提交者:
GitHub
8月 31, 2022
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
[ROCM] fix bmm_kernel (#45530)
上级
4a25b60d
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
102 addition
and
0 deletion
+102
-0
paddle/phi/kernels/funcs/blas/blas_impl.hip.h
paddle/phi/kernels/funcs/blas/blas_impl.hip.h
+102
-0
未找到文件。
paddle/phi/kernels/funcs/blas/blas_impl.hip.h
浏览文件 @
4478389c
...
...
@@ -1128,6 +1128,108 @@ void Blas<phi::GPUContext>::BatchedGEMM(CBLAS_TRANSPOSE transA,
});
}
// note(wangran16): unknown bug. parameters dislocation when calling
// GEMM_STRIDED_BATCH<float> and GEMM_STRIDED_BATCH<double>
template
<
>
template
<
>
inline
void
Blas
<
phi
::
GPUContext
>::
BatchedGEMM
(
CBLAS_TRANSPOSE
transA
,
CBLAS_TRANSPOSE
transB
,
int
M
,
int
N
,
int
K
,
float
alpha
,
const
float
*
A
,
const
float
*
B
,
float
beta
,
float
*
C
,
int
batchCount
,
int64_t
strideA
,
int64_t
strideB
)
const
{
// Note that cublas follows fortran order, so the order is different from
// the cblas convention.
int
lda
=
(
transA
==
CblasNoTrans
)
?
K
:
M
;
int
ldb
=
(
transB
==
CblasNoTrans
)
?
N
:
K
;
int
ldc
=
N
;
rocblas_operation
cuTransA
=
(
transA
==
CblasNoTrans
)
?
rocblas_operation_none
:
rocblas_operation_transpose
;
rocblas_operation
cuTransB
=
(
transB
==
CblasNoTrans
)
?
rocblas_operation_none
:
rocblas_operation_transpose
;
const
int64_t
strideC
=
M
*
N
;
context_
.
CublasCall
([
&
](
rocblas_handle
handle
)
{
PADDLE_ENFORCE_GPU_SUCCESS
(
paddle
::
platform
::
dynload
::
rocblas_sgemm_strided_batched
(
handle
,
cuTransB
,
cuTransA
,
N
,
M
,
K
,
&
alpha
,
B
,
ldb
,
strideB
,
A
,
lda
,
strideA
,
&
beta
,
C
,
ldc
,
strideC
,
batchCount
));
});
}
template
<
>
template
<
>
inline
void
Blas
<
phi
::
GPUContext
>::
BatchedGEMM
(
CBLAS_TRANSPOSE
transA
,
CBLAS_TRANSPOSE
transB
,
int
M
,
int
N
,
int
K
,
double
alpha
,
const
double
*
A
,
const
double
*
B
,
double
beta
,
double
*
C
,
int
batchCount
,
int64_t
strideA
,
int64_t
strideB
)
const
{
// Note that cublas follows fortran order, so the order is different from
// the cblas convention.
int
lda
=
(
transA
==
CblasNoTrans
)
?
K
:
M
;
int
ldb
=
(
transB
==
CblasNoTrans
)
?
N
:
K
;
int
ldc
=
N
;
rocblas_operation
cuTransA
=
(
transA
==
CblasNoTrans
)
?
rocblas_operation_none
:
rocblas_operation_transpose
;
rocblas_operation
cuTransB
=
(
transB
==
CblasNoTrans
)
?
rocblas_operation_none
:
rocblas_operation_transpose
;
const
int64_t
strideC
=
M
*
N
;
context_
.
CublasCall
([
&
](
rocblas_handle
handle
)
{
PADDLE_ENFORCE_GPU_SUCCESS
(
paddle
::
platform
::
dynload
::
rocblas_dgemm_strided_batched
(
handle
,
cuTransB
,
cuTransA
,
N
,
M
,
K
,
&
alpha
,
B
,
ldb
,
strideB
,
A
,
lda
,
strideA
,
&
beta
,
C
,
ldc
,
strideC
,
batchCount
));
});
}
template
<
>
template
<
>
inline
void
Blas
<
phi
::
GPUContext
>::
BatchedGEMM
(
CBLAS_TRANSPOSE
transA
,
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录