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
兽拳
Paddle
提交
d9b202e7
P
Paddle
项目概览
兽拳
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
1
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1
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
1
Issue
1
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 >>
提交
d9b202e7
编写于
10月 15, 2018
作者:
M
minqiyang
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Move tensor copy src_ptr and dst_ptr check to TensorCopy function
test=develop
上级
f4084882
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
43 addition
and
45 deletion
+43
-45
paddle/fluid/framework/tensor_util.cc
paddle/fluid/framework/tensor_util.cc
+11
-0
paddle/fluid/operators/reshape_op.cc
paddle/fluid/operators/reshape_op.cc
+32
-45
未找到文件。
paddle/fluid/framework/tensor_util.cc
浏览文件 @
d9b202e7
...
...
@@ -114,6 +114,11 @@ void TensorCopySync(const Tensor& src, const platform::Place& dst_place,
auto
dst_ptr
=
dst
->
mutable_data
(
dst_place
,
src
.
type
());
auto
size
=
src
.
numel
()
*
SizeOfType
(
src
.
type
());
if
(
platform
::
is_cpu_place
(
src_place
)
&&
platform
::
is_cpu_place
(
dst_place
))
{
if
(
src_ptr
==
dst_ptr
)
{
VLOG
(
3
)
<<
"Skip copy the same data from "
<<
src
.
place
()
<<
" to "
<<
dst_place
;
return
;
}
memory
::
Copy
(
boost
::
get
<
platform
::
CPUPlace
>
(
dst_place
),
dst_ptr
,
boost
::
get
<
platform
::
CPUPlace
>
(
src_place
),
src_ptr
,
size
);
}
...
...
@@ -132,6 +137,12 @@ void TensorCopySync(const Tensor& src, const platform::Place& dst_place,
platform
::
is_gpu_place
(
dst_place
))
{
auto
src_gpu_place
=
boost
::
get
<
platform
::
CUDAPlace
>
(
src_place
);
auto
dst_gpu_place
=
boost
::
get
<
platform
::
CUDAPlace
>
(
dst_place
);
if
(
src_ptr
==
dst_ptr
&&
src_gpu_place
.
GetDeviceId
()
==
dst_gpu_place
.
GetDeviceId
())
{
VLOG
(
3
)
<<
"Skip copy the same data from "
<<
src
.
place
()
<<
" to "
<<
dst_place
;
return
;
}
memory
::
Copy
(
dst_gpu_place
,
dst_ptr
,
src_gpu_place
,
src_ptr
,
size
,
nullptr
);
}
#endif
...
...
paddle/fluid/operators/reshape_op.cc
浏览文件 @
d9b202e7
...
...
@@ -195,7 +195,6 @@ class ReshapeGradOp : public framework::OperatorWithKernel {
}
};
template
<
typename
T
>
class
ReshapeKernel
{
public:
void
operator
()(
const
framework
::
ExecutionContext
&
ctx
)
const
{
...
...
@@ -228,15 +227,12 @@ class ReshapeKernel {
"sequence_reshape op."
);
}
if
(
in
->
data
<
T
>
()
!=
reinterpret_cast
<
T
*>
(
out
->
mutable_data
(
ctx
.
GetPlace
(),
in
->
type
())))
{
framework
::
TensorCopySync
(
*
in
,
ctx
.
GetPlace
(),
out
);
}
out
->
mutable_data
(
ctx
.
GetPlace
(),
in
->
type
());
framework
::
TensorCopySync
(
*
in
,
ctx
.
GetPlace
(),
out
);
out
->
Resize
(
out_dims
);
}
};
template
<
typename
T
>
class
ReshapeGradKernel
{
public:
void
operator
()(
const
framework
::
ExecutionContext
&
ctx
)
const
{
...
...
@@ -244,9 +240,8 @@ class ReshapeGradKernel {
auto
*
d_x
=
ctx
.
Output
<
framework
::
Tensor
>
(
framework
::
GradVarName
(
"X"
));
auto
in_dims
=
d_x
->
dims
();
if
(
d_out
->
data
<
T
>
()
!=
d_x
->
mutable_data
(
ctx
.
GetPlace
(),
d_out
->
type
()))
{
framework
::
TensorCopySync
(
*
d_out
,
ctx
.
GetPlace
(),
d_x
);
}
d_x
->
mutable_data
(
ctx
.
GetPlace
(),
d_out
->
type
());
framework
::
TensorCopySync
(
*
d_out
,
ctx
.
GetPlace
(),
d_x
);
d_x
->
Resize
(
in_dims
);
}
};
...
...
@@ -341,46 +336,38 @@ namespace ops = paddle::operators;
REGISTER_OPERATOR
(
reshape
,
ops
::
ReshapeOp
,
ops
::
ReshapeOpMaker
,
paddle
::
framework
::
DefaultGradOpDescMaker
<
true
>
);
REGISTER_OPERATOR
(
reshape_grad
,
ops
::
ReshapeGradOp
);
REGISTER_OP_CPU_KERNEL_FUNCTOR
(
reshape
,
float
,
ops
::
ReshapeKernel
<
float
>
,
double
,
ops
::
ReshapeKernel
<
double
>
,
int
,
ops
::
ReshapeKernel
<
int
>
,
int64_t
,
ops
::
ReshapeKernel
<
int64_t
>
);
REGISTER_OP_CPU_KERNEL_FUNCTOR
(
reshape_grad
,
float
,
ops
::
ReshapeGradKernel
<
float
>
,
double
,
ops
::
ReshapeGradKernel
<
double
>
,
int
,
ops
::
ReshapeGradKernel
<
int
>
,
int64_t
,
ops
::
ReshapeGradKernel
<
int64_t
>
);
REGISTER_OP_CPU_KERNEL_FUNCTOR
(
reshape
,
float
,
ops
::
ReshapeKernel
,
double
,
ops
::
ReshapeKernel
,
int
,
ops
::
ReshapeKernel
,
int64_t
,
ops
::
ReshapeKernel
);
REGISTER_OP_CPU_KERNEL_FUNCTOR
(
reshape_grad
,
float
,
ops
::
ReshapeGradKernel
,
double
,
ops
::
ReshapeGradKernel
,
int
,
ops
::
ReshapeGradKernel
,
int64_t
,
ops
::
ReshapeGradKernel
);
REGISTER_OPERATOR
(
reshape2
,
ops
::
Reshape2Op
,
ops
::
Reshape2OpMaker
,
ops
::
Reshape2GradMaker
);
REGISTER_OPERATOR
(
reshape2_grad
,
ops
::
Reshape2GradOp
);
REGISTER_OP_CPU_KERNEL_FUNCTOR
(
reshape2
,
float
,
ops
::
ReshapeKernel
<
float
>
,
double
,
ops
::
ReshapeKernel
<
double
>
,
int
,
ops
::
ReshapeKernel
<
int
>
,
int64_t
,
ops
::
ReshapeKernel
<
int64_t
>
);
REGISTER_OP_CPU_KERNEL_FUNCTOR
(
reshape2_grad
,
float
,
ops
::
ReshapeGradKernel
<
float
>
,
double
,
ops
::
ReshapeGradKernel
<
double
>
,
int
,
ops
::
ReshapeGradKernel
<
int
>
,
int64_t
,
ops
::
ReshapeGradKernel
<
int64_t
>
);
REGISTER_OP_CPU_KERNEL_FUNCTOR
(
reshape2
,
float
,
ops
::
ReshapeKernel
,
double
,
ops
::
ReshapeKernel
,
int
,
ops
::
ReshapeKernel
,
int64_t
,
ops
::
ReshapeKernel
);
REGISTER_OP_CPU_KERNEL_FUNCTOR
(
reshape2_grad
,
float
,
ops
::
ReshapeGradKernel
,
double
,
ops
::
ReshapeGradKernel
,
int
,
ops
::
ReshapeGradKernel
,
int64_t
,
ops
::
ReshapeGradKernel
);
#ifdef PADDLE_WITH_CUDA
REGISTER_OP_CUDA_KERNEL_FUNCTOR
(
reshape
,
float
,
ops
::
ReshapeKernel
<
float
>
,
double
,
ops
::
ReshapeKernel
<
double
>
,
int
,
ops
::
ReshapeKernel
<
int
>
,
int64_t
,
ops
::
ReshapeKernel
<
int64_t
>
);
REGISTER_OP_CUDA_KERNEL_FUNCTOR
(
reshape_grad
,
float
,
ops
::
ReshapeGradKernel
<
float
>
,
double
,
ops
::
ReshapeGradKernel
<
double
>
,
int
,
ops
::
ReshapeGradKernel
<
int
>
,
int64_t
,
ops
::
ReshapeGradKernel
<
int64_t
>
);
REGISTER_OP_CUDA_KERNEL_FUNCTOR
(
reshape2
,
float
,
ops
::
ReshapeKernel
<
float
>
,
double
,
ops
::
ReshapeKernel
<
double
>
,
int
,
ops
::
ReshapeKernel
<
int
>
,
int64_t
,
ops
::
ReshapeKernel
<
int64_t
>
);
REGISTER_OP_CUDA_KERNEL_FUNCTOR
(
reshape2_grad
,
float
,
ops
::
ReshapeGradKernel
<
float
>
,
double
,
ops
::
ReshapeGradKernel
<
double
>
,
int
,
ops
::
ReshapeGradKernel
<
int
>
,
int64_t
,
ops
::
ReshapeGradKernel
<
int64_t
>
);
REGISTER_OP_CUDA_KERNEL_FUNCTOR
(
reshape
,
float
,
ops
::
ReshapeKernel
,
double
,
ops
::
ReshapeKernel
,
int
,
ops
::
ReshapeKernel
,
int64_t
,
ops
::
ReshapeKernel
);
REGISTER_OP_CUDA_KERNEL_FUNCTOR
(
reshape_grad
,
float
,
ops
::
ReshapeGradKernel
,
double
,
ops
::
ReshapeGradKernel
,
int
,
ops
::
ReshapeGradKernel
,
int64_t
,
ops
::
ReshapeGradKernel
);
REGISTER_OP_CUDA_KERNEL_FUNCTOR
(
reshape2
,
float
,
ops
::
ReshapeKernel
,
double
,
ops
::
ReshapeKernel
,
int
,
ops
::
ReshapeKernel
,
int64_t
,
ops
::
ReshapeKernel
);
REGISTER_OP_CUDA_KERNEL_FUNCTOR
(
reshape2_grad
,
float
,
ops
::
ReshapeGradKernel
,
double
,
ops
::
ReshapeGradKernel
,
int
,
ops
::
ReshapeGradKernel
,
int64_t
,
ops
::
ReshapeGradKernel
);
#endif
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