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
飞不过沧海
Paddle
提交
9773f38f
P
Paddle
项目概览
飞不过沧海
/
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
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 >>
提交
9773f38f
编写于
3月 01, 2019
作者:
L
luotao1
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
cache runtime_context
test=develop
上级
187cffd0
变更
2
隐藏空白更改
内联
并排
Showing
2 changed file
with
15 addition
and
7 deletion
+15
-7
paddle/fluid/framework/operator.cc
paddle/fluid/framework/operator.cc
+14
-7
paddle/fluid/framework/operator.h
paddle/fluid/framework/operator.h
+1
-0
未找到文件。
paddle/fluid/framework/operator.cc
浏览文件 @
9773f38f
...
...
@@ -916,7 +916,14 @@ std::vector<KernelConfig>* OperatorWithKernel::GetKernelConfig(
void
OperatorWithKernel
::
RunImpl
(
const
Scope
&
scope
,
const
platform
::
Place
&
place
)
const
{
RuntimeContext
ctx
(
Inputs
(),
Outputs
(),
scope
);
if
(
!
runtime_ctx_
)
{
// RuntimeContext is used to relate input/output names of Operator with
// the corresponding variables in Scope.
// Since the input/output names of Operator do not change in the execution,
// RuntimeContext could be created only at the first iteration of
// the execution to save the elapsed time.
runtime_ctx_
=
new
RuntimeContext
(
Inputs
(),
Outputs
(),
scope
);
}
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
auto
*
dev_ctx
=
pool
.
Get
(
place
);
...
...
@@ -931,7 +938,7 @@ void OperatorWithKernel::RunImpl(const Scope& scope,
OpKernelMap
&
kernels
=
kernels_iter
->
second
;
auto
expected_kernel_key
=
this
->
GetExpectedKernelType
(
ExecutionContext
(
*
this
,
scope
,
*
dev_ctx
,
ctx
,
nullptr
));
ExecutionContext
(
*
this
,
scope
,
*
dev_ctx
,
*
runtime_ctx_
,
nullptr
));
VLOG
(
3
)
<<
"expected_kernel_key:"
<<
expected_kernel_key
;
auto
kernel_iter
=
kernels
.
find
(
expected_kernel_key
);
...
...
@@ -955,8 +962,8 @@ void OperatorWithKernel::RunImpl(const Scope& scope,
// do data transformScope &transfer_scope;
std
::
vector
<
std
::
string
>
transfered_inplace_vars
;
auto
*
transfer_scope
=
PrepareData
(
scope
,
expected_kernel_key
,
&
transfered_inplace_vars
,
&
ctx
);
auto
*
transfer_scope
=
PrepareData
(
scope
,
expected_kernel_key
,
&
transfered_inplace_vars
,
runtime_ctx_
);
// exec scope is the scope that kernel actually executed on.
const
Scope
&
exec_scope
=
...
...
@@ -966,12 +973,12 @@ void OperatorWithKernel::RunImpl(const Scope& scope,
dev_ctx
=
pool
.
Get
(
expected_kernel_key
.
place_
);
}
RuntimeInferShapeContext
infer_shape_ctx
(
*
this
,
exec_scope
,
ctx
);
RuntimeInferShapeContext
infer_shape_ctx
(
*
this
,
exec_scope
,
*
runtime_ctx_
);
this
->
InferShape
(
&
infer_shape_ctx
);
// TODO(panyx0718): ExecutionContext should only depend on RuntimeContext
// not Scope. Imperative mode only pass inputs and get outputs.
kernel_iter
->
second
(
ExecutionContext
(
*
this
,
exec_scope
,
*
dev_ctx
,
ctx
,
kernel_configs
));
kernel_iter
->
second
(
ExecutionContext
(
*
this
,
exec_scope
,
*
dev_ctx
,
*
runtime_ctx_
,
kernel_configs
));
if
(
!
transfered_inplace_vars
.
empty
())
{
// there is inplace variable has been transfered.
...
...
paddle/fluid/framework/operator.h
浏览文件 @
9773f38f
...
...
@@ -541,6 +541,7 @@ class OperatorWithKernel : public OperatorBase {
protected:
mutable
OpKernelConfigsMap
kernel_configs_map_
;
mutable
RuntimeContext
*
runtime_ctx_
=
nullptr
;
};
extern
bool
OpSupportGPU
(
const
std
::
string
&
op_type
);
...
...
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