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
机器未来
Paddle
提交
5685a48c
P
Paddle
项目概览
机器未来
/
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
提交
5685a48c
编写于
3月 12, 2019
作者:
Z
Zhen Wang
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Add some fixme. test=develop
上级
4ab06076
变更
4
隐藏空白更改
内联
并排
Showing
4 changed file
with
18 addition
and
2 deletion
+18
-2
paddle/fluid/framework/details/inplace_op_pass.cc
paddle/fluid/framework/details/inplace_op_pass.cc
+5
-0
paddle/fluid/framework/details/memory_optimize_pass.cc
paddle/fluid/framework/details/memory_optimize_pass.cc
+5
-0
paddle/fluid/framework/ir/graph.cc
paddle/fluid/framework/ir/graph.cc
+1
-1
paddle/fluid/framework/ir/graph.h
paddle/fluid/framework/ir/graph.h
+7
-1
未找到文件。
paddle/fluid/framework/details/inplace_op_pass.cc
浏览文件 @
5685a48c
...
...
@@ -16,6 +16,7 @@
#include <algorithm>
#include <deque>
#include <iterator>
#include <memory>
#include <stack>
#include <string>
#include <unordered_map>
...
...
@@ -263,6 +264,10 @@ void InplacePass::WithdrawModify(const NodeSwapQueue& nodes,
void
InplacePass
::
TryInplaceOpInputOutput
(
ir
::
Node
*
op
,
ir
::
Graph
*
graph
)
const
{
VLOG
(
4
)
<<
"Try to inplace op "
<<
op
->
Name
();
// FIXME(liuwei1031): Graph is not aware of the existence of BlockDescs and
// ProgramDescs.
// The operations related to BlockDesc or ProgramDesc should perform on Graph
// or Node directly!
PADDLE_ENFORCE
(
op
->
Op
()
!=
nullptr
&&
op
->
Op
()
->
Block
()
!=
nullptr
,
"op_desc is nullptr"
);
// some pre-requirments need to meet if the op want to inplaced.
...
...
paddle/fluid/framework/details/memory_optimize_pass.cc
浏览文件 @
5685a48c
...
...
@@ -24,6 +24,7 @@
#include <sstream>
#include <string>
#include <type_traits>
#include <unordered_set>
#include <vector>
#include "gflags/gflags.h"
#include "paddle/fluid/framework/data_type.h"
...
...
@@ -191,6 +192,10 @@ void MemoryOptimizePass::SubGraphOptimize(OpDesc* op_desc) const {
// immediately to make the subblock variable reuse strategy take
// effect. Because it is a single op in graph. No need to
// update the ir nodes.
// FIXME(liuwei1031): Graph is not aware of the existence of
// BlockDescs and ProgramDescs.
// The operations related to BlockDesc or ProgramDesc should perform
// on Graph or Node directly!
sub_op_desc
->
Rename
(
var
->
Name
(),
cache
->
Name
());
if
(
sub_op_desc
->
Block
()
!=
nullptr
&&
sub_op_desc
->
Block
()
->
HasVar
(
var
->
Name
()))
{
...
...
paddle/fluid/framework/ir/graph.cc
浏览文件 @
5685a48c
...
...
@@ -13,7 +13,7 @@ See the License for the specific language governing permissions and
limitations under the License. */
#include <algorithm>
#include <unordered_
set
>
#include <unordered_
map
>
#include "paddle/fluid/framework/ir/graph.h"
#include "paddle/fluid/framework/op_proto_maker.h"
...
...
paddle/fluid/framework/ir/graph.h
浏览文件 @
5685a48c
...
...
@@ -17,6 +17,7 @@ limitations under the License. */
#include <map>
#include <memory>
#include <string>
#include <unordered_set>
#include <vector>
#include "paddle/fluid/framework/ir/node.h"
...
...
@@ -199,7 +200,12 @@ class Graph {
// WARN: After a series of passes, the current graph can be quite
// different from OriginProgram. Caller shouldn't assume much from
// the returned OriginProgram.
const
ProgramDesc
&
OriginProgram
()
const
{
return
program_
;
}
const
ProgramDesc
&
OriginProgram
()
const
{
LOG
(
WARNING
)
<<
"WARN: After a series of passes, the current graph can be "
"quite different from OriginProgram. So, please avoid "
"using the `OriginProgram()` method!"
;
return
program_
;
}
// This method takes ownership of `node`.
ir
::
Node
*
AddNode
(
ir
::
Node
*
node
)
{
...
...
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