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
PaddlePaddle
Paddle
提交
e19032fb
P
Paddle
项目概览
PaddlePaddle
/
Paddle
大约 1 年 前同步成功
通知
2299
Star
20931
Fork
5422
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
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
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
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
未验证
提交
e19032fb
编写于
12月 21, 2017
作者:
C
chengduo
提交者:
GitHub
12月 21, 2017
浏览文件
操作
浏览文件
下载
差异文件
Merge pull request #6743 from chengduoZH/profiling/02.recognize_digits
Refine elementwiseAdd and im2col
上级
82eff9dc
cb3a74e4
变更
2
显示空白变更内容
内联
并排
Showing
2 changed file
with
19 addition
and
17 deletion
+19
-17
paddle/operators/elementwise_op_function.h
paddle/operators/elementwise_op_function.h
+7
-5
paddle/operators/math/im2col.cc
paddle/operators/math/im2col.cc
+12
-12
未找到文件。
paddle/operators/elementwise_op_function.h
浏览文件 @
e19032fb
...
@@ -103,11 +103,13 @@ class MidWiseTransformIterator<T, platform::CPUDeviceContext> {
...
@@ -103,11 +103,13 @@ class MidWiseTransformIterator<T, platform::CPUDeviceContext> {
MidWiseTransformIterator
<
T
,
platform
::
CPUDeviceContext
>&
operator
++
()
{
MidWiseTransformIterator
<
T
,
platform
::
CPUDeviceContext
>&
operator
++
()
{
++
j_
;
++
j_
;
i
_
=
j_
/
post_
;
i
f
(
UNLIKELY
(
j_
==
post_
))
{
if
(
UNLIKELY
(
i_
==
n_
))
{
++
i_
;
j_
=
0
;
j_
=
0
;
if
(
UNLIKELY
(
i_
==
n_
))
{
i_
=
0
;
i_
=
0
;
}
}
}
return
*
this
;
return
*
this
;
}
}
...
@@ -125,10 +127,10 @@ class MidWiseTransformIterator<T, platform::CPUDeviceContext> {
...
@@ -125,10 +127,10 @@ class MidWiseTransformIterator<T, platform::CPUDeviceContext> {
private:
private:
const
T
*
ptr_
;
const
T
*
ptr_
;
int
i_
;
int
64_t
i_
;
int64_t
j_
;
int64_t
j_
;
int64_t
n_
;
int64_t
n_
;
int
post_
;
int
64_t
post_
;
};
};
#ifdef __NVCC__
#ifdef __NVCC__
...
...
paddle/operators/math/im2col.cc
浏览文件 @
e19032fb
...
@@ -61,14 +61,13 @@ class Im2ColFunctor<paddle::operators::math::ColFormat::kCFO,
...
@@ -61,14 +61,13 @@ class Im2ColFunctor<paddle::operators::math::ColFormat::kCFO,
const
T
*
im_data
=
im
.
data
<
T
>
();
const
T
*
im_data
=
im
.
data
<
T
>
();
T
*
col_data
=
col
->
data
<
T
>
();
T
*
col_data
=
col
->
data
<
T
>
();
for
(
int
c
=
0
;
c
<
channels_col
;
++
c
)
{
for
(
int
c
=
0
;
c
<
channels_col
;
++
c
)
{
int
w_offset
=
c
%
filter_width
;
int
w_offset
=
c
%
filter_width
;
int
h_offset
=
(
c
/
filter_width
)
%
filter_height
;
int
h_offset
=
(
c
/
filter_width
)
%
filter_height
;
int
c_im
=
c
/
filter_width
/
filter_height
;
int
c_im
=
c
/
(
filter_width
*
filter_height
)
;
for
(
int
h
=
0
;
h
<
col_height
;
++
h
)
{
for
(
int
h
=
0
;
h
<
col_height
;
++
h
)
{
for
(
int
w
=
0
;
w
<
col_width
;
++
w
)
{
int
im_row_idx
=
h
*
stride
[
0
]
-
padding
[
0
]
+
h_offset
*
dilation
[
0
];
int
im_row_idx
=
h
*
stride
[
0
]
-
padding
[
0
]
+
h_offset
*
dilation
[
0
];
for
(
int
w
=
0
;
w
<
col_width
;
++
w
)
{
int
im_col_idx
=
w
*
stride
[
1
]
-
padding
[
1
]
+
w_offset
*
dilation
[
1
];
int
im_col_idx
=
w
*
stride
[
1
]
-
padding
[
1
]
+
w_offset
*
dilation
[
1
];
int
col_idx
=
(
c
*
col_height
+
h
)
*
col_width
+
w
;
int
col_idx
=
(
c
*
col_height
+
h
)
*
col_width
+
w
;
int
im_idx
=
(
im_row_idx
+
c_im
*
im_height
)
*
im_width
+
im_col_idx
;
int
im_idx
=
(
im_row_idx
+
c_im
*
im_height
)
*
im_width
+
im_col_idx
;
...
@@ -130,16 +129,14 @@ class Col2ImFunctor<paddle::operators::math::ColFormat::kCFO,
...
@@ -130,16 +129,14 @@ class Col2ImFunctor<paddle::operators::math::ColFormat::kCFO,
for
(
int
c
=
0
;
c
<
channels_col
;
++
c
)
{
for
(
int
c
=
0
;
c
<
channels_col
;
++
c
)
{
int
w_offset
=
c
%
filter_width
;
int
w_offset
=
c
%
filter_width
;
int
h_offset
=
(
c
/
filter_width
)
%
filter_height
;
int
h_offset
=
(
c
/
filter_width
)
%
filter_height
;
int
c_im
=
c
/
filter_width
/
filter_height
;
int
c_im
=
c
/
(
filter_width
*
filter_height
)
;
for
(
int
h
=
0
;
h
<
col_height
;
++
h
)
{
for
(
int
h
=
0
;
h
<
col_height
;
++
h
)
{
for
(
int
w
=
0
;
w
<
col_width
;
++
w
)
{
int
im_row_idx
=
h
*
stride
[
0
]
-
padding
[
0
]
+
h_offset
*
dilation
[
0
];
int
im_row_idx
=
h
*
stride
[
0
]
-
padding
[
0
]
+
h_offset
*
dilation
[
0
];
for
(
int
w
=
0
;
w
<
col_width
;
++
w
)
{
int
im_col_idx
=
w
*
stride
[
1
]
-
padding
[
1
]
+
w_offset
*
dilation
[
1
];
int
im_col_idx
=
w
*
stride
[
1
]
-
padding
[
1
]
+
w_offset
*
dilation
[
1
];
if
((
im_row_idx
)
>=
0
&&
(
im_row_idx
)
<
im_height
&&
if
((
im_row_idx
)
>=
0
&&
(
im_row_idx
)
<
im_height
&&
(
im_col_idx
)
>=
0
&&
(
im_col_idx
)
<
im_width
)
{
(
im_col_idx
)
>=
0
&&
(
im_col_idx
)
<
im_width
)
{
im_row_idx
+=
c_im
*
im_height
;
im_data
[(
im_row_idx
+
c_im
*
im_height
)
*
im_width
+
im_col_idx
]
+=
im_data
[
im_row_idx
*
im_width
+
im_col_idx
]
+=
col_data
[(
c
*
col_height
+
h
)
*
col_width
+
w
];
col_data
[(
c
*
col_height
+
h
)
*
col_width
+
w
];
}
}
}
}
...
@@ -199,12 +196,13 @@ class Im2ColFunctor<paddle::operators::math::ColFormat::kOCF,
...
@@ -199,12 +196,13 @@ class Im2ColFunctor<paddle::operators::math::ColFormat::kOCF,
for
(
int
channel
=
0
;
channel
<
im_channels
;
++
channel
)
{
for
(
int
channel
=
0
;
channel
<
im_channels
;
++
channel
)
{
for
(
int
filter_row_idx
=
0
;
filter_row_idx
<
filter_height
;
for
(
int
filter_row_idx
=
0
;
filter_row_idx
<
filter_height
;
++
filter_row_idx
)
{
++
filter_row_idx
)
{
for
(
int
filter_col_idx
=
0
;
filter_col_idx
<
filter_width
;
++
filter_col_idx
)
{
int
im_row_offset
=
int
im_row_offset
=
col_row_idx
*
stride
[
0
]
+
filter_row_idx
-
padding
[
0
];
col_row_idx
*
stride
[
0
]
+
filter_row_idx
-
padding
[
0
];
for
(
int
filter_col_idx
=
0
;
filter_col_idx
<
filter_width
;
++
filter_col_idx
)
{
int
im_col_offset
=
int
im_col_offset
=
col_col_idx
*
stride
[
1
]
+
filter_col_idx
-
padding
[
1
];
col_col_idx
*
stride
[
1
]
+
filter_col_idx
-
padding
[
1
];
int
col_offset
=
int
col_offset
=
((((
col_row_idx
)
*
col_width
+
col_col_idx
)
*
im_channels
+
((((
col_row_idx
)
*
col_width
+
col_col_idx
)
*
im_channels
+
channel
)
*
channel
)
*
...
@@ -271,12 +269,13 @@ class Col2ImFunctor<paddle::operators::math::ColFormat::kOCF,
...
@@ -271,12 +269,13 @@ class Col2ImFunctor<paddle::operators::math::ColFormat::kOCF,
for
(
int
channel
=
0
;
channel
<
im_channels
;
++
channel
)
{
for
(
int
channel
=
0
;
channel
<
im_channels
;
++
channel
)
{
for
(
int
filter_row_idx
=
0
;
filter_row_idx
<
filter_height
;
for
(
int
filter_row_idx
=
0
;
filter_row_idx
<
filter_height
;
++
filter_row_idx
)
{
++
filter_row_idx
)
{
for
(
int
filter_col_idx
=
0
;
filter_col_idx
<
filter_width
;
++
filter_col_idx
)
{
int
im_row_offset
=
int
im_row_offset
=
col_row_idx
*
stride
[
0
]
+
filter_row_idx
-
padding
[
0
];
col_row_idx
*
stride
[
0
]
+
filter_row_idx
-
padding
[
0
];
for
(
int
filter_col_idx
=
0
;
filter_col_idx
<
filter_width
;
++
filter_col_idx
)
{
int
im_col_offset
=
int
im_col_offset
=
col_col_idx
*
stride
[
1
]
+
filter_col_idx
-
padding
[
1
];
col_col_idx
*
stride
[
1
]
+
filter_col_idx
-
padding
[
1
];
int
col_offset
=
int
col_offset
=
(((
col_row_idx
*
col_width
+
col_col_idx
)
*
im_channels
+
(((
col_row_idx
*
col_width
+
col_col_idx
)
*
im_channels
+
channel
)
*
channel
)
*
...
@@ -284,6 +283,7 @@ class Col2ImFunctor<paddle::operators::math::ColFormat::kOCF,
...
@@ -284,6 +283,7 @@ class Col2ImFunctor<paddle::operators::math::ColFormat::kOCF,
filter_row_idx
)
*
filter_row_idx
)
*
filter_width
+
filter_width
+
filter_col_idx
;
filter_col_idx
;
if
(
im_row_offset
>=
0
&&
im_row_offset
<
im_height
&&
if
(
im_row_offset
>=
0
&&
im_row_offset
<
im_height
&&
im_col_offset
>=
0
&&
im_col_offset
<
im_width
)
{
im_col_offset
>=
0
&&
im_col_offset
<
im_width
)
{
int
im_offset
=
int
im_offset
=
...
...
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