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-Lite
提交
bc0bd998
P
Paddle-Lite
项目概览
PaddlePaddle
/
Paddle-Lite
通知
331
Star
4
Fork
1
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
271
列表
看板
标记
里程碑
合并请求
78
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle-Lite
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
271
Issue
271
列表
看板
标记
里程碑
合并请求
78
合并请求
78
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
bc0bd998
编写于
8月 12, 2020
作者:
C
chenjiaoAngel
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
fix asm error
上级
d3813bd9
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
13 addition
and
13 deletion
+13
-13
lite/backends/arm/math/conv5x5s1_depthwise_fp32.cc
lite/backends/arm/math/conv5x5s1_depthwise_fp32.cc
+13
-13
未找到文件。
lite/backends/arm/math/conv5x5s1_depthwise_fp32.cc
浏览文件 @
bc0bd998
...
...
@@ -2812,7 +2812,7 @@ inline void compute_all_padding_post_relu6(float* dout,
*
dout
++
=
sum
>
0.
f
?
(
sum
<
six
[
0
]
?
sum
:
six
[
0
])
:
0.
f
;
}
for
(
int
w
=
pad_right
;
w
>
4
;
w
--
)
{
*
dout
++
=
bias
[
0
]
>
0.
f
?
?
(
bias
[
0
]
<
six
[
0
]
?
bias
[
0
]
:
six
[
0
])
:
0.
f
;
*
dout
++
=
bias
[
0
]
>
0.
f
?
(
bias
[
0
]
<
six
[
0
]
?
bias
[
0
]
:
six
[
0
])
:
0.
f
;
}
}
inline
void
compute_all_padding_pre_leakyRelu
(
float
*
dout
,
...
...
@@ -3444,7 +3444,7 @@ inline void compute_all_padding_post_leakyRelu(float* dout,
*
dout
++
=
sum
>
0.
f
?
sum
:
sum
*
scale
[
0
];
}
for
(
int
w
=
pad_right
;
w
>
4
;
w
--
)
{
*
dout
++
=
bias
[
0
]
>
0.
f
?
?
bias
[
0
]
:
bias
[
0
]
*
scale
[
0
];
*
dout
++
=
bias
[
0
]
>
0.
f
?
bias
[
0
]
:
bias
[
0
]
*
scale
[
0
];
}
}
void
conv_depthwise_5x5s1_bias
(
float
*
dout
,
...
...
@@ -3483,14 +3483,14 @@ void conv_depthwise_5x5s1_bias(float* dout,
for
(
int
c
=
0
;
c
<
chin
;
c
++
)
{
const
float
*
din_ch
=
din_batch
+
c
*
in_size
;
const
float
*
weights_ch
=
weights
+
c
*
weights_size
;
float
*
dout_ch
=
dout_
_
batch
+
c
*
out_size
;
float
*
dout_ch
=
dout_batch
+
c
*
out_size
;
float
bias_val
=
flag_bias
?
bias
[
c
]
:
0.
f
;
const
float
*
din_ptr0
=
din_ch
;
const
float
*
din_ptr1
=
din_ptr0
+
win
;
const
float
*
din_ptr2
=
din_ptr1
+
win
;
const
float
*
din_ptr3
=
din_ptr2
+
win
;
const
float
*
din_ptr4
=
din_ptr3
+
win
;
float
vbias
[
4
]
=
{
bias_
c
,
bias_c
,
bias_c
,
bias_c
};
float
vbias
[
4
]
=
{
bias_
val
,
bias_val
,
bias_val
,
bias_val
};
float
*
dout_ptr
=
dout_ch
;
float32x4_t
wr5
;
float32x4_t
wr6
;
...
...
@@ -3576,7 +3576,7 @@ void conv_depthwise_5x5s1_bias_relu(float* dout,
int
in_channel_size
=
chin
*
in_size
;
int
out_channel_size
=
chin
*
out_size
;
int
weights_size
=
25
;
float32x4_t
vzero
=
vdupq_f32
(
0.
f
);
float32x4_t
vzero
=
vdupq_
n_
f32
(
0.
f
);
for
(
int
n
=
0
;
n
<
num
;
n
++
)
{
const
float
*
din_batch
=
din
+
n
*
in_channel_size
;
float
*
dout_batch
=
dout
+
n
*
out_channel_size
;
...
...
@@ -3584,14 +3584,14 @@ void conv_depthwise_5x5s1_bias_relu(float* dout,
for
(
int
c
=
0
;
c
<
chin
;
c
++
)
{
const
float
*
din_ch
=
din_batch
+
c
*
in_size
;
const
float
*
weights_ch
=
weights
+
c
*
weights_size
;
float
*
dout_ch
=
dout_
_
batch
+
c
*
out_size
;
float
*
dout_ch
=
dout_batch
+
c
*
out_size
;
float
bias_val
=
flag_bias
?
bias
[
c
]
:
0.
f
;
const
float
*
din_ptr0
=
din_ch
;
const
float
*
din_ptr1
=
din_ptr0
+
win
;
const
float
*
din_ptr2
=
din_ptr1
+
win
;
const
float
*
din_ptr3
=
din_ptr2
+
win
;
const
float
*
din_ptr4
=
din_ptr3
+
win
;
float
vbias
[
4
]
=
{
bias_
c
,
bias_c
,
bias_c
,
bias_c
};
float
vbias
[
4
]
=
{
bias_
val
,
bias_val
,
bias_val
,
bias_val
};
float
*
dout_ptr
=
dout_ch
;
float32x4_t
wr5
;
float32x4_t
wr6
;
...
...
@@ -3678,7 +3678,7 @@ void conv_depthwise_5x5s1_bias_relu6(float* dout,
int
in_channel_size
=
chin
*
in_size
;
int
out_channel_size
=
chin
*
out_size
;
int
weights_size
=
25
;
float32x4_t
vzero
=
vdupq_f32
(
0.
f
);
float32x4_t
vzero
=
vdupq_
n_
f32
(
0.
f
);
for
(
int
n
=
0
;
n
<
num
;
n
++
)
{
const
float
*
din_batch
=
din
+
n
*
in_channel_size
;
float
*
dout_batch
=
dout
+
n
*
out_channel_size
;
...
...
@@ -3686,14 +3686,14 @@ void conv_depthwise_5x5s1_bias_relu6(float* dout,
for
(
int
c
=
0
;
c
<
chin
;
c
++
)
{
const
float
*
din_ch
=
din_batch
+
c
*
in_size
;
const
float
*
weights_ch
=
weights
+
c
*
weights_size
;
float
*
dout_ch
=
dout_
_
batch
+
c
*
out_size
;
float
*
dout_ch
=
dout_batch
+
c
*
out_size
;
float
bias_val
=
flag_bias
?
bias
[
c
]
:
0.
f
;
const
float
*
din_ptr0
=
din_ch
;
const
float
*
din_ptr1
=
din_ptr0
+
win
;
const
float
*
din_ptr2
=
din_ptr1
+
win
;
const
float
*
din_ptr3
=
din_ptr2
+
win
;
const
float
*
din_ptr4
=
din_ptr3
+
win
;
float
vbias
[
4
]
=
{
bias_
c
,
bias_c
,
bias_c
,
bias_c
};
float
vbias
[
4
]
=
{
bias_
val
,
bias_val
,
bias_val
,
bias_val
};
float
*
dout_ptr
=
dout_ch
;
float32x4_t
wr5
;
float32x4_t
wr6
;
...
...
@@ -3783,7 +3783,7 @@ void conv_depthwise_5x5s1_bias_leakyRelu(float* dout,
int
in_channel_size
=
chin
*
in_size
;
int
out_channel_size
=
chin
*
out_size
;
int
weights_size
=
25
;
float32x4_t
vzero
=
vdupq_f32
(
0.
f
);
float32x4_t
vzero
=
vdupq_
n_
f32
(
0.
f
);
for
(
int
n
=
0
;
n
<
num
;
n
++
)
{
const
float
*
din_batch
=
din
+
n
*
in_channel_size
;
float
*
dout_batch
=
dout
+
n
*
out_channel_size
;
...
...
@@ -3791,14 +3791,14 @@ void conv_depthwise_5x5s1_bias_leakyRelu(float* dout,
for
(
int
c
=
0
;
c
<
chin
;
c
++
)
{
const
float
*
din_ch
=
din_batch
+
c
*
in_size
;
const
float
*
weights_ch
=
weights
+
c
*
weights_size
;
float
*
dout_ch
=
dout_
_
batch
+
c
*
out_size
;
float
*
dout_ch
=
dout_batch
+
c
*
out_size
;
float
bias_val
=
flag_bias
?
bias
[
c
]
:
0.
f
;
const
float
*
din_ptr0
=
din_ch
;
const
float
*
din_ptr1
=
din_ptr0
+
win
;
const
float
*
din_ptr2
=
din_ptr1
+
win
;
const
float
*
din_ptr3
=
din_ptr2
+
win
;
const
float
*
din_ptr4
=
din_ptr3
+
win
;
float
vbias
[
4
]
=
{
bias_
c
,
bias_c
,
bias_c
,
bias_c
};
float
vbias
[
4
]
=
{
bias_
val
,
bias_val
,
bias_val
,
bias_val
};
float
*
dout_ptr
=
dout_ch
;
float32x4_t
wr5
;
float32x4_t
wr6
;
...
...
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