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
Greenplum
Opencv
提交
a2a8ba17
O
Opencv
项目概览
Greenplum
/
Opencv
大约 1 年 前同步成功
通知
7
Star
0
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
DevOps
流水线
流水线任务
计划
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
O
Opencv
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
DevOps
DevOps
流水线
流水线任务
计划
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
流水线任务
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
提交
a2a8ba17
编写于
1月 12, 2015
作者:
I
Ilya Lavrenov
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
compare
上级
8d48632e
变更
1
显示空白变更内容
内联
并排
Showing
1 changed file
with
124 addition
and
0 deletion
+124
-0
modules/core/src/arithm.cpp
modules/core/src/arithm.cpp
+124
-0
未找到文件。
modules/core/src/arithm.cpp
浏览文件 @
a2a8ba17
...
...
@@ -3268,6 +3268,130 @@ struct Cmp_SIMD<float>
uint8x8_t
v_mask
;
};
#elif CV_SSE2
template
<
>
struct
Cmp_SIMD
<
schar
>
{
explicit
Cmp_SIMD
(
int
code_
)
:
code
(
code_
)
{
CV_Assert
(
code
==
CMP_GT
||
code
==
CMP_LE
||
code
==
CMP_EQ
||
code
==
CMP_NE
);
haveSSE
=
checkHardwareSupport
(
CV_CPU_SSE2
);
v_mask
=
_mm_set1_epi8
(
0xff
);
}
int
operator
()
(
const
schar
*
src1
,
const
schar
*
src2
,
uchar
*
dst
,
int
width
)
const
{
int
x
=
0
;
if
(
!
haveSSE
)
return
x
;
if
(
code
==
CMP_GT
)
for
(
;
x
<=
width
-
16
;
x
+=
16
)
_mm_storeu_si128
((
__m128i
*
)(
dst
+
x
),
_mm_cmpgt_epi8
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
))));
else
if
(
code
==
CMP_LE
)
for
(
;
x
<=
width
-
16
;
x
+=
16
)
{
__m128i
v_gt
=
_mm_cmpgt_epi8
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
)));
_mm_storeu_si128
((
__m128i
*
)(
dst
+
x
),
_mm_xor_si128
(
v_mask
,
v_gt
));
}
else
if
(
code
==
CMP_EQ
)
for
(
;
x
<=
width
-
16
;
x
+=
16
)
_mm_storeu_si128
((
__m128i
*
)(
dst
+
x
),
_mm_cmpeq_epi8
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
))));
else
if
(
code
==
CMP_NE
)
for
(
;
x
<=
width
-
16
;
x
+=
16
)
{
__m128i
v_eq
=
_mm_cmpeq_epi8
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
)));
_mm_storeu_si128
((
__m128i
*
)(
dst
+
x
),
_mm_xor_si128
(
v_mask
,
v_eq
));
}
return
x
;
}
int
code
;
__m128i
v_mask
;
bool
haveSSE
;
};
template
<
>
struct
Cmp_SIMD
<
int
>
{
explicit
Cmp_SIMD
(
int
code_
)
:
code
(
code_
)
{
CV_Assert
(
code
==
CMP_GT
||
code
==
CMP_LE
||
code
==
CMP_EQ
||
code
==
CMP_NE
);
haveSSE
=
checkHardwareSupport
(
CV_CPU_SSE2
);
v_mask
=
_mm_set1_epi32
(
0xffffffff
);
}
int
operator
()
(
const
int
*
src1
,
const
int
*
src2
,
uchar
*
dst
,
int
width
)
const
{
int
x
=
0
;
if
(
!
haveSSE
)
return
x
;
if
(
code
==
CMP_GT
)
for
(
;
x
<=
width
-
8
;
x
+=
8
)
{
__m128i
v_dst0
=
_mm_cmpgt_epi32
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
)));
__m128i
v_dst1
=
_mm_cmpgt_epi32
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
+
4
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
+
4
)));
_mm_storel_epi64
((
__m128i
*
)(
dst
+
x
),
_mm_packs_epi16
(
_mm_packs_epi32
(
v_dst0
,
v_dst1
),
v_mask
));
}
else
if
(
code
==
CMP_LE
)
for
(
;
x
<=
width
-
8
;
x
+=
8
)
{
__m128i
v_dst0
=
_mm_cmpgt_epi32
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
)));
__m128i
v_dst1
=
_mm_cmpgt_epi32
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
+
4
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
+
4
)));
_mm_storel_epi64
((
__m128i
*
)(
dst
+
x
),
_mm_xor_si128
(
_mm_packs_epi16
(
_mm_packs_epi32
(
v_dst0
,
v_dst1
),
v_mask
),
v_mask
));
}
else
if
(
code
==
CMP_EQ
)
for
(
;
x
<=
width
-
8
;
x
+=
8
)
{
__m128i
v_dst0
=
_mm_cmpeq_epi32
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
)));
__m128i
v_dst1
=
_mm_cmpeq_epi32
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
+
4
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
+
4
)));
_mm_storel_epi64
((
__m128i
*
)(
dst
+
x
),
_mm_packs_epi16
(
_mm_packs_epi32
(
v_dst0
,
v_dst1
),
v_mask
));
}
else
if
(
code
==
CMP_NE
)
for
(
;
x
<=
width
-
8
;
x
+=
8
)
{
__m128i
v_dst0
=
_mm_cmpeq_epi32
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
)));
__m128i
v_dst1
=
_mm_cmpeq_epi32
(
_mm_loadu_si128
((
const
__m128i
*
)(
src1
+
x
+
4
)),
_mm_loadu_si128
((
const
__m128i
*
)(
src2
+
x
+
4
)));
_mm_storel_epi64
((
__m128i
*
)(
dst
+
x
),
_mm_xor_si128
(
v_mask
,
_mm_packs_epi16
(
_mm_packs_epi32
(
v_dst0
,
v_dst1
),
v_mask
)));
}
return
x
;
}
int
code
;
__m128i
v_mask
;
bool
haveSSE
;
};
#endif
template
<
typename
T
>
static
void
...
...
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