Skip to content
Projects
Groups
Snippets
Help
Loading...
Sign in / Register
Toggle navigation
O
opencv
Project
Project
Details
Activity
Cycle Analytics
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Charts
Issues
0
Issues
0
List
Board
Labels
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Charts
Packages
Packages
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Charts
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
submodule
opencv
Commits
34d68edd
Commit
34d68edd
authored
Jul 18, 2016
by
Vadim Pisarevsky
Browse files
Options
Browse Files
Download
Plain Diff
Merge pull request #6924 from K-Shinotsuka:issue3
parents
b6f21a6c
6f8695c1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
22 additions
and
26 deletions
+22
-26
filter.cpp
modules/imgproc/src/filter.cpp
+22
-26
No files found.
modules/imgproc/src/filter.cpp
View file @
34d68edd
...
@@ -506,56 +506,52 @@ struct RowVec_8u32s
...
@@ -506,56 +506,52 @@ struct RowVec_8u32s
if
(
smallValues
)
if
(
smallValues
)
{
{
for
(
;
i
<=
width
-
16
;
i
+=
16
)
__m128i
z
=
_mm_setzero_si128
();
for
(
;
i
<=
width
-
8
;
i
+=
8
)
{
{
const
uchar
*
src
=
_src
+
i
;
const
uchar
*
src
=
_src
+
i
;
__m128i
f
,
z
=
_mm_setzero_si128
(),
s0
=
z
,
s1
=
z
,
s2
=
z
,
s3
=
z
;
__m128i
s0
=
z
,
s1
=
z
;
__m128i
x0
,
x1
,
x2
,
x3
;
for
(
k
=
0
;
k
<
_ksize
;
k
++
,
src
+=
cn
)
for
(
k
=
0
;
k
<
_ksize
;
k
++
,
src
+=
cn
)
{
{
f
=
_mm_cvtsi32_si128
(
_kx
[
k
]);
__m128i
f
=
_mm_cvtsi32_si128
(
_kx
[
k
]);
f
=
_mm_shuffle_epi32
(
f
,
0
);
f
=
_mm_shuffle_epi32
(
f
,
0
);
f
=
_mm_packs_epi32
(
f
,
f
);
x0
=
_mm_loadu_si128
((
const
__m128i
*
)
src
);
__m128i
x0
=
_mm_loadl_epi64
((
const
__m128i
*
)
src
);
x2
=
_mm_unpackhi_epi8
(
x0
,
z
);
x0
=
_mm_unpacklo_epi8
(
x0
,
z
);
x0
=
_mm_unpacklo_epi8
(
x0
,
z
);
x1
=
_mm_mulhi_epi16
(
x0
,
f
);
x3
=
_mm_mulhi_epi16
(
x2
,
f
);
__m128i
x1
=
_mm_unpackhi_epi16
(
x0
,
z
);
x0
=
_mm_
mullo_epi16
(
x0
,
f
);
x0
=
_mm_
unpacklo_epi16
(
x0
,
z
);
x2
=
_mm_mullo_epi16
(
x2
,
f
);
x0
=
_mm_madd_epi16
(
x0
,
f
);
s0
=
_mm_add_epi32
(
s0
,
_mm_unpacklo_epi16
(
x0
,
x1
)
);
x1
=
_mm_madd_epi16
(
x1
,
f
);
s1
=
_mm_add_epi32
(
s1
,
_mm_unpackhi_epi16
(
x0
,
x1
));
s
2
=
_mm_add_epi32
(
s2
,
_mm_unpacklo_epi16
(
x2
,
x3
)
);
s
0
=
_mm_add_epi32
(
s0
,
x0
);
s
3
=
_mm_add_epi32
(
s3
,
_mm_unpackhi_epi16
(
x2
,
x3
)
);
s
1
=
_mm_add_epi32
(
s1
,
x1
);
}
}
_mm_store_si128
((
__m128i
*
)(
dst
+
i
),
s0
);
_mm_store_si128
((
__m128i
*
)(
dst
+
i
),
s0
);
_mm_store_si128
((
__m128i
*
)(
dst
+
i
+
4
),
s1
);
_mm_store_si128
((
__m128i
*
)(
dst
+
i
+
4
),
s1
);
_mm_store_si128
((
__m128i
*
)(
dst
+
i
+
8
),
s2
);
_mm_store_si128
((
__m128i
*
)(
dst
+
i
+
12
),
s3
);
}
}
for
(
;
i
<=
width
-
4
;
i
+=
4
)
if
(
i
<=
width
-
4
)
{
{
const
uchar
*
src
=
_src
+
i
;
const
uchar
*
src
=
_src
+
i
;
__m128i
f
,
z
=
_mm_setzero_si128
(),
s0
=
z
,
x0
,
x1
;
__m128i
s0
=
z
;
for
(
k
=
0
;
k
<
_ksize
;
k
++
,
src
+=
cn
)
for
(
k
=
0
;
k
<
_ksize
;
k
++
,
src
+=
cn
)
{
{
f
=
_mm_cvtsi32_si128
(
_kx
[
k
]);
__m128i
f
=
_mm_cvtsi32_si128
(
_kx
[
k
]);
f
=
_mm_shuffle_epi32
(
f
,
0
);
f
=
_mm_shuffle_epi32
(
f
,
0
);
f
=
_mm_packs_epi32
(
f
,
f
);
x0
=
_mm_cvtsi32_si128
(
*
(
const
int
*
)
src
);
__m128i
x0
=
_mm_cvtsi32_si128
(
*
(
const
int
*
)
src
);
x0
=
_mm_unpacklo_epi8
(
x0
,
z
);
x0
=
_mm_unpacklo_epi8
(
x0
,
z
);
x
1
=
_mm_mulhi_epi16
(
x0
,
f
);
x
0
=
_mm_unpacklo_epi16
(
x0
,
z
);
x0
=
_mm_m
ullo
_epi16
(
x0
,
f
);
x0
=
_mm_m
add
_epi16
(
x0
,
f
);
s0
=
_mm_add_epi32
(
s0
,
_mm_unpacklo_epi16
(
x0
,
x1
)
);
s0
=
_mm_add_epi32
(
s0
,
x0
);
}
}
_mm_store_si128
((
__m128i
*
)(
dst
+
i
),
s0
);
_mm_store_si128
((
__m128i
*
)(
dst
+
i
),
s0
);
i
+=
4
;
}
}
}
}
return
i
;
return
i
;
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment