Add support for odd height YUVA alpha blending.

Add support for odd height YUVA alpha blending.

R=dhrosa@google.com, harryjin@google.com
BUG=libyuv:527
Committed: https://chromium.googlesource.com/libyuv/libyuv/+/2657688e701709a5af935e6ea27f4f8967208f2d

[Diony Rosa, 2015-12-07 18:40:25]

https://codereview.chromium.org/1507683003/diff/1/source/planar_functions.cc
File source/planar_functions.cc (right):

https://codereview.chromium.org/1507683003/diff/1/source/planar_functions.cc#...
source/planar_functions.cc:724: // last row of odd height image use 1 row of
alpha instead of 2.
This skips the last row if the height is odd. e.g. if the height is 5, you'll
get:

y = 0
y = 2

I think you want to use halfheight here, and check if y == (halfheight - 1) AND
(halfheight & 1)

[fbarchard, 2015-12-07 18:58:10]

fbarchard@chromium.org changed reviewers:
+ fbarchard@chromium.org

[fbarchard, 2015-12-07 18:58:11]

drmemory passes
d:\src\libyuv\libyuv>drmemory out\debug\libyuv_unittest.exe
--gtest_catch_exceptions=0 --gtest_filter=*I420Blend_Opt
~~Dr.M~~ Dr. Memory version 1.9.0
~~Dr.M~~ Running "out\debug\libyuv_unittest.exe --gtest_catch_exceptions=0
--gtest_filter=*I420Blend_Opt"
Note: Google Test filter = *I420Blend_Opt
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from LibYUVPlanarTest
[ RUN      ] LibYUVPlanarTest.I420Blend_Opt
[       OK ] LibYUVPlanarTest.I420Blend_Opt (76 ms)
[----------] 1 test from LibYUVPlanarTest (114 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test case ran. (450 ms total)
[  PASSED  ] 1 test.
~~Dr.M~~
~~Dr.M~~ NO ERRORS FOUND:
~~Dr.M~~       0 unique,     0 total unaddressable access(es)
~~Dr.M~~       0 unique,     0 total uninitialized access(es)
~~Dr.M~~       0 unique,     0 total invalid heap argument(s)
~~Dr.M~~       0 unique,     0 total GDI usage error(s)
~~Dr.M~~       0 unique,     0 total handle leak(s)
~~Dr.M~~       0 unique,     0 total warning(s)
~~Dr.M~~       0 unique,     0 total,      0 byte(s) of leak(s)
~~Dr.M~~       0 unique,     0 total,      0 byte(s) of possible leak(s)
~~Dr.M~~ ERRORS IGNORED:
~~Dr.M~~      10 potential leak(s) (suspected false positives)
~~Dr.M~~          (details: C:\Users\fbarchard\AppData\Roaming\Dr.
Memory\DrMemory-libyuv_unittest.exe.7768.000\potential_errors.txt)
~~Dr.M~~     123 unique,   195 total,  12505 byte(s) of still-reachable
allocation(s)
~~Dr.M~~          (re-run with "-show_reachable" for details)
~~Dr.M~~ Details: C:\Users\fbarchard\AppData\Roaming\Dr.
Memory\DrMemory-libyuv_unittest.exe.7768.000\results.txt

https://codereview.chromium.org/1507683003/diff/1/source/planar_functions.cc
File source/planar_functions.cc (right):

https://codereview.chromium.org/1507683003/diff/1/source/planar_functions.cc#...
source/planar_functions.cc:724: // last row of odd height image use 1 row of
alpha instead of 2.
On 2015/12/07 18:40:24, Diony Rosa wrote:
> This skips the last row if the height is odd. e.g. if the height is 5, you'll
> get:
> 
> y = 0
> y = 2
> 
> I think you want to use halfheight here, and check if y == (halfheight - 1)
AND
> (halfheight & 1)

In the example of height = 5, y will be
0,
2,
4,
4 < 5, so it does this last loop width y = 4, height = 5;
The code should work as is but it adds an if to the Y loop.
Another way to do it is do the even number of rows

for (y = 0; y < height - 1; y += 2) {
}
then check for odd height, and do the last row with similar code but 1 row of
alpha.

perf shows the I420Blend function taking
2.15%  libyuv_unittest  libyuv_unittest    [.] libyuv::TestI420Blend(int, int,
int, int, int, int, int) [clone .isra.28]
which is calling overhead and this loop, for 720p, so hoisting the if and
duplicating the loop isnt worth the code size.

[fbarchard, 2015-12-07 19:15:06]

added odd width (any) and inverted image tests

out\release\libyuv_unittest --libyuv_width=1280 --libyuv_height=720
--libyuv_repeat=999 --libyuv_flags=-1 --gtest_filter=*Blend*
ARGBBlend_Any (626 ms)
ARGBBlend_Unaligned (631 ms)
ARGBBlend_Invert (499 ms)
ARGBBlend_Opt (494 ms)
BlendPlaneRow_Opt (167 ms)
BlendPlaneRow_Unaligned (179 ms)
BlendPlane_Opt (163 ms)
BlendPlane_Unaligned (183 ms)
BlendPlane_Any (177 ms)
BlendPlane_Invert (179 ms)
I420Blend_Opt (288 ms)
I420Blend_Unaligned (312 ms)
I420Blend_Any (1058 ms)
I420Blend_Invert (289 ms)

[Diony Rosa, 2015-12-07 19:45:15]

lgtm

[fbarchard1, 2015-12-07 20:03:27]

Description was changed from

==========
Add support for odd height YUVA alpha blending.

R=dhrosa@google.com, harryjin@google.com
BUG=libyuv:527
==========

to

==========
Add support for odd height YUVA alpha blending.

R=dhrosa@google.com, harryjin@google.com
BUG=libyuv:527

Committed:
https://chromium.googlesource.com/libyuv/libyuv/+/2657688e701709a5af935e6ea27...
==========

[fbarchard1, 2015-12-07 20:03:28]

Committed patchset #2 (id:20001) manually as
2657688e701709a5af935e6ea27f4f8967208f2d (presubmit successful).