Closed GoogleCodeExporter closed 9 years ago
r538 checkes width == stride and does single CopyRow() for full plane.
out\release\libyuv_unittest --gtest_filter=*I420ToI420_* | grep ms
Before
I420ToI420_Any (10768 ms)
I420ToI420_Unaligned (16541 ms)
I420ToI420_Invert (10117 ms)
I420ToI420_Opt (10157 ms)
After
I420ToI420_Any (8900 ms)
I420ToI420_Unaligned (13356 ms)
I420ToI420_Invert (10005 ms)
I420ToI420_Opt (9241 ms)
Original comment by fbarch...@google.com
on 16 Jan 2013 at 3:05
Linux (z430)
LIBYUV_DISABLE_ASM=1 LIBYUV_REPEAT=100000 out/Release/libyuv_unittest
--gtest_filter=*I420ToI420_* | grep ms
[ OK ] libyuvTest.I420ToI420_Any (10859 ms)
[ OK ] libyuvTest.I420ToI420_Unaligned (10979 ms)
[ OK ] libyuvTest.I420ToI420_Invert (9347 ms)
[ OK ] libyuvTest.I420ToI420_Opt (9134 ms)
[----------] 4 tests from libyuvTest (40319 ms total)
LIBYUV_REPEAT=100000 out/Release/libyuv_unittest --gtest_filter=*I420ToI420_* |
grep ms
[ OK ] libyuvTest.I420ToI420_Any (10102 ms)
[ OK ] libyuvTest.I420ToI420_Unaligned (13822 ms)
[ OK ] libyuvTest.I420ToI420_Invert (9516 ms)
[ OK ] libyuvTest.I420ToI420_Opt (9004 ms)
[----------] 4 tests from libyuvTest (42444 ms total)
[==========] 4 tests from 1 test case ran. (42444 ms total)
Original comment by fbarch...@google.com
on 16 Jan 2013 at 7:18
r552 avoids memcpy for mjpeg internals - calls CopyPlane, which coalesces rows
into one copy and optimizes for alignment and cpu
Original comment by fbarch...@chromium.org
on 26 Jan 2013 at 7:39
r555 cleans up memcpy/memset headers/lint.
Original comment by fbarch...@chromium.org
on 28 Jan 2013 at 8:05
r556 uses CopyRow for mips in all instances.
All known issues resolved.
Original comment by fbarch...@chromium.org
on 28 Jan 2013 at 11:13
Original issue reported on code.google.com by
fbarch...@google.com
on 16 Jan 2013 at 1:39