GitLab

AVX2 version of H010ToAR30_Opt is 42% faster on Broadwell.
Add ABGRToAR30 fully optimized for AVX2 and SSSE3.
Add AR30ToABGR, unoptimized,

1000 iterations of 720p:
Was SSSE3 H010ToAR30_Opt (635 ms)
Was SSSE3 I420ToAR30_Opt (635 ms)

Now AVX2  H010ToAR30_Opt (448 ms)
Now AVX2  I420ToAR30_Opt (446 ms)

ffec313 ABGRToAR30 used AVX2 with reversed shuffler
ff8ab9b AR30ToABGR for 10 to 8 bit RGB on Android
ed96b7b AVX2 port of H010ToAR30_AVX2
c95fd57 AVX2 port of I010ToAR30_AVX2
3f43ecc Add H420ToAR30 and a test that does a histogram

Bug: 797347, libyuv:751
Test: try bots passed for local libyuv build
Change-Id: I2ea965b65b6757dc31ab1cdb706dc02c68afeab3
Reviewed-on: https://chromium-review.googlesource.com/892340
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Reviewed-by: Miguel Casas <mcasas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#532876}