tree 7505fbddb477813e7aafdf5613a1eb62ad136e4d
parent c0ff7a3b07b5de129a549251a49a8b87a5a813b7
author Peng Bin <binpengsmail@gmail.com> 1520595652 +0800
committer Zoe Liu <zoeliu@google.com> 1520866987 +0000

Implement  aom_variance{128,64,32}_sse2/avx2

Add sse2 and avx2 version of aom_variance for size with 128
128x128, 128x64, 64x128, 128x32, 32x128

The speed test in  unittest shows sse2 version is 5x~6x
faster than C version, and avx2 version is 9x~11x faster
For encoder time, 20 frames of foreman_cip shows 2% speedup.

Change-Id: Ibf91514f6c876d47d56081fab54614f7c3609666