~10% performance gain. Fixed the cosmetics noted in the previous commit. Change-Id: Iddf475f34d0d0a3e356b2143682aeabac459ed13
This version is ~1.91x faster than the sse2 version. When highbitdepth is enabled, it is ~1.74x. Change-Id: I2b0e92ede9f55c6259ca07bf1f8c8a5d0d0955bd