ref: 82d504b50f5dbc81ba1e1e1c1b07bb76dddde43f
parent: a32a086d23c2061344af7653892456bde3fffd0d
author: Jingning Han <jingning@google.com>
date: Mon Jun 24 15:52:55 EDT 2013
Use aligned buffer operations in 8x8/16x16 2D-DCT This reduces 16x16 2D-DCT runtime from 865 cycles to 837 cycles. Change-Id: I137758b81cd127b936175284310e81378db64552