Use aligned buffer operations in 8x8/16x16 2D-DCT
authorJingning Han <jingning@google.com>
Tue, 25 Jun 2013 02:52:55 +0000 (19:52 -0700)
committerJingning Han <jingning@google.com>
Tue, 25 Jun 2013 02:56:23 +0000 (19:56 -0700)
commit82d504b50f5dbc81ba1e1e1c1b07bb76dddde43f
treedab9cacdc48732c1d2fec435ee8ce1f16a8fe8a7
parenta32a086d23c2061344af7653892456bde3fffd0d
Use aligned buffer operations in 8x8/16x16 2D-DCT

This reduces 16x16 2D-DCT runtime from 865 cycles to 837 cycles.

Change-Id: I137758b81cd127b936175284310e81378db64552
vp9/encoder/x86/vp9_dct_sse2.c