RFC: Reorganize MFQE loops
authorJohann <johannkoenig@google.com>
Fri, 27 Jan 2012 18:23:52 +0000 (10:23 -0800)
committerJohann <johannkoenig@google.com>
Tue, 6 Mar 2012 23:20:03 +0000 (15:20 -0800)
commitfd903902ef2c6fef505829dffa4aad6b4d85950e
tree5c2208962c89bbac383d99e7a91e515fb9a9f832
parente50f96a4a3c4974f784fa764a29c873a58cca150
RFC: Reorganize MFQE loops

Break MFQE code into it's own file.

It is currently only valid for 16x16 and 8x8 Y blocks. It also filters
4x4 U/V blocks.

Refactor filtering and add associated assembly. Limited test cases show
--mfqe introduces a penalty of ~20% with HD content. The assembly
reduces the penalty to ~15%

Change-Id: I4b8de6b5cdff5413037de5b6c42f437033ee55bf
vp8/common/asm_com_offsets.c
vp8/common/mfqe.c [new file with mode: 0644]
vp8/common/postproc.c
vp8/common/postproc.h
vp8/common/rtcd_defs.sh
vp8/common/x86/mfqe_sse2.asm [new file with mode: 0644]
vp8/vp8_common.mk