h264: integrate clear_blocks calls with IDCT.
authorRonald S. Bultje <rsbultje@gmail.com>
Tue, 19 Feb 2013 05:03:02 +0000 (21:03 -0800)
committerMichael Niedermayer <michaelni@gmx.at>
Tue, 19 Feb 2013 15:25:50 +0000 (16:25 +0100)
commit1acd7d594c15aa491729c837ad3519d3469e620a
tree4a515aa31c7483d8a2fdfa1d4399c9a9d2b29cae
parenta1f1ca96b42698002920467c0aab9e636893088a
h264: integrate clear_blocks calls with IDCT.

The non-intra-pcm branch in hl_decode_mb (simple, 8bpp) goes from 700
to 672 cycles, and the complete loop of decode_mb_cabac and hl_decode_mb
(in the decode_slice loop) goes from 1759 to 1733 cycles on the clip
tested (cathedral), i.e. almost 30 cycles per mb faster.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
13 files changed:
libavcodec/arm/h264idct_neon.S
libavcodec/h264.c
libavcodec/h264_mb_template.c
libavcodec/h264addpx_template.c
libavcodec/h264dsp.c
libavcodec/h264dsp.h
libavcodec/h264idct_template.c
libavcodec/h264pred.h
libavcodec/h264pred_template.c
libavcodec/ppc/h264_altivec.c
libavcodec/svq3.c
libavcodec/x86/h264_idct.asm
libavcodec/x86/h264_idct_10bit.asm