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>

1 files changed, 1 insertions, 10 deletions

diff --git a/libavcodec/h264_mb_template.c b/libavcodec/h264_mb_template.c
index b617029a9f..6c62a7608b 100644
--- a/libavcodec/h264_mb_template.c
+++ b/libavcodec/h264_mb_template.c
@@ -204,7 +204,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
                                                             h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
                                                             uvlinesize);
                 } else {
-                    idct_add = h->h264dsp.h264_add_pixels4;
+                    idct_add = h->h264dsp.h264_add_pixels4_clear;
                     for (j = 1; j < 3; j++) {
                         for (i = j * 16; i < j * 16 + 4; i++)
                             if (h->non_zero_count_cache[scan8[i]] ||
@@ -258,10 +258,6 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
                 }
             }
         }
-        if (h->cbp || IS_INTRA(mb_type)) {
-            h->dsp.clear_blocks(h->mb);
-            h->dsp.clear_blocks(h->mb + (24 * 16 << PIXEL_SHIFT));
-        }
     }
 }
 
@@ -365,11 +361,6 @@ static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h)
             hl_decode_mb_idct_luma(h, mb_type, 1, SIMPLE, transform_bypass,
                                    PIXEL_SHIFT, block_offset, linesize,
                                    dest[p], p);
-
-        if (h->cbp || IS_INTRA(mb_type)) {
-            h->dsp.clear_blocks(h->mb);
-            h->dsp.clear_blocks(h->mb + (24 * 16 << PIXEL_SHIFT));
-        }
     }
 }