diff options
| author | Ronald S. Bultje <rsbultje@gmail.com> | 2013-02-18 21:03:02 -0800 |
|---|---|---|
| committer | Martin Storsjö <martin@martin.st> | 2013-04-10 11:03:06 +0300 |
| commit | 62844c3fd66940c7747e9b2bb7804e265319f43f (patch) | |
| tree | b0e5a05644457aa5d7598d1fefa1a41f83550753 /libavcodec/h264pred_template.c | |
| parent | e8cafd2773bc56455c8816593cbd9368f2d69a80 (diff) | |
| download | ffmpeg-62844c3fd66940c7747e9b2bb7804e265319f43f.tar.gz | |
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: Martin Storsjö <martin@martin.st>
Diffstat (limited to 'libavcodec/h264pred_template.c')
| -rw-r--r-- | libavcodec/h264pred_template.c | 28 |
1 files changed, 18 insertions, 10 deletions
diff --git a/libavcodec/h264pred_template.c b/libavcodec/h264pred_template.c index 8e97d11104..e15c76c849 100644 --- a/libavcodec/h264pred_template.c +++ b/libavcodec/h264pred_template.c @@ -1131,7 +1131,7 @@ static void FUNCC(pred8x8l_horizontal_up)(uint8_t *_src, int has_topleft, #undef PL #undef SRC -static void FUNCC(pred4x4_vertical_add)(uint8_t *_pix, const int16_t *_block, +static void FUNCC(pred4x4_vertical_add)(uint8_t *_pix, int16_t *_block, ptrdiff_t stride) { int i; @@ -1148,9 +1148,11 @@ static void FUNCC(pred4x4_vertical_add)(uint8_t *_pix, const int16_t *_block, pix++; block++; } + + memset(_block, 0, sizeof(dctcoef) * 16); } -static void FUNCC(pred4x4_horizontal_add)(uint8_t *_pix, const int16_t *_block, +static void FUNCC(pred4x4_horizontal_add)(uint8_t *_pix, int16_t *_block, ptrdiff_t stride) { int i; @@ -1166,9 +1168,11 @@ static void FUNCC(pred4x4_horizontal_add)(uint8_t *_pix, const int16_t *_block, pix+= stride; block+= 4; } + + memset(_block, 0, sizeof(dctcoef) * 16); } -static void FUNCC(pred8x8l_vertical_add)(uint8_t *_pix, const int16_t *_block, +static void FUNCC(pred8x8l_vertical_add)(uint8_t *_pix, int16_t *_block, ptrdiff_t stride) { int i; @@ -1189,9 +1193,11 @@ static void FUNCC(pred8x8l_vertical_add)(uint8_t *_pix, const int16_t *_block, pix++; block++; } + + memset(_block, 0, sizeof(dctcoef) * 64); } -static void FUNCC(pred8x8l_horizontal_add)(uint8_t *_pix, const int16_t *_block, +static void FUNCC(pred8x8l_horizontal_add)(uint8_t *_pix, int16_t *_block, ptrdiff_t stride) { int i; @@ -1211,10 +1217,12 @@ static void FUNCC(pred8x8l_horizontal_add)(uint8_t *_pix, const int16_t *_block, pix+= stride; block+= 8; } + + memset(_block, 0, sizeof(dctcoef) * 64); } static void FUNCC(pred16x16_vertical_add)(uint8_t *pix, const int *block_offset, - const int16_t *block, + int16_t *block, ptrdiff_t stride) { int i; @@ -1224,7 +1232,7 @@ static void FUNCC(pred16x16_vertical_add)(uint8_t *pix, const int *block_offset, static void FUNCC(pred16x16_horizontal_add)(uint8_t *pix, const int *block_offset, - const int16_t *block, + int16_t *block, ptrdiff_t stride) { int i; @@ -1233,7 +1241,7 @@ static void FUNCC(pred16x16_horizontal_add)(uint8_t *pix, } static void FUNCC(pred8x8_vertical_add)(uint8_t *pix, const int *block_offset, - const int16_t *block, ptrdiff_t stride) + int16_t *block, ptrdiff_t stride) { int i; for(i=0; i<4; i++) @@ -1241,7 +1249,7 @@ static void FUNCC(pred8x8_vertical_add)(uint8_t *pix, const int *block_offset, } static void FUNCC(pred8x16_vertical_add)(uint8_t *pix, const int *block_offset, - const int16_t *block, ptrdiff_t stride) + int16_t *block, ptrdiff_t stride) { int i; for(i=0; i<4; i++) @@ -1251,7 +1259,7 @@ static void FUNCC(pred8x16_vertical_add)(uint8_t *pix, const int *block_offset, } static void FUNCC(pred8x8_horizontal_add)(uint8_t *pix, const int *block_offset, - const int16_t *block, + int16_t *block, ptrdiff_t stride) { int i; @@ -1261,7 +1269,7 @@ static void FUNCC(pred8x8_horizontal_add)(uint8_t *pix, const int *block_offset, static void FUNCC(pred8x16_horizontal_add)(uint8_t *pix, const int *block_offset, - const int16_t *block, ptrdiff_t stride) + int16_t *block, ptrdiff_t stride) { int i; for(i=0; i<4; i++) |