Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

Store data in direct_table interleaved.

Browse Source 
seems 20cpu cycles faster

Originally committed as revision 22055 to svn://svn.ffmpeg.org/ffmpeg/trunk
release/0.6
Michael Niedermayer 15 years ago
parent 013202d720
commit 5e350863cc
1 changed files with 9 additions and 9 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 18
      libavcodec/h264.h

18
libavcodec/h264.h
Unescape View File

@ -1088,9 +1088,9 @@ static void fill_decode_caches(H264Context *h, int mb_type){
if(IS_DIRECT(top_type)){ if(IS_DIRECT(top_type)){
AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_DIRECT2>>1)); AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_DIRECT2>>1));
}else if(IS_8X8(top_type)){ }else if(IS_8X8(top_type)){
int b8_xy = h->mb2b8_xy[top_xy] + h->b8_stride; int b8_xy = 4*top_xy;
h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy]; h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy + 2];
h->direct_cache[scan8[0] + 2 - 1*8]= h->direct_table[b8_xy + 1]; h->direct_cache[scan8[0] + 2 - 1*8]= h->direct_table[b8_xy + 3];
}else{ }else{
AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_16x16>>1)); AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_16x16>>1));
} }
@ -1098,14 +1098,14 @@ static void fill_decode_caches(H264Context *h, int mb_type){
if(IS_DIRECT(left_type[0])) if(IS_DIRECT(left_type[0]))
h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_DIRECT2>>1; h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_DIRECT2>>1;
else if(IS_8X8(left_type[0])) else if(IS_8X8(left_type[0]))
h->direct_cache[scan8[0] - 1 + 0*8]= h->direct_table[h->mb2b8_xy[left_xy[0]] + 1 + h->b8_stride*(left_block[0]>>1)]; h->direct_cache[scan8[0] - 1 + 0*8]= h->direct_table[4*left_xy[0] + 1 + (left_block[0]&~1)];
else else
h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_16x16>>1; h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_16x16>>1;
if(IS_DIRECT(left_type[1])) if(IS_DIRECT(left_type[1]))
h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_DIRECT2>>1; h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_DIRECT2>>1;
else if(IS_8X8(left_type[1])) else if(IS_8X8(left_type[1]))
h->direct_cache[scan8[0] - 1 + 2*8]= h->direct_table[h->mb2b8_xy[left_xy[1]] + 1 + h->b8_stride*(left_block[2]>>1)]; h->direct_cache[scan8[0] - 1 + 2*8]= h->direct_table[4*left_xy[1] + 1 + (left_block[2]&~1)];
else else
h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_16x16>>1; h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_16x16>>1;
} }
@ -1433,10 +1433,10 @@ static inline void write_back_motion(H264Context *h, int mb_type){
if(h->slice_type_nos == FF_B_TYPE && CABAC){ if(h->slice_type_nos == FF_B_TYPE && CABAC){
if(IS_8X8(mb_type)){ if(IS_8X8(mb_type)){
uint8_t *direct_table = &h->direct_table[b8_xy]; uint8_t *direct_table = &h->direct_table[4*h->mb_xy];
direct_table[1+0*h->b8_stride] = h->sub_mb_type[1]>>1; direct_table[1] = h->sub_mb_type[1]>>1;
direct_table[0+1*h->b8_stride] = h->sub_mb_type[2]>>1; direct_table[2] = h->sub_mb_type[2]>>1;
direct_table[1+1*h->b8_stride] = h->sub_mb_type[3]>>1; direct_table[3] = h->sub_mb_type[3]>>1;
} }
} }
} }

Write Preview
Loading…
Cancel
Save