eessppeelllloo/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2025-12-13 02:20:07 +01:00
Code Issues Packages Projects Releases Wiki Activity

H264: use alias-safe macros

Browse Source 
This eliminates all aliasing violation warnings in h264 code.
No measurable speed difference with gcc-4.4.3 on i7.

Originally committed as revision 21881 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
Måns Rullgård
2010-02-18 16:24:31 +00:00
parent f4a7434f16
commit 19769ece3b
5 changed files with 86 additions and 82 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
libavcodec/h264_loopfilter.c
View File

@@ -25,6 +25,7 @@
* @author Michael Niedermayer <michaelni@gmx.at>
*/
#include "libavutil/intreadwrite.h"
#include "internal.h"
#include "dsputil.h"
#include "avcodec.h"
@@ -368,11 +369,13 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
return;
} else {
LOCAL_ALIGNED_8(int16_t, bS, [2], [4][4]);
uint64_t (*bSv)[4] = (uint64_t(*)[4])bS;
int edges;
if( IS_8x8DCT(mb_type) && (h->cbp&7) == 7 ) {
edges = 4;
bSv[0][0] = bSv[0][2] = bSv[1][0] = bSv[1][2] = 0x0002000200020002ULL;
AV_WN64A(bS[0][0], 0x0002000200020002ULL);
AV_WN64A(bS[0][2], 0x0002000200020002ULL);
AV_WN64A(bS[1][0], 0x0002000200020002ULL);
AV_WN64A(bS[1][2], 0x0002000200020002ULL);
} else {
int mask_edge1 = (3*(((5*mb_type)>>5)&1)) | (mb_type>>4); //(mb_type & (MB_TYPE_16x16 | MB_TYPE_8x16)) ? 3 : (mb_type & MB_TYPE_16x8) ? 1 : 0;
int mask_edge0 = 3*((mask_edge1>>1) & ((5*left_type)>>5)&1); // (mb_type & (MB_TYPE_16x16 | MB_TYPE_8x16)) && (h->left_type[0] & (MB_TYPE_16x16 | MB_TYPE_8x16)) ? 3 : 0;
@@ -382,12 +385,12 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
h->list_count==2, edges, step, mask_edge0, mask_edge1, FIELD_PICTURE);
}
if( IS_INTRA(left_type) )
bSv[0][0] = 0x0004000400040004ULL;
AV_WN64A(bS[0][0], 0x0004000400040004ULL);
if( IS_INTRA(h->top_type) )
bSv[1][0] = FIELD_PICTURE ? 0x0003000300030003ULL : 0x0004000400040004ULL;
AV_WN64A(bS[1][0], FIELD_PICTURE ? 0x0003000300030003ULL : 0x0004000400040004ULL);
#define FILTER(hv,dir,edge)\
if(bSv[dir][edge]) {\
if(AV_RN64A(bS[dir][edge])) { \
filter_mb_edge##hv( &img_y[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qp : qp##dir, h );\
if(!(edge&1)) {\
filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h );\
@@ -477,7 +480,7 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
DECLARE_ALIGNED_8(int16_t, bS)[4];
int qp;
if( IS_INTRA(mb_type|s->current_picture.mb_type[mbn_xy]) ) {
*(uint64_t*)bS= 0x0003000300030003ULL;
AV_WN64A(bS, 0x0003000300030003ULL);
} else {
if(!CABAC && IS_8x8DCT(s->current_picture.mb_type[mbn_xy])){
bS[0]= 1+((h->cbp_table[mbn_xy] & 4)||h->non_zero_count_cache[scan8[0]+0]);
@@ -508,17 +511,17 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
int qp;
if( IS_INTRA(mb_type|mbm_type)) {
*(uint64_t*)bS= 0x0003000300030003ULL;
AV_WN64A(bS, 0x0003000300030003ULL);
if ( (!IS_INTERLACED(mb_type|mbm_type))
|| ((FRAME_MBAFF || (s->picture_structure != PICT_FRAME)) && (dir == 0))
)
*(uint64_t*)bS= 0x0004000400040004ULL;
AV_WN64A(bS, 0x0004000400040004ULL);
} else {
int i;
int mv_done;
if( dir && FRAME_MBAFF && IS_INTERLACED(mb_type ^ mbm_type)) {
*(uint64_t*)bS= 0x0001000100010001ULL;
AV_WN64A(bS, 0x0001000100010001ULL);
mv_done = 1;
}
else if( mask_par0 && ((mbm_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)))) ) {
@@ -588,13 +591,13 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
continue;
if( IS_INTRA(mb_type)) {
*(uint64_t*)bS= 0x0003000300030003ULL;
AV_WN64A(bS, 0x0003000300030003ULL);
} else {
int i;
int mv_done;
if( edge & mask_edge ) {
*(uint64_t*)bS= 0;
AV_ZERO64(bS);
mv_done = 1;
}
else if( mask_par0 ) {
@@ -674,10 +677,10 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
int i;
first_vertical_edge_done = 1;
if( IS_INTRA(mb_type) )
*(uint64_t*)&bS[0]=
*(uint64_t*)&bS[4]= 0x0004000400040004ULL;
else {
if( IS_INTRA(mb_type) ) {
AV_WN64A(&bS[0], 0x0004000400040004ULL);
AV_WN64A(&bS[4], 0x0004000400040004ULL);
} else {
static const uint8_t offset[2][2][8]={
{
{7+8*0, 7+8*0, 7+8*0, 7+8*0, 7+8*1, 7+8*1, 7+8*1, 7+8*1},