virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-01-24 13:56:33 +02:00
Code Issues Releases Activity

H.264: avoid redundant alpha/beta calculations in loopfilter

Browse Source
This commit is contained in:
Jason Garrett-Glaser 2011-07-10 02:25:53 -07:00
parent a4f6be86d6
commit f6b7f72461
1 changed files with 122 additions and 121 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

243
libavcodec/h264_loopfilter.c
View File

@ -101,11 +101,10 @@ static const uint8_t tc0_table[52*3][4] = {
}; };
/* intra: 0 if this loopfilter call is guaranteed to be inter (bS < 4), 1 if it might be intra (bS == 4) */ /* intra: 0 if this loopfilter call is guaranteed to be inter (bS < 4), 1 if it might be intra (bS == 4) */
static void av_always_inline filter_mb_edgev( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, H264Context *h, int intra ) { static void av_always_inline filter_mb_edgev( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra ) {
const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8); const unsigned int index_a = qp + a;
const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
const int alpha = alpha_table[index_a]; const int alpha = alpha_table[index_a];
const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset]; const int beta = beta_table[qp + b];
if (alpha ==0 || beta == 0) return; if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 || !intra ) { if( bS[0] < 4 || !intra ) {
@ -119,11 +118,10 @@ static void av_always_inline filter_mb_edgev( uint8_t *pix, int stride, const in
h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta); h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);
} }
} }
static void av_always_inline filter_mb_edgecv( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, H264Context *h, int intra ) { static void av_always_inline filter_mb_edgecv( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra ) {
const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8); const unsigned int index_a = qp + a;
const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
const int alpha = alpha_table[index_a]; const int alpha = alpha_table[index_a];
const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset]; const int beta = beta_table[qp + b];
if (alpha ==0 || beta == 0) return; if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 || !intra ) { if( bS[0] < 4 || !intra ) {
@ -138,11 +136,10 @@ static void av_always_inline filter_mb_edgecv( uint8_t *pix, int stride, const i
} }
} }
static void av_always_inline filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int intra ) { static void av_always_inline filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int a, int b, int intra ) {
const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8); const unsigned int index_a = qp + a;
int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset; const int alpha = alpha_table[index_a];
int alpha = alpha_table[index_a]; const int beta = beta_table[qp + b];
int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset];
if (alpha ==0 || beta == 0) return; if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 || !intra ) { if( bS[0] < 4 || !intra ) {
@ -156,11 +153,10 @@ static void av_always_inline filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix
h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta); h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta);
} }
} }
static void av_always_inline filter_mb_mbaff_edgecv( H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int intra ) { static void av_always_inline filter_mb_mbaff_edgecv( H264Context *h, uint8_t *pix, int stride, const int16_t bS[7], int bsi, int qp, int a, int b, int intra ) {
const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8); const unsigned int index_a = qp + a;
int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset; const int alpha = alpha_table[index_a];
int alpha = alpha_table[index_a]; const int beta = beta_table[qp + b];
int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset];
if (alpha ==0 || beta == 0) return; if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 || !intra ) { if( bS[0] < 4 || !intra ) {
@ -175,11 +171,10 @@ static void av_always_inline filter_mb_mbaff_edgecv( H264Context *h, uint8_t *pi
} }
} }
static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, H264Context *h, int intra ) { static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra ) {
const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8); const unsigned int index_a = qp + a;
const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
const int alpha = alpha_table[index_a]; const int alpha = alpha_table[index_a];
const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset]; const int beta = beta_table[qp + b];
if (alpha ==0 || beta == 0) return; if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 || !intra ) { if( bS[0] < 4 || !intra ) {
@ -194,11 +189,10 @@ static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, const in
} }
} }
static void av_always_inline filter_mb_edgech( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, H264Context *h, int intra ) { static void av_always_inline filter_mb_edgech( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra ) {
const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8); const unsigned int index_a = qp + a;
const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
const int alpha = alpha_table[index_a]; const int alpha = alpha_table[index_a];
const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset]; const int beta = beta_table[qp + b];
if (alpha ==0 || beta == 0) return; if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 || !intra ) { if( bS[0] < 4 || !intra ) {
@ -220,6 +214,7 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
int qp, qp0, qp1, qpc, qpc0, qpc1; int qp, qp0, qp1, qpc, qpc0, qpc1;
int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY)); int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
int chroma444 = CHROMA444; int chroma444 = CHROMA444;
int qp_bd_offset, a, b;
mb_xy = h->mb_xy; mb_xy = h->mb_xy;
@ -242,76 +237,79 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
qp1 = (qp + qp1 + 1) >> 1; qp1 = (qp + qp1 + 1) >> 1;
qpc0 = (qpc + qpc0 + 1) >> 1; qpc0 = (qpc + qpc0 + 1) >> 1;
qpc1 = (qpc + qpc1 + 1) >> 1; qpc1 = (qpc + qpc1 + 1) >> 1;
qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
a = h->slice_alpha_c0_offset - qp_bd_offset;
b = h->slice_beta_offset - qp_bd_offset;
if( IS_INTRA(mb_type) ) { if( IS_INTRA(mb_type) ) {
static const int16_t bS4[4] = {4,4,4,4}; static const int16_t bS4[4] = {4,4,4,4};
static const int16_t bS3[4] = {3,3,3,3}; static const int16_t bS3[4] = {3,3,3,3};
const int16_t *bSH = FIELD_PICTURE ? bS3 : bS4; const int16_t *bSH = FIELD_PICTURE ? bS3 : bS4;
if(left_type) if(left_type)
filter_mb_edgev( &img_y[4*0], linesize, bS4, qp0, h, 1); filter_mb_edgev( &img_y[4*0], linesize, bS4, qp0, a, b, h, 1);
if( IS_8x8DCT(mb_type) ) { if( IS_8x8DCT(mb_type) ) {
filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, h, 0); filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, a, b, h, 0);
if(top_type){ if(top_type){
filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, h, 1); filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);
} }
filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, h, 0); filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);
} else { } else {
filter_mb_edgev( &img_y[4*1], linesize, bS3, qp, h, 0); filter_mb_edgev( &img_y[4*1], linesize, bS3, qp, a, b, h, 0);
filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, h, 0); filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, a, b, h, 0);
filter_mb_edgev( &img_y[4*3], linesize, bS3, qp, h, 0); filter_mb_edgev( &img_y[4*3], linesize, bS3, qp, a, b, h, 0);
if(top_type){ if(top_type){
filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, h, 1); filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);
} }
filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, h, 0); filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, a, b, h, 0);
filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, h, 0); filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);
filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, h, 0); filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, a, b, h, 0);
} }
if(chroma){ if(chroma){
if(chroma444){ if(chroma444){
if(left_type){ if(left_type){
filter_mb_edgev( &img_cb[4*0], linesize, bS4, qpc0, h, 1); filter_mb_edgev( &img_cb[4*0], linesize, bS4, qpc0, a, b, h, 1);
filter_mb_edgev( &img_cr[4*0], linesize, bS4, qpc0, h, 1); filter_mb_edgev( &img_cr[4*0], linesize, bS4, qpc0, a, b, h, 1);
} }
if( IS_8x8DCT(mb_type) ) { if( IS_8x8DCT(mb_type) ) {
filter_mb_edgev( &img_cb[4*2], linesize, bS3, qpc, h, 0); filter_mb_edgev( &img_cb[4*2], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgev( &img_cr[4*2], linesize, bS3, qpc, h, 0); filter_mb_edgev( &img_cr[4*2], linesize, bS3, qpc, a, b, h, 0);
if(top_type){ if(top_type){
filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, h,1 ); filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );
filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, h,1 ); filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );
} }
filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, h, 0); filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, h, 0); filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
} else { } else {
filter_mb_edgev( &img_cb[4*1], linesize, bS3, qpc, h, 0); filter_mb_edgev( &img_cb[4*1], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgev( &img_cr[4*1], linesize, bS3, qpc, h, 0); filter_mb_edgev( &img_cr[4*1], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgev( &img_cb[4*2], linesize, bS3, qpc, h, 0); filter_mb_edgev( &img_cb[4*2], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgev( &img_cr[4*2], linesize, bS3, qpc, h, 0); filter_mb_edgev( &img_cr[4*2], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgev( &img_cb[4*3], linesize, bS3, qpc, h, 0); filter_mb_edgev( &img_cb[4*3], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgev( &img_cr[4*3], linesize, bS3, qpc, h, 0); filter_mb_edgev( &img_cr[4*3], linesize, bS3, qpc, a, b, h, 0);
if(top_type){ if(top_type){
filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, h, 1); filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);
filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, h, 1); filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);
} }
filter_mb_edgeh( &img_cb[4*1*linesize], linesize, bS3, qpc, h, 0); filter_mb_edgeh( &img_cb[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgeh( &img_cr[4*1*linesize], linesize, bS3, qpc, h, 0); filter_mb_edgeh( &img_cr[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, h, 0); filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, h, 0); filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, h, 0); filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, h, 0); filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
} }
}else{ }else{
if(left_type){ if(left_type){
filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, h, 1); filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, a, b, h, 1);
filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, h, 1); filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, a, b, h, 1);
} }
filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, h, 0); filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, a, b, h, 0);
filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, h, 0); filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, a, b, h, 0);
if(top_type){ if(top_type){
filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, h, 1); filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, h, 1); filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
} }
filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, h, 0); filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, h, 0); filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
} }
} }
return; return;
@ -339,14 +337,14 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
#define FILTER(hv,dir,edge,intra)\ #define FILTER(hv,dir,edge,intra)\
if(AV_RN64A(bS[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, intra );\ filter_mb_edge##hv( &img_y[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qp : qp##dir, a, b, h, intra );\
if(chroma){\ if(chroma){\
if(chroma444){\ if(chroma444){\
filter_mb_edge##hv( &img_cb[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, h, intra );\ filter_mb_edge##hv( &img_cb[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
filter_mb_edge##hv( &img_cr[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, h, intra );\ filter_mb_edge##hv( &img_cr[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
} else if(!(edge&1)) {\ } else if(!(edge&1)) {\
filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h, intra );\ filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
filter_mb_edgec##hv( &img_cr[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, h, intra );\ filter_mb_edgec##hv( &img_cr[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
}\ }\
}\ }\
} }
@ -403,7 +401,7 @@ static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){
return v; return v;
} }
static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int chroma, int chroma444, int dir) { static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int a, int b, int chroma, int chroma444, int dir) {
MpegEncContext * const s = &h->s; MpegEncContext * const s = &h->s;
int edge; int edge;
int chroma_qp_avg[2]; int chroma_qp_avg[2];
@ -457,16 +455,16 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbn_xy] + 1) >> 1; qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbn_xy] + 1) >> 1;
tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize); tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize);
{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); } { int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, h, 0 ); filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, a, b, h, 0 );
chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1; chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;
chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1; chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;
if (chroma) { if (chroma) {
if (chroma444) { if (chroma444) {
filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], h, 0); filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h, 0); filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);
} else { } else {
filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], h, 0); filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h, 0); filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);
} }
} }
} }
@ -526,25 +524,25 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1; chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;
chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1; chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;
if( dir == 0 ) { if( dir == 0 ) {
filter_mb_edgev( &img_y[0], linesize, bS, qp, h, 1 ); filter_mb_edgev( &img_y[0], linesize, bS, qp, a, b, h, 1 );
if (chroma) { if (chroma) {
if (chroma444) { if (chroma444) {
filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h, 1); filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h, 1); filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
} else { } else {
filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h, 1); filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h, 1); filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
} }
} }
} else { } else {
filter_mb_edgeh( &img_y[0], linesize, bS, qp, h, 1 ); filter_mb_edgeh( &img_y[0], linesize, bS, qp, a, b, h, 1 );
if (chroma) { if (chroma) {
if (chroma444) { if (chroma444) {
filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h, 1); filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h, 1); filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
} else { } else {
filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h, 1); filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], h, 1); filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
} }
} }
} }
@ -608,25 +606,25 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize); tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
//{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); } //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
if( dir == 0 ) { if( dir == 0 ) {
filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, h, 0 ); filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, a, b, h, 0 );
if (chroma) { if (chroma) {
if (chroma444) { if (chroma444) {
filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h, 0); filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h, 0); filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
} else if( (edge&1) == 0 ) { } else if( (edge&1) == 0 ) {
filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h, 0); filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h, 0); filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
} }
} }
} else { } else {
filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, h, 0 ); filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0 );
if (chroma) { if (chroma) {
if (chroma444) { if (chroma444) {
filter_mb_edgeh ( &img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h, 0); filter_mb_edgeh ( &img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
filter_mb_edgeh ( &img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h, 0); filter_mb_edgeh ( &img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
} else if( (edge&1) == 0 ) { } else if( (edge&1) == 0 ) {
filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], h, 0); filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], h, 0); filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
} }
} }
} }
@ -641,6 +639,9 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
int first_vertical_edge_done = 0; int first_vertical_edge_done = 0;
av_unused int dir; av_unused int dir;
int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY)); int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
int a = h->slice_alpha_c0_offset - qp_bd_offset;
int b = h->slice_beta_offset - qp_bd_offset;
if (FRAME_MBAFF if (FRAME_MBAFF
// and current and left pair do not have the same interlaced type // and current and left pair do not have the same interlaced type
@ -707,35 +708,35 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
tprintf(s->avctx, "filter mb:%d/%d MBAFF, QPy:%d/%d, QPb:%d/%d QPr:%d/%d ls:%d uvls:%d", mb_x, mb_y, qp[0], qp[1], bqp[0], bqp[1], rqp[0], rqp[1], linesize, uvlinesize); tprintf(s->avctx, "filter mb:%d/%d MBAFF, QPy:%d/%d, QPb:%d/%d QPr:%d/%d ls:%d uvls:%d", mb_x, mb_y, qp[0], qp[1], bqp[0], bqp[1], rqp[0], rqp[1], linesize, uvlinesize);
{ int i; for (i = 0; i < 8; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); } { int i; for (i = 0; i < 8; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
if(MB_FIELD){ if(MB_FIELD){
filter_mb_mbaff_edgev ( h, img_y , linesize, bS , 1, qp [0], 1 ); filter_mb_mbaff_edgev ( h, img_y , linesize, bS , 1, qp [0], a, b, 1 );
filter_mb_mbaff_edgev ( h, img_y + 8* linesize, linesize, bS+4, 1, qp [1], 1 ); filter_mb_mbaff_edgev ( h, img_y + 8* linesize, linesize, bS+4, 1, qp [1], a, b, 1 );
if (chroma){ if (chroma){
if (CHROMA444) { if (CHROMA444) {
filter_mb_mbaff_edgev ( h, img_cb, uvlinesize, bS , 1, bqp[0], 1 ); filter_mb_mbaff_edgev ( h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1 );
filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], 1 ); filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
filter_mb_mbaff_edgev ( h, img_cr, uvlinesize, bS , 1, rqp[0], 1 ); filter_mb_mbaff_edgev ( h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1 );
filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], 1 ); filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
}else{ }else{
filter_mb_mbaff_edgecv( h, img_cb, uvlinesize, bS , 1, bqp[0], 1 ); filter_mb_mbaff_edgecv( h, img_cb, uvlinesize, bS , 1, bqp[0], a, b, 1 );
filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], 1 ); filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
filter_mb_mbaff_edgecv( h, img_cr, uvlinesize, bS , 1, rqp[0], 1 ); filter_mb_mbaff_edgecv( h, img_cr, uvlinesize, bS , 1, rqp[0], a, b, 1 );
filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1], 1 ); filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
} }
} }
}else{ }else{
filter_mb_mbaff_edgev ( h, img_y , 2* linesize, bS , 2, qp [0], 1 ); filter_mb_mbaff_edgev ( h, img_y , 2* linesize, bS , 2, qp [0], a, b, 1 );
filter_mb_mbaff_edgev ( h, img_y + linesize, 2* linesize, bS+1, 2, qp [1], 1 ); filter_mb_mbaff_edgev ( h, img_y + linesize, 2* linesize, bS+1, 2, qp [1], a, b, 1 );
if (chroma){ if (chroma){
if (CHROMA444) { if (CHROMA444) {
filter_mb_mbaff_edgev ( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], 1 ); filter_mb_mbaff_edgev ( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], a, b, 1 );
filter_mb_mbaff_edgev ( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], 1 ); filter_mb_mbaff_edgev ( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );
filter_mb_mbaff_edgev ( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], 1 ); filter_mb_mbaff_edgev ( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], a, b, 1 );
filter_mb_mbaff_edgev ( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], 1 ); filter_mb_mbaff_edgev ( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );
}else{ }else{
filter_mb_mbaff_edgecv( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], 1 ); filter_mb_mbaff_edgecv( h, img_cb, 2*uvlinesize, bS , 2, bqp[0], a, b, 1 );
filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], 1 ); filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );
filter_mb_mbaff_edgecv( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], 1 ); filter_mb_mbaff_edgecv( h, img_cr, 2*uvlinesize, bS , 2, rqp[0], a, b, 1 );
filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], 1 ); filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );
} }
} }
} }
@ -743,9 +744,9 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint
#if CONFIG_SMALL #if CONFIG_SMALL
for( dir = 0; dir < 2; dir++ ) for( dir = 0; dir < 2; dir++ )
filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, chroma, CHROMA444, dir); filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, a, b, chroma, CHROMA444, dir);
#else #else
filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, chroma, CHROMA444, 0); filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, a, b, chroma, CHROMA444, 0);
filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, chroma, CHROMA444, 1); filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, a, b, chroma, CHROMA444, 1);
#endif #endif
} }