Mercurial > libavcodec.hg
view h263.c @ 10852:86d7ab878805 libavcodec
Get rid of #include "svq3.c"
functions called more than per mb are moved into the header, scan8 is also
as it must be known at compiletime.
The code after this patch duplicates h264data.h, this has been done to minimize
the changes in this step and allow more fine grained benchmarking.
Speedwise this is 1% faster on my pentium dual core with diegos cursed cathedral
sample.
author | michael |
---|---|
date | Tue, 12 Jan 2010 05:30:31 +0000 |
parents | d0657e337f91 |
children | 31033caa5344 |
line wrap: on
line source
/* * H263/MPEG4 backend for ffmpeg encoder and decoder * Copyright (c) 2000,2001 Fabrice Bellard * H263+ support. * Copyright (c) 2001 Juan J. Sierralta P * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at> * * This file is part of FFmpeg. * * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ /** * @file libavcodec/h263.c * h263/mpeg4 codec. */ //#define DEBUG #include <limits.h> #include "dsputil.h" #include "avcodec.h" #include "mpegvideo.h" #include "h263.h" #include "h263data.h" #include "mathops.h" #include "unary.h" #include "flv.h" #include "mpeg4video.h" //#undef NDEBUG //#include <assert.h> uint8_t ff_h263_static_rl_table_store[2][2][2*MAX_RUN + MAX_LEVEL + 3]; void ff_h263_update_motion_val(MpegEncContext * s){ const int mb_xy = s->mb_y * s->mb_stride + s->mb_x; //FIXME a lot of that is only needed for !low_delay const int wrap = s->b8_stride; const int xy = s->block_index[0]; s->current_picture.mbskip_table[mb_xy]= s->mb_skipped; if(s->mv_type != MV_TYPE_8X8){ int motion_x, motion_y; if (s->mb_intra) { motion_x = 0; motion_y = 0; } else if (s->mv_type == MV_TYPE_16X16) { motion_x = s->mv[0][0][0]; motion_y = s->mv[0][0][1]; } else /*if (s->mv_type == MV_TYPE_FIELD)*/ { int i; motion_x = s->mv[0][0][0] + s->mv[0][1][0]; motion_y = s->mv[0][0][1] + s->mv[0][1][1]; motion_x = (motion_x>>1) | (motion_x&1); for(i=0; i<2; i++){ s->p_field_mv_table[i][0][mb_xy][0]= s->mv[0][i][0]; s->p_field_mv_table[i][0][mb_xy][1]= s->mv[0][i][1]; } s->current_picture.ref_index[0][xy ]= s->current_picture.ref_index[0][xy + 1]= s->field_select[0][0]; s->current_picture.ref_index[0][xy + wrap ]= s->current_picture.ref_index[0][xy + wrap + 1]= s->field_select[0][1]; } /* no update if 8X8 because it has been done during parsing */ s->current_picture.motion_val[0][xy][0] = motion_x; s->current_picture.motion_val[0][xy][1] = motion_y; s->current_picture.motion_val[0][xy + 1][0] = motion_x; s->current_picture.motion_val[0][xy + 1][1] = motion_y; s->current_picture.motion_val[0][xy + wrap][0] = motion_x; s->current_picture.motion_val[0][xy + wrap][1] = motion_y; s->current_picture.motion_val[0][xy + 1 + wrap][0] = motion_x; s->current_picture.motion_val[0][xy + 1 + wrap][1] = motion_y; } if(s->encoding){ //FIXME encoding MUST be cleaned up if (s->mv_type == MV_TYPE_8X8) s->current_picture.mb_type[mb_xy]= MB_TYPE_L0 | MB_TYPE_8x8; else if(s->mb_intra) s->current_picture.mb_type[mb_xy]= MB_TYPE_INTRA; else s->current_picture.mb_type[mb_xy]= MB_TYPE_L0 | MB_TYPE_16x16; } } int h263_pred_dc(MpegEncContext * s, int n, int16_t **dc_val_ptr) { int x, y, wrap, a, c, pred_dc; int16_t *dc_val; /* find prediction */ if (n < 4) { x = 2 * s->mb_x + (n & 1); y = 2 * s->mb_y + ((n & 2) >> 1); wrap = s->b8_stride; dc_val = s->dc_val[0]; } else { x = s->mb_x; y = s->mb_y; wrap = s->mb_stride; dc_val = s->dc_val[n - 4 + 1]; } /* B C * A X */ a = dc_val[(x - 1) + (y) * wrap]; c = dc_val[(x) + (y - 1) * wrap]; /* No prediction outside GOB boundary */ if(s->first_slice_line && n!=3){ if(n!=2) c= 1024; if(n!=1 && s->mb_x == s->resync_mb_x) a= 1024; } /* just DC prediction */ if (a != 1024 && c != 1024) pred_dc = (a + c) >> 1; else if (a != 1024) pred_dc = a; else pred_dc = c; /* we assume pred is positive */ *dc_val_ptr = &dc_val[x + y * wrap]; return pred_dc; } void ff_h263_loop_filter(MpegEncContext * s){ int qp_c; const int linesize = s->linesize; const int uvlinesize= s->uvlinesize; const int xy = s->mb_y * s->mb_stride + s->mb_x; uint8_t *dest_y = s->dest[0]; uint8_t *dest_cb= s->dest[1]; uint8_t *dest_cr= s->dest[2]; // if(s->pict_type==FF_B_TYPE && !s->readable) return; /* Diag Top Left Center */ if(!IS_SKIP(s->current_picture.mb_type[xy])){ qp_c= s->qscale; s->dsp.h263_v_loop_filter(dest_y+8*linesize , linesize, qp_c); s->dsp.h263_v_loop_filter(dest_y+8*linesize+8, linesize, qp_c); }else qp_c= 0; if(s->mb_y){ int qp_dt, qp_tt, qp_tc; if(IS_SKIP(s->current_picture.mb_type[xy-s->mb_stride])) qp_tt=0; else qp_tt= s->current_picture.qscale_table[xy-s->mb_stride]; if(qp_c) qp_tc= qp_c; else qp_tc= qp_tt; if(qp_tc){ const int chroma_qp= s->chroma_qscale_table[qp_tc]; s->dsp.h263_v_loop_filter(dest_y , linesize, qp_tc); s->dsp.h263_v_loop_filter(dest_y+8, linesize, qp_tc); s->dsp.h263_v_loop_filter(dest_cb , uvlinesize, chroma_qp); s->dsp.h263_v_loop_filter(dest_cr , uvlinesize, chroma_qp); } if(qp_tt) s->dsp.h263_h_loop_filter(dest_y-8*linesize+8 , linesize, qp_tt); if(s->mb_x){ if(qp_tt || IS_SKIP(s->current_picture.mb_type[xy-1-s->mb_stride])) qp_dt= qp_tt; else qp_dt= s->current_picture.qscale_table[xy-1-s->mb_stride]; if(qp_dt){ const int chroma_qp= s->chroma_qscale_table[qp_dt]; s->dsp.h263_h_loop_filter(dest_y -8*linesize , linesize, qp_dt); s->dsp.h263_h_loop_filter(dest_cb-8*uvlinesize, uvlinesize, chroma_qp); s->dsp.h263_h_loop_filter(dest_cr-8*uvlinesize, uvlinesize, chroma_qp); } } } if(qp_c){ s->dsp.h263_h_loop_filter(dest_y +8, linesize, qp_c); if(s->mb_y + 1 == s->mb_height) s->dsp.h263_h_loop_filter(dest_y+8*linesize+8, linesize, qp_c); } if(s->mb_x){ int qp_lc; if(qp_c || IS_SKIP(s->current_picture.mb_type[xy-1])) qp_lc= qp_c; else qp_lc= s->current_picture.qscale_table[xy-1]; if(qp_lc){ s->dsp.h263_h_loop_filter(dest_y, linesize, qp_lc); if(s->mb_y + 1 == s->mb_height){ const int chroma_qp= s->chroma_qscale_table[qp_lc]; s->dsp.h263_h_loop_filter(dest_y +8* linesize, linesize, qp_lc); s->dsp.h263_h_loop_filter(dest_cb , uvlinesize, chroma_qp); s->dsp.h263_h_loop_filter(dest_cr , uvlinesize, chroma_qp); } } } } void h263_pred_acdc(MpegEncContext * s, DCTELEM *block, int n) { int x, y, wrap, a, c, pred_dc, scale, i; int16_t *dc_val, *ac_val, *ac_val1; /* find prediction */ if (n < 4) { x = 2 * s->mb_x + (n & 1); y = 2 * s->mb_y + (n>> 1); wrap = s->b8_stride; dc_val = s->dc_val[0]; ac_val = s->ac_val[0][0]; scale = s->y_dc_scale; } else { x = s->mb_x; y = s->mb_y; wrap = s->mb_stride; dc_val = s->dc_val[n - 4 + 1]; ac_val = s->ac_val[n - 4 + 1][0]; scale = s->c_dc_scale; } ac_val += ((y) * wrap + (x)) * 16; ac_val1 = ac_val; /* B C * A X */ a = dc_val[(x - 1) + (y) * wrap]; c = dc_val[(x) + (y - 1) * wrap]; /* No prediction outside GOB boundary */ if(s->first_slice_line && n!=3){ if(n!=2) c= 1024; if(n!=1 && s->mb_x == s->resync_mb_x) a= 1024; } if (s->ac_pred) { pred_dc = 1024; if (s->h263_aic_dir) { /* left prediction */ if (a != 1024) { ac_val -= 16; for(i=1;i<8;i++) { block[s->dsp.idct_permutation[i<<3]] += ac_val[i]; } pred_dc = a; } } else { /* top prediction */ if (c != 1024) { ac_val -= 16 * wrap; for(i=1;i<8;i++) { block[s->dsp.idct_permutation[i ]] += ac_val[i + 8]; } pred_dc = c; } } } else { /* just DC prediction */ if (a != 1024 && c != 1024) pred_dc = (a + c) >> 1; else if (a != 1024) pred_dc = a; else pred_dc = c; } /* we assume pred is positive */ block[0]=block[0]*scale + pred_dc; if (block[0] < 0) block[0] = 0; else block[0] |= 1; /* Update AC/DC tables */ dc_val[(x) + (y) * wrap] = block[0]; /* left copy */ for(i=1;i<8;i++) ac_val1[i ] = block[s->dsp.idct_permutation[i<<3]]; /* top copy */ for(i=1;i<8;i++) ac_val1[8 + i] = block[s->dsp.idct_permutation[i ]]; } int16_t *h263_pred_motion(MpegEncContext * s, int block, int dir, int *px, int *py) { int wrap; int16_t *A, *B, *C, (*mot_val)[2]; static const int off[4]= {2, 1, 1, -1}; wrap = s->b8_stride; mot_val = s->current_picture.motion_val[dir] + s->block_index[block]; A = mot_val[ - 1]; /* special case for first (slice) line */ if (s->first_slice_line && block<3) { // we can't just change some MVs to simulate that as we need them for the B frames (and ME) // and if we ever support non rectangular objects than we need to do a few ifs here anyway :( if(block==0){ //most common case if(s->mb_x == s->resync_mb_x){ //rare *px= *py = 0; }else if(s->mb_x + 1 == s->resync_mb_x && s->h263_pred){ //rare C = mot_val[off[block] - wrap]; if(s->mb_x==0){ *px = C[0]; *py = C[1]; }else{ *px = mid_pred(A[0], 0, C[0]); *py = mid_pred(A[1], 0, C[1]); } }else{ *px = A[0]; *py = A[1]; } }else if(block==1){ if(s->mb_x + 1 == s->resync_mb_x && s->h263_pred){ //rare C = mot_val[off[block] - wrap]; *px = mid_pred(A[0], 0, C[0]); *py = mid_pred(A[1], 0, C[1]); }else{ *px = A[0]; *py = A[1]; } }else{ /* block==2*/ B = mot_val[ - wrap]; C = mot_val[off[block] - wrap]; if(s->mb_x == s->resync_mb_x) //rare A[0]=A[1]=0; *px = mid_pred(A[0], B[0], C[0]); *py = mid_pred(A[1], B[1], C[1]); } } else { B = mot_val[ - wrap]; C = mot_val[off[block] - wrap]; *px = mid_pred(A[0], B[0], C[0]); *py = mid_pred(A[1], B[1], C[1]); } return *mot_val; } /** * Get the GOB height based on picture height. */ int ff_h263_get_gob_height(MpegEncContext *s){ if (s->height <= 400) return 1; else if (s->height <= 800) return 2; else return 4; }