Libav 0.7.1
|
00001 /* 00002 * The simplest mpeg encoder (well, it was the simplest!) 00003 * Copyright (c) 2000,2001 Fabrice Bellard 00004 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at> 00005 * 00006 * 4MV & hq & B-frame encoding stuff by Michael Niedermayer <michaelni@gmx.at> 00007 * 00008 * This file is part of Libav. 00009 * 00010 * Libav is free software; you can redistribute it and/or 00011 * modify it under the terms of the GNU Lesser General Public 00012 * License as published by the Free Software Foundation; either 00013 * version 2.1 of the License, or (at your option) any later version. 00014 * 00015 * Libav is distributed in the hope that it will be useful, 00016 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00017 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00018 * Lesser General Public License for more details. 00019 * 00020 * You should have received a copy of the GNU Lesser General Public 00021 * License along with Libav; if not, write to the Free Software 00022 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00023 */ 00024 00030 #ifndef AVCODEC_MPEGVIDEO_COMMON_H 00031 #define AVCODEC_MPEGVIDEO_COMMON_H 00032 00033 #include <string.h> 00034 #include "avcodec.h" 00035 #include "dsputil.h" 00036 #include "mpegvideo.h" 00037 #include "mjpegenc.h" 00038 #include "msmpeg4.h" 00039 #include "faandct.h" 00040 #include <limits.h> 00041 00042 int dct_quantize_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow); 00043 00048 int alloc_picture(MpegEncContext *s, Picture *pic, int shared); 00049 00054 void MPV_common_defaults(MpegEncContext *s); 00055 00056 static inline void gmc1_motion(MpegEncContext *s, 00057 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, 00058 uint8_t **ref_picture) 00059 { 00060 uint8_t *ptr; 00061 int offset, src_x, src_y, linesize, uvlinesize; 00062 int motion_x, motion_y; 00063 int emu=0; 00064 00065 motion_x= s->sprite_offset[0][0]; 00066 motion_y= s->sprite_offset[0][1]; 00067 src_x = s->mb_x * 16 + (motion_x >> (s->sprite_warping_accuracy+1)); 00068 src_y = s->mb_y * 16 + (motion_y >> (s->sprite_warping_accuracy+1)); 00069 motion_x<<=(3-s->sprite_warping_accuracy); 00070 motion_y<<=(3-s->sprite_warping_accuracy); 00071 src_x = av_clip(src_x, -16, s->width); 00072 if (src_x == s->width) 00073 motion_x =0; 00074 src_y = av_clip(src_y, -16, s->height); 00075 if (src_y == s->height) 00076 motion_y =0; 00077 00078 linesize = s->linesize; 00079 uvlinesize = s->uvlinesize; 00080 00081 ptr = ref_picture[0] + (src_y * linesize) + src_x; 00082 00083 if(s->flags&CODEC_FLAG_EMU_EDGE){ 00084 if( (unsigned)src_x >= s->h_edge_pos - 17 00085 || (unsigned)src_y >= s->v_edge_pos - 17){ 00086 s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, linesize, 17, 17, src_x, src_y, s->h_edge_pos, s->v_edge_pos); 00087 ptr= s->edge_emu_buffer; 00088 } 00089 } 00090 00091 if((motion_x|motion_y)&7){ 00092 s->dsp.gmc1(dest_y , ptr , linesize, 16, motion_x&15, motion_y&15, 128 - s->no_rounding); 00093 s->dsp.gmc1(dest_y+8, ptr+8, linesize, 16, motion_x&15, motion_y&15, 128 - s->no_rounding); 00094 }else{ 00095 int dxy; 00096 00097 dxy= ((motion_x>>3)&1) | ((motion_y>>2)&2); 00098 if (s->no_rounding){ 00099 s->dsp.put_no_rnd_pixels_tab[0][dxy](dest_y, ptr, linesize, 16); 00100 }else{ 00101 s->dsp.put_pixels_tab [0][dxy](dest_y, ptr, linesize, 16); 00102 } 00103 } 00104 00105 if(CONFIG_GRAY && s->flags&CODEC_FLAG_GRAY) return; 00106 00107 motion_x= s->sprite_offset[1][0]; 00108 motion_y= s->sprite_offset[1][1]; 00109 src_x = s->mb_x * 8 + (motion_x >> (s->sprite_warping_accuracy+1)); 00110 src_y = s->mb_y * 8 + (motion_y >> (s->sprite_warping_accuracy+1)); 00111 motion_x<<=(3-s->sprite_warping_accuracy); 00112 motion_y<<=(3-s->sprite_warping_accuracy); 00113 src_x = av_clip(src_x, -8, s->width>>1); 00114 if (src_x == s->width>>1) 00115 motion_x =0; 00116 src_y = av_clip(src_y, -8, s->height>>1); 00117 if (src_y == s->height>>1) 00118 motion_y =0; 00119 00120 offset = (src_y * uvlinesize) + src_x; 00121 ptr = ref_picture[1] + offset; 00122 if(s->flags&CODEC_FLAG_EMU_EDGE){ 00123 if( (unsigned)src_x >= (s->h_edge_pos>>1) - 9 00124 || (unsigned)src_y >= (s->v_edge_pos>>1) - 9){ 00125 s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); 00126 ptr= s->edge_emu_buffer; 00127 emu=1; 00128 } 00129 } 00130 s->dsp.gmc1(dest_cb, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding); 00131 00132 ptr = ref_picture[2] + offset; 00133 if(emu){ 00134 s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); 00135 ptr= s->edge_emu_buffer; 00136 } 00137 s->dsp.gmc1(dest_cr, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding); 00138 00139 return; 00140 } 00141 00142 static inline void gmc_motion(MpegEncContext *s, 00143 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, 00144 uint8_t **ref_picture) 00145 { 00146 uint8_t *ptr; 00147 int linesize, uvlinesize; 00148 const int a= s->sprite_warping_accuracy; 00149 int ox, oy; 00150 00151 linesize = s->linesize; 00152 uvlinesize = s->uvlinesize; 00153 00154 ptr = ref_picture[0]; 00155 00156 ox= s->sprite_offset[0][0] + s->sprite_delta[0][0]*s->mb_x*16 + s->sprite_delta[0][1]*s->mb_y*16; 00157 oy= s->sprite_offset[0][1] + s->sprite_delta[1][0]*s->mb_x*16 + s->sprite_delta[1][1]*s->mb_y*16; 00158 00159 s->dsp.gmc(dest_y, ptr, linesize, 16, 00160 ox, 00161 oy, 00162 s->sprite_delta[0][0], s->sprite_delta[0][1], 00163 s->sprite_delta[1][0], s->sprite_delta[1][1], 00164 a+1, (1<<(2*a+1)) - s->no_rounding, 00165 s->h_edge_pos, s->v_edge_pos); 00166 s->dsp.gmc(dest_y+8, ptr, linesize, 16, 00167 ox + s->sprite_delta[0][0]*8, 00168 oy + s->sprite_delta[1][0]*8, 00169 s->sprite_delta[0][0], s->sprite_delta[0][1], 00170 s->sprite_delta[1][0], s->sprite_delta[1][1], 00171 a+1, (1<<(2*a+1)) - s->no_rounding, 00172 s->h_edge_pos, s->v_edge_pos); 00173 00174 if(CONFIG_GRAY && s->flags&CODEC_FLAG_GRAY) return; 00175 00176 ox= s->sprite_offset[1][0] + s->sprite_delta[0][0]*s->mb_x*8 + s->sprite_delta[0][1]*s->mb_y*8; 00177 oy= s->sprite_offset[1][1] + s->sprite_delta[1][0]*s->mb_x*8 + s->sprite_delta[1][1]*s->mb_y*8; 00178 00179 ptr = ref_picture[1]; 00180 s->dsp.gmc(dest_cb, ptr, uvlinesize, 8, 00181 ox, 00182 oy, 00183 s->sprite_delta[0][0], s->sprite_delta[0][1], 00184 s->sprite_delta[1][0], s->sprite_delta[1][1], 00185 a+1, (1<<(2*a+1)) - s->no_rounding, 00186 s->h_edge_pos>>1, s->v_edge_pos>>1); 00187 00188 ptr = ref_picture[2]; 00189 s->dsp.gmc(dest_cr, ptr, uvlinesize, 8, 00190 ox, 00191 oy, 00192 s->sprite_delta[0][0], s->sprite_delta[0][1], 00193 s->sprite_delta[1][0], s->sprite_delta[1][1], 00194 a+1, (1<<(2*a+1)) - s->no_rounding, 00195 s->h_edge_pos>>1, s->v_edge_pos>>1); 00196 } 00197 00198 static inline int hpel_motion(MpegEncContext *s, 00199 uint8_t *dest, uint8_t *src, 00200 int field_based, int field_select, 00201 int src_x, int src_y, 00202 int width, int height, int stride, 00203 int h_edge_pos, int v_edge_pos, 00204 int w, int h, op_pixels_func *pix_op, 00205 int motion_x, int motion_y) 00206 { 00207 int dxy; 00208 int emu=0; 00209 00210 dxy = ((motion_y & 1) << 1) | (motion_x & 1); 00211 src_x += motion_x >> 1; 00212 src_y += motion_y >> 1; 00213 00214 /* WARNING: do no forget half pels */ 00215 src_x = av_clip(src_x, -16, width); //FIXME unneeded for emu? 00216 if (src_x == width) 00217 dxy &= ~1; 00218 src_y = av_clip(src_y, -16, height); 00219 if (src_y == height) 00220 dxy &= ~2; 00221 src += src_y * stride + src_x; 00222 00223 if(s->unrestricted_mv && (s->flags&CODEC_FLAG_EMU_EDGE)){ 00224 if( (unsigned)src_x > h_edge_pos - (motion_x&1) - w 00225 || (unsigned)src_y > v_edge_pos - (motion_y&1) - h){ 00226 s->dsp.emulated_edge_mc(s->edge_emu_buffer, src, s->linesize, w+1, (h+1)<<field_based, 00227 src_x, src_y<<field_based, h_edge_pos, s->v_edge_pos); 00228 src= s->edge_emu_buffer; 00229 emu=1; 00230 } 00231 } 00232 if(field_select) 00233 src += s->linesize; 00234 pix_op[dxy](dest, src, stride, h); 00235 return emu; 00236 } 00237 00238 static av_always_inline 00239 void mpeg_motion_internal(MpegEncContext *s, 00240 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, 00241 int field_based, int bottom_field, int field_select, 00242 uint8_t **ref_picture, op_pixels_func (*pix_op)[4], 00243 int motion_x, int motion_y, int h, int is_mpeg12, int mb_y) 00244 { 00245 uint8_t *ptr_y, *ptr_cb, *ptr_cr; 00246 int dxy, uvdxy, mx, my, src_x, src_y, 00247 uvsrc_x, uvsrc_y, v_edge_pos, uvlinesize, linesize; 00248 00249 #if 0 00250 if(s->quarter_sample) 00251 { 00252 motion_x>>=1; 00253 motion_y>>=1; 00254 } 00255 #endif 00256 00257 v_edge_pos = s->v_edge_pos >> field_based; 00258 linesize = s->current_picture.linesize[0] << field_based; 00259 uvlinesize = s->current_picture.linesize[1] << field_based; 00260 00261 dxy = ((motion_y & 1) << 1) | (motion_x & 1); 00262 src_x = s->mb_x* 16 + (motion_x >> 1); 00263 src_y =( mb_y<<(4-field_based)) + (motion_y >> 1); 00264 00265 if (!is_mpeg12 && s->out_format == FMT_H263) { 00266 if((s->workaround_bugs & FF_BUG_HPEL_CHROMA) && field_based){ 00267 mx = (motion_x>>1)|(motion_x&1); 00268 my = motion_y >>1; 00269 uvdxy = ((my & 1) << 1) | (mx & 1); 00270 uvsrc_x = s->mb_x* 8 + (mx >> 1); 00271 uvsrc_y =( mb_y<<(3-field_based))+ (my >> 1); 00272 }else{ 00273 uvdxy = dxy | (motion_y & 2) | ((motion_x & 2) >> 1); 00274 uvsrc_x = src_x>>1; 00275 uvsrc_y = src_y>>1; 00276 } 00277 }else if(!is_mpeg12 && s->out_format == FMT_H261){//even chroma mv's are full pel in H261 00278 mx = motion_x / 4; 00279 my = motion_y / 4; 00280 uvdxy = 0; 00281 uvsrc_x = s->mb_x*8 + mx; 00282 uvsrc_y = mb_y*8 + my; 00283 } else { 00284 if(s->chroma_y_shift){ 00285 mx = motion_x / 2; 00286 my = motion_y / 2; 00287 uvdxy = ((my & 1) << 1) | (mx & 1); 00288 uvsrc_x = s->mb_x* 8 + (mx >> 1); 00289 uvsrc_y =( mb_y<<(3-field_based))+ (my >> 1); 00290 } else { 00291 if(s->chroma_x_shift){ 00292 //Chroma422 00293 mx = motion_x / 2; 00294 uvdxy = ((motion_y & 1) << 1) | (mx & 1); 00295 uvsrc_x = s->mb_x* 8 + (mx >> 1); 00296 uvsrc_y = src_y; 00297 } else { 00298 //Chroma444 00299 uvdxy = dxy; 00300 uvsrc_x = src_x; 00301 uvsrc_y = src_y; 00302 } 00303 } 00304 } 00305 00306 ptr_y = ref_picture[0] + src_y * linesize + src_x; 00307 ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x; 00308 ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x; 00309 00310 if( (unsigned)src_x > s->h_edge_pos - (motion_x&1) - 16 00311 || (unsigned)src_y > v_edge_pos - (motion_y&1) - h){ 00312 if(is_mpeg12 || s->codec_id == CODEC_ID_MPEG2VIDEO || 00313 s->codec_id == CODEC_ID_MPEG1VIDEO){ 00314 av_log(s->avctx,AV_LOG_DEBUG, 00315 "MPEG motion vector out of boundary (%d %d)\n", src_x, src_y); 00316 return; 00317 } 00318 s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 00319 17, 17+field_based, 00320 src_x, src_y<<field_based, 00321 s->h_edge_pos, s->v_edge_pos); 00322 ptr_y = s->edge_emu_buffer; 00323 if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ 00324 uint8_t *uvbuf= s->edge_emu_buffer+18*s->linesize; 00325 s->dsp.emulated_edge_mc(uvbuf , 00326 ptr_cb, s->uvlinesize, 00327 9, 9+field_based, 00328 uvsrc_x, uvsrc_y<<field_based, 00329 s->h_edge_pos>>1, s->v_edge_pos>>1); 00330 s->dsp.emulated_edge_mc(uvbuf+16, 00331 ptr_cr, s->uvlinesize, 00332 9, 9+field_based, 00333 uvsrc_x, uvsrc_y<<field_based, 00334 s->h_edge_pos>>1, s->v_edge_pos>>1); 00335 ptr_cb= uvbuf; 00336 ptr_cr= uvbuf+16; 00337 } 00338 } 00339 00340 if(bottom_field){ //FIXME use this for field pix too instead of the obnoxious hack which changes picture.data 00341 dest_y += s->linesize; 00342 dest_cb+= s->uvlinesize; 00343 dest_cr+= s->uvlinesize; 00344 } 00345 00346 if(field_select){ 00347 ptr_y += s->linesize; 00348 ptr_cb+= s->uvlinesize; 00349 ptr_cr+= s->uvlinesize; 00350 } 00351 00352 pix_op[0][dxy](dest_y, ptr_y, linesize, h); 00353 00354 if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ 00355 pix_op[s->chroma_x_shift][uvdxy] 00356 (dest_cb, ptr_cb, uvlinesize, h >> s->chroma_y_shift); 00357 pix_op[s->chroma_x_shift][uvdxy] 00358 (dest_cr, ptr_cr, uvlinesize, h >> s->chroma_y_shift); 00359 } 00360 if(!is_mpeg12 && (CONFIG_H261_ENCODER || CONFIG_H261_DECODER) && 00361 s->out_format == FMT_H261){ 00362 ff_h261_loop_filter(s); 00363 } 00364 } 00365 /* apply one mpeg motion vector to the three components */ 00366 static av_always_inline 00367 void mpeg_motion(MpegEncContext *s, 00368 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, 00369 int field_based, int bottom_field, int field_select, 00370 uint8_t **ref_picture, op_pixels_func (*pix_op)[4], 00371 int motion_x, int motion_y, int h, int mb_y) 00372 { 00373 #if !CONFIG_SMALL 00374 if(s->out_format == FMT_MPEG1) 00375 mpeg_motion_internal(s, dest_y, dest_cb, dest_cr, field_based, 00376 bottom_field, field_select, ref_picture, pix_op, 00377 motion_x, motion_y, h, 1, mb_y); 00378 else 00379 #endif 00380 mpeg_motion_internal(s, dest_y, dest_cb, dest_cr, field_based, 00381 bottom_field, field_select, ref_picture, pix_op, 00382 motion_x, motion_y, h, 0, mb_y); 00383 } 00384 00385 //FIXME move to dsputil, avg variant, 16x16 version 00386 static inline void put_obmc(uint8_t *dst, uint8_t *src[5], int stride){ 00387 int x; 00388 uint8_t * const top = src[1]; 00389 uint8_t * const left = src[2]; 00390 uint8_t * const mid = src[0]; 00391 uint8_t * const right = src[3]; 00392 uint8_t * const bottom= src[4]; 00393 #define OBMC_FILTER(x, t, l, m, r, b)\ 00394 dst[x]= (t*top[x] + l*left[x] + m*mid[x] + r*right[x] + b*bottom[x] + 4)>>3 00395 #define OBMC_FILTER4(x, t, l, m, r, b)\ 00396 OBMC_FILTER(x , t, l, m, r, b);\ 00397 OBMC_FILTER(x+1 , t, l, m, r, b);\ 00398 OBMC_FILTER(x +stride, t, l, m, r, b);\ 00399 OBMC_FILTER(x+1+stride, t, l, m, r, b); 00400 00401 x=0; 00402 OBMC_FILTER (x , 2, 2, 4, 0, 0); 00403 OBMC_FILTER (x+1, 2, 1, 5, 0, 0); 00404 OBMC_FILTER4(x+2, 2, 1, 5, 0, 0); 00405 OBMC_FILTER4(x+4, 2, 0, 5, 1, 0); 00406 OBMC_FILTER (x+6, 2, 0, 5, 1, 0); 00407 OBMC_FILTER (x+7, 2, 0, 4, 2, 0); 00408 x+= stride; 00409 OBMC_FILTER (x , 1, 2, 5, 0, 0); 00410 OBMC_FILTER (x+1, 1, 2, 5, 0, 0); 00411 OBMC_FILTER (x+6, 1, 0, 5, 2, 0); 00412 OBMC_FILTER (x+7, 1, 0, 5, 2, 0); 00413 x+= stride; 00414 OBMC_FILTER4(x , 1, 2, 5, 0, 0); 00415 OBMC_FILTER4(x+2, 1, 1, 6, 0, 0); 00416 OBMC_FILTER4(x+4, 1, 0, 6, 1, 0); 00417 OBMC_FILTER4(x+6, 1, 0, 5, 2, 0); 00418 x+= 2*stride; 00419 OBMC_FILTER4(x , 0, 2, 5, 0, 1); 00420 OBMC_FILTER4(x+2, 0, 1, 6, 0, 1); 00421 OBMC_FILTER4(x+4, 0, 0, 6, 1, 1); 00422 OBMC_FILTER4(x+6, 0, 0, 5, 2, 1); 00423 x+= 2*stride; 00424 OBMC_FILTER (x , 0, 2, 5, 0, 1); 00425 OBMC_FILTER (x+1, 0, 2, 5, 0, 1); 00426 OBMC_FILTER4(x+2, 0, 1, 5, 0, 2); 00427 OBMC_FILTER4(x+4, 0, 0, 5, 1, 2); 00428 OBMC_FILTER (x+6, 0, 0, 5, 2, 1); 00429 OBMC_FILTER (x+7, 0, 0, 5, 2, 1); 00430 x+= stride; 00431 OBMC_FILTER (x , 0, 2, 4, 0, 2); 00432 OBMC_FILTER (x+1, 0, 1, 5, 0, 2); 00433 OBMC_FILTER (x+6, 0, 0, 5, 1, 2); 00434 OBMC_FILTER (x+7, 0, 0, 4, 2, 2); 00435 } 00436 00437 /* obmc for 1 8x8 luma block */ 00438 static inline void obmc_motion(MpegEncContext *s, 00439 uint8_t *dest, uint8_t *src, 00440 int src_x, int src_y, 00441 op_pixels_func *pix_op, 00442 int16_t mv[5][2]/* mid top left right bottom*/) 00443 #define MID 0 00444 { 00445 int i; 00446 uint8_t *ptr[5]; 00447 00448 assert(s->quarter_sample==0); 00449 00450 for(i=0; i<5; i++){ 00451 if(i && mv[i][0]==mv[MID][0] && mv[i][1]==mv[MID][1]){ 00452 ptr[i]= ptr[MID]; 00453 }else{ 00454 ptr[i]= s->obmc_scratchpad + 8*(i&1) + s->linesize*8*(i>>1); 00455 hpel_motion(s, ptr[i], src, 0, 0, 00456 src_x, src_y, 00457 s->width, s->height, s->linesize, 00458 s->h_edge_pos, s->v_edge_pos, 00459 8, 8, pix_op, 00460 mv[i][0], mv[i][1]); 00461 } 00462 } 00463 00464 put_obmc(dest, ptr, s->linesize); 00465 } 00466 00467 static inline void qpel_motion(MpegEncContext *s, 00468 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, 00469 int field_based, int bottom_field, int field_select, 00470 uint8_t **ref_picture, op_pixels_func (*pix_op)[4], 00471 qpel_mc_func (*qpix_op)[16], 00472 int motion_x, int motion_y, int h) 00473 { 00474 uint8_t *ptr_y, *ptr_cb, *ptr_cr; 00475 int dxy, uvdxy, mx, my, src_x, src_y, uvsrc_x, uvsrc_y, v_edge_pos, linesize, uvlinesize; 00476 00477 dxy = ((motion_y & 3) << 2) | (motion_x & 3); 00478 src_x = s->mb_x * 16 + (motion_x >> 2); 00479 src_y = s->mb_y * (16 >> field_based) + (motion_y >> 2); 00480 00481 v_edge_pos = s->v_edge_pos >> field_based; 00482 linesize = s->linesize << field_based; 00483 uvlinesize = s->uvlinesize << field_based; 00484 00485 if(field_based){ 00486 mx= motion_x/2; 00487 my= motion_y>>1; 00488 }else if(s->workaround_bugs&FF_BUG_QPEL_CHROMA2){ 00489 static const int rtab[8]= {0,0,1,1,0,0,0,1}; 00490 mx= (motion_x>>1) + rtab[motion_x&7]; 00491 my= (motion_y>>1) + rtab[motion_y&7]; 00492 }else if(s->workaround_bugs&FF_BUG_QPEL_CHROMA){ 00493 mx= (motion_x>>1)|(motion_x&1); 00494 my= (motion_y>>1)|(motion_y&1); 00495 }else{ 00496 mx= motion_x/2; 00497 my= motion_y/2; 00498 } 00499 mx= (mx>>1)|(mx&1); 00500 my= (my>>1)|(my&1); 00501 00502 uvdxy= (mx&1) | ((my&1)<<1); 00503 mx>>=1; 00504 my>>=1; 00505 00506 uvsrc_x = s->mb_x * 8 + mx; 00507 uvsrc_y = s->mb_y * (8 >> field_based) + my; 00508 00509 ptr_y = ref_picture[0] + src_y * linesize + src_x; 00510 ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x; 00511 ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x; 00512 00513 if( (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 16 00514 || (unsigned)src_y > v_edge_pos - (motion_y&3) - h ){ 00515 s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 00516 17, 17+field_based, src_x, src_y<<field_based, 00517 s->h_edge_pos, s->v_edge_pos); 00518 ptr_y= s->edge_emu_buffer; 00519 if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ 00520 uint8_t *uvbuf= s->edge_emu_buffer + 18*s->linesize; 00521 s->dsp.emulated_edge_mc(uvbuf, ptr_cb, s->uvlinesize, 00522 9, 9 + field_based, 00523 uvsrc_x, uvsrc_y<<field_based, 00524 s->h_edge_pos>>1, s->v_edge_pos>>1); 00525 s->dsp.emulated_edge_mc(uvbuf + 16, ptr_cr, s->uvlinesize, 00526 9, 9 + field_based, 00527 uvsrc_x, uvsrc_y<<field_based, 00528 s->h_edge_pos>>1, s->v_edge_pos>>1); 00529 ptr_cb= uvbuf; 00530 ptr_cr= uvbuf + 16; 00531 } 00532 } 00533 00534 if(!field_based) 00535 qpix_op[0][dxy](dest_y, ptr_y, linesize); 00536 else{ 00537 if(bottom_field){ 00538 dest_y += s->linesize; 00539 dest_cb+= s->uvlinesize; 00540 dest_cr+= s->uvlinesize; 00541 } 00542 00543 if(field_select){ 00544 ptr_y += s->linesize; 00545 ptr_cb += s->uvlinesize; 00546 ptr_cr += s->uvlinesize; 00547 } 00548 //damn interlaced mode 00549 //FIXME boundary mirroring is not exactly correct here 00550 qpix_op[1][dxy](dest_y , ptr_y , linesize); 00551 qpix_op[1][dxy](dest_y+8, ptr_y+8, linesize); 00552 } 00553 if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ 00554 pix_op[1][uvdxy](dest_cr, ptr_cr, uvlinesize, h >> 1); 00555 pix_op[1][uvdxy](dest_cb, ptr_cb, uvlinesize, h >> 1); 00556 } 00557 } 00558 00562 static inline void chroma_4mv_motion(MpegEncContext *s, 00563 uint8_t *dest_cb, uint8_t *dest_cr, 00564 uint8_t **ref_picture, 00565 op_pixels_func *pix_op, 00566 int mx, int my){ 00567 int dxy, emu=0, src_x, src_y, offset; 00568 uint8_t *ptr; 00569 00570 /* In case of 8X8, we construct a single chroma motion vector 00571 with a special rounding */ 00572 mx= ff_h263_round_chroma(mx); 00573 my= ff_h263_round_chroma(my); 00574 00575 dxy = ((my & 1) << 1) | (mx & 1); 00576 mx >>= 1; 00577 my >>= 1; 00578 00579 src_x = s->mb_x * 8 + mx; 00580 src_y = s->mb_y * 8 + my; 00581 src_x = av_clip(src_x, -8, (s->width >> 1)); 00582 if (src_x == (s->width >> 1)) 00583 dxy &= ~1; 00584 src_y = av_clip(src_y, -8, (s->height >> 1)); 00585 if (src_y == (s->height >> 1)) 00586 dxy &= ~2; 00587 00588 offset = (src_y * (s->uvlinesize)) + src_x; 00589 ptr = ref_picture[1] + offset; 00590 if(s->flags&CODEC_FLAG_EMU_EDGE){ 00591 if( (unsigned)src_x > (s->h_edge_pos>>1) - (dxy &1) - 8 00592 || (unsigned)src_y > (s->v_edge_pos>>1) - (dxy>>1) - 8){ 00593 s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 00594 9, 9, src_x, src_y, 00595 s->h_edge_pos>>1, s->v_edge_pos>>1); 00596 ptr= s->edge_emu_buffer; 00597 emu=1; 00598 } 00599 } 00600 pix_op[dxy](dest_cb, ptr, s->uvlinesize, 8); 00601 00602 ptr = ref_picture[2] + offset; 00603 if(emu){ 00604 s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 00605 9, 9, src_x, src_y, 00606 s->h_edge_pos>>1, s->v_edge_pos>>1); 00607 ptr= s->edge_emu_buffer; 00608 } 00609 pix_op[dxy](dest_cr, ptr, s->uvlinesize, 8); 00610 } 00611 00612 static inline void prefetch_motion(MpegEncContext *s, uint8_t **pix, int dir){ 00613 /* fetch pixels for estimated mv 4 macroblocks ahead 00614 * optimized for 64byte cache lines */ 00615 const int shift = s->quarter_sample ? 2 : 1; 00616 const int mx= (s->mv[dir][0][0]>>shift) + 16*s->mb_x + 8; 00617 const int my= (s->mv[dir][0][1]>>shift) + 16*s->mb_y; 00618 int off= mx + (my + (s->mb_x&3)*4)*s->linesize + 64; 00619 s->dsp.prefetch(pix[0]+off, s->linesize, 4); 00620 off= (mx>>1) + ((my>>1) + (s->mb_x&7))*s->uvlinesize + 64; 00621 s->dsp.prefetch(pix[1]+off, pix[2]-pix[1], 2); 00622 } 00623 00636 static av_always_inline void MPV_motion_internal(MpegEncContext *s, 00637 uint8_t *dest_y, uint8_t *dest_cb, 00638 uint8_t *dest_cr, int dir, 00639 uint8_t **ref_picture, 00640 op_pixels_func (*pix_op)[4], 00641 qpel_mc_func (*qpix_op)[16], int is_mpeg12) 00642 { 00643 int dxy, mx, my, src_x, src_y, motion_x, motion_y; 00644 int mb_x, mb_y, i; 00645 uint8_t *ptr, *dest; 00646 00647 mb_x = s->mb_x; 00648 mb_y = s->mb_y; 00649 00650 prefetch_motion(s, ref_picture, dir); 00651 00652 if(!is_mpeg12 && s->obmc && s->pict_type != AV_PICTURE_TYPE_B){ 00653 int16_t mv_cache[4][4][2]; 00654 const int xy= s->mb_x + s->mb_y*s->mb_stride; 00655 const int mot_stride= s->b8_stride; 00656 const int mot_xy= mb_x*2 + mb_y*2*mot_stride; 00657 00658 assert(!s->mb_skipped); 00659 00660 memcpy(mv_cache[1][1], s->current_picture.motion_val[0][mot_xy ], sizeof(int16_t)*4); 00661 memcpy(mv_cache[2][1], s->current_picture.motion_val[0][mot_xy+mot_stride], sizeof(int16_t)*4); 00662 memcpy(mv_cache[3][1], s->current_picture.motion_val[0][mot_xy+mot_stride], sizeof(int16_t)*4); 00663 00664 if(mb_y==0 || IS_INTRA(s->current_picture.mb_type[xy-s->mb_stride])){ 00665 memcpy(mv_cache[0][1], mv_cache[1][1], sizeof(int16_t)*4); 00666 }else{ 00667 memcpy(mv_cache[0][1], s->current_picture.motion_val[0][mot_xy-mot_stride], sizeof(int16_t)*4); 00668 } 00669 00670 if(mb_x==0 || IS_INTRA(s->current_picture.mb_type[xy-1])){ 00671 AV_COPY32(mv_cache[1][0], mv_cache[1][1]); 00672 AV_COPY32(mv_cache[2][0], mv_cache[2][1]); 00673 }else{ 00674 AV_COPY32(mv_cache[1][0], s->current_picture.motion_val[0][mot_xy-1]); 00675 AV_COPY32(mv_cache[2][0], s->current_picture.motion_val[0][mot_xy-1+mot_stride]); 00676 } 00677 00678 if(mb_x+1>=s->mb_width || IS_INTRA(s->current_picture.mb_type[xy+1])){ 00679 AV_COPY32(mv_cache[1][3], mv_cache[1][2]); 00680 AV_COPY32(mv_cache[2][3], mv_cache[2][2]); 00681 }else{ 00682 AV_COPY32(mv_cache[1][3], s->current_picture.motion_val[0][mot_xy+2]); 00683 AV_COPY32(mv_cache[2][3], s->current_picture.motion_val[0][mot_xy+2+mot_stride]); 00684 } 00685 00686 mx = 0; 00687 my = 0; 00688 for(i=0;i<4;i++) { 00689 const int x= (i&1)+1; 00690 const int y= (i>>1)+1; 00691 int16_t mv[5][2]= { 00692 {mv_cache[y][x ][0], mv_cache[y][x ][1]}, 00693 {mv_cache[y-1][x][0], mv_cache[y-1][x][1]}, 00694 {mv_cache[y][x-1][0], mv_cache[y][x-1][1]}, 00695 {mv_cache[y][x+1][0], mv_cache[y][x+1][1]}, 00696 {mv_cache[y+1][x][0], mv_cache[y+1][x][1]}}; 00697 //FIXME cleanup 00698 obmc_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize, 00699 ref_picture[0], 00700 mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8, 00701 pix_op[1], 00702 mv); 00703 00704 mx += mv[0][0]; 00705 my += mv[0][1]; 00706 } 00707 if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)) 00708 chroma_4mv_motion(s, dest_cb, dest_cr, ref_picture, pix_op[1], mx, my); 00709 00710 return; 00711 } 00712 00713 switch(s->mv_type) { 00714 case MV_TYPE_16X16: 00715 if(s->mcsel){ 00716 if(s->real_sprite_warping_points==1){ 00717 gmc1_motion(s, dest_y, dest_cb, dest_cr, 00718 ref_picture); 00719 }else{ 00720 gmc_motion(s, dest_y, dest_cb, dest_cr, 00721 ref_picture); 00722 } 00723 }else if(!is_mpeg12 && s->quarter_sample){ 00724 qpel_motion(s, dest_y, dest_cb, dest_cr, 00725 0, 0, 0, 00726 ref_picture, pix_op, qpix_op, 00727 s->mv[dir][0][0], s->mv[dir][0][1], 16); 00728 }else if(!is_mpeg12 && (CONFIG_WMV2_DECODER || CONFIG_WMV2_ENCODER) && s->mspel){ 00729 ff_mspel_motion(s, dest_y, dest_cb, dest_cr, 00730 ref_picture, pix_op, 00731 s->mv[dir][0][0], s->mv[dir][0][1], 16); 00732 }else 00733 { 00734 mpeg_motion(s, dest_y, dest_cb, dest_cr, 00735 0, 0, 0, 00736 ref_picture, pix_op, 00737 s->mv[dir][0][0], s->mv[dir][0][1], 16, mb_y); 00738 } 00739 break; 00740 case MV_TYPE_8X8: 00741 if (!is_mpeg12) { 00742 mx = 0; 00743 my = 0; 00744 if(s->quarter_sample){ 00745 for(i=0;i<4;i++) { 00746 motion_x = s->mv[dir][i][0]; 00747 motion_y = s->mv[dir][i][1]; 00748 00749 dxy = ((motion_y & 3) << 2) | (motion_x & 3); 00750 src_x = mb_x * 16 + (motion_x >> 2) + (i & 1) * 8; 00751 src_y = mb_y * 16 + (motion_y >> 2) + (i >>1) * 8; 00752 00753 /* WARNING: do no forget half pels */ 00754 src_x = av_clip(src_x, -16, s->width); 00755 if (src_x == s->width) 00756 dxy &= ~3; 00757 src_y = av_clip(src_y, -16, s->height); 00758 if (src_y == s->height) 00759 dxy &= ~12; 00760 00761 ptr = ref_picture[0] + (src_y * s->linesize) + (src_x); 00762 if(s->flags&CODEC_FLAG_EMU_EDGE){ 00763 if( (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 8 00764 || (unsigned)src_y > s->v_edge_pos - (motion_y&3) - 8 ){ 00765 s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, 00766 s->linesize, 9, 9, 00767 src_x, src_y, 00768 s->h_edge_pos, s->v_edge_pos); 00769 ptr= s->edge_emu_buffer; 00770 } 00771 } 00772 dest = dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize; 00773 qpix_op[1][dxy](dest, ptr, s->linesize); 00774 00775 mx += s->mv[dir][i][0]/2; 00776 my += s->mv[dir][i][1]/2; 00777 } 00778 }else{ 00779 for(i=0;i<4;i++) { 00780 hpel_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize, 00781 ref_picture[0], 0, 0, 00782 mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8, 00783 s->width, s->height, s->linesize, 00784 s->h_edge_pos, s->v_edge_pos, 00785 8, 8, pix_op[1], 00786 s->mv[dir][i][0], s->mv[dir][i][1]); 00787 00788 mx += s->mv[dir][i][0]; 00789 my += s->mv[dir][i][1]; 00790 } 00791 } 00792 00793 if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)) 00794 chroma_4mv_motion(s, dest_cb, dest_cr, ref_picture, pix_op[1], mx, my); 00795 } 00796 break; 00797 case MV_TYPE_FIELD: 00798 if (s->picture_structure == PICT_FRAME) { 00799 if(!is_mpeg12 && s->quarter_sample){ 00800 for(i=0; i<2; i++){ 00801 qpel_motion(s, dest_y, dest_cb, dest_cr, 00802 1, i, s->field_select[dir][i], 00803 ref_picture, pix_op, qpix_op, 00804 s->mv[dir][i][0], s->mv[dir][i][1], 8); 00805 } 00806 }else{ 00807 /* top field */ 00808 mpeg_motion(s, dest_y, dest_cb, dest_cr, 00809 1, 0, s->field_select[dir][0], 00810 ref_picture, pix_op, 00811 s->mv[dir][0][0], s->mv[dir][0][1], 8, mb_y); 00812 /* bottom field */ 00813 mpeg_motion(s, dest_y, dest_cb, dest_cr, 00814 1, 1, s->field_select[dir][1], 00815 ref_picture, pix_op, 00816 s->mv[dir][1][0], s->mv[dir][1][1], 8, mb_y); 00817 } 00818 } else { 00819 if(s->picture_structure != s->field_select[dir][0] + 1 && s->pict_type != AV_PICTURE_TYPE_B && !s->first_field){ 00820 ref_picture= s->current_picture_ptr->data; 00821 } 00822 00823 mpeg_motion(s, dest_y, dest_cb, dest_cr, 00824 0, 0, s->field_select[dir][0], 00825 ref_picture, pix_op, 00826 s->mv[dir][0][0], s->mv[dir][0][1], 16, mb_y>>1); 00827 } 00828 break; 00829 case MV_TYPE_16X8: 00830 for(i=0; i<2; i++){ 00831 uint8_t ** ref2picture; 00832 00833 if(s->picture_structure == s->field_select[dir][i] + 1 00834 || s->pict_type == AV_PICTURE_TYPE_B || s->first_field){ 00835 ref2picture= ref_picture; 00836 }else{ 00837 ref2picture= s->current_picture_ptr->data; 00838 } 00839 00840 mpeg_motion(s, dest_y, dest_cb, dest_cr, 00841 0, 0, s->field_select[dir][i], 00842 ref2picture, pix_op, 00843 s->mv[dir][i][0], s->mv[dir][i][1] + 16*i, 8, mb_y>>1); 00844 00845 dest_y += 16*s->linesize; 00846 dest_cb+= (16>>s->chroma_y_shift)*s->uvlinesize; 00847 dest_cr+= (16>>s->chroma_y_shift)*s->uvlinesize; 00848 } 00849 break; 00850 case MV_TYPE_DMV: 00851 if(s->picture_structure == PICT_FRAME){ 00852 for(i=0; i<2; i++){ 00853 int j; 00854 for(j=0; j<2; j++){ 00855 mpeg_motion(s, dest_y, dest_cb, dest_cr, 00856 1, j, j^i, 00857 ref_picture, pix_op, 00858 s->mv[dir][2*i + j][0], s->mv[dir][2*i + j][1], 8, mb_y); 00859 } 00860 pix_op = s->dsp.avg_pixels_tab; 00861 } 00862 }else{ 00863 for(i=0; i<2; i++){ 00864 mpeg_motion(s, dest_y, dest_cb, dest_cr, 00865 0, 0, s->picture_structure != i+1, 00866 ref_picture, pix_op, 00867 s->mv[dir][2*i][0],s->mv[dir][2*i][1],16, mb_y>>1); 00868 00869 // after put we make avg of the same block 00870 pix_op=s->dsp.avg_pixels_tab; 00871 00872 //opposite parity is always in the same frame if this is second field 00873 if(!s->first_field){ 00874 ref_picture = s->current_picture_ptr->data; 00875 } 00876 } 00877 } 00878 break; 00879 default: assert(0); 00880 } 00881 } 00882 00883 static inline void MPV_motion(MpegEncContext *s, 00884 uint8_t *dest_y, uint8_t *dest_cb, 00885 uint8_t *dest_cr, int dir, 00886 uint8_t **ref_picture, 00887 op_pixels_func (*pix_op)[4], 00888 qpel_mc_func (*qpix_op)[16]) 00889 { 00890 #if !CONFIG_SMALL 00891 if(s->out_format == FMT_MPEG1) 00892 MPV_motion_internal(s, dest_y, dest_cb, dest_cr, dir, 00893 ref_picture, pix_op, qpix_op, 1); 00894 else 00895 #endif 00896 MPV_motion_internal(s, dest_y, dest_cb, dest_cr, dir, 00897 ref_picture, pix_op, qpix_op, 0); 00898 } 00899 #endif /* AVCODEC_MPEGVIDEO_COMMON_H */