mpegvideo_common.h 33.7 KB
Newer Older
1 2
/*
 * The simplest mpeg encoder (well, it was the simplest!)
3
 * Copyright (c) 2000,2001 Fabrice Bellard
4 5
 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
 *
6 7
 * 4MV & hq & B-frame encoding stuff by Michael Niedermayer <michaelni@gmx.at>
 *
8
 * This file is part of Libav.
9
 *
10
 * Libav is free software; you can redistribute it and/or
11 12 13 14
 * 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.
 *
15
 * Libav is distributed in the hope that it will be useful,
16 17 18 19 20
 * 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
21
 * License along with Libav; if not, write to the Free Software
22 23 24 25
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
26
 * @file
27 28 29
 * The simplest mpeg encoder (well, it was the simplest!).
 */

30 31
#ifndef AVCODEC_MPEGVIDEO_COMMON_H
#define AVCODEC_MPEGVIDEO_COMMON_H
32

33
#include <string.h>
34 35 36 37 38 39 40 41
#include "avcodec.h"
#include "dsputil.h"
#include "mpegvideo.h"
#include "mjpegenc.h"
#include "msmpeg4.h"
#include "faandct.h"
#include <limits.h>

42
int ff_dct_quantize_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow);
43 44

/**
45 46
 * Set the given MpegEncContext to common defaults (same for encoding and decoding).
 * The changed fields will not depend upon the prior state of the MpegEncContext.
47
 */
48
void ff_MPV_common_defaults(MpegEncContext *s);
49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77

static inline void gmc1_motion(MpegEncContext *s,
                               uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                               uint8_t **ref_picture)
{
    uint8_t *ptr;
    int offset, src_x, src_y, linesize, uvlinesize;
    int motion_x, motion_y;
    int emu=0;

    motion_x= s->sprite_offset[0][0];
    motion_y= s->sprite_offset[0][1];
    src_x = s->mb_x * 16 + (motion_x >> (s->sprite_warping_accuracy+1));
    src_y = s->mb_y * 16 + (motion_y >> (s->sprite_warping_accuracy+1));
    motion_x<<=(3-s->sprite_warping_accuracy);
    motion_y<<=(3-s->sprite_warping_accuracy);
    src_x = av_clip(src_x, -16, s->width);
    if (src_x == s->width)
        motion_x =0;
    src_y = av_clip(src_y, -16, s->height);
    if (src_y == s->height)
        motion_y =0;

    linesize = s->linesize;
    uvlinesize = s->uvlinesize;

    ptr = ref_picture[0] + (src_y * linesize) + src_x;

    if(s->flags&CODEC_FLAG_EMU_EDGE){
78 79
        if(   (unsigned)src_x >= FFMAX(s->h_edge_pos - 17, 0)
           || (unsigned)src_y >= FFMAX(s->v_edge_pos - 17, 0)){
80
            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);
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98
            ptr= s->edge_emu_buffer;
        }
    }

    if((motion_x|motion_y)&7){
        s->dsp.gmc1(dest_y  , ptr  , linesize, 16, motion_x&15, motion_y&15, 128 - s->no_rounding);
        s->dsp.gmc1(dest_y+8, ptr+8, linesize, 16, motion_x&15, motion_y&15, 128 - s->no_rounding);
    }else{
        int dxy;

        dxy= ((motion_x>>3)&1) | ((motion_y>>2)&2);
        if (s->no_rounding){
            s->dsp.put_no_rnd_pixels_tab[0][dxy](dest_y, ptr, linesize, 16);
        }else{
            s->dsp.put_pixels_tab       [0][dxy](dest_y, ptr, linesize, 16);
        }
    }

99
    if(CONFIG_GRAY && s->flags&CODEC_FLAG_GRAY) return;
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116

    motion_x= s->sprite_offset[1][0];
    motion_y= s->sprite_offset[1][1];
    src_x = s->mb_x * 8 + (motion_x >> (s->sprite_warping_accuracy+1));
    src_y = s->mb_y * 8 + (motion_y >> (s->sprite_warping_accuracy+1));
    motion_x<<=(3-s->sprite_warping_accuracy);
    motion_y<<=(3-s->sprite_warping_accuracy);
    src_x = av_clip(src_x, -8, s->width>>1);
    if (src_x == s->width>>1)
        motion_x =0;
    src_y = av_clip(src_y, -8, s->height>>1);
    if (src_y == s->height>>1)
        motion_y =0;

    offset = (src_y * uvlinesize) + src_x;
    ptr = ref_picture[1] + offset;
    if(s->flags&CODEC_FLAG_EMU_EDGE){
117 118
        if(   (unsigned)src_x >= FFMAX((s->h_edge_pos>>1) - 9, 0)
           || (unsigned)src_y >= FFMAX((s->v_edge_pos>>1) - 9, 0)){
119
            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);
120 121 122 123 124 125 126 127
            ptr= s->edge_emu_buffer;
            emu=1;
        }
    }
    s->dsp.gmc1(dest_cb, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding);

    ptr = ref_picture[2] + offset;
    if(emu){
128
        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);
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167
        ptr= s->edge_emu_buffer;
    }
    s->dsp.gmc1(dest_cr, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding);

    return;
}

static inline void gmc_motion(MpegEncContext *s,
                               uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                               uint8_t **ref_picture)
{
    uint8_t *ptr;
    int linesize, uvlinesize;
    const int a= s->sprite_warping_accuracy;
    int ox, oy;

    linesize = s->linesize;
    uvlinesize = s->uvlinesize;

    ptr = ref_picture[0];

    ox= s->sprite_offset[0][0] + s->sprite_delta[0][0]*s->mb_x*16 + s->sprite_delta[0][1]*s->mb_y*16;
    oy= s->sprite_offset[0][1] + s->sprite_delta[1][0]*s->mb_x*16 + s->sprite_delta[1][1]*s->mb_y*16;

    s->dsp.gmc(dest_y, ptr, linesize, 16,
           ox,
           oy,
           s->sprite_delta[0][0], s->sprite_delta[0][1],
           s->sprite_delta[1][0], s->sprite_delta[1][1],
           a+1, (1<<(2*a+1)) - s->no_rounding,
           s->h_edge_pos, s->v_edge_pos);
    s->dsp.gmc(dest_y+8, ptr, linesize, 16,
           ox + s->sprite_delta[0][0]*8,
           oy + s->sprite_delta[1][0]*8,
           s->sprite_delta[0][0], s->sprite_delta[0][1],
           s->sprite_delta[1][0], s->sprite_delta[1][1],
           a+1, (1<<(2*a+1)) - s->no_rounding,
           s->h_edge_pos, s->v_edge_pos);

168
    if(CONFIG_GRAY && s->flags&CODEC_FLAG_GRAY) return;
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217

    ox= s->sprite_offset[1][0] + s->sprite_delta[0][0]*s->mb_x*8 + s->sprite_delta[0][1]*s->mb_y*8;
    oy= s->sprite_offset[1][1] + s->sprite_delta[1][0]*s->mb_x*8 + s->sprite_delta[1][1]*s->mb_y*8;

    ptr = ref_picture[1];
    s->dsp.gmc(dest_cb, ptr, uvlinesize, 8,
           ox,
           oy,
           s->sprite_delta[0][0], s->sprite_delta[0][1],
           s->sprite_delta[1][0], s->sprite_delta[1][1],
           a+1, (1<<(2*a+1)) - s->no_rounding,
           s->h_edge_pos>>1, s->v_edge_pos>>1);

    ptr = ref_picture[2];
    s->dsp.gmc(dest_cr, ptr, uvlinesize, 8,
           ox,
           oy,
           s->sprite_delta[0][0], s->sprite_delta[0][1],
           s->sprite_delta[1][0], s->sprite_delta[1][1],
           a+1, (1<<(2*a+1)) - s->no_rounding,
           s->h_edge_pos>>1, s->v_edge_pos>>1);
}

static inline int hpel_motion(MpegEncContext *s,
                                  uint8_t *dest, uint8_t *src,
                                  int field_based, int field_select,
                                  int src_x, int src_y,
                                  int width, int height, int stride,
                                  int h_edge_pos, int v_edge_pos,
                                  int w, int h, op_pixels_func *pix_op,
                                  int motion_x, int motion_y)
{
    int dxy;
    int emu=0;

    dxy = ((motion_y & 1) << 1) | (motion_x & 1);
    src_x += motion_x >> 1;
    src_y += motion_y >> 1;

    /* WARNING: do no forget half pels */
    src_x = av_clip(src_x, -16, width); //FIXME unneeded for emu?
    if (src_x == width)
        dxy &= ~1;
    src_y = av_clip(src_y, -16, height);
    if (src_y == height)
        dxy &= ~2;
    src += src_y * stride + src_x;

    if(s->unrestricted_mv && (s->flags&CODEC_FLAG_EMU_EDGE)){
218 219
        if(   (unsigned)src_x > FFMAX(h_edge_pos - (motion_x&1) - w, 0)
           || (unsigned)src_y > FFMAX(v_edge_pos - (motion_y&1) - h, 0)){
220
            s->dsp.emulated_edge_mc(s->edge_emu_buffer, src, s->linesize, w+1, (h+1)<<field_based,
221 222 223 224 225 226 227 228 229 230 231
                             src_x, src_y<<field_based, h_edge_pos, s->v_edge_pos);
            src= s->edge_emu_buffer;
            emu=1;
        }
    }
    if(field_select)
        src += s->linesize;
    pix_op[dxy](dest, src, stride, h);
    return emu;
}

232
static av_always_inline
Keiji Costantini's avatar
Keiji Costantini committed
233
void mpeg_motion_internal(MpegEncContext *s,
234 235 236
                 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                 int field_based, int bottom_field, int field_select,
                 uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
237
                 int motion_x, int motion_y, int h, int is_mpeg12, int mb_y)
238 239
{
    uint8_t *ptr_y, *ptr_cb, *ptr_cr;
240 241
    int dxy, uvdxy, mx, my, src_x, src_y,
        uvsrc_x, uvsrc_y, v_edge_pos, uvlinesize, linesize;
242 243 244 245 246 247 248 249 250 251

#if 0
if(s->quarter_sample)
{
    motion_x>>=1;
    motion_y>>=1;
}
#endif

    v_edge_pos = s->v_edge_pos >> field_based;
252 253
    linesize   = s->current_picture.f.linesize[0] << field_based;
    uvlinesize = s->current_picture.f.linesize[1] << field_based;
254 255 256

    dxy = ((motion_y & 1) << 1) | (motion_x & 1);
    src_x = s->mb_x* 16               + (motion_x >> 1);
257
    src_y =(   mb_y<<(4-field_based)) + (motion_y >> 1);
258

Keiji Costantini's avatar
Keiji Costantini committed
259
    if (!is_mpeg12 && s->out_format == FMT_H263) {
260 261 262 263 264
        if((s->workaround_bugs & FF_BUG_HPEL_CHROMA) && field_based){
            mx = (motion_x>>1)|(motion_x&1);
            my = motion_y >>1;
            uvdxy = ((my & 1) << 1) | (mx & 1);
            uvsrc_x = s->mb_x* 8               + (mx >> 1);
265
            uvsrc_y =(   mb_y<<(3-field_based))+ (my >> 1);
266 267 268 269 270
        }else{
            uvdxy = dxy | (motion_y & 2) | ((motion_x & 2) >> 1);
            uvsrc_x = src_x>>1;
            uvsrc_y = src_y>>1;
        }
Keiji Costantini's avatar
Keiji Costantini committed
271
    }else if(!is_mpeg12 && s->out_format == FMT_H261){//even chroma mv's are full pel in H261
272 273 274 275
        mx = motion_x / 4;
        my = motion_y / 4;
        uvdxy = 0;
        uvsrc_x = s->mb_x*8 + mx;
276
        uvsrc_y =    mb_y*8 + my;
277 278 279 280 281 282
    } else {
        if(s->chroma_y_shift){
            mx = motion_x / 2;
            my = motion_y / 2;
            uvdxy = ((my & 1) << 1) | (mx & 1);
            uvsrc_x = s->mb_x* 8               + (mx >> 1);
283
            uvsrc_y =(   mb_y<<(3-field_based))+ (my >> 1);
284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303
        } else {
            if(s->chroma_x_shift){
            //Chroma422
                mx = motion_x / 2;
                uvdxy = ((motion_y & 1) << 1) | (mx & 1);
                uvsrc_x = s->mb_x* 8           + (mx >> 1);
                uvsrc_y = src_y;
            } else {
            //Chroma444
                uvdxy = dxy;
                uvsrc_x = src_x;
                uvsrc_y = src_y;
            }
        }
    }

    ptr_y  = ref_picture[0] + src_y * linesize + src_x;
    ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x;
    ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x;

304 305
    if(   (unsigned)src_x > FFMAX(s->h_edge_pos - (motion_x&1) - 16, 0)
       || (unsigned)src_y > FFMAX(   v_edge_pos - (motion_y&1) - h , 0)){
Keiji Costantini's avatar
Keiji Costantini committed
306
            if(is_mpeg12 || s->codec_id == CODEC_ID_MPEG2VIDEO ||
307
               s->codec_id == CODEC_ID_MPEG1VIDEO){
308
                av_log(s->avctx,AV_LOG_DEBUG,
309
                        "MPEG motion vector out of boundary (%d %d)\n", src_x, src_y);
310
                return;
311
            }
312
            s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize,
313 314 315
                                17, 17+field_based,
                                src_x, src_y<<field_based,
                                s->h_edge_pos, s->v_edge_pos);
316
            ptr_y = s->edge_emu_buffer;
317
            if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
318
                uint8_t *uvbuf= s->edge_emu_buffer+18*s->linesize;
319
                s->dsp.emulated_edge_mc(uvbuf ,
320 321 322 323
                                    ptr_cb, s->uvlinesize,
                                    9, 9+field_based,
                                    uvsrc_x, uvsrc_y<<field_based,
                                    s->h_edge_pos>>1, s->v_edge_pos>>1);
324
                s->dsp.emulated_edge_mc(uvbuf+16,
325 326 327 328
                                    ptr_cr, s->uvlinesize,
                                    9, 9+field_based,
                                    uvsrc_x, uvsrc_y<<field_based,
                                    s->h_edge_pos>>1, s->v_edge_pos>>1);
329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347
                ptr_cb= uvbuf;
                ptr_cr= uvbuf+16;
            }
    }

    if(bottom_field){ //FIXME use this for field pix too instead of the obnoxious hack which changes picture.data
        dest_y += s->linesize;
        dest_cb+= s->uvlinesize;
        dest_cr+= s->uvlinesize;
    }

    if(field_select){
        ptr_y += s->linesize;
        ptr_cb+= s->uvlinesize;
        ptr_cr+= s->uvlinesize;
    }

    pix_op[0][dxy](dest_y, ptr_y, linesize, h);

348
    if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
349 350 351 352
        pix_op[s->chroma_x_shift][uvdxy]
                (dest_cb, ptr_cb, uvlinesize, h >> s->chroma_y_shift);
        pix_op[s->chroma_x_shift][uvdxy]
                (dest_cr, ptr_cr, uvlinesize, h >> s->chroma_y_shift);
353
    }
354
    if(!is_mpeg12 && (CONFIG_H261_ENCODER || CONFIG_H261_DECODER) &&
355
         s->out_format == FMT_H261){
356 357 358
        ff_h261_loop_filter(s);
    }
}
Keiji Costantini's avatar
Keiji Costantini committed
359 360 361 362 363 364
/* apply one mpeg motion vector to the three components */
static av_always_inline
void mpeg_motion(MpegEncContext *s,
                 uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                 int field_based, int bottom_field, int field_select,
                 uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
365
                 int motion_x, int motion_y, int h, int mb_y)
Keiji Costantini's avatar
Keiji Costantini committed
366
{
367
#if !CONFIG_SMALL
Keiji Costantini's avatar
Keiji Costantini committed
368 369 370
    if(s->out_format == FMT_MPEG1)
        mpeg_motion_internal(s, dest_y, dest_cb, dest_cr, field_based,
                    bottom_field, field_select, ref_picture, pix_op,
371
                    motion_x, motion_y, h, 1, mb_y);
Keiji Costantini's avatar
Keiji Costantini committed
372 373 374 375
    else
#endif
        mpeg_motion_internal(s, dest_y, dest_cb, dest_cr, field_based,
                    bottom_field, field_select, ref_picture, pix_op,
376
                    motion_x, motion_y, h, 0, mb_y);
Keiji Costantini's avatar
Keiji Costantini committed
377
}
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506

//FIXME move to dsputil, avg variant, 16x16 version
static inline void put_obmc(uint8_t *dst, uint8_t *src[5], int stride){
    int x;
    uint8_t * const top   = src[1];
    uint8_t * const left  = src[2];
    uint8_t * const mid   = src[0];
    uint8_t * const right = src[3];
    uint8_t * const bottom= src[4];
#define OBMC_FILTER(x, t, l, m, r, b)\
    dst[x]= (t*top[x] + l*left[x] + m*mid[x] + r*right[x] + b*bottom[x] + 4)>>3
#define OBMC_FILTER4(x, t, l, m, r, b)\
    OBMC_FILTER(x         , t, l, m, r, b);\
    OBMC_FILTER(x+1       , t, l, m, r, b);\
    OBMC_FILTER(x  +stride, t, l, m, r, b);\
    OBMC_FILTER(x+1+stride, t, l, m, r, b);

    x=0;
    OBMC_FILTER (x  , 2, 2, 4, 0, 0);
    OBMC_FILTER (x+1, 2, 1, 5, 0, 0);
    OBMC_FILTER4(x+2, 2, 1, 5, 0, 0);
    OBMC_FILTER4(x+4, 2, 0, 5, 1, 0);
    OBMC_FILTER (x+6, 2, 0, 5, 1, 0);
    OBMC_FILTER (x+7, 2, 0, 4, 2, 0);
    x+= stride;
    OBMC_FILTER (x  , 1, 2, 5, 0, 0);
    OBMC_FILTER (x+1, 1, 2, 5, 0, 0);
    OBMC_FILTER (x+6, 1, 0, 5, 2, 0);
    OBMC_FILTER (x+7, 1, 0, 5, 2, 0);
    x+= stride;
    OBMC_FILTER4(x  , 1, 2, 5, 0, 0);
    OBMC_FILTER4(x+2, 1, 1, 6, 0, 0);
    OBMC_FILTER4(x+4, 1, 0, 6, 1, 0);
    OBMC_FILTER4(x+6, 1, 0, 5, 2, 0);
    x+= 2*stride;
    OBMC_FILTER4(x  , 0, 2, 5, 0, 1);
    OBMC_FILTER4(x+2, 0, 1, 6, 0, 1);
    OBMC_FILTER4(x+4, 0, 0, 6, 1, 1);
    OBMC_FILTER4(x+6, 0, 0, 5, 2, 1);
    x+= 2*stride;
    OBMC_FILTER (x  , 0, 2, 5, 0, 1);
    OBMC_FILTER (x+1, 0, 2, 5, 0, 1);
    OBMC_FILTER4(x+2, 0, 1, 5, 0, 2);
    OBMC_FILTER4(x+4, 0, 0, 5, 1, 2);
    OBMC_FILTER (x+6, 0, 0, 5, 2, 1);
    OBMC_FILTER (x+7, 0, 0, 5, 2, 1);
    x+= stride;
    OBMC_FILTER (x  , 0, 2, 4, 0, 2);
    OBMC_FILTER (x+1, 0, 1, 5, 0, 2);
    OBMC_FILTER (x+6, 0, 0, 5, 1, 2);
    OBMC_FILTER (x+7, 0, 0, 4, 2, 2);
}

/* obmc for 1 8x8 luma block */
static inline void obmc_motion(MpegEncContext *s,
                               uint8_t *dest, uint8_t *src,
                               int src_x, int src_y,
                               op_pixels_func *pix_op,
                               int16_t mv[5][2]/* mid top left right bottom*/)
#define MID    0
{
    int i;
    uint8_t *ptr[5];

    assert(s->quarter_sample==0);

    for(i=0; i<5; i++){
        if(i && mv[i][0]==mv[MID][0] && mv[i][1]==mv[MID][1]){
            ptr[i]= ptr[MID];
        }else{
            ptr[i]= s->obmc_scratchpad + 8*(i&1) + s->linesize*8*(i>>1);
            hpel_motion(s, ptr[i], src, 0, 0,
                        src_x, src_y,
                        s->width, s->height, s->linesize,
                        s->h_edge_pos, s->v_edge_pos,
                        8, 8, pix_op,
                        mv[i][0], mv[i][1]);
        }
    }

    put_obmc(dest, ptr, s->linesize);
}

static inline void qpel_motion(MpegEncContext *s,
                               uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
                               int field_based, int bottom_field, int field_select,
                               uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
                               qpel_mc_func (*qpix_op)[16],
                               int motion_x, int motion_y, int h)
{
    uint8_t *ptr_y, *ptr_cb, *ptr_cr;
    int dxy, uvdxy, mx, my, src_x, src_y, uvsrc_x, uvsrc_y, v_edge_pos, linesize, uvlinesize;

    dxy = ((motion_y & 3) << 2) | (motion_x & 3);
    src_x = s->mb_x *  16                 + (motion_x >> 2);
    src_y = s->mb_y * (16 >> field_based) + (motion_y >> 2);

    v_edge_pos = s->v_edge_pos >> field_based;
    linesize = s->linesize << field_based;
    uvlinesize = s->uvlinesize << field_based;

    if(field_based){
        mx= motion_x/2;
        my= motion_y>>1;
    }else if(s->workaround_bugs&FF_BUG_QPEL_CHROMA2){
        static const int rtab[8]= {0,0,1,1,0,0,0,1};
        mx= (motion_x>>1) + rtab[motion_x&7];
        my= (motion_y>>1) + rtab[motion_y&7];
    }else if(s->workaround_bugs&FF_BUG_QPEL_CHROMA){
        mx= (motion_x>>1)|(motion_x&1);
        my= (motion_y>>1)|(motion_y&1);
    }else{
        mx= motion_x/2;
        my= motion_y/2;
    }
    mx= (mx>>1)|(mx&1);
    my= (my>>1)|(my&1);

    uvdxy= (mx&1) | ((my&1)<<1);
    mx>>=1;
    my>>=1;

    uvsrc_x = s->mb_x *  8                 + mx;
    uvsrc_y = s->mb_y * (8 >> field_based) + my;

    ptr_y  = ref_picture[0] +   src_y *   linesize +   src_x;
    ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x;
    ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x;

507 508
    if(   (unsigned)src_x > FFMAX(s->h_edge_pos - (motion_x&3) - 16, 0)
       || (unsigned)src_y > FFMAX(   v_edge_pos - (motion_y&3) - h , 0)){
509
        s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize,
510 511
                            17, 17+field_based, src_x, src_y<<field_based,
                            s->h_edge_pos, s->v_edge_pos);
512
        ptr_y= s->edge_emu_buffer;
513
        if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
514
            uint8_t *uvbuf= s->edge_emu_buffer + 18*s->linesize;
515
            s->dsp.emulated_edge_mc(uvbuf, ptr_cb, s->uvlinesize,
516 517 518
                                9, 9 + field_based,
                                uvsrc_x, uvsrc_y<<field_based,
                                s->h_edge_pos>>1, s->v_edge_pos>>1);
519
            s->dsp.emulated_edge_mc(uvbuf + 16, ptr_cr, s->uvlinesize,
520 521 522
                                9, 9 + field_based,
                                uvsrc_x, uvsrc_y<<field_based,
                                s->h_edge_pos>>1, s->v_edge_pos>>1);
523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546
            ptr_cb= uvbuf;
            ptr_cr= uvbuf + 16;
        }
    }

    if(!field_based)
        qpix_op[0][dxy](dest_y, ptr_y, linesize);
    else{
        if(bottom_field){
            dest_y += s->linesize;
            dest_cb+= s->uvlinesize;
            dest_cr+= s->uvlinesize;
        }

        if(field_select){
            ptr_y  += s->linesize;
            ptr_cb += s->uvlinesize;
            ptr_cr += s->uvlinesize;
        }
        //damn interlaced mode
        //FIXME boundary mirroring is not exactly correct here
        qpix_op[1][dxy](dest_y  , ptr_y  , linesize);
        qpix_op[1][dxy](dest_y+8, ptr_y+8, linesize);
    }
547
    if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
548 549 550 551 552 553
        pix_op[1][uvdxy](dest_cr, ptr_cr, uvlinesize, h >> 1);
        pix_op[1][uvdxy](dest_cb, ptr_cb, uvlinesize, h >> 1);
    }
}

/**
Panagiotis Issaris's avatar
Panagiotis Issaris committed
554
 * h263 chroma 4mv motion compensation.
555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
 */
static inline void chroma_4mv_motion(MpegEncContext *s,
                                     uint8_t *dest_cb, uint8_t *dest_cr,
                                     uint8_t **ref_picture,
                                     op_pixels_func *pix_op,
                                     int mx, int my){
    int dxy, emu=0, src_x, src_y, offset;
    uint8_t *ptr;

    /* In case of 8X8, we construct a single chroma motion vector
       with a special rounding */
    mx= ff_h263_round_chroma(mx);
    my= ff_h263_round_chroma(my);

    dxy = ((my & 1) << 1) | (mx & 1);
    mx >>= 1;
    my >>= 1;

    src_x = s->mb_x * 8 + mx;
    src_y = s->mb_y * 8 + my;
575 576
    src_x = av_clip(src_x, -8, (s->width >> 1));
    if (src_x == (s->width >> 1))
577
        dxy &= ~1;
578 579
    src_y = av_clip(src_y, -8, (s->height >> 1));
    if (src_y == (s->height >> 1))
580 581
        dxy &= ~2;

582
    offset = src_y * s->uvlinesize + src_x;
583 584
    ptr = ref_picture[1] + offset;
    if(s->flags&CODEC_FLAG_EMU_EDGE){
585 586
        if(   (unsigned)src_x > FFMAX((s->h_edge_pos>>1) - (dxy &1) - 8, 0)
           || (unsigned)src_y > FFMAX((s->v_edge_pos>>1) - (dxy>>1) - 8, 0)){
587
            s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize,
588 589
                                9, 9, src_x, src_y,
                                s->h_edge_pos>>1, s->v_edge_pos>>1);
590 591 592 593 594 595 596 597
            ptr= s->edge_emu_buffer;
            emu=1;
        }
    }
    pix_op[dxy](dest_cb, ptr, s->uvlinesize, 8);

    ptr = ref_picture[2] + offset;
    if(emu){
598
        s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize,
599 600
                            9, 9, src_x, src_y,
                            s->h_edge_pos>>1, s->v_edge_pos>>1);
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625
        ptr= s->edge_emu_buffer;
    }
    pix_op[dxy](dest_cr, ptr, s->uvlinesize, 8);
}

static inline void prefetch_motion(MpegEncContext *s, uint8_t **pix, int dir){
    /* fetch pixels for estimated mv 4 macroblocks ahead
     * optimized for 64byte cache lines */
    const int shift = s->quarter_sample ? 2 : 1;
    const int mx= (s->mv[dir][0][0]>>shift) + 16*s->mb_x + 8;
    const int my= (s->mv[dir][0][1]>>shift) + 16*s->mb_y;
    int off= mx + (my + (s->mb_x&3)*4)*s->linesize + 64;
    s->dsp.prefetch(pix[0]+off, s->linesize, 4);
    off= (mx>>1) + ((my>>1) + (s->mb_x&7))*s->uvlinesize + 64;
    s->dsp.prefetch(pix[1]+off, pix[2]-pix[1], 2);
}

/**
 * motion compensation of a single macroblock
 * @param s context
 * @param dest_y luma destination pointer
 * @param dest_cb chroma cb/u destination pointer
 * @param dest_cr chroma cr/v destination pointer
 * @param dir direction (0->forward, 1->backward)
 * @param ref_picture array[3] of pointers to the 3 planes of the reference picture
626 627
 * @param pix_op halfpel motion compensation function (average or put normally)
 * @param qpix_op qpel motion compensation function (average or put normally)
628 629
 * the motion vectors are taken from s->mv and the MV type from s->mv_type
 */
630
static av_always_inline void MPV_motion_internal(MpegEncContext *s,
631 632 633 634
                              uint8_t *dest_y, uint8_t *dest_cb,
                              uint8_t *dest_cr, int dir,
                              uint8_t **ref_picture,
                              op_pixels_func (*pix_op)[4],
Keiji Costantini's avatar
Keiji Costantini committed
635
                              qpel_mc_func (*qpix_op)[16], int is_mpeg12)
636 637 638 639 640 641 642 643 644 645
{
    int dxy, mx, my, src_x, src_y, motion_x, motion_y;
    int mb_x, mb_y, i;
    uint8_t *ptr, *dest;

    mb_x = s->mb_x;
    mb_y = s->mb_y;

    prefetch_motion(s, ref_picture, dir);

646
    if(!is_mpeg12 && s->obmc && s->pict_type != AV_PICTURE_TYPE_B){
647 648 649 650 651 652 653
        int16_t mv_cache[4][4][2];
        const int xy= s->mb_x + s->mb_y*s->mb_stride;
        const int mot_stride= s->b8_stride;
        const int mot_xy= mb_x*2 + mb_y*2*mot_stride;

        assert(!s->mb_skipped);

654 655 656
        memcpy(mv_cache[1][1], s->current_picture.f.motion_val[0][mot_xy             ], sizeof(int16_t) * 4);
        memcpy(mv_cache[2][1], s->current_picture.f.motion_val[0][mot_xy + mot_stride], sizeof(int16_t) * 4);
        memcpy(mv_cache[3][1], s->current_picture.f.motion_val[0][mot_xy + mot_stride], sizeof(int16_t) * 4);
657

658
        if (mb_y == 0 || IS_INTRA(s->current_picture.f.mb_type[xy - s->mb_stride])) {
659 660
            memcpy(mv_cache[0][1], mv_cache[1][1], sizeof(int16_t)*4);
        }else{
661
            memcpy(mv_cache[0][1], s->current_picture.f.motion_val[0][mot_xy - mot_stride], sizeof(int16_t) * 4);
662 663
        }

664
        if (mb_x == 0 || IS_INTRA(s->current_picture.f.mb_type[xy - 1])) {
665 666
            AV_COPY32(mv_cache[1][0], mv_cache[1][1]);
            AV_COPY32(mv_cache[2][0], mv_cache[2][1]);
667
        }else{
668 669
            AV_COPY32(mv_cache[1][0], s->current_picture.f.motion_val[0][mot_xy - 1]);
            AV_COPY32(mv_cache[2][0], s->current_picture.f.motion_val[0][mot_xy - 1 + mot_stride]);
670 671
        }

672
        if (mb_x + 1 >= s->mb_width || IS_INTRA(s->current_picture.f.mb_type[xy + 1])) {
673 674
            AV_COPY32(mv_cache[1][3], mv_cache[1][2]);
            AV_COPY32(mv_cache[2][3], mv_cache[2][2]);
675
        }else{
676 677
            AV_COPY32(mv_cache[1][3], s->current_picture.f.motion_val[0][mot_xy + 2]);
            AV_COPY32(mv_cache[2][3], s->current_picture.f.motion_val[0][mot_xy + 2 + mot_stride]);
678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700
        }

        mx = 0;
        my = 0;
        for(i=0;i<4;i++) {
            const int x= (i&1)+1;
            const int y= (i>>1)+1;
            int16_t mv[5][2]= {
                {mv_cache[y][x  ][0], mv_cache[y][x  ][1]},
                {mv_cache[y-1][x][0], mv_cache[y-1][x][1]},
                {mv_cache[y][x-1][0], mv_cache[y][x-1][1]},
                {mv_cache[y][x+1][0], mv_cache[y][x+1][1]},
                {mv_cache[y+1][x][0], mv_cache[y+1][x][1]}};
            //FIXME cleanup
            obmc_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize,
                        ref_picture[0],
                        mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8,
                        pix_op[1],
                        mv);

            mx += mv[0][0];
            my += mv[0][1];
        }
701
        if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY))
702 703 704 705 706 707 708 709 710 711 712 713 714 715 716
            chroma_4mv_motion(s, dest_cb, dest_cr, ref_picture, pix_op[1], mx, my);

        return;
    }

    switch(s->mv_type) {
    case MV_TYPE_16X16:
        if(s->mcsel){
            if(s->real_sprite_warping_points==1){
                gmc1_motion(s, dest_y, dest_cb, dest_cr,
                            ref_picture);
            }else{
                gmc_motion(s, dest_y, dest_cb, dest_cr,
                            ref_picture);
            }
Keiji Costantini's avatar
Keiji Costantini committed
717
        }else if(!is_mpeg12 && s->quarter_sample){
718 719 720 721
            qpel_motion(s, dest_y, dest_cb, dest_cr,
                        0, 0, 0,
                        ref_picture, pix_op, qpix_op,
                        s->mv[dir][0][0], s->mv[dir][0][1], 16);
722
        }else if(!is_mpeg12 && (CONFIG_WMV2_DECODER || CONFIG_WMV2_ENCODER) && s->mspel){
723 724 725 726 727 728 729 730
            ff_mspel_motion(s, dest_y, dest_cb, dest_cr,
                        ref_picture, pix_op,
                        s->mv[dir][0][0], s->mv[dir][0][1], 16);
        }else
        {
            mpeg_motion(s, dest_y, dest_cb, dest_cr,
                        0, 0, 0,
                        ref_picture, pix_op,
731
                        s->mv[dir][0][0], s->mv[dir][0][1], 16, mb_y);
732 733 734
        }
        break;
    case MV_TYPE_8X8:
Keiji Costantini's avatar
Keiji Costantini committed
735
    if (!is_mpeg12) {
736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756
        mx = 0;
        my = 0;
        if(s->quarter_sample){
            for(i=0;i<4;i++) {
                motion_x = s->mv[dir][i][0];
                motion_y = s->mv[dir][i][1];

                dxy = ((motion_y & 3) << 2) | (motion_x & 3);
                src_x = mb_x * 16 + (motion_x >> 2) + (i & 1) * 8;
                src_y = mb_y * 16 + (motion_y >> 2) + (i >>1) * 8;

                /* WARNING: do no forget half pels */
                src_x = av_clip(src_x, -16, s->width);
                if (src_x == s->width)
                    dxy &= ~3;
                src_y = av_clip(src_y, -16, s->height);
                if (src_y == s->height)
                    dxy &= ~12;

                ptr = ref_picture[0] + (src_y * s->linesize) + (src_x);
                if(s->flags&CODEC_FLAG_EMU_EDGE){
757 758
                    if(   (unsigned)src_x > FFMAX(s->h_edge_pos - (motion_x&3) - 8, 0)
                       || (unsigned)src_y > FFMAX(s->v_edge_pos - (motion_y&3) - 8, 0)){
759
                        s->dsp.emulated_edge_mc(s->edge_emu_buffer, ptr,
760 761 762
                                            s->linesize, 9, 9,
                                            src_x, src_y,
                                            s->h_edge_pos, s->v_edge_pos);
763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786
                        ptr= s->edge_emu_buffer;
                    }
                }
                dest = dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize;
                qpix_op[1][dxy](dest, ptr, s->linesize);

                mx += s->mv[dir][i][0]/2;
                my += s->mv[dir][i][1]/2;
            }
        }else{
            for(i=0;i<4;i++) {
                hpel_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize,
                            ref_picture[0], 0, 0,
                            mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8,
                            s->width, s->height, s->linesize,
                            s->h_edge_pos, s->v_edge_pos,
                            8, 8, pix_op[1],
                            s->mv[dir][i][0], s->mv[dir][i][1]);

                mx += s->mv[dir][i][0];
                my += s->mv[dir][i][1];
            }
        }

787
        if(!CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY))
788
            chroma_4mv_motion(s, dest_cb, dest_cr, ref_picture, pix_op[1], mx, my);
Keiji Costantini's avatar
Keiji Costantini committed
789
    }
790 791 792
        break;
    case MV_TYPE_FIELD:
        if (s->picture_structure == PICT_FRAME) {
Keiji Costantini's avatar
Keiji Costantini committed
793
            if(!is_mpeg12 && s->quarter_sample){
794 795 796 797 798 799 800 801 802 803 804
                for(i=0; i<2; i++){
                    qpel_motion(s, dest_y, dest_cb, dest_cr,
                                1, i, s->field_select[dir][i],
                                ref_picture, pix_op, qpix_op,
                                s->mv[dir][i][0], s->mv[dir][i][1], 8);
                }
            }else{
                /* top field */
                mpeg_motion(s, dest_y, dest_cb, dest_cr,
                            1, 0, s->field_select[dir][0],
                            ref_picture, pix_op,
805
                            s->mv[dir][0][0], s->mv[dir][0][1], 8, mb_y);
806 807 808 809
                /* bottom field */
                mpeg_motion(s, dest_y, dest_cb, dest_cr,
                            1, 1, s->field_select[dir][1],
                            ref_picture, pix_op,
810
                            s->mv[dir][1][0], s->mv[dir][1][1], 8, mb_y);
811 812
            }
        } else {
813
            if(s->picture_structure != s->field_select[dir][0] + 1 && s->pict_type != AV_PICTURE_TYPE_B && !s->first_field){
814
                ref_picture = s->current_picture_ptr->f.data;
815 816 817 818 819
            }

            mpeg_motion(s, dest_y, dest_cb, dest_cr,
                        0, 0, s->field_select[dir][0],
                        ref_picture, pix_op,
820
                        s->mv[dir][0][0], s->mv[dir][0][1], 16, mb_y>>1);
821 822 823 824 825 826
        }
        break;
    case MV_TYPE_16X8:
        for(i=0; i<2; i++){
            uint8_t ** ref2picture;

827
            if(s->picture_structure == s->field_select[dir][i] + 1
828
               || s->pict_type == AV_PICTURE_TYPE_B || s->first_field){
829 830
                ref2picture= ref_picture;
            }else{
831
                ref2picture = s->current_picture_ptr->f.data;
832 833 834 835 836
            }

            mpeg_motion(s, dest_y, dest_cb, dest_cr,
                        0, 0, s->field_select[dir][i],
                        ref2picture, pix_op,
837
                        s->mv[dir][i][0], s->mv[dir][i][1] + 16*i, 8, mb_y>>1);
838 839 840 841 842 843 844 845 846 847 848 849 850 851

            dest_y += 16*s->linesize;
            dest_cb+= (16>>s->chroma_y_shift)*s->uvlinesize;
            dest_cr+= (16>>s->chroma_y_shift)*s->uvlinesize;
        }
        break;
    case MV_TYPE_DMV:
        if(s->picture_structure == PICT_FRAME){
            for(i=0; i<2; i++){
                int j;
                for(j=0; j<2; j++){
                    mpeg_motion(s, dest_y, dest_cb, dest_cr,
                                1, j, j^i,
                                ref_picture, pix_op,
852
                                s->mv[dir][2*i + j][0], s->mv[dir][2*i + j][1], 8, mb_y);
853 854 855 856 857 858 859 860
                }
                pix_op = s->dsp.avg_pixels_tab;
            }
        }else{
            for(i=0; i<2; i++){
                mpeg_motion(s, dest_y, dest_cb, dest_cr,
                            0, 0, s->picture_structure != i+1,
                            ref_picture, pix_op,
861
                            s->mv[dir][2*i][0],s->mv[dir][2*i][1],16, mb_y>>1);
862 863 864 865 866 867

                // after put we make avg of the same block
                pix_op=s->dsp.avg_pixels_tab;

                //opposite parity is always in the same frame if this is second field
                if(!s->first_field){
868
                    ref_picture = s->current_picture_ptr->f.data;
869 870 871 872 873 874 875 876
                }
            }
        }
    break;
    default: assert(0);
    }
}

Keiji Costantini's avatar
Keiji Costantini committed
877 878 879 880 881 882 883
static inline void MPV_motion(MpegEncContext *s,
                              uint8_t *dest_y, uint8_t *dest_cb,
                              uint8_t *dest_cr, int dir,
                              uint8_t **ref_picture,
                              op_pixels_func (*pix_op)[4],
                              qpel_mc_func (*qpix_op)[16])
{
884
#if !CONFIG_SMALL
Keiji Costantini's avatar
Keiji Costantini committed
885 886 887 888 889 890 891 892
    if(s->out_format == FMT_MPEG1)
        MPV_motion_internal(s, dest_y, dest_cb, dest_cr, dir,
                            ref_picture, pix_op, qpix_op, 1);
    else
#endif
        MPV_motion_internal(s, dest_y, dest_cb, dest_cr, dir,
                            ref_picture, pix_op, qpix_op, 0);
}
893
#endif /* AVCODEC_MPEGVIDEO_COMMON_H */