20 |
* along with this program ; if not, write to the Free Software |
* along with this program ; if not, write to the Free Software |
21 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
23 |
* $Id: estimation_rd_based.c,v 1.10 2004-12-09 04:58:12 syskin Exp $ |
* $Id: estimation_rd_based.c,v 1.11 2004-12-18 06:51:14 syskin Exp $ |
24 |
* |
* |
25 |
****************************************************************************/ |
****************************************************************************/ |
26 |
|
|
183 |
{ |
{ |
184 |
|
|
185 |
int16_t *in = data->dctSpace, *coeff = data->dctSpace + 64; |
int16_t *in = data->dctSpace, *coeff = data->dctSpace + 64; |
186 |
int32_t rd = 0; |
/* minimum nuber of bits INTER can take is 1 (mcbpc) + 2 (cby) + 2 (vector) */ |
187 |
|
int32_t rd = BITS_MULT * (1+2+2); |
188 |
VECTOR * current; |
VECTOR * current; |
189 |
const uint8_t * ptr; |
const uint8_t * ptr; |
190 |
int i, t, xc, yc; |
int i, t, xc, yc; |
211 |
data->mpeg_quant_matrices, data->quant_sq); |
data->mpeg_quant_matrices, data->quant_sq); |
212 |
} |
} |
213 |
|
|
214 |
rd += t = BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision); |
rd += t = BITS_MULT * (d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision) - 2); |
215 |
|
|
216 |
if (data->temp[0] + t < data->iMinSAD[1]) { |
if (data->temp[0] + t < data->iMinSAD[1]) { |
217 |
data->iMinSAD[1] = data->temp[0] + t; current[1].x = x; current[1].y = y; data->cbp[1] = (data->cbp[1]&~32) | (cbp&32); } |
data->iMinSAD[1] = data->temp[0] + t; current[1].x = x; current[1].y = y; data->cbp[1] = (data->cbp[1]&~32) | (cbp&32); } |
222 |
if (data->temp[3] < data->iMinSAD[4]) { |
if (data->temp[3] < data->iMinSAD[4]) { |
223 |
data->iMinSAD[4] = data->temp[3]; current[4].x = x; current[4].y = y; data->cbp[1] = (data->cbp[1]&~4) | (cbp&4); } |
data->iMinSAD[4] = data->temp[3]; current[4].x = x; current[4].y = y; data->cbp[1] = (data->cbp[1]&~4) | (cbp&4); } |
224 |
|
|
225 |
rd += BITS_MULT*xvid_cbpy_tab[15-(cbp>>2)].len; |
rd += BITS_MULT * (xvid_cbpy_tab[15-(cbp>>2)].len - 2); |
226 |
|
|
227 |
if (rd >= data->iMinSAD[0]) return; |
if (rd >= data->iMinSAD[0]) return; |
228 |
|
|
245 |
&cbp, 5, data->scan_table, data->lambda[5], |
&cbp, 5, data->scan_table, data->lambda[5], |
246 |
data->mpeg_quant_matrices, data->quant_sq); |
data->mpeg_quant_matrices, data->quant_sq); |
247 |
|
|
248 |
rd += BITS_MULT*mcbpc_inter_tab[(MODE_INTER & 7) | ((cbp & 3) << 3)].len; |
rd += BITS_MULT * (mcbpc_inter_tab[(MODE_INTER & 7) | ((cbp & 3) << 3)].len - 1); /* one was added before */ |
249 |
|
|
250 |
if (rd < data->iMinSAD[0]) { |
if (rd < data->iMinSAD[0]) { |
251 |
data->iMinSAD[0] = rd; |
data->iMinSAD[0] = rd; |
280 |
rd = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, |
rd = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, |
281 |
&cbp, 5, data->scan_table, data->lambda[0], |
&cbp, 5, data->scan_table, data->lambda[0], |
282 |
data->mpeg_quant_matrices, data->quant_sq); |
data->mpeg_quant_matrices, data->quant_sq); |
283 |
rd += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision); |
/* we took 2 bits into account before */ |
284 |
|
rd += BITS_MULT * (d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision) - 2); |
285 |
|
|
286 |
|
|
287 |
if (rd < data->iMinSAD[0]) { |
if (rd < data->iMinSAD[0]) { |
288 |
*data->cbp = cbp; |
*data->cbp = cbp; |
366 |
const VECTOR * const backup) |
const VECTOR * const backup) |
367 |
{ |
{ |
368 |
|
|
369 |
unsigned int cbp = 0, bits = 0, t = 0, i; |
unsigned int cbp = 0, t = 0, i; |
370 |
|
|
371 |
|
/* minimum number of bits INTER4V can take is 2 (cbpy) + 3 (mcbpc) + 4*2 (vectors)*/ |
372 |
|
int bits = (2+3+4*2)*BITS_MULT; |
373 |
SearchData Data2, *Data8 = &Data2; |
SearchData Data2, *Data8 = &Data2; |
374 |
int sumx = 0, sumy = 0; |
int sumx = 0, sumy = 0; |
375 |
int16_t *in = Data->dctSpace, *coeff = Data->dctSpace + 64; |
int16_t *in = Data->dctSpace, *coeff = Data->dctSpace + 64; |
393 |
if(Data->qpel) { |
if(Data->qpel) { |
394 |
Data8->predMV = get_qpmv2(pMBs, pParam->mb_width, 0, x, y, i); |
Data8->predMV = get_qpmv2(pMBs, pParam->mb_width, 0, x, y, i); |
395 |
if (i != 0) t = d_mv_bits( Data8->currentQMV->x, Data8->currentQMV->y, |
if (i != 0) t = d_mv_bits( Data8->currentQMV->x, Data8->currentQMV->y, |
396 |
Data8->predMV, Data8->iFcode, 0); |
Data8->predMV, Data8->iFcode, 0) - 2; |
397 |
} else { |
} else { |
398 |
Data8->predMV = get_pmv2(pMBs, pParam->mb_width, 0, x, y, i); |
Data8->predMV = get_pmv2(pMBs, pParam->mb_width, 0, x, y, i); |
399 |
if (i != 0) t = d_mv_bits( Data8->currentMV->x, Data8->currentMV->y, |
if (i != 0) t = d_mv_bits( Data8->currentMV->x, Data8->currentMV->y, |
400 |
Data8->predMV, Data8->iFcode, 0); |
Data8->predMV, Data8->iFcode, 0) - 2; |
401 |
} |
} |
402 |
|
|
403 |
get_range(&Data8->min_dx, &Data8->max_dx, &Data8->min_dy, &Data8->max_dy, 2*x + (i&1), 2*y + (i>>1), 3, |
get_range(&Data8->min_dx, &Data8->max_dx, &Data8->min_dy, &Data8->max_dy, 2*x + (i&1), 2*y + (i>>1), 3, |
498 |
|
|
499 |
} /* end - for all luma blocks */ |
} /* end - for all luma blocks */ |
500 |
|
|
501 |
bits += BITS_MULT*xvid_cbpy_tab[15-(cbp>>2)].len; |
bits += BITS_MULT * (xvid_cbpy_tab[15-(cbp>>2)].len - 2); /* 2 were added before */ |
502 |
|
|
503 |
/* let's check chroma */ |
/* let's check chroma */ |
504 |
sumx = (sumx >> 3) + roundtab_76[sumx & 0xf]; |
sumx = (sumx >> 3) + roundtab_76[sumx & 0xf]; |
518 |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, |
519 |
Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq); |
Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq); |
520 |
|
|
521 |
bits += BITS_MULT*mcbpc_inter_tab[(MODE_INTER4V & 7) | ((cbp & 3) << 3)].len; |
bits += BITS_MULT*(mcbpc_inter_tab[(MODE_INTER4V & 7) | ((cbp & 3) << 3)].len - 3); /* 3 were added before */ |
522 |
|
|
523 |
*Data->cbp = cbp; |
*Data->cbp = cbp; |
524 |
return bits; |
return bits; |
529 |
const int x, const int y, const int mb_width) |
const int x, const int y, const int mb_width) |
530 |
{ |
{ |
531 |
unsigned int cbp[2] = {0, 0}, bits[2], i; |
unsigned int cbp[2] = {0, 0}, bits[2], i; |
532 |
unsigned int bits1 = BITS_MULT*1, bits2 = BITS_MULT*1; /* this one is ac/dc prediction flag bit */ |
/* minimum number of bits that WILL be coded in intra - MODE 5, CBP 2 and AC/DC pred - 1 */ |
533 |
|
int bits1 = BITS_MULT*(5+2+1), bits2 = BITS_MULT*(5+2+1); |
534 |
unsigned int distortion = 0; |
unsigned int distortion = 0; |
535 |
|
|
536 |
int16_t *in = Data->dctSpace, * coeff = Data->dctSpace + 64, * dqcoeff = Data->dctSpace + 128; |
int16_t *in = Data->dctSpace, * coeff = Data->dctSpace + 64, * dqcoeff = Data->dctSpace + 128; |
552 |
return bits1; |
return bits1; |
553 |
} |
} |
554 |
|
|
555 |
bits1 += BITS_MULT*xvid_cbpy_tab[cbp[0]>>2].len; |
bits1 += BITS_MULT * (xvid_cbpy_tab[cbp[0]>>2].len - 2); /* two bits were added before */ |
556 |
bits2 += BITS_MULT*xvid_cbpy_tab[cbp[1]>>2].len; |
bits2 += BITS_MULT * (xvid_cbpy_tab[cbp[1]>>2].len - 2); |
557 |
|
|
558 |
/*chroma U */ |
/*chroma U */ |
559 |
transfer_8to16copy(in, Data->CurU, Data->iEdgedWidth/2); |
transfer_8to16copy(in, Data->CurU, Data->iEdgedWidth/2); |
575 |
bits1 += distortion + BITS_MULT * bits[0]; |
bits1 += distortion + BITS_MULT * bits[0]; |
576 |
bits2 += distortion + BITS_MULT * bits[1]; |
bits2 += distortion + BITS_MULT * bits[1]; |
577 |
|
|
578 |
bits1 += BITS_MULT*mcbpc_inter_tab[(MODE_INTRA & 7) | ((cbp[0] & 3) << 3)].len; |
bits1 += BITS_MULT * (mcbpc_inter_tab[(MODE_INTRA & 7) | ((cbp[0] & 3) << 3)].len - 5); /* 5 bits were added before */ |
579 |
bits2 += BITS_MULT*mcbpc_inter_tab[(MODE_INTRA & 7) | ((cbp[1] & 3) << 3)].len; |
bits2 += BITS_MULT * (mcbpc_inter_tab[(MODE_INTRA & 7) | ((cbp[1] & 3) << 3)].len - 5); |
580 |
|
|
581 |
*Data->cbp = bits1 <= bits2 ? cbp[0] : cbp[1]; |
*Data->cbp = bits1 <= bits2 ? cbp[0] : cbp[1]; |
582 |
|
|
587 |
static int |
static int |
588 |
findRD_gmc(SearchData * const Data, const IMAGE * const vGMC, const int x, const int y) |
findRD_gmc(SearchData * const Data, const IMAGE * const vGMC, const int x, const int y) |
589 |
{ |
{ |
590 |
int bits = BITS_MULT*1; /* this one is mcsel */ |
/* minimum nubler of bits - 1 (mcbpc) + 2 (cby) + 1 (mcsel) */ |
591 |
|
int bits = BITS_MULT * (1+2+1); |
592 |
unsigned int cbp = 0, i; |
unsigned int cbp = 0, i; |
593 |
int16_t *in = Data->dctSpace, * coeff = Data->dctSpace + 64; |
int16_t *in = Data->dctSpace, * coeff = Data->dctSpace + 64; |
594 |
|
|
600 |
if (bits >= Data->iMinSAD[0]) return bits; |
if (bits >= Data->iMinSAD[0]) return bits; |
601 |
} |
} |
602 |
|
|
603 |
bits += BITS_MULT*xvid_cbpy_tab[15-(cbp>>2)].len; |
bits += BITS_MULT * (xvid_cbpy_tab[15-(cbp>>2)].len - 2); |
604 |
|
|
605 |
/*chroma U */ |
/*chroma U */ |
606 |
transfer_8to16subro(in, Data->CurU, vGMC->u + 8*(x+y*(Data->iEdgedWidth/2)), Data->iEdgedWidth/2); |
transfer_8to16subro(in, Data->CurU, vGMC->u + 8*(x+y*(Data->iEdgedWidth/2)), Data->iEdgedWidth/2); |
614 |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, |
615 |
Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq); |
Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq); |
616 |
|
|
617 |
bits += BITS_MULT*mcbpc_inter_tab[(MODE_INTER & 7) | ((cbp & 3) << 3)].len; |
bits += BITS_MULT * (mcbpc_inter_tab[(MODE_INTER & 7) | ((cbp & 3) << 3)].len - 1); |
618 |
|
|
619 |
*Data->cbp = cbp; |
*Data->cbp = cbp; |
620 |
|
|
639 |
int mcsel = 0; |
int mcsel = 0; |
640 |
int inter4v = (VopFlags & XVID_VOP_INTER4V) && (pMB->dquant == 0); |
int inter4v = (VopFlags & XVID_VOP_INTER4V) && (pMB->dquant == 0); |
641 |
const uint32_t iQuant = pMB->quant; |
const uint32_t iQuant = pMB->quant; |
|
|
|
642 |
int min_rd, intra_rd, i, cbp; |
int min_rd, intra_rd, i, cbp; |
643 |
VECTOR backup[5], *v; |
VECTOR backup[5], *v; |
644 |
Data->iQuant = iQuant; |
Data->iQuant = iQuant; |