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 1.9 2004-12-05 04:53:01 syskin Exp $ |
* $Id: estimation_rd_based.c,v 1.10 2004-12-09 04:58:12 syskin Exp $ |
24 |
* |
* |
25 |
****************************************************************************/ |
****************************************************************************/ |
26 |
|
|
57 |
const int block, |
const int block, |
58 |
const uint16_t * scan_table, |
const uint16_t * scan_table, |
59 |
const unsigned int lambda, |
const unsigned int lambda, |
60 |
const uint16_t * mpeg_quant_matrices) |
const uint16_t * mpeg_quant_matrices, |
61 |
|
const unsigned int quant_sq) |
62 |
{ |
{ |
63 |
int sum; |
int sum; |
64 |
int bits; |
int bits; |
94 |
} |
} |
95 |
|
|
96 |
|
|
97 |
return bits + (lambda*distortion)/(quant*quant); |
return bits + (lambda*distortion)/quant_sq; |
98 |
} |
} |
99 |
|
|
100 |
static __inline unsigned int |
static __inline unsigned int |
112 |
unsigned int bits[2], |
unsigned int bits[2], |
113 |
unsigned int cbp[2], |
unsigned int cbp[2], |
114 |
unsigned int lambda, |
unsigned int lambda, |
115 |
const uint16_t * mpeg_quant_matrices) |
const uint16_t * mpeg_quant_matrices, |
116 |
|
const unsigned int quant_sq) |
117 |
{ |
{ |
118 |
int direction; |
int direction; |
119 |
int16_t *pCurrent; |
int16_t *pCurrent; |
173 |
|
|
174 |
distortion = sse8_16bit(coeff, dqcoeff, 8*sizeof(int16_t)); |
distortion = sse8_16bit(coeff, dqcoeff, 8*sizeof(int16_t)); |
175 |
|
|
176 |
return (lambda*distortion)/(quant*quant); |
return (lambda*distortion)/quant_sq; |
177 |
} |
} |
178 |
|
|
179 |
|
|
205 |
for(i = 0; i < 4; i++) { |
for(i = 0; i < 4; i++) { |
206 |
int s = 8*((i&1) + (i>>1)*data->iEdgedWidth); |
int s = 8*((i&1) + (i>>1)*data->iEdgedWidth); |
207 |
transfer_8to16subro(in, data->Cur + s, ptr + s, data->iEdgedWidth); |
transfer_8to16subro(in, data->Cur + s, ptr + s, data->iEdgedWidth); |
208 |
rd += data->temp[i] = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, i, data->scan_table, data->lambda[i], data->mpeg_quant_matrices); |
rd += data->temp[i] = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, |
209 |
|
data->quant_type, &cbp, i, data->scan_table, data->lambda[i], |
210 |
|
data->mpeg_quant_matrices, data->quant_sq); |
211 |
} |
} |
212 |
|
|
213 |
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); |
232 |
/* chroma U */ |
/* chroma U */ |
233 |
ptr = interpolate8x8_switch2(data->RefQ, data->RefP[4], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); |
ptr = interpolate8x8_switch2(data->RefQ, data->RefP[4], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); |
234 |
transfer_8to16subro(in, data->CurU, ptr, data->iEdgedWidth/2); |
transfer_8to16subro(in, data->CurU, ptr, data->iEdgedWidth/2); |
235 |
rd += Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 4, data->scan_table, data->lambda[4], data->mpeg_quant_matrices); |
rd += Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, |
236 |
|
&cbp, 4, data->scan_table, data->lambda[4], |
237 |
|
data->mpeg_quant_matrices, data->quant_sq); |
238 |
if (rd >= data->iMinSAD[0]) return; |
if (rd >= data->iMinSAD[0]) return; |
239 |
|
|
240 |
/* chroma V */ |
/* chroma V */ |
241 |
ptr = interpolate8x8_switch2(data->RefQ, data->RefP[5], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); |
ptr = interpolate8x8_switch2(data->RefQ, data->RefP[5], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); |
242 |
transfer_8to16subro(in, data->CurV, ptr, data->iEdgedWidth/2); |
transfer_8to16subro(in, data->CurV, ptr, data->iEdgedWidth/2); |
243 |
rd += Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 5, data->scan_table, data->lambda[5], data->mpeg_quant_matrices); |
rd += Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, |
244 |
|
&cbp, 5, data->scan_table, data->lambda[5], |
245 |
|
data->mpeg_quant_matrices, data->quant_sq); |
246 |
|
|
247 |
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; |
248 |
|
|
276 |
} |
} |
277 |
|
|
278 |
transfer_8to16subro(in, data->Cur, ptr, data->iEdgedWidth); |
transfer_8to16subro(in, data->Cur, ptr, data->iEdgedWidth); |
279 |
rd = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 5, data->scan_table, data->lambda[0], data->mpeg_quant_matrices); |
rd = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, |
280 |
|
&cbp, 5, data->scan_table, data->lambda[0], |
281 |
|
data->mpeg_quant_matrices, data->quant_sq); |
282 |
rd += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision); |
rd += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision); |
283 |
|
|
284 |
if (rd < data->iMinSAD[0]) { |
if (rd < data->iMinSAD[0]) { |
501 |
/* chroma U */ |
/* chroma U */ |
502 |
ptr = interpolate8x8_switch2(Data->RefQ + 64, Data->RefP[4], 0, 0, sumx, sumy, Data->iEdgedWidth/2, Data->rounding); |
ptr = interpolate8x8_switch2(Data->RefQ + 64, Data->RefP[4], 0, 0, sumx, sumy, Data->iEdgedWidth/2, Data->rounding); |
503 |
transfer_8to16subro(in, Data->CurU, ptr, Data->iEdgedWidth/2); |
transfer_8to16subro(in, Data->CurU, ptr, Data->iEdgedWidth/2); |
504 |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 4, Data->scan_table, Data->lambda[4], Data->mpeg_quant_matrices); |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 4, |
505 |
|
Data->scan_table, Data->lambda[4], Data->mpeg_quant_matrices, Data->quant_sq); |
506 |
|
|
507 |
if (bits >= *Data->iMinSAD) return bits; |
if (bits >= *Data->iMinSAD) return bits; |
508 |
|
|
509 |
/* chroma V */ |
/* chroma V */ |
510 |
ptr = interpolate8x8_switch2(Data->RefQ + 64, Data->RefP[5], 0, 0, sumx, sumy, Data->iEdgedWidth/2, Data->rounding); |
ptr = interpolate8x8_switch2(Data->RefQ + 64, Data->RefP[5], 0, 0, sumx, sumy, Data->iEdgedWidth/2, Data->rounding); |
511 |
transfer_8to16subro(in, Data->CurV, ptr, Data->iEdgedWidth/2); |
transfer_8to16subro(in, Data->CurV, ptr, Data->iEdgedWidth/2); |
512 |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices); |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, |
513 |
|
Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq); |
514 |
|
|
515 |
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; |
516 |
|
|
536 |
|
|
537 |
|
|
538 |
distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, i, in, coeff, dqcoeff, |
distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, i, in, coeff, dqcoeff, |
539 |
predictors[i], iQuant, Data->quant_type, bits, cbp, Data->lambda[i], Data->mpeg_quant_matrices); |
predictors[i], iQuant, Data->quant_type, bits, cbp, |
540 |
|
Data->lambda[i], Data->mpeg_quant_matrices, Data->quant_sq); |
541 |
bits1 += distortion + BITS_MULT * bits[0]; |
bits1 += distortion + BITS_MULT * bits[0]; |
542 |
bits2 += distortion + BITS_MULT * bits[1]; |
bits2 += distortion + BITS_MULT * bits[1]; |
543 |
|
|
551 |
/*chroma U */ |
/*chroma U */ |
552 |
transfer_8to16copy(in, Data->CurU, Data->iEdgedWidth/2); |
transfer_8to16copy(in, Data->CurU, Data->iEdgedWidth/2); |
553 |
distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, 4, in, coeff, dqcoeff, |
distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, 4, in, coeff, dqcoeff, |
554 |
predictors[4], iQuant, Data->quant_type, bits, cbp, Data->lambda[4], Data->mpeg_quant_matrices); |
predictors[4], iQuant, Data->quant_type, bits, cbp, |
555 |
|
Data->lambda[4], Data->mpeg_quant_matrices, Data->quant_sq); |
556 |
bits1 += distortion + BITS_MULT * bits[0]; |
bits1 += distortion + BITS_MULT * bits[0]; |
557 |
bits2 += distortion + BITS_MULT * bits[1]; |
bits2 += distortion + BITS_MULT * bits[1]; |
558 |
|
|
562 |
/* chroma V */ |
/* chroma V */ |
563 |
transfer_8to16copy(in, Data->CurV, Data->iEdgedWidth/2); |
transfer_8to16copy(in, Data->CurV, Data->iEdgedWidth/2); |
564 |
distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, 5, in, coeff, dqcoeff, |
distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, 5, in, coeff, dqcoeff, |
565 |
predictors[5], iQuant, Data->quant_type, bits, cbp, Data->lambda[5], Data->mpeg_quant_matrices); |
predictors[5], iQuant, Data->quant_type, bits, cbp, |
566 |
|
Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq); |
567 |
|
|
568 |
bits1 += distortion + BITS_MULT * bits[0]; |
bits1 += distortion + BITS_MULT * bits[0]; |
569 |
bits2 += distortion + BITS_MULT * bits[1]; |
bits2 += distortion + BITS_MULT * bits[1]; |
587 |
for(i = 0; i < 4; i++) { |
for(i = 0; i < 4; i++) { |
588 |
int s = 8*((i&1) + (i>>1)*Data->iEdgedWidth); |
int s = 8*((i&1) + (i>>1)*Data->iEdgedWidth); |
589 |
transfer_8to16subro(in, Data->Cur + s, vGMC->y + s + 16*(x+y*Data->iEdgedWidth), Data->iEdgedWidth); |
transfer_8to16subro(in, Data->Cur + s, vGMC->y + s + 16*(x+y*Data->iEdgedWidth), Data->iEdgedWidth); |
590 |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, i, Data->scan_table, Data->lambda[i], Data->mpeg_quant_matrices); |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, i, |
591 |
|
Data->scan_table, Data->lambda[i], Data->mpeg_quant_matrices, Data->quant_sq); |
592 |
if (bits >= Data->iMinSAD[0]) return bits; |
if (bits >= Data->iMinSAD[0]) return bits; |
593 |
} |
} |
594 |
|
|
596 |
|
|
597 |
/*chroma U */ |
/*chroma U */ |
598 |
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); |
599 |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 4, Data->scan_table, Data->lambda[4], Data->mpeg_quant_matrices); |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 4, |
600 |
|
Data->scan_table, Data->lambda[4], Data->mpeg_quant_matrices, Data->quant_sq); |
601 |
|
|
602 |
if (bits >= Data->iMinSAD[0]) return bits; |
if (bits >= Data->iMinSAD[0]) return bits; |
603 |
|
|
604 |
/* chroma V */ |
/* chroma V */ |
605 |
transfer_8to16subro(in, Data->CurV , vGMC->v + 8*(x+y*(Data->iEdgedWidth/2)), Data->iEdgedWidth/2); |
transfer_8to16subro(in, Data->CurV , vGMC->v + 8*(x+y*(Data->iEdgedWidth/2)), Data->iEdgedWidth/2); |
606 |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices); |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, |
607 |
|
Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq); |
608 |
|
|
609 |
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; |
610 |
|
|
635 |
int min_rd, intra_rd, i, cbp; |
int min_rd, intra_rd, i, cbp; |
636 |
VECTOR backup[5], *v; |
VECTOR backup[5], *v; |
637 |
Data->iQuant = iQuant; |
Data->iQuant = iQuant; |
638 |
|
Data->quant_sq = iQuant*iQuant; |
639 |
Data->scan_table = VopFlags & XVID_VOP_ALTERNATESCAN ? |
Data->scan_table = VopFlags & XVID_VOP_ALTERNATESCAN ? |
640 |
scan_tables[2] : scan_tables[0]; |
scan_tables[2] : scan_tables[0]; |
641 |
|
|
748 |
scan_tables[2] : scan_tables[0]; |
scan_tables[2] : scan_tables[0]; |
749 |
|
|
750 |
pMB->mcsel = 0; |
pMB->mcsel = 0; |
751 |
|
Data->iQuant = iQuant; |
752 |
|
Data->quant_sq = iQuant*iQuant; |
753 |
|
|
754 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
755 |
/* HVS models, anyone ? */ |
/* HVS models, anyone ? */ |