56 |
uint32_t * cbp, |
uint32_t * cbp, |
57 |
const int block, |
const int block, |
58 |
const uint16_t * scan_table, |
const uint16_t * scan_table, |
59 |
const uint16_t * mpeg_quant_matrices) |
const unsigned int lambda, |
60 |
|
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 |
111 |
const int quant_type, |
const int quant_type, |
112 |
unsigned int bits[2], |
unsigned int bits[2], |
113 |
unsigned int cbp[2], |
unsigned int cbp[2], |
114 |
const uint16_t * mpeg_quant_matrices) |
unsigned int lambda, |
115 |
|
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; |
132 |
} |
} |
133 |
|
|
134 |
predict_acdc(pMB-(x+mb_width*y), x, y, mb_width, block, qcoeff, |
predict_acdc(pMB-(x+mb_width*y), x, y, mb_width, block, qcoeff, |
135 |
quant, iDcScaler, predictors, 0, 0); |
quant, iDcScaler, predictors, 0); |
136 |
|
|
137 |
direction = pMB->acpred_directions[block]; |
direction = pMB->acpred_directions[block]; |
138 |
pCurrent = pMB->pred_values[block]; |
pCurrent = pMB->pred_values[block]; |
139 |
|
|
140 |
/* store current coeffs to pred_values[] for future prediction */ |
/* store current coeffs to pred_values[] for future prediction */ |
141 |
pCurrent[0] = qcoeff[0] * iDcScaler; |
pCurrent[0] = qcoeff[0] * iDcScaler; |
142 |
|
pCurrent[0] = CLIP(pCurrent[0], -2048, 2047); |
143 |
for (i = 1; i < 8; i++) { |
for (i = 1; i < 8; i++) { |
144 |
pCurrent[i] = qcoeff[i]; |
pCurrent[i] = qcoeff[i]; |
145 |
pCurrent[i + 7] = qcoeff[i * 8]; |
pCurrent[i + 7] = qcoeff[i * 8]; |
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->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, 0); |
rd += t = BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision); |
214 |
|
|
215 |
if (data->temp[0] + t < data->iMinSAD[1]) { |
if (data->temp[0] + t < data->iMinSAD[1]) { |
216 |
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); } |
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->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->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->mpeg_quant_matrices); |
rd = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, |
280 |
rd += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision, 0); |
&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); |
283 |
|
|
284 |
if (rd < data->iMinSAD[0]) { |
if (rd < data->iMinSAD[0]) { |
285 |
*data->cbp = cbp; |
*data->cbp = cbp; |
310 |
if (MotionFlags & (XVID_ME_HALFPELREFINE16_RD | XVID_ME_EXTSEARCH_RD)) { /* we have to prepare for halfpixel-precision search */ |
if (MotionFlags & (XVID_ME_HALFPELREFINE16_RD | XVID_ME_EXTSEARCH_RD)) { /* we have to prepare for halfpixel-precision search */ |
311 |
for(i = 0; i < 5; i++) bsad[i] = Data->iMinSAD[i]; |
for(i = 0; i < 5; i++) bsad[i] = Data->iMinSAD[i]; |
312 |
get_range(&Data->min_dx, &Data->max_dx, &Data->min_dy, &Data->max_dy, x, y, 4, |
get_range(&Data->min_dx, &Data->max_dx, &Data->min_dy, &Data->max_dy, x, y, 4, |
313 |
pParam->width, pParam->height, Data->iFcode - Data->qpel, 1, Data->rrv); |
pParam->width, pParam->height, Data->iFcode - Data->qpel, 1); |
314 |
Data->qpel_precision = 0; |
Data->qpel_precision = 0; |
315 |
if (Data->currentQMV->x & 1 || Data->currentQMV->y & 1) |
if (Data->currentQMV->x & 1 || Data->currentQMV->y & 1) |
316 |
CheckCandidateRD16(Data->currentMV[0].x, Data->currentMV[0].y, Data, 255); |
CheckCandidateRD16(Data->currentMV[0].x, Data->currentMV[0].y, Data, 255); |
325 |
xvid_me_SquareSearch(Data->currentMV->x, Data->currentMV->y, Data, 255, CheckCandidateRD16); |
xvid_me_SquareSearch(Data->currentMV->x, Data->currentMV->y, Data, 255, CheckCandidateRD16); |
326 |
|
|
327 |
if (MotionFlags&XVID_ME_HALFPELREFINE16_RD) |
if (MotionFlags&XVID_ME_HALFPELREFINE16_RD) |
328 |
xvid_me_SubpelRefine(Data, CheckCandidateRD16, 0); |
xvid_me_SubpelRefine(Data->currentMV[0], Data, CheckCandidateRD16, 0); |
329 |
|
|
330 |
if (Data->qpel) { |
if (Data->qpel) { |
331 |
if (MotionFlags&(XVID_ME_EXTSEARCH_RD | XVID_ME_HALFPELREFINE16_RD)) { /* there was halfpel-precision search */ |
if (MotionFlags&(XVID_ME_EXTSEARCH_RD | XVID_ME_HALFPELREFINE16_RD)) { /* there was halfpel-precision search */ |
337 |
/* preparing for qpel-precision search */ |
/* preparing for qpel-precision search */ |
338 |
Data->qpel_precision = 1; |
Data->qpel_precision = 1; |
339 |
get_range(&Data->min_dx, &Data->max_dx, &Data->min_dy, &Data->max_dy, x, y, 4, |
get_range(&Data->min_dx, &Data->max_dx, &Data->min_dy, &Data->max_dy, x, y, 4, |
340 |
pParam->width, pParam->height, Data->iFcode, 2, 0); |
pParam->width, pParam->height, Data->iFcode, 2); |
341 |
} |
} |
342 |
if (MotionFlags & XVID_ME_QUARTERPELREFINE16_RD) { |
if (MotionFlags & XVID_ME_QUARTERPELREFINE16_RD) { |
343 |
if (MotionFlags & XVID_ME_FASTREFINE16) |
if (MotionFlags & XVID_ME_FASTREFINE16) |
344 |
FullRefine_Fast(Data, CheckCandidateRD16, 0); |
FullRefine_Fast(Data, CheckCandidateRD16, 0); |
345 |
else |
else |
346 |
xvid_me_SubpelRefine(Data, CheckCandidateRD16, 0); |
xvid_me_SubpelRefine(Data->currentQMV[0], Data, CheckCandidateRD16, 0); |
347 |
} |
} |
348 |
} |
} |
349 |
|
|
382 |
Data8->RefP[1] = Data->RefP[1] + 8*((i&1) + (i>>1)*Data->iEdgedWidth); |
Data8->RefP[1] = Data->RefP[1] + 8*((i&1) + (i>>1)*Data->iEdgedWidth); |
383 |
Data8->RefP[3] = Data->RefP[3] + 8*((i&1) + (i>>1)*Data->iEdgedWidth); |
Data8->RefP[3] = Data->RefP[3] + 8*((i&1) + (i>>1)*Data->iEdgedWidth); |
384 |
*Data8->cbp = (Data->cbp[1] & (1<<(5-i))) ? 1:0; /* copy corresponding cbp bit */ |
*Data8->cbp = (Data->cbp[1] & (1<<(5-i))) ? 1:0; /* copy corresponding cbp bit */ |
385 |
|
Data8->lambda[0] = Data->lambda[i]; |
386 |
|
|
387 |
if(Data->qpel) { |
if(Data->qpel) { |
388 |
Data8->predMV = get_qpmv2(pMBs, pParam->mb_width, 0, x, y, i); |
Data8->predMV = get_qpmv2(pMBs, pParam->mb_width, 0, x, y, i); |
389 |
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, |
390 |
Data8->predMV, Data8->iFcode, 0, 0); |
Data8->predMV, Data8->iFcode, 0); |
391 |
} else { |
} else { |
392 |
Data8->predMV = get_pmv2(pMBs, pParam->mb_width, 0, x, y, i); |
Data8->predMV = get_pmv2(pMBs, pParam->mb_width, 0, x, y, i); |
393 |
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, |
394 |
Data8->predMV, Data8->iFcode, 0, 0); |
Data8->predMV, Data8->iFcode, 0); |
395 |
} |
} |
396 |
|
|
397 |
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, |
398 |
pParam->width, pParam->height, Data8->iFcode, Data8->qpel+1, 0); |
pParam->width, pParam->height, Data8->iFcode, Data8->qpel+1); |
399 |
|
|
400 |
*Data8->iMinSAD += BITS_MULT*t; |
*Data8->iMinSAD += BITS_MULT*t; |
401 |
|
|
418 |
if (MotionFlags&XVID_ME_HALFPELREFINE8_RD || (MotionFlags&XVID_ME_EXTSEARCH8 && MotionFlags&XVID_ME_EXTSEARCH_RD)) { /* halfpixel motion search follows */ |
if (MotionFlags&XVID_ME_HALFPELREFINE8_RD || (MotionFlags&XVID_ME_EXTSEARCH8 && MotionFlags&XVID_ME_EXTSEARCH_RD)) { /* halfpixel motion search follows */ |
419 |
Data8->qpel_precision = 0; |
Data8->qpel_precision = 0; |
420 |
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, |
421 |
pParam->width, pParam->height, Data8->iFcode - 1, 1, 0); |
pParam->width, pParam->height, Data8->iFcode - 1, 1); |
422 |
|
|
423 |
if (Data8->currentQMV->x & 1 || Data8->currentQMV->y & 1) |
if (Data8->currentQMV->x & 1 || Data8->currentQMV->y & 1) |
424 |
CheckCandidateRD8(Data8->currentMV->x, Data8->currentMV->y, Data8, 255); |
CheckCandidateRD8(Data8->currentMV->x, Data8->currentMV->y, Data8, 255); |
427 |
xvid_me_SquareSearch(Data8->currentMV->x, Data8->currentMV->x, Data8, 255, CheckCandidateRD8); |
xvid_me_SquareSearch(Data8->currentMV->x, Data8->currentMV->x, Data8, 255, CheckCandidateRD8); |
428 |
|
|
429 |
if (MotionFlags & XVID_ME_HALFPELREFINE8_RD) |
if (MotionFlags & XVID_ME_HALFPELREFINE8_RD) |
430 |
xvid_me_SubpelRefine(Data8, CheckCandidateRD8, 0); |
xvid_me_SubpelRefine(Data->currentMV[0], Data8, CheckCandidateRD8, 0); |
431 |
|
|
432 |
if(bsad > *Data8->iMinSAD) { /* we have found a better match */ |
if(bsad > *Data8->iMinSAD) { /* we have found a better match */ |
433 |
bx = Data8->currentQMV->x = 2*Data8->currentMV->x; |
bx = Data8->currentQMV->x = 2*Data8->currentMV->x; |
437 |
|
|
438 |
Data8->qpel_precision = 1; |
Data8->qpel_precision = 1; |
439 |
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, |
440 |
pParam->width, pParam->height, Data8->iFcode, 2, 0); |
pParam->width, pParam->height, Data8->iFcode, 2); |
441 |
|
|
442 |
} |
} |
443 |
|
|
444 |
if (MotionFlags & XVID_ME_QUARTERPELREFINE8_RD) { |
if (MotionFlags & XVID_ME_QUARTERPELREFINE8_RD) { |
445 |
if (MotionFlags & XVID_ME_FASTREFINE8) |
if (MotionFlags & XVID_ME_FASTREFINE8) |
446 |
FullRefine_Fast(Data8, CheckCandidateRD8, 0); |
FullRefine_Fast(Data8, CheckCandidateRD8, 0); |
447 |
else xvid_me_SubpelRefine(Data8, CheckCandidateRD8, 0); |
else xvid_me_SubpelRefine(Data->currentQMV[0], Data8, CheckCandidateRD8, 0); |
448 |
} |
} |
449 |
|
|
450 |
if (bsad <= Data->iMinSAD[0]) { |
if (bsad <= Data->iMinSAD[0]) { |
460 |
xvid_me_SquareSearch(Data8->currentMV->x, Data8->currentMV->x, Data8, 255, CheckCandidateRD8); |
xvid_me_SquareSearch(Data8->currentMV->x, Data8->currentMV->x, Data8, 255, CheckCandidateRD8); |
461 |
|
|
462 |
if (MotionFlags & XVID_ME_HALFPELREFINE8_RD) |
if (MotionFlags & XVID_ME_HALFPELREFINE8_RD) |
463 |
xvid_me_SubpelRefine(Data8, CheckCandidateRD8, 0); /* halfpel refinement */ |
xvid_me_SubpelRefine(Data->currentMV[0], Data8, CheckCandidateRD8, 0); /* halfpel refinement */ |
464 |
} |
} |
465 |
|
|
466 |
/* checking vector equal to predicion */ |
/* checking vector equal to predicion */ |
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->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->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->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->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->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->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->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->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 |
|
|
647 |
backup[i] = v[i]; |
backup[i] = v[i]; |
648 |
} |
} |
649 |
|
|
650 |
|
for (i = 0; i < 6; i++) { |
651 |
|
/* HVS models, anyone ? */ |
652 |
|
Data->lambda[i] = LAMBDA; |
653 |
|
} |
654 |
|
|
655 |
min_rd = findRD_inter(Data, x, y, pParam, MotionFlags); |
min_rd = findRD_inter(Data, x, y, pParam, MotionFlags); |
656 |
cbp = *Data->cbp; |
cbp = *Data->cbp; |
657 |
|
|
687 |
pMB->sad16 = pMB->sad8[0] = pMB->sad8[1] = pMB->sad8[2] = pMB->sad8[3] = 0; |
pMB->sad16 = pMB->sad8[0] = pMB->sad8[1] = pMB->sad8[2] = pMB->sad8[3] = 0; |
688 |
pMB->cbp = cbp; |
pMB->cbp = cbp; |
689 |
|
|
|
|
|
|
if (Data->rrv) { |
|
|
Data->currentMV[0].x = RRV_MV_SCALEDOWN(Data->currentMV[0].x); |
|
|
Data->currentMV[0].y = RRV_MV_SCALEDOWN(Data->currentMV[0].y); |
|
|
} |
|
|
|
|
690 |
if (mode == MODE_INTER && mcsel == 0) { |
if (mode == MODE_INTER && mcsel == 0) { |
691 |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = Data->currentMV[0]; |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = Data->currentMV[0]; |
692 |
|
|
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++) { |
755 |
|
/* HVS models, anyone ? */ |
756 |
|
Data->lambda[i] = LAMBDA; |
757 |
|
} |
758 |
|
|
759 |
/* INTER <-> INTER4V decision */ |
/* INTER <-> INTER4V decision */ |
760 |
if ((Data->iMinSAD[0] + 75 < Data->iMinSAD[1] + |
if ((Data->iMinSAD[0] + 75 < Data->iMinSAD[1] + |
773 |
/* final skip decision, a.k.a. "the vector you found, really that good?" */ |
/* final skip decision, a.k.a. "the vector you found, really that good?" */ |
774 |
if (skip_possible && (pMB->sad16 < (int)iQuant * MAX_SAD00_FOR_SKIP)) |
if (skip_possible && (pMB->sad16 < (int)iQuant * MAX_SAD00_FOR_SKIP)) |
775 |
if ( (100*sad)/(pMB->sad16+1) > FINAL_SKIP_THRESH) |
if ( (100*sad)/(pMB->sad16+1) > FINAL_SKIP_THRESH) |
776 |
if (Data->chroma || xvid_me_SkipDecisionP(pCurrent, pRef, x, y, Data->iEdgedWidth/2, iQuant, Data->rrv)) { |
if (Data->chroma || xvid_me_SkipDecisionP(pCurrent, pRef, x, y, Data->iEdgedWidth/2, iQuant)) { |
777 |
mode = MODE_NOT_CODED; |
mode = MODE_NOT_CODED; |
778 |
sad = 0; /* Compiler warning */ |
sad = 0; /* Compiler warning */ |
779 |
goto early_out; |
goto early_out; |
804 |
/* final skip decision, a.k.a. "the vector you found, really that good?" */ |
/* final skip decision, a.k.a. "the vector you found, really that good?" */ |
805 |
if (skip_possible && (pMB->sad16 < (int)iQuant * MAX_SAD00_FOR_SKIP)) |
if (skip_possible && (pMB->sad16 < (int)iQuant * MAX_SAD00_FOR_SKIP)) |
806 |
if ( (100*Data->iMinSAD[0])/(pMB->sad16+1) > FINAL_SKIP_THRESH) |
if ( (100*Data->iMinSAD[0])/(pMB->sad16+1) > FINAL_SKIP_THRESH) |
807 |
if (Data->chroma || xvid_me_SkipDecisionP(pCurrent, pRef, x, y, Data->iEdgedWidth/2, iQuant, Data->rrv)) { |
if (Data->chroma || xvid_me_SkipDecisionP(pCurrent, pRef, x, y, Data->iEdgedWidth/2, iQuant)) { |
808 |
mode = MODE_NOT_CODED; |
mode = MODE_NOT_CODED; |
809 |
sad = 0; /* Compiler warning */ |
sad = 0; /* Compiler warning */ |
810 |
goto early_out; |
goto early_out; |
880 |
if ((pMB - 1)->mode == MODE_INTRA ) InterBias -= 80; |
if ((pMB - 1)->mode == MODE_INTRA ) InterBias -= 80; |
881 |
|
|
882 |
if (Data->chroma) InterBias += 50; /* dev8(chroma) ??? <-- yes, we need dev8 (no big difference though) */ |
if (Data->chroma) InterBias += 50; /* dev8(chroma) ??? <-- yes, we need dev8 (no big difference though) */ |
|
if (Data->rrv) InterBias *= 4; |
|
883 |
|
|
884 |
if (InterBias < sad) { |
if (InterBias < sad) { |
885 |
int32_t deviation; |
int32_t deviation = dev16(Data->Cur, Data->iEdgedWidth); |
|
if (!Data->rrv) |
|
|
deviation = dev16(Data->Cur, Data->iEdgedWidth); |
|
|
else |
|
|
deviation = dev16(Data->Cur, Data->iEdgedWidth) + /* dev32() */ |
|
|
dev16(Data->Cur+16, Data->iEdgedWidth) + |
|
|
dev16(Data->Cur + 16*Data->iEdgedWidth, Data->iEdgedWidth) + |
|
|
dev16(Data->Cur+16+16*Data->iEdgedWidth, Data->iEdgedWidth); |
|
|
|
|
886 |
if (deviation < (sad - InterBias)) mode = MODE_INTRA; |
if (deviation < (sad - InterBias)) mode = MODE_INTRA; |
887 |
} |
} |
888 |
|
|
938 |
early_out: |
early_out: |
939 |
pMB->sad16 = pMB->sad8[0] = pMB->sad8[1] = pMB->sad8[2] = pMB->sad8[3] = sad; |
pMB->sad16 = pMB->sad8[0] = pMB->sad8[1] = pMB->sad8[2] = pMB->sad8[3] = sad; |
940 |
|
|
|
if (Data->rrv) { |
|
|
Data->currentMV[0].x = RRV_MV_SCALEDOWN(Data->currentMV[0].x); |
|
|
Data->currentMV[0].y = RRV_MV_SCALEDOWN(Data->currentMV[0].y); |
|
|
} |
|
|
|
|
941 |
if (mode == MODE_INTER && mcsel == 0) { |
if (mode == MODE_INTER && mcsel == 0) { |
942 |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = Data->currentMV[0]; |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = Data->currentMV[0]; |
943 |
|
|