667 |
for(i = 0; i < 4; i++) { |
for(i = 0; i < 4; i++) { |
668 |
int s = 8*((i&1) + (i>>1)*data->iEdgedWidth); |
int s = 8*((i&1) + (i>>1)*data->iEdgedWidth); |
669 |
transfer_8to16subro(in, data->Cur + s, ptr + s, data->iEdgedWidth); |
transfer_8to16subro(in, data->Cur + s, ptr + s, data->iEdgedWidth); |
670 |
bits += data->temp[i] = Block_CalcBits(coeff, in, data->iQuant, data->quant_type, &cbp, i); |
bits += data->temp[i] = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, i); |
671 |
} |
} |
672 |
|
|
673 |
bits += t = BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision, 0); |
bits += t = BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision, 0); |
692 |
/* chroma U */ |
/* chroma U */ |
693 |
ptr = interpolate8x8_switch2(data->RefQ + 64, data->RefP[4], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); |
ptr = interpolate8x8_switch2(data->RefQ + 64, data->RefP[4], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); |
694 |
transfer_8to16subro(in, ptr, data->CurU, data->iEdgedWidth/2); |
transfer_8to16subro(in, ptr, data->CurU, data->iEdgedWidth/2); |
695 |
bits += Block_CalcBits(coeff, in, data->iQuant, data->quant_type, &cbp, 4); |
bits += Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 4); |
696 |
if (bits >= data->iMinSAD[0]) return; |
if (bits >= data->iMinSAD[0]) return; |
697 |
|
|
698 |
/* chroma V */ |
/* chroma V */ |
699 |
ptr = interpolate8x8_switch2(data->RefQ + 64, data->RefP[5], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); |
ptr = interpolate8x8_switch2(data->RefQ + 64, data->RefP[5], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding); |
700 |
transfer_8to16subro(in, ptr, data->CurV, data->iEdgedWidth/2); |
transfer_8to16subro(in, ptr, data->CurV, data->iEdgedWidth/2); |
701 |
bits += Block_CalcBits(coeff, in, data->iQuant, data->quant_type, &cbp, 5); |
bits += Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 5); |
702 |
|
|
703 |
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; |
704 |
|
|
731 |
} |
} |
732 |
|
|
733 |
transfer_8to16subro(in, data->Cur, ptr, data->iEdgedWidth); |
transfer_8to16subro(in, data->Cur, ptr, data->iEdgedWidth); |
734 |
bits = Block_CalcBits(coeff, in, data->iQuant, data->quant_type, &cbp, 5); |
bits = Block_CalcBits(coeff, in, data->dctSpace + 128, data->iQuant, data->quant_type, &cbp, 5); |
735 |
bits += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision, 0); |
bits += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision, 0); |
736 |
|
|
737 |
if (bits < data->iMinSAD[0]) { |
if (bits < data->iMinSAD[0]) { |
1089 |
VECTOR currentMV[5]; |
VECTOR currentMV[5]; |
1090 |
VECTOR currentQMV[5]; |
VECTOR currentQMV[5]; |
1091 |
int32_t iMinSAD[5]; |
int32_t iMinSAD[5]; |
1092 |
DECLARE_ALIGNED_MATRIX(dct_space, 2, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(dct_space, 3, 64, int16_t, CACHE_LINE); |
1093 |
SearchData Data; |
SearchData Data; |
1094 |
memset(&Data, 0, sizeof(SearchData)); |
memset(&Data, 0, sizeof(SearchData)); |
1095 |
Data.iEdgedWidth = iEdgedWidth; |
Data.iEdgedWidth = iEdgedWidth; |
2161 |
} |
} |
2162 |
} |
} |
2163 |
|
|
2164 |
#define INTRA_THRESH 1700 |
#define INTRA_THRESH 2200 |
2165 |
#define INTER_THRESH 1200 |
#define INTER_THRESH 50 |
2166 |
|
#define INTRA_THRESH2 95 |
2167 |
|
|
2168 |
int |
int |
2169 |
MEanalysis( const IMAGE * const pRef, |
MEanalysis( const IMAGE * const pRef, |
2170 |
const FRAMEINFO * const Current, |
const FRAMEINFO * const Current, |
2171 |
const MBParam * const pParam, |
const MBParam * const pParam, |
2172 |
const int maxIntra, /* maximum number if non-I frames */ |
const int maxIntra, //maximum number if non-I frames |
2173 |
const int intraCount, /* number of non-I frames after last I frame; 0 if we force P/B frame */ |
const int intraCount, //number of non-I frames after last I frame; 0 if we force P/B frame |
2174 |
const int bCount, /* number of B frames in a row */ |
const int bCount, // number of B frames in a row |
2175 |
const int b_thresh) |
const int b_thresh) |
2176 |
{ |
{ |
2177 |
uint32_t x, y, intra = 0; |
uint32_t x, y, intra = 0; |
2178 |
int sSAD = 0; |
int sSAD = 0; |
2179 |
MACROBLOCK * const pMBs = Current->mbs; |
MACROBLOCK * const pMBs = Current->mbs; |
2180 |
const IMAGE * const pCurrent = &Current->image; |
const IMAGE * const pCurrent = &Current->image; |
2181 |
int IntraThresh = INTRA_THRESH, InterThresh = INTER_THRESH + 10*b_thresh; |
int IntraThresh = INTRA_THRESH, InterThresh = INTER_THRESH + b_thresh; |
2182 |
int s = 0, blocks = 0; |
int s = 0, blocks = 0; |
2183 |
|
int complexity = 0; |
2184 |
|
|
2185 |
int32_t iMinSAD[5], temp[5]; |
int32_t iMinSAD[5], temp[5]; |
2186 |
VECTOR currentMV[5]; |
VECTOR currentMV[5]; |
2192 |
Data.temp = temp; |
Data.temp = temp; |
2193 |
CheckCandidate = CheckCandidate32I; |
CheckCandidate = CheckCandidate32I; |
2194 |
|
|
2195 |
|
|
2196 |
if (intraCount != 0) { |
if (intraCount != 0) { |
2197 |
if (intraCount < 10) /* we're right after an I frame */ |
if (intraCount < 10) // we're right after an I frame |
2198 |
IntraThresh += 15* (intraCount - 10) * (intraCount - 10); |
IntraThresh += 15* (intraCount - 10) * (intraCount - 10); |
2199 |
else |
else |
2200 |
if ( 5*(maxIntra - intraCount) < maxIntra) /* we're close to maximum. 2 sec when max is 10 sec */ |
if ( 5*(maxIntra - intraCount) < maxIntra) // we're close to maximum. 2 sec when max is 10 sec |
2201 |
IntraThresh -= (IntraThresh * (maxIntra - 8*(maxIntra - intraCount)))/maxIntra; |
IntraThresh -= (IntraThresh * (maxIntra - 8*(maxIntra - intraCount)))/maxIntra; |
2202 |
} |
} |
2203 |
|
|
2204 |
InterThresh -= (350 - 8*b_thresh) * bCount; |
InterThresh -= 12 * bCount; |
2205 |
if (InterThresh < 300 + 5*b_thresh) InterThresh = 300 + 5*b_thresh; |
if (InterThresh < 15 + b_thresh) InterThresh = 15 + b_thresh; |
2206 |
|
|
2207 |
if (sadInit) (*sadInit) (); |
if (sadInit) (*sadInit) (); |
2208 |
|
|
2209 |
for (y = 1; y < pParam->mb_height-1; y += 2) { |
for (y = 1; y < pParam->mb_height-1; y += 2) { |
2210 |
for (x = 1; x < pParam->mb_width-1; x += 2) { |
for (x = 1; x < pParam->mb_width-1; x += 2) { |
2211 |
int i; |
int i; |
2212 |
blocks += 4; |
blocks += 10; |
2213 |
|
|
2214 |
if (bCount == 0) pMBs[x + y * pParam->mb_width].mvs[0] = zeroMV; |
if (bCount == 0) pMBs[x + y * pParam->mb_width].mvs[0] = zeroMV; |
2215 |
else { /* extrapolation of the vector found for last frame */ |
else { //extrapolation of the vector found for last frame |
2216 |
pMBs[x + y * pParam->mb_width].mvs[0].x = |
pMBs[x + y * pParam->mb_width].mvs[0].x = |
2217 |
(pMBs[x + y * pParam->mb_width].mvs[0].x * (bCount+1) ) / bCount; |
(pMBs[x + y * pParam->mb_width].mvs[0].x * (bCount+1) ) / bCount; |
2218 |
pMBs[x + y * pParam->mb_width].mvs[0].y = |
pMBs[x + y * pParam->mb_width].mvs[0].y = |
2224 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
2225 |
int dev; |
int dev; |
2226 |
MACROBLOCK *pMB = &pMBs[x+(i&1) + (y+(i>>1)) * pParam->mb_width]; |
MACROBLOCK *pMB = &pMBs[x+(i&1) + (y+(i>>1)) * pParam->mb_width]; |
|
if (pMB->sad16 > IntraThresh) { |
|
2227 |
dev = dev16(pCurrent->y + (x + (i&1) + (y + (i>>1)) * pParam->edged_width) * 16, |
dev = dev16(pCurrent->y + (x + (i&1) + (y + (i>>1)) * pParam->edged_width) * 16, |
2228 |
pParam->edged_width); |
pParam->edged_width); |
2229 |
|
|
2230 |
|
complexity += dev; |
2231 |
if (dev + IntraThresh < pMB->sad16) { |
if (dev + IntraThresh < pMB->sad16) { |
2232 |
pMB->mode = MODE_INTRA; |
pMB->mode = MODE_INTRA; |
2233 |
if (++intra > ((pParam->mb_height-2)*(pParam->mb_width-2))/2) return I_VOP; |
if (++intra > ((pParam->mb_height-2)*(pParam->mb_width-2))/2) return I_VOP; |
2234 |
} |
} |
2235 |
} |
|
2236 |
if (pMB->mvs[0].x == 0 && pMB->mvs[0].y == 0) s++; |
if (pMB->mvs[0].x == 0 && pMB->mvs[0].y == 0) |
2237 |
|
if (dev > 500 && pMB->sad16 < 1000) |
2238 |
|
sSAD += 1000; |
2239 |
|
|
2240 |
sSAD += pMB->sad16; |
sSAD += pMB->sad16; |
2241 |
} |
} |
2242 |
} |
} |
2243 |
} |
} |
2244 |
|
complexity >>= 7; |
2245 |
|
|
2246 |
sSAD /= blocks; |
sSAD /= complexity + 4*blocks; |
|
|
|
|
if (b_thresh < 20) { |
|
|
s = (10*s) / blocks; |
|
|
if (s > 4) sSAD += (s - 2) * (40 - 2*b_thresh); /* static block - looks bad when in bframe... */ |
|
|
} |
|
2247 |
|
|
2248 |
|
if (intraCount > 12 && sSAD > INTRA_THRESH2 ) return I_VOP; |
2249 |
if (sSAD > InterThresh ) return P_VOP; |
if (sSAD > InterThresh ) return P_VOP; |
2250 |
emms(); |
emms(); |
2251 |
return B_VOP; |
return B_VOP; |
2252 |
} |
} |
2253 |
|
|
|
|
|
2254 |
static WARPPOINTS |
static WARPPOINTS |
2255 |
GlobalMotionEst(const MACROBLOCK * const pMBs, |
GlobalMotionEst(const MACROBLOCK * const pMBs, |
2256 |
const MBParam * const pParam, |
const MBParam * const pParam, |
2623 |
/* chroma U */ |
/* chroma U */ |
2624 |
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); |
2625 |
transfer_8to16subro(in, Data->CurU, ptr, Data->iEdgedWidth/2); |
transfer_8to16subro(in, Data->CurU, ptr, Data->iEdgedWidth/2); |
2626 |
bits += Block_CalcBits(coeff, in, Data->iQuant, Data->quant_type, &cbp, 4); |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 4); |
2627 |
|
|
2628 |
if (bits >= *Data->iMinSAD) return bits; |
if (bits >= *Data->iMinSAD) return bits; |
2629 |
|
|
2630 |
/* chroma V */ |
/* chroma V */ |
2631 |
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); |
2632 |
transfer_8to16subro(in, Data->CurV, ptr, Data->iEdgedWidth/2); |
transfer_8to16subro(in, Data->CurV, ptr, Data->iEdgedWidth/2); |
2633 |
bits += Block_CalcBits(coeff, in, Data->iQuant, Data->quant_type, &cbp, 5); |
bits += Block_CalcBits(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5); |
2634 |
|
|
2635 |
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; |
2636 |
|
|
2647 |
for(i = 0; i < 4; i++) { |
for(i = 0; i < 4; i++) { |
2648 |
int s = 8*((i&1) + (i>>1)*Data->iEdgedWidth); |
int s = 8*((i&1) + (i>>1)*Data->iEdgedWidth); |
2649 |
transfer_8to16copy(in, Data->Cur + s, Data->iEdgedWidth); |
transfer_8to16copy(in, Data->Cur + s, Data->iEdgedWidth); |
2650 |
bits += Block_CalcBitsIntra(coeff, in, Data->iQuant, Data->quant_type, &cbp, i, &dc); |
bits += Block_CalcBitsIntra(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, i, &dc); |
2651 |
|
|
2652 |
if (bits >= Data->iMinSAD[0]) return bits; |
if (bits >= Data->iMinSAD[0]) return bits; |
2653 |
} |
} |
2656 |
|
|
2657 |
/*chroma U */ |
/*chroma U */ |
2658 |
transfer_8to16copy(in, Data->CurU, Data->iEdgedWidth/2); |
transfer_8to16copy(in, Data->CurU, Data->iEdgedWidth/2); |
2659 |
bits += Block_CalcBitsIntra(coeff, in, Data->iQuant, Data->quant_type, &cbp, 4, &dc); |
bits += Block_CalcBitsIntra(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 4, &dc); |
2660 |
|
|
2661 |
if (bits >= Data->iMinSAD[0]) return bits; |
if (bits >= Data->iMinSAD[0]) return bits; |
2662 |
|
|
2663 |
/* chroma V */ |
/* chroma V */ |
2664 |
transfer_8to16copy(in, Data->CurV, Data->iEdgedWidth/2); |
transfer_8to16copy(in, Data->CurV, Data->iEdgedWidth/2); |
2665 |
bits += Block_CalcBitsIntra(coeff, in, Data->iQuant, Data->quant_type, &cbp, 5, &dc); |
bits += Block_CalcBitsIntra(coeff, in, Data->dctSpace + 128, Data->iQuant, Data->quant_type, &cbp, 5, &dc); |
2666 |
|
|
2667 |
bits += BITS_MULT*mcbpc_inter_tab[(MODE_INTRA & 7) | ((cbp & 3) << 3)].len; |
bits += BITS_MULT*mcbpc_inter_tab[(MODE_INTRA & 7) | ((cbp & 3) << 3)].len; |
2668 |
|
|