--- estimation_rd_based.c	2004/07/18 12:23:45	1.8
+++ estimation_rd_based.c	2004/12/09 04:58:12	1.10
@@ -20,7 +20,7 @@
  *  along with this program ; if not, write to the Free Software
  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  *
- * $Id: estimation_rd_based.c,v 1.8 2004/07/18 12:23:45 syskin Exp $
+ * $Id: estimation_rd_based.c,v 1.10 2004/12/09 04:58:12 syskin Exp $
  *
  ****************************************************************************/
 
@@ -57,7 +57,8 @@
 				const int block,
 				const uint16_t * scan_table,
 				const unsigned int lambda,
-				const uint16_t * mpeg_quant_matrices)
+				const uint16_t * mpeg_quant_matrices,
+				const unsigned int quant_sq)
 {
 	int sum;
 	int bits;
@@ -93,7 +94,7 @@
 	}
 
 
-	return bits + (lambda*distortion)/(quant*quant);
+	return bits + (lambda*distortion)/quant_sq;
 }
 
 static __inline unsigned int
@@ -111,7 +112,8 @@
 					unsigned int bits[2],
 					unsigned int cbp[2],
 					unsigned int lambda,
-					const uint16_t * mpeg_quant_matrices)
+					const uint16_t * mpeg_quant_matrices,
+					const unsigned int quant_sq)
 {
 	int direction;
 	int16_t *pCurrent;
@@ -171,7 +173,7 @@
 
 	distortion = sse8_16bit(coeff, dqcoeff, 8*sizeof(int16_t));
 
-	return (lambda*distortion)/(quant*quant);
+	return (lambda*distortion)/quant_sq;
 }
 
 
@@ -203,10 +205,12 @@
 	for(i = 0; i < 4; i++) {
 		int s = 8*((i&1) + (i>>1)*data->iEdgedWidth);
 		transfer_8to16subro(in, data->Cur + s, ptr + s, data->iEdgedWidth);
-		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, 
+								data->quant_type, &cbp, i, data->scan_table, data->lambda[i], 
+								data->mpeg_quant_matrices, data->quant_sq);
 	}
 
-	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);
 
 	if (data->temp[0] + t < data->iMinSAD[1]) {
 		data->iMinSAD[1] = data->temp[0] + t; current[1].x = x; current[1].y = y; data->cbp[1] = (data->cbp[1]&~32) | (cbp&32); }
@@ -228,13 +232,17 @@
 	/* chroma U */
 	ptr = interpolate8x8_switch2(data->RefQ, data->RefP[4], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding);
 	transfer_8to16subro(in, data->CurU, ptr, data->iEdgedWidth/2);
-	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,
+							&cbp, 4, data->scan_table, data->lambda[4], 
+							data->mpeg_quant_matrices, data->quant_sq);
 	if (rd >= data->iMinSAD[0]) return;
 
 	/* chroma V */
 	ptr = interpolate8x8_switch2(data->RefQ, data->RefP[5], 0, 0, xc, yc, data->iEdgedWidth/2, data->rounding);
 	transfer_8to16subro(in, data->CurV, ptr, data->iEdgedWidth/2);
-	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,
+							&cbp, 5, data->scan_table, data->lambda[5], 
+							data->mpeg_quant_matrices, data->quant_sq);
 
 	rd += BITS_MULT*mcbpc_inter_tab[(MODE_INTER & 7) | ((cbp & 3) << 3)].len;
 
@@ -268,8 +276,10 @@
 	}
 
 	transfer_8to16subro(in, data->Cur, ptr, data->iEdgedWidth);
-	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 += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision, 0);
+	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, data->quant_sq);
+	rd += BITS_MULT*d_mv_bits(x, y, data->predMV, data->iFcode, data->qpel^data->qpel_precision);
 
 	if (rd < data->iMinSAD[0]) {
 		*data->cbp = cbp;
@@ -300,7 +310,7 @@
 		if (MotionFlags & (XVID_ME_HALFPELREFINE16_RD | XVID_ME_EXTSEARCH_RD)) { /* we have to prepare for halfpixel-precision search */
 			for(i = 0; i < 5; i++) bsad[i] = Data->iMinSAD[i];
 			get_range(&Data->min_dx, &Data->max_dx, &Data->min_dy, &Data->max_dy, x, y, 4,
-						pParam->width, pParam->height, Data->iFcode - Data->qpel, 1, Data->rrv);
+						pParam->width, pParam->height, Data->iFcode - Data->qpel, 1);
 			Data->qpel_precision = 0;
 			if (Data->currentQMV->x & 1 || Data->currentQMV->y & 1)
 				CheckCandidateRD16(Data->currentMV[0].x, Data->currentMV[0].y, Data, 255);
@@ -327,7 +337,7 @@
 			/* preparing for qpel-precision search */
 			Data->qpel_precision = 1;
 			get_range(&Data->min_dx, &Data->max_dx, &Data->min_dy, &Data->max_dy, x, y, 4,
-					pParam->width, pParam->height, Data->iFcode, 2, 0);
+					pParam->width, pParam->height, Data->iFcode, 2);
 		}
 		if (MotionFlags & XVID_ME_QUARTERPELREFINE16_RD) {
 			if (MotionFlags & XVID_ME_FASTREFINE16)
@@ -377,15 +387,15 @@
 		if(Data->qpel) {
 			Data8->predMV = get_qpmv2(pMBs, pParam->mb_width, 0, x, y, i);
 			if (i != 0)	t = d_mv_bits(	Data8->currentQMV->x, Data8->currentQMV->y,
-										Data8->predMV, Data8->iFcode, 0, 0);
+										Data8->predMV, Data8->iFcode, 0);
 		} else {
 			Data8->predMV = get_pmv2(pMBs, pParam->mb_width, 0, x, y, i);
 			if (i != 0)	t = d_mv_bits(	Data8->currentMV->x, Data8->currentMV->y,
-										Data8->predMV, Data8->iFcode, 0, 0);
+										Data8->predMV, Data8->iFcode, 0);
 		}
 
 		get_range(&Data8->min_dx, &Data8->max_dx, &Data8->min_dy, &Data8->max_dy, 2*x + (i&1), 2*y + (i>>1), 3,
-					pParam->width, pParam->height, Data8->iFcode, Data8->qpel+1, 0);
+					pParam->width, pParam->height, Data8->iFcode, Data8->qpel+1);
 
 		*Data8->iMinSAD += BITS_MULT*t;
 
@@ -408,7 +418,7 @@
 			if (MotionFlags&XVID_ME_HALFPELREFINE8_RD || (MotionFlags&XVID_ME_EXTSEARCH8 && MotionFlags&XVID_ME_EXTSEARCH_RD)) { /* halfpixel motion search follows */
 				Data8->qpel_precision = 0;
 				get_range(&Data8->min_dx, &Data8->max_dx, &Data8->min_dy, &Data8->max_dy, 2*x + (i&1), 2*y + (i>>1), 3,
-							pParam->width, pParam->height, Data8->iFcode - 1, 1, 0);
+							pParam->width, pParam->height, Data8->iFcode - 1, 1);
 
 				if (Data8->currentQMV->x & 1 || Data8->currentQMV->y & 1)
 					CheckCandidateRD8(Data8->currentMV->x, Data8->currentMV->y, Data8, 255);
@@ -427,7 +437,7 @@
 
 				Data8->qpel_precision = 1;
 				get_range(&Data8->min_dx, &Data8->max_dx, &Data8->min_dy, &Data8->max_dy, 2*x + (i&1), 2*y + (i>>1), 3,
-							pParam->width, pParam->height, Data8->iFcode, 2, 0);
+							pParam->width, pParam->height, Data8->iFcode, 2);
 
 			}
 
@@ -491,14 +501,16 @@
 	/* chroma U */
 	ptr = interpolate8x8_switch2(Data->RefQ + 64, Data->RefP[4], 0, 0, sumx, sumy, Data->iEdgedWidth/2, Data->rounding);
 	transfer_8to16subro(in, Data->CurU, ptr, Data->iEdgedWidth/2);
-	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,
+							Data->scan_table, Data->lambda[4], Data->mpeg_quant_matrices, Data->quant_sq);
 
 	if (bits >= *Data->iMinSAD) return bits;
 
 	/* chroma V */
 	ptr = interpolate8x8_switch2(Data->RefQ + 64, Data->RefP[5], 0, 0, sumx, sumy, Data->iEdgedWidth/2, Data->rounding);
 	transfer_8to16subro(in, Data->CurV, ptr, Data->iEdgedWidth/2);
-	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,
+							Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq);
 
 	bits += BITS_MULT*mcbpc_inter_tab[(MODE_INTER4V & 7) | ((cbp & 3) << 3)].len;
 
@@ -524,7 +536,8 @@
 		
 
 		distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, i, in, coeff, dqcoeff,
-								predictors[i], iQuant, Data->quant_type, bits, cbp, Data->lambda[i], Data->mpeg_quant_matrices);
+								predictors[i], iQuant, Data->quant_type, bits, cbp, 
+								Data->lambda[i], Data->mpeg_quant_matrices, Data->quant_sq);
 		bits1 += distortion + BITS_MULT * bits[0];
 		bits2 += distortion + BITS_MULT * bits[1];
 
@@ -538,7 +551,8 @@
 	/*chroma U */
 	transfer_8to16copy(in, Data->CurU, Data->iEdgedWidth/2);
 	distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, 4, in, coeff, dqcoeff,
-									predictors[4], iQuant, Data->quant_type, bits, cbp, Data->lambda[4], Data->mpeg_quant_matrices);
+									predictors[4], iQuant, Data->quant_type, bits, cbp, 
+									Data->lambda[4], Data->mpeg_quant_matrices, Data->quant_sq);
 	bits1 += distortion + BITS_MULT * bits[0];
 	bits2 += distortion + BITS_MULT * bits[1];
 
@@ -548,7 +562,8 @@
 	/* chroma V */
 	transfer_8to16copy(in, Data->CurV, Data->iEdgedWidth/2);
 	distortion = Block_CalcBitsIntra(pMB, x, y, mb_width, 5, in, coeff, dqcoeff,
-									predictors[5], iQuant, Data->quant_type, bits, cbp, Data->lambda[5], Data->mpeg_quant_matrices);
+									predictors[5], iQuant, Data->quant_type, bits, cbp, 
+									Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq);
 
 	bits1 += distortion + BITS_MULT * bits[0];
 	bits2 += distortion + BITS_MULT * bits[1];
@@ -572,7 +587,8 @@
 	for(i = 0; i < 4; i++) {
 		int s = 8*((i&1) + (i>>1)*Data->iEdgedWidth);
 		transfer_8to16subro(in, Data->Cur + s, vGMC->y + s + 16*(x+y*Data->iEdgedWidth), Data->iEdgedWidth);
-		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,
+								Data->scan_table, Data->lambda[i], Data->mpeg_quant_matrices, Data->quant_sq);
 		if (bits >= Data->iMinSAD[0]) return bits;
 	}
 
@@ -580,13 +596,15 @@
 
 	/*chroma U */
 	transfer_8to16subro(in, Data->CurU, vGMC->u + 8*(x+y*(Data->iEdgedWidth/2)), Data->iEdgedWidth/2);
-	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,
+							Data->scan_table, Data->lambda[4], Data->mpeg_quant_matrices, Data->quant_sq);
 
 	if (bits >= Data->iMinSAD[0]) return bits;
 
 	/* chroma V */
 	transfer_8to16subro(in, Data->CurV , vGMC->v + 8*(x+y*(Data->iEdgedWidth/2)), Data->iEdgedWidth/2);
-	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,
+							Data->scan_table, Data->lambda[5], Data->mpeg_quant_matrices, Data->quant_sq);
 
 	bits += BITS_MULT*mcbpc_inter_tab[(MODE_INTER & 7) | ((cbp & 3) << 3)].len;
 
@@ -617,6 +635,7 @@
 	int min_rd, intra_rd, i, cbp;
 	VECTOR backup[5], *v;
 	Data->iQuant = iQuant;
+	Data->quant_sq = iQuant*iQuant;
 	Data->scan_table = VopFlags & XVID_VOP_ALTERNATESCAN ?
 						scan_tables[2] : scan_tables[0];
 
@@ -668,12 +687,6 @@
 	pMB->sad16 = pMB->sad8[0] = pMB->sad8[1] = pMB->sad8[2] = pMB->sad8[3] = 0;
 	pMB->cbp = cbp;
 
-
-	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);
-	}
-
 	if (mode == MODE_INTER && mcsel == 0) {
 		pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = Data->currentMV[0];
 
@@ -735,6 +748,8 @@
 						scan_tables[2] : scan_tables[0];
 
 	pMB->mcsel = 0;
+	Data->iQuant = iQuant;
+	Data->quant_sq = iQuant*iQuant;
 
 	for (i = 0; i < 6; i++) {
 		/* HVS models, anyone ? */
@@ -758,7 +773,7 @@
 		/* final skip decision, a.k.a. "the vector you found, really that good?" */
 		if (skip_possible && (pMB->sad16 < (int)iQuant * MAX_SAD00_FOR_SKIP))
 			if ( (100*sad)/(pMB->sad16+1) > FINAL_SKIP_THRESH)
-				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)) {
 					mode = MODE_NOT_CODED;
 					sad = 0;  /* Compiler warning */
 					goto early_out;
@@ -789,7 +804,7 @@
 		/* final skip decision, a.k.a. "the vector you found, really that good?" */
 		if (skip_possible && (pMB->sad16 < (int)iQuant * MAX_SAD00_FOR_SKIP))
 			if ( (100*Data->iMinSAD[0])/(pMB->sad16+1) > FINAL_SKIP_THRESH)
-				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)) {
 					mode = MODE_NOT_CODED;
 					sad = 0; /* Compiler warning */
 					goto early_out;
@@ -865,18 +880,9 @@
 			if ((pMB - 1)->mode == MODE_INTRA ) InterBias -= 80;
 
 		if (Data->chroma) InterBias += 50; /* dev8(chroma) ??? <-- yes, we need dev8 (no big difference though) */
-		if (Data->rrv) InterBias *= 4;
 
 		if (InterBias < sad) {
-			int32_t deviation;
-			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);
-
+			int32_t deviation = dev16(Data->Cur, Data->iEdgedWidth);
 			if (deviation < (sad - InterBias)) mode = MODE_INTRA;
 		}
 
@@ -932,11 +938,6 @@
 early_out:
 	pMB->sad16 = pMB->sad8[0] = pMB->sad8[1] = pMB->sad8[2] = pMB->sad8[3] = sad;
 
-	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);
-	}
-
 	if (mode == MODE_INTER && mcsel == 0) {
 		pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = Data->currentMV[0];