--- trunk/xvidcore/src/encoder.c	2004/04/04 11:47:21	1406
+++ trunk/xvidcore/src/encoder.c	2010/03/09 10:00:30	1883
@@ -3,9 +3,9 @@
  *  XVID MPEG-4 VIDEO CODEC
  *  - Encoder main module -
  *
- *  Copyright(C) 2002	  Michael Militzer <isibaar@xvid.org>
- *			   2002-2003 Peter Ross <pross@xvid.org>
- *			   2002	  Daniel Smith <danielsmith@astroboymail.com>
+ *  Copyright(C) 2002-2010 Michael Militzer <isibaar@xvid.org>
+ *               2002-2003 Peter Ross <pross@xvid.org>
+ *               2002      Daniel Smith <danielsmith@astroboymail.com>
  *
  *  This program is free software ; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
@@ -21,7 +21,7 @@
  *  along with this program ; if not, write to the Free Software
  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  *
- * $Id: encoder.c,v 1.105 2004-04-04 11:47:21 syskin Exp $
+ * $Id: encoder.c,v 1.131 2010-03-09 10:00:14 Isibaar Exp $
  *
  ****************************************************************************/
 
@@ -49,6 +49,9 @@
 #include "quant/quant_matrix.h"
 #include "utils/mem_align.h"
 
+# include "motion/motion_smp.h"
+
+
 /*****************************************************************************
  * Local function prototypes
  ****************************************************************************/
@@ -85,26 +88,48 @@
 /*
  * Simplify the "fincr/fbase" fraction
 */
+static int
+gcd(int a, int b)
+{
+	int r ;
+
+	if (b > a) {
+		r = a;
+		a = b;
+		b = r;
+	}
+
+	while ((r = a % b)) {
+		a = b;
+		b = r;
+	}
+	return b;
+}
+
 static void
 simplify_time(int *inc, int *base)
 {
 	/* common factor */
-	int i = *inc;
-	while (i > 1) {
-		if (*inc % i == 0 && *base % i == 0) {
-			*inc /= i;
-			*base /= i;
-			i = *inc;
-			continue;
-		}
-		i--;
-	}
-
-	/* if neccessary, round to 65535 accuracy */
-	if (*base > 65535) {
-		float div = (float) *base / 65535;
-		*base = (int) (*base / div);
-		*inc = (int) (*inc / div);
+	const int s = gcd(*inc, *base);
+  *inc  /= s;
+  *base /= s;
+
+	if (*base > 65535 || *inc > 65535) {
+		int *biggest;
+		int *other;
+		float div;
+
+		if (*base > *inc) {
+			biggest = base;
+			other = inc;
+		} else {
+			biggest = inc;
+			other = base;
+		}
+		
+		div = ((float)*biggest)/((float)65535);
+		*biggest = (unsigned int)(((float)*biggest)/div);
+		*other = (unsigned int)(((float)*other)/div);
 	}
 }
 
@@ -113,7 +138,7 @@
 enc_create(xvid_enc_create_t * create)
 {
 	Encoder *pEnc;
-  int n;
+	int n;
 
 	if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */
 		return XVID_ERR_VERSION;
@@ -135,6 +160,8 @@
 
 	/* global flags */
 	pEnc->mbParam.global_flags = create->global;
+  if ((pEnc->mbParam.global_flags & XVID_GLOBAL_PACKED))
+    pEnc->mbParam.global_flags |= XVID_GLOBAL_DIVX5_USERDATA;
 
 	/* width, height */
 	pEnc->mbParam.width = create->width;
@@ -148,7 +175,7 @@
 	pEnc->mbParam.fincr = MAX(create->fincr, 0);
 	pEnc->mbParam.fbase = create->fincr <= 0 ? 25 : create->fbase;
 	if (pEnc->mbParam.fincr>0)
-		simplify_time(&pEnc->mbParam.fincr, &pEnc->mbParam.fbase);
+		simplify_time((int*)&pEnc->mbParam.fincr, (int*)&pEnc->mbParam.fbase);
 
 	/* zones */
 	if(create->num_zones > 0) {
@@ -179,7 +206,7 @@
 
 		memset(&pinfo, 0, sizeof(xvid_plg_info_t));
 		pinfo.version = XVID_VERSION;
-		if (create->plugins[n].func(0, XVID_PLG_INFO, &pinfo, 0) >= 0) {
+		if (create->plugins[n].func(NULL, XVID_PLG_INFO, &pinfo, NULL) >= 0) {
 			pEnc->mbParam.plugin_flags |= pinfo.flags;
 		}
 
@@ -196,7 +223,7 @@
 		pcreate.param = create->plugins[n].param;
 
 		pEnc->plugins[n].func = NULL;   /* disable plugins that fail */
-		if (create->plugins[n].func(0, XVID_PLG_CREATE, &pcreate, &pEnc->plugins[n].param) >= 0) {
+		if (create->plugins[n].func(NULL, XVID_PLG_CREATE, &pcreate, &pEnc->plugins[n].param) >= 0) {
 			pEnc->plugins[n].func = create->plugins[n].func;
 		}
 	}
@@ -214,6 +241,14 @@
 			goto xvid_err_memory1a;
 	}
 
+	/* temp lambdas */
+	if (pEnc->mbParam.plugin_flags & XVID_REQLAMBDA) {
+		pEnc->temp_lambda = (float *) xvid_malloc(pEnc->mbParam.mb_width *
+						pEnc->mbParam.mb_height * 6 * sizeof(float), CACHE_LINE);
+		if (pEnc->temp_lambda == NULL)
+			goto xvid_err_memory1a;
+	}
+
 	/* bframes */
 	pEnc->mbParam.max_bframes = MAX(create->max_bframes, 0);
 	pEnc->mbParam.bquant_ratio = MAX(create->bquant_ratio, 0);
@@ -400,7 +435,7 @@
 	/* timestamp stuff */
 
 	pEnc->mbParam.m_stamp = 0;
-	pEnc->m_framenum = 0;
+	pEnc->m_framenum = create->start_frame_num;
 	pEnc->current->stamp = 0;
 	pEnc->reference->stamp = 0;
 
@@ -409,6 +444,36 @@
 	pEnc->iFrameNum = 0;
 	pEnc->fMvPrevSigma = -1;
 
+	/* multithreaded stuff */
+	if (create->num_threads > 0) {
+		int t = create->num_threads;
+		int rows_per_thread = (pEnc->mbParam.mb_height+t-1)/t;
+		pEnc->num_threads = t;
+		pEnc->motionData = xvid_malloc(t*sizeof(SMPmotionData), CACHE_LINE);
+		if (!pEnc->motionData) 
+			goto xvid_err_nosmp;
+		
+		for (n = 0; n < t; n++) {
+			pEnc->motionData[n].complete_count_self =
+				xvid_malloc(rows_per_thread * sizeof(int), CACHE_LINE);
+
+			if (!pEnc->motionData[n].complete_count_self)
+				goto xvid_err_nosmp;
+		
+			if (n != 0) 
+				pEnc->motionData[n].complete_count_above = 
+					pEnc->motionData[n-1].complete_count_self;
+		}
+		pEnc->motionData[0].complete_count_above =
+			pEnc->motionData[t-1].complete_count_self - 1;
+
+	} else {
+  xvid_err_nosmp:
+		/* no SMP */
+		create->num_threads = 0;
+		pEnc->motionData = NULL;
+	}
+
 	create->handle = (void *) pEnc;
 
 	init_timer();
@@ -495,10 +560,14 @@
 		xvid_free(pEnc->temp_dquants);
 	}
 
+	if(pEnc->mbParam.plugin_flags & XVID_REQLAMBDA) {
+		xvid_free(pEnc->temp_lambda);
+	}
+
   xvid_err_memory0:
 	for (n=0; n<pEnc->num_plugins;n++) {
 		if (pEnc->plugins[n].func) {
-			pEnc->plugins[n].func(pEnc->plugins[n].param, XVID_PLG_DESTROY, 0, 0);
+			pEnc->plugins[n].func(pEnc->plugins[n].param, XVID_PLG_DESTROY, NULL, NULL);
 		}
 	}
 	xvid_free(pEnc->plugins);
@@ -593,6 +662,9 @@
 		xvid_free(pEnc->temp_dquants);
 	}
 
+	if ((pEnc->mbParam.plugin_flags & XVID_REQLAMBDA)) {
+		xvid_free(pEnc->temp_lambda);
+	}
 
 	if (pEnc->num_plugins>0) {
 		xvid_plg_destroy_t pdestroy;
@@ -603,7 +675,7 @@
 
 		for (i=0; i<pEnc->num_plugins;i++) {
 			if (pEnc->plugins[i].func) {
-				pEnc->plugins[i].func(pEnc->plugins[i].param, XVID_PLG_DESTROY, &pdestroy, 0);
+				pEnc->plugins[i].func(pEnc->plugins[i].param, XVID_PLG_DESTROY, &pdestroy, NULL);
 			}
 		}
 		xvid_free(pEnc->plugins);
@@ -611,9 +683,16 @@
 
 	xvid_free(pEnc->mbParam.mpeg_quant_matrices);
 
-	if (pEnc->num_plugins>0)
+	if (pEnc->num_zones > 0)
 		xvid_free(pEnc->zones);
 
+	if (pEnc->num_threads > 0) {
+		for (i = 0; i < pEnc->num_threads; i++)
+			xvid_free(pEnc->motionData[i].complete_count_self);
+
+		xvid_free(pEnc->motionData);
+	}
+
 	xvid_free(pEnc);
 
 	return 0;  /* ok */
@@ -627,7 +706,7 @@
 static void call_plugins(Encoder * pEnc, FRAMEINFO * frame, IMAGE * original,
 						 int opt, int * type, int * quant, xvid_enc_stats_t * stats)
 {
-	unsigned int i, j;
+	unsigned int i, j, k;
 	xvid_plg_data_t data;
 
 	/* set data struct */
@@ -686,9 +765,19 @@
 		if ((pEnc->mbParam.plugin_flags & XVID_REQDQUANTS)) {
 			data.dquant = pEnc->temp_dquants;
 			data.dquant_stride = pEnc->mbParam.mb_width;
-			memset(data.dquant, 0, data.mb_width*data.mb_height);
+			memset(data.dquant, 0, data.mb_width*data.mb_height*sizeof(int));
 		}
-	
+
+		if(pEnc->mbParam.plugin_flags & XVID_REQLAMBDA) {
+			int block = 0;
+			emms();
+			data.lambda = pEnc->temp_lambda;
+			for(i = 0;i < pEnc->mbParam.mb_height; i++)
+				for(j = 0;j < pEnc->mbParam.mb_width; j++)
+					for (k = 0; k < 6; k++) 
+						data.lambda[block++] = 1.0f;
+		}
+
 	} else { /* XVID_PLG_AFTER */
 		if ((pEnc->mbParam.plugin_flags & XVID_REQORIGINAL)) {
 			data.original.csp = XVID_CSP_PLANAR;
@@ -763,7 +852,7 @@
 	for (i=0; i<(unsigned int)pEnc->num_plugins;i++) {
 		emms();
 		if (pEnc->plugins[i].func) {
-			if (pEnc->plugins[i].func(pEnc->plugins[i].param, opt, &data, 0) < 0) {
+			if (pEnc->plugins[i].func(pEnc->plugins[i].param, opt, &data, NULL) < 0) {
 				continue;
 			}
 		}
@@ -792,6 +881,23 @@
 				frame->mbs[j*pEnc->mbParam.mb_width + i].dquant = 0;
 			}
 		}
+
+		if (pEnc->mbParam.plugin_flags & XVID_REQLAMBDA) {
+			for (j = 0; j < pEnc->mbParam.mb_height; j++)
+				for (i = 0; i < pEnc->mbParam.mb_width; i++)
+					for (k = 0; k < 6; k++) {
+						frame->mbs[j*pEnc->mbParam.mb_width + i].lambda[k] = 
+							(int) ((float)(1<<LAMBDA_EXP) * data.lambda[6 * (j * data.mb_width + i) + k]);
+			}
+		} else {
+			for (j = 0; j<pEnc->mbParam.mb_height; j++)
+				for (i = 0; i<pEnc->mbParam.mb_width; i++)
+					for (k = 0; k < 6; k++) {
+						frame->mbs[j*pEnc->mbParam.mb_width + i].lambda[k] = 1<<LAMBDA_EXP;
+			}
+		}
+
+
 		frame->mbs[0].quant = data.quant; /* FRAME will not affect the quant in stats */
 	}
 
@@ -861,24 +967,6 @@
 #endif
 }
 
-static int
-gcd(int a, int b)
-{
-	int r ;
-
-	if (b > a) {
-		r = a;
-		a = b;
-		b = r;
-	}
-
-	while ((r = a % b)) {
-		a = b;
-		b = r;
-	}
-	return b;
-}
-
 static void
 simplify_par(int *par_width, int *par_height)
 {
@@ -1004,7 +1092,7 @@
 			}
 
 			FrameCodeB(pEnc, pEnc->bframes[pEnc->bframenum_head], &bs);
-			call_plugins(pEnc, pEnc->bframes[pEnc->bframenum_head], &pEnc->sOriginal2, XVID_PLG_AFTER, 0, 0, stats);
+			call_plugins(pEnc, pEnc->bframes[pEnc->bframenum_head], &pEnc->sOriginal2, XVID_PLG_AFTER, NULL, NULL, stats);
 			pEnc->bframenum_head++;
 
 			goto done;
@@ -1036,7 +1124,7 @@
 
 			/* add the not-coded length to the reference frame size */
 			pEnc->current->length += (BitstreamPos(&bs) - bits) / 8;
-			call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, 0, 0, stats);
+			call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, NULL, NULL, stats);
 
 			/* flush complete: reset counters */
 			pEnc->flush_bframes = 0;
@@ -1064,7 +1152,7 @@
 				pEnc->queue_head, pEnc->queue_tail, pEnc->queue_size);
 
 			if (!(pEnc->mbParam.global_flags & XVID_GLOBAL_PACKED) && pEnc->mbParam.max_bframes > 0) {
-				call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, 0, 0, stats);
+				call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, NULL, NULL, stats);
 			}
 
 			/* if the very last frame is to be b-vop, we must change it to a p-vop */
@@ -1093,7 +1181,7 @@
 
 
 				if ((pEnc->mbParam.global_flags & XVID_GLOBAL_PACKED) && pEnc->bframenum_tail==0) {
-					call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, 0, 0, stats);
+					call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, NULL, NULL, stats);
 				}else{
 					pEnc->flush_bframes = 1;
 					goto done;
@@ -1142,7 +1230,7 @@
 	type = frame->type;
 	pEnc->current->quant = frame->quant;
 
-	call_plugins(pEnc, pEnc->current, NULL, XVID_PLG_BEFORE, &type, &pEnc->current->quant, stats);
+	call_plugins(pEnc, pEnc->current, NULL, XVID_PLG_BEFORE, &type, (int*)&pEnc->current->quant, stats);
 
 	if (type > 0){ 	/* XVID_TYPE_?VOP */
 		type = type2coding(type);	/* convert XVID_TYPE_?VOP to bitstream coding type */
@@ -1217,10 +1305,11 @@
 	if (!(pEnc->mbParam.global_flags & XVID_GLOBAL_PACKED) && pEnc->mbParam.max_bframes > 0)
 	{
 		if (pEnc->current->stamp > 0) {
-			call_plugins(pEnc, pEnc->reference, &pEnc->sOriginal, XVID_PLG_AFTER, 0, 0, stats);
+			call_plugins(pEnc, pEnc->reference, &pEnc->sOriginal, XVID_PLG_AFTER, NULL, NULL, stats);
 		}
-		else
-			stats->type = XVID_TYPE_NOTHING;
+        else if (stats) {
+            stats->type = XVID_TYPE_NOTHING;
+        }
 	}
 
 	/* %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
@@ -1245,7 +1334,7 @@
 		SWAP(FRAMEINFO*, pEnc->current, pEnc->bframes[pEnc->bframenum_tail]);
 
 		if ((pEnc->current->vop_flags & XVID_VOP_DEBUG)) {
-			image_printf(&pEnc->current->image, pEnc->mbParam.edged_width, pEnc->mbParam.height, 5, 100, "DX50 BVOP->PVOP");
+			image_printf(&pEnc->current->image, pEnc->mbParam.edged_width, pEnc->mbParam.height, 5, 100, "CLOSED GOP BVOP->PVOP");
 		}
 
 		/* convert B-VOP quant to P-VOP */
@@ -1306,9 +1395,6 @@
 
 		/* prevent vol/vop misuse */
 
-		if (!(pEnc->current->vol_flags & XVID_VOL_REDUCED_ENABLE))
-			pEnc->current->vop_flags &= ~XVID_VOP_REDUCED;
-
 		if (!(pEnc->current->vol_flags & XVID_VOL_INTERLACING))
 			pEnc->current->vop_flags &= ~(XVID_VOP_TOPFIELDFIRST|XVID_VOP_ALTERNATESCAN);
 
@@ -1343,7 +1429,9 @@
 
 		if ( FrameCodeP(pEnc, &bs) == 0 ) {
 			/* N-VOP, we mustn't code b-frames yet */
-			call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, 0, 0, stats);
+			if ((pEnc->mbParam.global_flags & XVID_GLOBAL_PACKED) || 
+				 pEnc->mbParam.max_bframes == 0)
+				call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, NULL, NULL, stats);
 			goto done;
 		}
 	}
@@ -1364,7 +1452,7 @@
 
 	/* packed or no-bframes or no-bframes-queued: output stats */
 	if ((pEnc->mbParam.global_flags & XVID_GLOBAL_PACKED) || pEnc->mbParam.max_bframes == 0 ) {
-		call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, 0, 0, stats);
+		call_plugins(pEnc, pEnc->current, &pEnc->sOriginal, XVID_PLG_AFTER, NULL, NULL, stats);
 	}
 
 	/* %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
@@ -1436,20 +1524,6 @@
 
 	uint16_t x, y;
 
-	if ((pEnc->current->vol_flags & XVID_VOL_REDUCED_ENABLE))
-	{
-		mb_width = (pEnc->mbParam.width + 31) / 32;
-		mb_height = (pEnc->mbParam.height + 31) / 32;
-
-		/* 16x16->8x8 downsample requires 1 additional edge pixel*/
-		/* XXX: setedges is overkill */
-		start_timer();
-		image_setedges(&pEnc->current->image,
-			pEnc->mbParam.edged_width, pEnc->mbParam.edged_height,
-			pEnc->mbParam.width, pEnc->mbParam.height, 0);
-		stop_edges_timer();
-	}
-
 	pEnc->mbParam.m_rounding_type = 1;
 	pEnc->current->rounding_type = pEnc->mbParam.m_rounding_type;
 	pEnc->current->coding_type = I_VOP;
@@ -1467,6 +1541,7 @@
 	BitstreamWriteVopHeader(bs, &pEnc->mbParam, pEnc->current, 1, pEnc->current->mbs[0].quant);
 
 	pEnc->current->sStat.iTextBits = 0;
+	pEnc->current->sStat.iMVBits = 0;
 	pEnc->current->sStat.kblks = mb_width * mb_height;
 	pEnc->current->sStat.mblks = pEnc->current->sStat.ublks = 0;
 
@@ -1485,21 +1560,10 @@
 			stop_prediction_timer();
 
 			start_timer();
-			if (pEnc->current->vop_flags & XVID_VOP_GREYSCALE)
-			{	pMB->cbp &= 0x3C;		/* keep only bits 5-2 */
-				qcoeff[4*64+0]=0;		/* zero, because for INTRA MBs DC value is saved */
-				qcoeff[5*64+0]=0;
-			}
 			MBCoding(pEnc->current, pMB, qcoeff, bs, &pEnc->current->sStat);
 			stop_coding_timer();
 		}
 
-	if ((pEnc->current->vop_flags & XVID_VOP_REDUCED))
-	{
-		image_deblock_rrv(&pEnc->current->image, pEnc->mbParam.edged_width,
-			pEnc->current->mbs, mb_width, mb_height, pEnc->mbParam.mb_width,
-			16, 0);
-	}
 	emms();
 
 	BitstreamPadAlways(bs); /* next_start_code() at the end of VideoObjectPlane() */
@@ -1515,25 +1579,43 @@
 	return 1;					/* intra */
 }
 
+static __inline void
+updateFcode(Statistics * sStat, Encoder * pEnc)
+{
+	float fSigma;
+	int iSearchRange;
 
-#define INTRA_THRESHOLD 0.5
-#define BFRAME_SKIP_THRESHHOLD 30
+	if (sStat->iMvCount == 0)
+		sStat->iMvCount = 1;
+
+	fSigma = (float) sqrt((float) sStat->iMvSum / sStat->iMvCount);
+
+	iSearchRange = 16 << pEnc->mbParam.m_fcode;
+
+	if ((3.0 * fSigma > iSearchRange) && (pEnc->mbParam.m_fcode <= 5) )
+		pEnc->mbParam.m_fcode++;
+
+	else if ((5.0 * fSigma < iSearchRange)
+			   && (4.0 * pEnc->fMvPrevSigma < iSearchRange)
+			   && (pEnc->mbParam.m_fcode >= 2) )
+		pEnc->mbParam.m_fcode--;
+
+	pEnc->fMvPrevSigma = fSigma;
+}
 
+#define BFRAME_SKIP_THRESHHOLD 30
 
 /* FrameCodeP also handles S(GMC)-VOPs */
 static int
 FrameCodeP(Encoder * pEnc,
 		   Bitstream * bs)
 {
-	float fSigma;
 	int bits = BitstreamPos(bs);
 
 	DECLARE_ALIGNED_MATRIX(dct_codes, 6, 64, int16_t, CACHE_LINE);
 	DECLARE_ALIGNED_MATRIX(qcoeff, 6, 64, int16_t, CACHE_LINE);
 
 	int x, y, k;
-	int iSearchRange;
-	int skip_possible;
 	FRAMEINFO *const current = pEnc->current;
 	FRAMEINFO *const reference = pEnc->reference;
 	MBParam * const pParam = &pEnc->mbParam;
@@ -1541,17 +1623,8 @@
 	int mb_height = pParam->mb_height;
 	int coded = 1;
 
-
-	/* IMAGE *pCurrent = &current->image; */
 	IMAGE *pRef = &reference->image;
 
-	if ((current->vop_flags & XVID_VOP_REDUCED))
-	{
-		mb_width = (pParam->width + 31) / 32;
-		mb_height = (pParam->height + 31) / 32;
-	}
-
-
 	if (!reference->is_edged) {	
 		start_timer();
 		image_setedges(pRef, pParam->edged_width, pParam->edged_height,
@@ -1567,8 +1640,8 @@
 	if ((current->vop_flags & XVID_VOP_HALFPEL)) {
 		if (reference->is_interpolated != current->rounding_type) {
 			start_timer();
-			image_interpolate(pRef, &pEnc->vInterH, &pEnc->vInterV,
-							  &pEnc->vInterHV, pParam->edged_width,
+			image_interpolate(pRef->y, pEnc->vInterH.y, pEnc->vInterV.y,
+							  pEnc->vInterHV.y, pParam->edged_width,
 							  pParam->edged_height,
 							  (pParam->vol_flags & XVID_VOL_QUARTERPEL),
 							  current->rounding_type);
@@ -1577,6 +1650,10 @@
 		}
 	}
 
+	current->sStat.iTextBits = current->sStat.iMvSum = current->sStat.iMvCount =
+		current->sStat.kblks = current->sStat.mblks = current->sStat.ublks = 
+		current->sStat.iMVBits = 0;
+
 	current->coding_type = P_VOP;
 
 	call_plugins(pEnc, pEnc->current, NULL, XVID_PLG_FRAME, NULL, NULL, NULL);
@@ -1634,10 +1711,54 @@
 		}
 	}
 
-	MotionEstimation(&pEnc->mbParam, current, reference,
-					 &pEnc->vInterH, &pEnc->vInterV, &pEnc->vInterHV,
-					 &pEnc->vGMC, 256*4096);
 
+	if (pEnc->num_threads > 0) {
+		/* multithreaded motion estimation - dispatch threads */
+
+		void * status;
+		int rows_per_thread = (pParam->mb_height + pEnc->num_threads - 1)/pEnc->num_threads;
+
+		for (k = 0; k < pEnc->num_threads; k++) {
+			memset(pEnc->motionData[k].complete_count_self, 0, rows_per_thread * sizeof(int));
+			pEnc->motionData[k].pParam = &pEnc->mbParam;
+			pEnc->motionData[k].current = current;
+			pEnc->motionData[k].reference = reference;
+			pEnc->motionData[k].pRefH = &pEnc->vInterH;
+			pEnc->motionData[k].pRefV = &pEnc->vInterV;
+			pEnc->motionData[k].pRefHV = &pEnc->vInterHV;
+			pEnc->motionData[k].pGMC = &pEnc->vGMC;
+			pEnc->motionData[k].y_step = pEnc->num_threads;
+			pEnc->motionData[k].start_y = k;
+			/* todo: sort out temp space once and for all */
+			pEnc->motionData[k].RefQ = pEnc->vInterH.u + 16*k*pParam->edged_width;
+		}
+
+		for (k = 1; k < pEnc->num_threads; k++) {
+			pthread_create(&pEnc->motionData[k].handle, NULL, 
+				(void*)MotionEstimateSMP, (void*)&pEnc->motionData[k]);
+		}
+
+		MotionEstimateSMP(&pEnc->motionData[0]);
+
+		for (k = 1; k < pEnc->num_threads; k++) {
+			pthread_join(pEnc->motionData[k].handle, &status);
+		}
+
+		current->fcode = 0;
+		for (k = 0; k < pEnc->num_threads; k++) {
+			current->sStat.iMvSum += pEnc->motionData[k].mvSum;
+			current->sStat.iMvCount += pEnc->motionData[k].mvCount;
+			if (pEnc->motionData[k].minfcode > current->fcode)
+				current->fcode = pEnc->motionData[k].minfcode;
+		}
+
+	} else {
+		/* regular ME */
+
+		MotionEstimation(&pEnc->mbParam, current, reference,
+						 &pEnc->vInterH, &pEnc->vInterV, &pEnc->vInterHV,
+						 &pEnc->vGMC, 256*4096);
+	}
 
 	stop_motion_timer();
 
@@ -1645,18 +1766,12 @@
 
 	BitstreamWriteVopHeader(bs, &pEnc->mbParam, current, 1, current->mbs[0].quant);
 
-	current->sStat.iTextBits = current->sStat.iMvSum = current->sStat.iMvCount =
-		current->sStat.kblks = current->sStat.mblks = current->sStat.ublks = 0;
-
-
 	for (y = 0; y < mb_height; y++) {
 		for (x = 0; x < mb_width; x++) {
-			MACROBLOCK *pMB =
-				&current->mbs[x + y * pParam->mb_width];
-
-			int bIntra = (pMB->mode == MODE_INTRA) || (pMB->mode == MODE_INTRA_Q);
+			MACROBLOCK *pMB = &current->mbs[x + y * pParam->mb_width];
+			int skip_possible;
 
-			if (bIntra) {
+			if (pMB->mode == MODE_INTRA || pMB->mode == MODE_INTRA_Q) {
 				CodeIntraMB(pEnc, pMB);
 				MBTransQuantIntra(&pEnc->mbParam, current, pMB, x, y,
 								  dct_codes, qcoeff);
@@ -1667,11 +1782,6 @@
 
 				current->sStat.kblks++;
 
-				if (pEnc->current->vop_flags & XVID_VOP_GREYSCALE)
-				{	pMB->cbp &= 0x3C;		/* keep only bits 5-2 */
-					qcoeff[4*64+0]=0;		/* zero, because for INTRA MBs DC value is saved */
-					qcoeff[5*64+0]=0;
-				}
 				MBCoding(current, pMB, qcoeff, bs, &current->sStat);
 				stop_coding_timer();
 				continue;
@@ -1686,16 +1796,14 @@
 								 pParam->height,
 								 pParam->edged_width,
 								 (current->vol_flags & XVID_VOL_QUARTERPEL),
-								 (current->vop_flags & XVID_VOP_REDUCED),
 								 current->rounding_type);
 
 			stop_comp_timer();
 
 			pMB->field_pred = 0;
 
-			if (pMB->mode != MODE_NOT_CODED)
-			{	pMB->cbp =
-					MBTransQuantInter(&pEnc->mbParam, current, pMB, x, y,
+			if (pMB->cbp != 0) {
+				pMB->cbp = MBTransQuantInter(&pEnc->mbParam, current, pMB, x, y,
 									  dct_codes, qcoeff);
 			}
 
@@ -1715,141 +1823,62 @@
 
 			/* Finished processing the MB, now check if to CODE or SKIP */
 
-			skip_possible = (pMB->cbp == 0) && (pMB->mode == MODE_INTER) &&
-							(pMB->dquant == 0);
+			skip_possible = (pMB->cbp == 0) && (pMB->mode == MODE_INTER);
 
 			if (current->coding_type == S_VOP)
 				skip_possible &= (pMB->mcsel == 1);
-			else if (current->coding_type == P_VOP) {
-				if ((pParam->vol_flags & XVID_VOL_QUARTERPEL))
-					skip_possible &= ( (pMB->qmvs[0].x == 0) && (pMB->qmvs[0].y == 0) );
-				else
-					skip_possible &= ( (pMB->mvs[0].x == 0) && (pMB->mvs[0].y == 0) );
+			else { /* PVOP */
+				const VECTOR * const mv = (pParam->vol_flags & XVID_VOL_QUARTERPEL) ?
+										pMB->qmvs : pMB->mvs;
+				skip_possible &= ((mv->x|mv->y) == 0);
 			}
 
-			if ( (pMB->mode == MODE_NOT_CODED) || (skip_possible)) {
-
-/* This is a candidate for SKIPping, but for P-VOPs check intermediate B-frames first */
+			if ((pMB->mode == MODE_NOT_CODED) || (skip_possible)) {
+				/* This is a candidate for SKIPping, but for P-VOPs check intermediate B-frames first */
+				int bSkip = 1;
 
-				if (current->coding_type == P_VOP)	/* special rule for P-VOP's SKIP */
-				{
-					int bSkip = 1;
+				if (current->coding_type == P_VOP) {	/* special rule for P-VOP's SKIP */
 
-					for (k=pEnc->bframenum_head; k< pEnc->bframenum_tail; k++)
-					{
+					for (k = pEnc->bframenum_head; k < pEnc->bframenum_tail; k++) {
 						int iSAD;
 						iSAD = sad16(reference->image.y + 16*y*pParam->edged_width + 16*x,
-									pEnc->bframes[k]->image.y + 16*y*pParam->edged_width + 16*x,
-								pParam->edged_width,BFRAME_SKIP_THRESHHOLD);
-						if (iSAD >= BFRAME_SKIP_THRESHHOLD * pMB->quant)
-						{	bSkip = 0;
+										pEnc->bframes[k]->image.y + 16*y*pParam->edged_width + 16*x,
+										pParam->edged_width, BFRAME_SKIP_THRESHHOLD * pMB->quant);
+						if (iSAD >= BFRAME_SKIP_THRESHHOLD * pMB->quant) {
+							bSkip = 0; /* could not SKIP */
+							if (pParam->vol_flags & XVID_VOL_QUARTERPEL) {
+								VECTOR predMV = get_qpmv2(current->mbs, pParam->mb_width, 0, x, y, 0);
+								pMB->pmvs[0].x = - predMV.x;
+								pMB->pmvs[0].y = - predMV.y;
+							} else {
+								VECTOR predMV = get_pmv2(current->mbs, pParam->mb_width, 0, x, y, 0);
+								pMB->pmvs[0].x = - predMV.x;
+								pMB->pmvs[0].y = - predMV.y;
+							}
+							pMB->mode = MODE_INTER;
+							pMB->cbp = 0;
 							break;
 						}
 					}
-
-					if (!bSkip) {	/* no SKIP, but trivial block */
-						if((pParam->vol_flags & XVID_VOL_QUARTERPEL)) {
-							VECTOR predMV = get_qpmv2(current->mbs, pParam->mb_width, 0, x, y, 0);
-							pMB->pmvs[0].x = - predMV.x;
-							pMB->pmvs[0].y = - predMV.y;
-						}
-						else {
-							VECTOR predMV = get_pmv2(current->mbs, pParam->mb_width, 0, x, y, 0);
-							pMB->pmvs[0].x = - predMV.x;
-							pMB->pmvs[0].y = - predMV.y;
-						}
-						pMB->mode = MODE_INTER;
-						pMB->cbp = 0;
-						MBCoding(current, pMB, qcoeff, bs, &current->sStat);
-						stop_coding_timer();
-
-						continue;	/* next MB */
-					}
 				}
-				/* do SKIP */
-
-				pMB->mode = MODE_NOT_CODED;
-				MBSkip(bs);
-				stop_coding_timer();
-				continue;	/* next MB */
-			}
-			/* ordinary case: normal coded INTER/INTER4V block */
-
-			if ((current->vop_flags & XVID_VOP_GREYSCALE))
-			{	pMB->cbp &= 0x3C;		/* keep only bits 5-2 */
-				qcoeff[4*64+0]=0;		/* zero, because DC for INTRA MBs DC value is saved */
-				qcoeff[5*64+0]=0;
-			}
-
-			if((pParam->vol_flags & XVID_VOL_QUARTERPEL)) {
-				VECTOR predMV = get_qpmv2(current->mbs, pParam->mb_width, 0, x, y, 0);
-				pMB->pmvs[0].x = pMB->qmvs[0].x - predMV.x;
-				pMB->pmvs[0].y = pMB->qmvs[0].y - predMV.y;
-				DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", pMB->pmvs[0].x, pMB->pmvs[0].y, predMV.x, predMV.y, pMB->mvs[0].x, pMB->mvs[0].y);
-			} else {
-				VECTOR predMV = get_pmv2(current->mbs, pParam->mb_width, 0, x, y, 0);
-				pMB->pmvs[0].x = pMB->mvs[0].x - predMV.x;
-				pMB->pmvs[0].y = pMB->mvs[0].y - predMV.y;
-				DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", pMB->pmvs[0].x, pMB->pmvs[0].y, predMV.x, predMV.y, pMB->mvs[0].x, pMB->mvs[0].y);
-			}
-
-
-			if (pMB->mode == MODE_INTER4V)
-			{	int k;
-				for (k=1;k<4;k++)
-				{
-					if((pParam->vol_flags & XVID_VOL_QUARTERPEL)) {
-						VECTOR predMV = get_qpmv2(current->mbs, pParam->mb_width, 0, x, y, k);
-						pMB->pmvs[k].x = pMB->qmvs[k].x - predMV.x;
-						pMB->pmvs[k].y = pMB->qmvs[k].y - predMV.y;
-				DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", pMB->pmvs[k].x, pMB->pmvs[k].y, predMV.x, predMV.y, pMB->mvs[k].x, pMB->mvs[k].y);
-					} else {
-						VECTOR predMV = get_pmv2(current->mbs, pParam->mb_width, 0, x, y, k);
-						pMB->pmvs[k].x = pMB->mvs[k].x - predMV.x;
-						pMB->pmvs[k].y = pMB->mvs[k].y - predMV.y;
-				DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", pMB->pmvs[k].x, pMB->pmvs[k].y, predMV.x, predMV.y, pMB->mvs[k].x, pMB->mvs[k].y);
-					}
 
+				if (bSkip) {
+					/* do SKIP */
+					pMB->mode = MODE_NOT_CODED;
+					MBSkip(bs);
+					stop_coding_timer();
+					continue;	/* next MB */
 				}
 			}
 
+			/* ordinary case: normal coded INTER/INTER4V block */
 			MBCoding(current, pMB, qcoeff, bs, &pEnc->current->sStat);
 			stop_coding_timer();
-
 		}
 	}
 
-	if ((current->vop_flags & XVID_VOP_REDUCED))
-	{
-		image_deblock_rrv(&current->image, pParam->edged_width,
-			current->mbs, mb_width, mb_height, pParam->mb_width,
-			16, 0);
-	}
-
 	emms();
-
-	if (current->sStat.iMvCount == 0)
-		current->sStat.iMvCount = 1;
-
-	fSigma = (float) sqrt((float) current->sStat.iMvSum / current->sStat.iMvCount);
-
-	iSearchRange = 1 << (3 + pParam->m_fcode);
-
-	if ((fSigma > iSearchRange / 3)
-		&& (pParam->m_fcode <= (3 +  (pParam->vol_flags & XVID_VOL_QUARTERPEL?1:0)  )))	/* maximum search range 128 */
-	{
-		pParam->m_fcode++;
-		iSearchRange *= 2;
-	} else if ((fSigma < iSearchRange / 6)
-			   && (pEnc->fMvPrevSigma >= 0)
-			   && (pEnc->fMvPrevSigma < iSearchRange / 6)
-			   && (pParam->m_fcode >= (2 + (pParam->vol_flags & XVID_VOL_QUARTERPEL?1:0) )))	/* minimum search range 16 */
-	{
-		pParam->m_fcode--;
-		iSearchRange /= 2;
-	}
-
-	pEnc->fMvPrevSigma = fSigma;
+	updateFcode(&current->sStat, pEnc);
 
 	/* frame drop code */
 #if 0
@@ -1859,7 +1888,7 @@
 		(pParam->frame_drop_ratio * mb_width * mb_height) / 100 &&
 		( (pEnc->bframenum_head >= pEnc->bframenum_tail) || !(pEnc->mbParam.global_flags & XVID_GLOBAL_CLOSED_GOP)) )
 	{
-		current->sStat.kblks = current->sStat.mblks = 0;
+		current->sStat.kblks = current->sStat.mblks = current->sStat.iTextBits = 0;
 		current->sStat.ublks = mb_width * mb_height;
 
 		BitstreamReset(bs);
@@ -1934,8 +1963,6 @@
 		fprintf(fp,"Y=%3d   X=%3d   MB=%2d   CBP=%02X\n",y,x,mb->mode,mb->cbp); \
 	}
 
-	/* XXX: pEnc->current->global_flags &= ~XVID_VOP_REDUCED;  reduced resoltion not yet supported */
-
 	if (!first){
 		fp=fopen("C:\\XVIDDBGE.TXT","w");
 	}
@@ -1951,7 +1978,7 @@
 
 	if (pEnc->reference->is_interpolated != 0) {
 		start_timer();
-		image_interpolate(f_ref, &pEnc->f_refh, &pEnc->f_refv, &pEnc->f_refhv,
+		image_interpolate(f_ref->y, pEnc->f_refh.y, pEnc->f_refv.y, pEnc->f_refhv.y,
 						  pEnc->mbParam.edged_width, pEnc->mbParam.edged_height,
 						  (pEnc->mbParam.vol_flags & XVID_VOL_QUARTERPEL), 0);
 		stop_inter_timer();
@@ -1968,7 +1995,7 @@
 
 	if (pEnc->current->is_interpolated != 0) {
 		start_timer();
-		image_interpolate(b_ref, &pEnc->vInterH, &pEnc->vInterV, &pEnc->vInterHV,
+		image_interpolate(b_ref->y, pEnc->vInterH.y, pEnc->vInterV.y, pEnc->vInterHV.y,
 						pEnc->mbParam.edged_width, pEnc->mbParam.edged_height,
 						(pEnc->mbParam.vol_flags & XVID_VOL_QUARTERPEL), 0);
 		stop_inter_timer();
@@ -1976,22 +2003,72 @@
 	}
 
 	frame->coding_type = B_VOP;
-	call_plugins(pEnc, pEnc->current, NULL, XVID_PLG_FRAME, NULL, NULL, NULL);
+	call_plugins(pEnc, frame, NULL, XVID_PLG_FRAME, NULL, NULL, NULL);
+
+	frame->fcode = frame->bcode = pEnc->current->fcode;
 
 	start_timer();
-	MotionEstimationBVOP(&pEnc->mbParam, frame,
-						 ((int32_t)(pEnc->current->stamp - frame->stamp)),				/* time_bp */
-						 ((int32_t)(pEnc->current->stamp - pEnc->reference->stamp)), 	/* time_pp */
-						 pEnc->reference->mbs, f_ref,
-						 &pEnc->f_refh, &pEnc->f_refv, &pEnc->f_refhv,
-						 pEnc->current, b_ref, &pEnc->vInterH,
-						 &pEnc->vInterV, &pEnc->vInterHV);
+	if (pEnc->num_threads > 0) {
+		void * status;
+		int k;
+		/* multithreaded motion estimation - dispatch threads */
+		int rows_per_thread = (pEnc->mbParam.mb_height + pEnc->num_threads - 1)/pEnc->num_threads;
+
+		for (k = 0; k < pEnc->num_threads; k++) {
+			memset(pEnc->motionData[k].complete_count_self, 0, rows_per_thread * sizeof(int));
+			pEnc->motionData[k].pParam = &pEnc->mbParam;
+			pEnc->motionData[k].current = frame;
+			pEnc->motionData[k].reference = pEnc->current;
+			pEnc->motionData[k].fRef = f_ref;
+			pEnc->motionData[k].fRefH = &pEnc->f_refh;
+			pEnc->motionData[k].fRefV = &pEnc->f_refv;
+			pEnc->motionData[k].fRefHV = &pEnc->f_refhv;
+			pEnc->motionData[k].pRef = b_ref;
+			pEnc->motionData[k].pRefH = &pEnc->vInterH;
+			pEnc->motionData[k].pRefV = &pEnc->vInterV;
+			pEnc->motionData[k].pRefHV = &pEnc->vInterHV;
+			pEnc->motionData[k].time_bp = (int32_t)(pEnc->current->stamp - frame->stamp);
+			pEnc->motionData[k].time_pp = (int32_t)(pEnc->current->stamp - pEnc->reference->stamp);
+			pEnc->motionData[k].y_step = pEnc->num_threads;
+			pEnc->motionData[k].start_y = k;
+			/* todo: sort out temp space once and for all */
+			pEnc->motionData[k].RefQ = pEnc->vInterH.u + 16*k*pEnc->mbParam.edged_width;
+		}
+
+		for (k = 1; k < pEnc->num_threads; k++) {
+			pthread_create(&pEnc->motionData[k].handle, NULL, 
+				(void*)SMPMotionEstimationBVOP, (void*)&pEnc->motionData[k]);
+		}
+
+		SMPMotionEstimationBVOP(&pEnc->motionData[0]);
+
+		for (k = 1; k < pEnc->num_threads; k++) {
+			pthread_join(pEnc->motionData[k].handle, &status);
+		}
+
+		frame->fcode = frame->bcode = 0;
+		for (k = 0; k < pEnc->num_threads; k++) {
+			if (pEnc->motionData[k].minfcode > frame->fcode)
+				frame->fcode = pEnc->motionData[k].minfcode;
+			if (pEnc->motionData[k].minbcode > frame->bcode)
+				frame->bcode = pEnc->motionData[k].minbcode;
+		}
+	} else {
+		MotionEstimationBVOP(&pEnc->mbParam, frame,
+							 ((int32_t)(pEnc->current->stamp - frame->stamp)),				/* time_bp */
+							 ((int32_t)(pEnc->current->stamp - pEnc->reference->stamp)), 	/* time_pp */
+							 pEnc->reference->mbs, f_ref,
+							 &pEnc->f_refh, &pEnc->f_refv, &pEnc->f_refhv,
+							 pEnc->current, b_ref, &pEnc->vInterH,
+							 &pEnc->vInterV, &pEnc->vInterHV);
+	}
 	stop_motion_timer();
 
 	set_timecodes(frame, pEnc->reference,pEnc->mbParam.fbase);
 	BitstreamWriteVopHeader(bs, &pEnc->mbParam, frame, 1, frame->quant);
 
 	frame->sStat.iTextBits = 0;
+	frame->sStat.iMVBits = 0;
 	frame->sStat.iMvSum = 0;
 	frame->sStat.iMvCount = 0;
 	frame->sStat.kblks = frame->sStat.mblks = frame->sStat.ublks = 0;
@@ -2006,36 +2083,35 @@
 			if (mb->mode == MODE_NOT_CODED) {
 				if (pEnc->mbParam.plugin_flags & XVID_REQORIGINAL) {
 					MBMotionCompensation(mb, x, y, f_ref, NULL, f_ref, NULL, NULL, &frame->image,
-											NULL, 0, 0, pEnc->mbParam.edged_width, 0, 0, 0);
+											NULL, 0, 0, pEnc->mbParam.edged_width, 0, 0);
 				}
-
 				continue;
 			}
 
-			if (mb->mode != MODE_DIRECT_NONE_MV || pEnc->mbParam.plugin_flags & XVID_REQORIGINAL) {
+			mb->quant = frame->quant;
+
+			if (mb->cbp != 0 || pEnc->mbParam.plugin_flags & XVID_REQORIGINAL) {
+				/* we have to motion-compensate, transfer etc, 
+					because there might be blocks to code */
+
 				MBMotionCompensationBVOP(&pEnc->mbParam, mb, x, y, &frame->image,
-									 f_ref, &pEnc->f_refh, &pEnc->f_refv,
-									 &pEnc->f_refhv, b_ref, &pEnc->vInterH,
-									 &pEnc->vInterV, &pEnc->vInterHV,
-									 dct_codes);
-
-				if (mb->mode == MODE_DIRECT_NO4V) mb->mode = MODE_DIRECT;
-				mb->quant = frame->quant;
-
-				if (mb->mode != MODE_DIRECT_NONE_MV)
-					mb->cbp = MBTransQuantInterBVOP(&pEnc->mbParam, frame, mb, x, y,  dct_codes, qcoeff);
-
-				if ( (mb->mode == MODE_DIRECT) && (mb->cbp == 0)
-					&& (mb->pmvs[3].x == 0) && (mb->pmvs[3].y == 0) ) {
-					mb->mode = MODE_DIRECT_NONE_MV;	/* skipped */
-				}
+										 f_ref, &pEnc->f_refh, &pEnc->f_refv,
+										 &pEnc->f_refhv, b_ref, &pEnc->vInterH,
+										 &pEnc->vInterV, &pEnc->vInterHV,
+										 dct_codes);
+
+				mb->cbp = MBTransQuantInterBVOP(&pEnc->mbParam, frame, mb, x, y,  dct_codes, qcoeff);
 			}
+			
+			if (mb->mode == MODE_DIRECT_NO4V)
+				mb->mode = MODE_DIRECT;
 
-			/* keep only bits 5-2 -- Chroma blocks will just be skipped by the
-			 * coding function for BFrames, that's why we don't zero teh DC
-			 * coeffs */
-			if ((frame->vop_flags & XVID_VOP_GREYSCALE))
-				mb->cbp &= 0x3C;
+			if (mb->mode == MODE_DIRECT && (mb->cbp | mb->pmvs[3].x | mb->pmvs[3].y) == 0)
+				mb->mode = MODE_DIRECT_NONE_MV;	/* skipped */
+			else 
+				if (frame->vop_flags & XVID_VOP_GREYSCALE)
+					/* keep only bits 5-2 -- Chroma blocks will just be skipped by MBCodingBVOP */
+					mb->cbp &= 0x3C;
 
 			start_timer();
 			MBCodingBVOP(frame, mb, qcoeff, frame->fcode, frame->bcode, bs,
@@ -2043,11 +2119,8 @@
 			stop_coding_timer();
 		}
 	}
-
 	emms();
 
-	/* TODO: dynamic fcode/bcode ??? */
-
 	BitstreamPadAlways(bs); /* next_start_code() at the end of VideoObjectPlane() */
 	frame->length = (BitstreamPos(bs) - bits) / 8;