--- decoder.c	2002/10/11 15:07:32	1.37.2.5
+++ decoder.c	2002/12/09 10:47:05	1.37.2.17
@@ -55,7 +55,7 @@
  *  22.12.2001  lock based interpolation
  *  01.12.2001  inital version; (c)2001 peter ross <pross@cs.rmit.edu.au>
  *
- *  $Id: decoder.c,v 1.37.2.5 2002/10/11 15:07:32 Isibaar Exp $
+ *  $Id: decoder.c,v 1.37.2.17 2002/12/09 10:47:05 suxen_drol Exp $
  *
  *************************************************************************/
 
@@ -79,36 +79,41 @@
 #include "dct/fdct.h"
 #include "utils/mem_transfer.h"
 #include "image/interpolate8x8.h"
+#include "image/reduced.h"
 
 #include "bitstream/mbcoding.h"
 #include "prediction/mbprediction.h"
 #include "utils/timer.h"
 #include "utils/emms.h"
+#include "motion/motion.h"
 
 #include "image/image.h"
 #include "image/colorspace.h"
 #include "utils/mem_align.h"
 
 int
-decoder_create(XVID_DEC_PARAM * param)
+decoder_resize(DECODER * dec)
 {
-	DECODER *dec;
+	/* free existing */
 
-	dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
-	if (dec == NULL) {
-		return XVID_ERR_MEMORY;
-	}
-	param->handle = dec;
+	image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
+	image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
+	image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height);
+	image_destroy(&dec->refh, dec->edged_width, dec->edged_height);
+	image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
 
-	dec->width = param->width;
-	dec->height = param->height;
+	if (dec->last_mbs) 
+		xvid_free(dec->last_mbs);
+	if (dec->mbs)
+		xvid_free(dec->mbs);
+
+	/* realloc */
 
 	dec->mb_width = (dec->width + 15) / 16;
 	dec->mb_height = (dec->height + 15) / 16;
 
 	dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE;
 	dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE;
-	dec->low_delay = 0;
 
 	if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) {
 		xvid_free(dec);
@@ -120,6 +125,7 @@
 		xvid_free(dec);
 		return XVID_ERR_MEMORY;
 	}
+
 	// add by chenm001 <chenm001@163.com>
 	// for support B-frame to reference last 2 frame
 	if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) {
@@ -157,7 +163,6 @@
 		xvid_free(dec);
 		return XVID_ERR_MEMORY;
 	}
-
 	memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
 
 	// add by chenm001 <chenm001@163.com>
@@ -178,14 +183,50 @@
 
 	memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
 
+	return XVID_ERR_OK;
+}
+
+
+int
+decoder_create(XVID_DEC_PARAM * param)
+{
+	DECODER *dec;
+
+	dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
+	if (dec == NULL) {
+		return XVID_ERR_MEMORY;
+	}
+	memset(dec, 0, sizeof(DECODER));
+
+	param->handle = dec;
+
+	dec->width = param->width;
+	dec->height = param->height;
+
+	image_null(&dec->cur);
+	image_null(&dec->refn[0]);
+	image_null(&dec->refn[1]);
+	image_null(&dec->refn[2]);
+	image_null(&dec->refh);
+
+	dec->mbs = NULL;
+	dec->last_mbs = NULL;
+
 	init_timer();
 
 	// add by chenm001 <chenm001@163.com>
 	// for support B-frame to save reference frame's time
 	dec->frames = -1;
 	dec->time = dec->time_base = dec->last_time_base = 0;
+	dec->low_delay = 0;
+	dec->packed_mode = 0;
 
-	return XVID_ERR_OK;
+	dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
+
+	if (dec->fixed_dimensions)
+		return decoder_resize(dec);
+	else
+		return XVID_ERR_OK;		
 }
 
 
@@ -226,7 +267,8 @@
 				Bitstream * bs,
 				const uint32_t quant,
 				const uint32_t intra_dc_threshold,
-				const unsigned int bound)
+				const unsigned int bound,
+				const int reduced_resolution)
 {
 
 	DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
@@ -239,9 +281,15 @@
 	uint32_t iQuant = pMB->quant;
 	uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
 
-	pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
-	pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
-	pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
+	if (reduced_resolution) {
+		pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5);
+		pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4);
+		pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4);
+	}else{
+		pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
+		pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
+		pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
+	}
 
 	memset(block, 0, 6 * 64 * sizeof(int16_t));	// clear
 
@@ -302,6 +350,7 @@
 		start_timer();
 		idct(&data[i * 64]);
 		stop_idct_timer();
+
 	}
 
 	if (dec->interlacing && pMB->field_dct) {
@@ -310,12 +359,24 @@
 	}
 
 	start_timer();
-	transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
-	transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
-	transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
-	transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
-	transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
-	transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
+
+	if (reduced_resolution)
+	{
+		next_block*=2;
+		copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
+		copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
+		copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
+		copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
+		copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
+		copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
+	}else{
+		transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
+		transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
+		transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
+		transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
+		transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
+		transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
+	}
 	stop_transfer_timer();
 }
 
@@ -325,12 +386,27 @@
 
 #define SIGN(X) (((X)>0)?1:-1)
 #define ABS(X) (((X)>0)?(X):-(X))
-static const uint32_t roundtab[16] =
-	{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 };
-
 
 // decode an inter macroblock
 
+static void
+rrv_mv_scaleup(VECTOR * mv)
+{
+	if (mv->x > 0) {
+		mv->x = 2*mv->x - 1;
+	} else if (mv->x < 0) {
+		mv->x = 2*mv->x + 1;
+	}
+
+	if (mv->y > 0) {
+		mv->y = 2*mv->y - 1;
+	} else if (mv->y < 0) {
+		mv->y = 2*mv->y + 1;
+	}
+}
+
+
+
 void
 decoder_mbinter(DECODER * dec,
 				const MACROBLOCK * pMB,
@@ -340,7 +416,8 @@
 				const uint32_t cbp,
 				Bitstream * bs,
 				const uint32_t quant,
-				const uint32_t rounding)
+				const uint32_t rounding,
+				const int reduced_resolution)
 {
 
 	DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
@@ -348,102 +425,144 @@
 
 	uint32_t stride = dec->edged_width;
 	uint32_t stride2 = stride / 2;
-	uint32_t next_block = stride * 8;
+	uint32_t next_block = stride * (reduced_resolution ? 16 : 8);
 	uint32_t i;
 	uint32_t iQuant = pMB->quant;
 	uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
+
 	int uv_dx, uv_dy;
+	VECTOR mv[4];
 
-	pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
-	pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
-	pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
+	for (i = 0; i < 4; i++)
+	{
+		mv[i] = pMB->mvs[i];
+		//DPRINTF(DPRINTF_MB, "mv[%i]   orig=%i,%i   local=%i", i, pMB->mvs[i].x, pMB->mvs[i].y,						mv[i].x, mv[i].y);
+	}
+
+	if (reduced_resolution) {
+		pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5);
+		pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4);
+		pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4);
+		rrv_mv_scaleup(&mv[0]);
+		rrv_mv_scaleup(&mv[1]);
+		rrv_mv_scaleup(&mv[2]);
+		rrv_mv_scaleup(&mv[3]);
+	}else{
+		pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
+		pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
+		pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
+	}
 
 	if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) {
-		uv_dx = pMB->mvs[0].x;
-		uv_dy = pMB->mvs[0].y;
+		uv_dx = mv[0].x;
+		uv_dy = mv[0].y;
 
 		if (dec->quarterpel)
 		{
-			uv_dx = (uv_dx >> 1) | (uv_dx & 1);
-			uv_dy = (uv_dy >> 1) | (uv_dy & 1);
+			uv_dx /= 2;
+			uv_dy /= 2;
 		}
 
-		uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2;
-		uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2;
+		uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
+		uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
 
 		start_timer();
-		if(dec->quarterpel) {
-			interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
-	 								    dec->refh.y + 128, 16*x_pos, 16*y_pos,
-									    pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
-		}
-		else {
-			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos,
-								  pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
-			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos,
-							      pMB->mvs[1].x, pMB->mvs[1].y, stride,  rounding);
-			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8,
-								  pMB->mvs[2].x, pMB->mvs[2].y, stride,  rounding);
-			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, 
-								  pMB->mvs[3].x, pMB->mvs[3].y, stride,  rounding);
+		if (reduced_resolution)
+		{
+			interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
+								  mv[0].x, mv[0].y, stride,  rounding);
+			interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos,
+								  uv_dx, uv_dy, stride2, rounding);
+			interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos,
+								  uv_dx, uv_dy, stride2, rounding);
+
 		}
+		else
+		{
+			if(dec->quarterpel) {
+				interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+	 										dec->refh.y + 128, 16*x_pos, 16*y_pos,
+											mv[0].x, mv[0].y, stride,  rounding);
+			}
+			else {
+				interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos,
+									  mv[0].x, mv[0].y, stride,  rounding);
+			}
 
-		interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos,
-							  uv_dx, uv_dy, stride2, rounding);
-		interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos,
-							  uv_dx, uv_dy, stride2, rounding);
+			interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos,
+								  uv_dx, uv_dy, stride2, rounding);
+			interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos,
+								  uv_dx, uv_dy, stride2, rounding);
+		}
 		stop_comp_timer();
 
-	} else {
+	} else {	/* MODE_INTER4V */
 		int sum;
-		sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
-
-		if (dec->quarterpel)
-		{
-			sum /= 2;
-		}
-
-		uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
+		
+		if(dec->quarterpel)
+			sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
+		else
+			sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
+
+		uv_dx = (sum >> 3) + roundtab_76[sum & 0xf];
+
+		if(dec->quarterpel)
+			sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
+		else
+			sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
 
-		sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
+		uv_dy = (sum >> 3) + roundtab_76[sum & 0xf];
 
-		if (dec->quarterpel)
+		start_timer();
+		if (reduced_resolution)
 		{
-			sum /= 2;
-		}
-
-		uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
+			interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
+								  mv[0].x, mv[0].y, stride,  rounding);
+			interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos + 16, 32*y_pos,
+								  mv[1].x, mv[1].y, stride,  rounding);
+			interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos + 16,
+								  mv[2].x, mv[2].y, stride,  rounding);
+			interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos + 16, 32*y_pos + 16, 
+								  mv[3].x, mv[3].y, stride,  rounding);
+			interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos,
+								  uv_dx, uv_dy, stride2, rounding);
+			interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos,
+								  uv_dx, uv_dy, stride2, rounding);
+
+			// set_block(pY_Cur, stride, 32, 32, 127);
+		} 
+		else
+		{
+			if(dec->quarterpel) {
+				interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+										  dec->refh.y + 128, 16*x_pos, 16*y_pos,
+										  mv[0].x, mv[0].y, stride,  rounding);
+				interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+										  dec->refh.y + 128, 16*x_pos + 8, 16*y_pos,
+										  mv[1].x, mv[1].y, stride,  rounding);
+				interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+										  dec->refh.y + 128, 16*x_pos, 16*y_pos + 8,
+										  mv[2].x, mv[2].y, stride,  rounding);
+				interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+										  dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8,
+										  mv[3].x, mv[3].y, stride,  rounding);
+			}
+			else {
+				interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos,
+									  mv[0].x, mv[0].y, stride,  rounding);
+				interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos,
+									  mv[1].x, mv[1].y, stride,  rounding);
+				interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8,
+									  mv[2].x, mv[2].y, stride,  rounding);
+				interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, 
+									  mv[3].x, mv[3].y, stride,  rounding);
+			}
 
-		start_timer();
-		if(dec->quarterpel) {
-			interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
-									  dec->refh.y + 128, 16*x_pos, 16*y_pos,
-									  pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
-			interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
-									  dec->refh.y + 128, 16*x_pos + 8, 16*y_pos,
-									  pMB->mvs[1].x, pMB->mvs[1].y, stride,  rounding);
-			interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
-									  dec->refh.y + 128, 16*x_pos, 16*y_pos + 8,
-									  pMB->mvs[2].x, pMB->mvs[2].y, stride,  rounding);
-			interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
-									  dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8,
-									  pMB->mvs[3].x, pMB->mvs[3].y, stride,  rounding);
-		}
-		else {
-			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos,
-								  pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
-			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos,
-								  pMB->mvs[1].x, pMB->mvs[1].y, stride,  rounding);
-			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8,
-								  pMB->mvs[2].x, pMB->mvs[2].y, stride,  rounding);
-			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, 
-								  pMB->mvs[3].x, pMB->mvs[3].y, stride,  rounding);
+			interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos,
+								  uv_dx, uv_dy, stride2, rounding);
+			interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos,
+								  uv_dx, uv_dy, stride2, rounding);
 		}
-
-		interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos,
-							  uv_dx, uv_dy, stride2, rounding);
-		interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos,
-							  uv_dx, uv_dy, stride2, rounding);
 		stop_comp_timer();
 	}
 
@@ -478,18 +597,36 @@
 	}
 
 	start_timer();
-	if (cbp & 32)
-		transfer_16to8add(pY_Cur, &data[0 * 64], stride);
-	if (cbp & 16)
-		transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride);
-	if (cbp & 8)
-		transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride);
-	if (cbp & 4)
-		transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride);
-	if (cbp & 2)
-		transfer_16to8add(pU_Cur, &data[4 * 64], stride2);
-	if (cbp & 1)
-		transfer_16to8add(pV_Cur, &data[5 * 64], stride2);
+	if (reduced_resolution)
+	{
+		if (cbp & 32)
+			add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
+		if (cbp & 16)
+			add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
+		if (cbp & 8)
+			add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
+		if (cbp & 4)
+			add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
+		if (cbp & 2)
+			add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
+		if (cbp & 1)
+			add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
+	}
+	else
+	{
+		if (cbp & 32)
+			transfer_16to8add(pY_Cur, &data[0 * 64], stride);
+		if (cbp & 16)
+			transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride);
+		if (cbp & 8)
+			transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride);
+		if (cbp & 4)
+			transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride);
+		if (cbp & 2)
+			transfer_16to8add(pU_Cur, &data[4 * 64], stride2);
+		if (cbp & 1)
+			transfer_16to8add(pV_Cur, &data[5 * 64], stride2);
+	}
 	stop_transfer_timer();
 }
 
@@ -497,16 +634,25 @@
 void
 decoder_iframe(DECODER * dec,
 			   Bitstream * bs,
+			   int reduced_resolution,
 			   int quant,
 			   int intra_dc_threshold)
 {
 	uint32_t bound;
 	uint32_t x, y;
+	int mb_width = dec->mb_width;
+	int mb_height = dec->mb_height;
+	
+	if (reduced_resolution)
+	{
+		mb_width = (dec->width + 31) / 32;
+		mb_height = (dec->height + 31) / 32;
+	}
 
 	bound = 0;
 
-	for (y = 0; y < dec->mb_height; y++) {
-		for (x = 0; x < dec->mb_width; x++) {
+	for (y = 0; y < mb_height; y++) {
+		for (x = 0; x < mb_width; x++) {
 			MACROBLOCK *mb;
 			uint32_t mcbpc;
 			uint32_t cbpc;
@@ -519,9 +665,10 @@
 
 			if (check_resync_marker(bs, 0))
 			{
-				bound = read_video_packet_header(bs, 0, &quant);
-				x = bound % dec->mb_width;
-				y = bound / dec->mb_width;
+				bound = read_video_packet_header(bs, dec, 0, 
+							&quant, NULL, NULL, &intra_dc_threshold);
+				x = bound % mb_width;
+				y = bound / mb_width;
 			}
 			mb = &dec->mbs[y * dec->mb_width + x];
 
@@ -556,11 +703,11 @@
 			}
 
 			decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
-							intra_dc_threshold, bound);
+							intra_dc_threshold, bound, reduced_resolution);
+
 		}
 		if(dec->out_frm)
-		  output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width);
-
+		  output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
 	}
 
 }
@@ -572,7 +719,7 @@
 				  int x,
 				  int y,
 				  int k,
-				  VECTOR * mv,
+				  VECTOR * ret_mv,
 				  int fcode,
 				  const int bound)
 {
@@ -583,48 +730,73 @@
 	int range = (64 * scale_fac);
 
 	VECTOR pmv;
-	int mv_x, mv_y;
+	VECTOR mv;
 
 	pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
 
-	mv_x = get_mv(bs, fcode);
-	mv_y = get_mv(bs, fcode);
+	mv.x = get_mv(bs, fcode);
+	mv.y = get_mv(bs, fcode);
 
-	DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y);
+	DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv.x, mv.y, pmv.x, pmv.y);
 
-	mv_x += pmv.x;
-	mv_y += pmv.y;
+	mv.x += pmv.x;
+	mv.y += pmv.y;
 
-	if (mv_x < low) {
-		mv_x += range;
-	} else if (mv_x > high) {
-		mv_x -= range;
+	if (mv.x < low) {
+		mv.x += range;
+	} else if (mv.x > high) {
+		mv.x -= range;
 	}
 
-	if (mv_y < low) {
-		mv_y += range;
-	} else if (mv_y > high) {
-		mv_y -= range;
+	if (mv.y < low) {
+		mv.y += range;
+	} else if (mv.y > high) {
+		mv.y -= range;
 	}
 
-	mv->x = mv_x;
-	mv->y = mv_y;
+	ret_mv->x = mv.x;
+	ret_mv->y = mv.y;
+}
 
+
+
+static __inline int gmc_sanitize(int value, int quarterpel, int fcode)
+{
+	int length = 1 << (fcode+4);
+
+	if (quarterpel) value *= 2;
+
+	if (value < -length) 
+		return -length;
+	else if (value >= length) 
+		return length-1;
+	else return value;
 }
 
 
+/* for P_VOP set gmc_mv to NULL */
 void
 decoder_pframe(DECODER * dec,
 			   Bitstream * bs,
 			   int rounding,
+			   int reduced_resolution,
 			   int quant,
 			   int fcode,
-			   int intra_dc_threshold)
+			   int intra_dc_threshold,
+			   VECTOR * gmc_mv)
 {
 
 	uint32_t x, y;
 	uint32_t bound;
 	int cp_mb, st_mb;
+	int mb_width = dec->mb_width;
+	int mb_height = dec->mb_height;
+	
+	if (reduced_resolution)
+	{
+		mb_width = (dec->width + 31) / 32;
+		mb_height = (dec->height + 31) / 32;
+	}
 
 	start_timer();
 	image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
@@ -633,9 +805,9 @@
 
 	bound = 0;
 
-	for (y = 0; y < dec->mb_height; y++) {
+	for (y = 0; y < mb_height; y++) {
 		cp_mb = st_mb = 0;
-		for (x = 0; x < dec->mb_width; x++) {
+		for (x = 0; x < mb_width; x++) {
 			MACROBLOCK *mb;
 
 			// skip stuffing
@@ -644,9 +816,10 @@
 
 			if (check_resync_marker(bs, fcode - 1))
 			{
-				bound = read_video_packet_header(bs, fcode - 1, &quant);
-				x = bound % dec->mb_width;
-				y = bound / dec->mb_width;
+				bound = read_video_packet_header(bs, dec, fcode - 1, 
+					&quant, &fcode, NULL, &intra_dc_threshold);
+				x = bound % mb_width;
+				y = bound / mb_width;
 			}
 			mb = &dec->mbs[y * dec->mb_width + x];
 
@@ -661,6 +834,7 @@
 				uint32_t cbpy;
 				uint32_t cbp;
 				uint32_t intra;
+				int mcsel = 0;		// mcsel: '0'=local motion, '1'=GMC
 
 				cp_mb++;
 				mcbpc = get_mcbpc_inter(bs);
@@ -677,6 +851,11 @@
 					acpred_flag = BitstreamGetBit(bs);
 				}
 
+				if (gmc_mv && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
+				{
+					mcsel = BitstreamGetBit(bs);
+				}
+
 				cbpy = get_cbpy(bs, intra);
 				DPRINTF(DPRINTF_MB, "cbpy %i", cbpy);
 
@@ -715,7 +894,13 @@
 				}
 
 				if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
-					if (dec->interlacing && mb->field_pred) {
+
+					if (mcsel)
+					{
+						mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode);
+						mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode);
+
+					} else if (dec->interlacing && mb->field_pred) {
 						get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0],
 										  fcode, bound);
 						get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1],
@@ -741,56 +926,62 @@
 					mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y =
 						0;
 					decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
-									intra_dc_threshold, bound);
+									intra_dc_threshold, bound, reduced_resolution);
 					continue;
 				}
 
 				decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant,
-								rounding);
-			} else				// not coded
+								rounding, reduced_resolution);
+
+			} 
+			else if (gmc_mv)	/* not coded S_VOP macroblock */
+			{
+				mb->mode = MODE_NOT_CODED;
+				mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode);
+				mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode);
+				decoder_mbinter(dec, mb, x, y, 0, 0, bs, quant, rounding, reduced_resolution);
+			}
+			else	/* not coded P_VOP macroblock */
 			{
 				mb->mode = MODE_NOT_CODED;
+
 				mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
 				mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
-
 				// copy macroblock directly from ref to cur
 
 				start_timer();
 
-				transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width +
-								 (16 * x),
-								 dec->refn[0].y + (16 * y) * dec->edged_width +
-								 (16 * x), dec->edged_width);
-
-				transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width +
-								 (16 * x + 8),
-								 dec->refn[0].y + (16 * y) * dec->edged_width +
-								 (16 * x + 8), dec->edged_width);
-
-				transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width +
-								 (16 * x),
-								 dec->refn[0].y + (16 * y +
-												   8) * dec->edged_width +
-								 (16 * x), dec->edged_width);
-
-				transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width +
-								 (16 * x + 8),
-								 dec->refn[0].y + (16 * y +
-												   8) * dec->edged_width +
-								 (16 * x + 8), dec->edged_width);
-
-				transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 +
-								 (8 * x),
-								 dec->refn[0].u +
-								 (8 * y) * dec->edged_width / 2 + (8 * x),
-								 dec->edged_width / 2);
-
-				transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 +
-								 (8 * x),
-								 dec->refn[0].v +
-								 (8 * y) * dec->edged_width / 2 + (8 * x),
-								 dec->edged_width / 2);
+				if (reduced_resolution)
+				{
+					transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x),
+									 dec->refn[0].y + (32*y)*dec->edged_width + (32*x), 
+									 dec->edged_width);
+
+					transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x),
+									dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x),
+									dec->edged_width/2);
+
+					transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x),
+									 dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x),
+									 dec->edged_width/2);
+				}
+				else
+				{
+					transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x),
+									 dec->refn[0].y + (16*y)*dec->edged_width + (16*x), 
+									 dec->edged_width);
+
+					transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x),
+									dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x),
+									dec->edged_width/2);
+
+					transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x),
+									 dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x),
+									 dec->edged_width/2);
+				}
+				
 				stop_transfer_timer();
+
 				if(dec->out_frm && cp_mb > 0) {
 				  output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
 				  cp_mb = 0;
@@ -882,35 +1073,49 @@
 		uv_dx = pMB->mvs[0].x;
 		uv_dy = pMB->mvs[0].y;
 
-		uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2;
-		uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2;
+		if (dec->quarterpel)
+		{
+			uv_dx /= 2;
+			uv_dy /= 2;
+		}
+
+		uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
+		uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
 	} else {
 		int sum;
 
-		sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
-		uv_dx =
-			(sum ==
-			 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] +
-								  (ABS(sum) / 16) * 2));
-
-		sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
-		uv_dy =
-			(sum ==
-			 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] +
-								  (ABS(sum) / 16) * 2));
+		if(dec->quarterpel)
+			sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
+		else
+			sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
+
+		uv_dx = (sum >> 3) + roundtab_76[sum & 0xf];
+
+		if(dec->quarterpel)
+			sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
+		else
+			sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
+
+		uv_dy = (sum >> 3) + roundtab_76[sum & 0xf];
 	}
 
 	start_timer();
-	interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos,
-						  pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
-	interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8,
-						  16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
-	interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos,
-						  16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride,
-						  0);
-	interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8,
-						  16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride,
-						  0);
+	if(dec->quarterpel) {
+		interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->refh.y, dec->refh.y + 64,
+ 								    dec->refh.y + 128, 16*x_pos, 16*y_pos,
+								    pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
+	}
+	else {
+		interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
+							  pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
+		interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos,
+						      pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
+		interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8,
+							  pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
+		interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, 
+							  pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
+	}
+
 	interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
 						  uv_dx, uv_dy, stride2, 0);
 	interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
@@ -997,74 +1202,130 @@
 		uv_dx = pMB->mvs[0].x;
 		uv_dy = pMB->mvs[0].y;
 
-		uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2;
-		uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2;
-
 		b_uv_dx = pMB->b_mvs[0].x;
 		b_uv_dy = pMB->b_mvs[0].y;
 
-		b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2;
-		b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2;
+		if (dec->quarterpel)
+		{
+			uv_dx /= 2;
+			uv_dy /= 2;
+
+			b_uv_dx /= 2;
+			b_uv_dy /= 2;
+		}
+
+		uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
+		uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
+
+		b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
+		b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
 	} else {
 		int sum;
 
-		sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
-		uv_dx =
-			(sum ==
-			 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] +
-								  (ABS(sum) / 16) * 2));
-
-		sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
-		uv_dy =
-			(sum ==
-			 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] +
-								  (ABS(sum) / 16) * 2));
-
-		sum =
-			pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x +
-			pMB->b_mvs[3].x;
-		b_uv_dx =
-			(sum ==
-			 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] +
-								  (ABS(sum) / 16) * 2));
-
-		sum =
-			pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y +
-			pMB->b_mvs[3].y;
-		b_uv_dy =
-			(sum ==
-			 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] +
-								  (ABS(sum) / 16) * 2));
+		if(dec->quarterpel)
+			sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
+		else
+			sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
+
+		uv_dx = (sum >> 3) + roundtab_76[sum & 0xf];
+
+		if(dec->quarterpel)
+			sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
+		else
+			sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
+
+		uv_dy = (sum >> 3) + roundtab_76[sum & 0xf];
+
+
+		if(dec->quarterpel)
+			sum = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2);
+		else
+			sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
+
+		b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf];
+
+		if(dec->quarterpel)
+			sum = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2);
+		else
+			sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
+
+		b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf];
 	}
 
 
 	start_timer();
-	interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
-						  pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
-	interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
-						  pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
-	interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
-						  pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
-	interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8,
-						  16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride,
-						  0);
+	if(dec->quarterpel) {
+		if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q))
+			interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos, 16*y_pos,
+									    pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
+		else {
+			interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos, 16*y_pos,
+									    pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
+			interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos + 8, 16*y_pos,
+									    pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
+			interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos, 16*y_pos + 8,
+									    pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
+			interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8,
+									    pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
+		}
+	}
+	else {
+		interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
+							  pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
+		interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
+							  pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
+		interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
+							  pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
+		interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8,
+							  16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride,
+							  0);
+	}
+
 	interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
 						  uv_dy, stride2, 0);
 	interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
 						  uv_dy, stride2, 0);
 
 
-	interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos,
-						  pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
-	interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8,
-						  16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride,
-						  0);
-	interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos,
-						  16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y,
-						  stride, 0);
-	interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8,
-						  16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y,
-						  stride, 0);
+	if(dec->quarterpel) {
+		if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q))
+			interpolate16x16_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos, 16*y_pos,
+									    pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
+		else {
+			interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos, 16*y_pos,
+									    pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
+			interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos + 8, 16*y_pos,
+									    pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
+			interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos, 16*y_pos + 8,
+									    pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
+			interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64,
+ 									    dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8,
+									    pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
+		}
+	}
+	else {
+		interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos,
+							  pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
+		interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8,
+							  16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride,
+							  0);
+		interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos,
+							  16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y,
+							  stride, 0);
+		interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8,
+							  16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y,
+							  stride, 0);
+	}
+
 	interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos,
 						  b_uv_dx, b_uv_dy, stride2, 0);
 	interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos,
@@ -1073,32 +1334,32 @@
 	interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
 						dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
 						dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos,
-						stride, 0);
+						stride, 1, 8);
 
 	interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
 						dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
 						dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos + 8,
-						stride, 0);
+						stride, 1, 8);
 
 	interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
 						dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
 						dec->refn[2].y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
-						stride, 0);
+						stride, 1, 8);
 
 	interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
 						dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
 						dec->refn[2].y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
-						stride, 0);
+						stride, 1, 8);
 
 	interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->refn[2].u + (8 * y_pos * stride2) + 8 * x_pos,
-						stride2, 0);
+						stride2, 1, 8);
 
 	interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
 						dec->refn[2].v + (8 * y_pos * stride2) + 8 * x_pos,
-						stride2, 0);
+						stride2, 1, 8);
 
 	stop_comp_timer();
 
@@ -1368,16 +1629,19 @@
 
 int
 decoder_decode(DECODER * dec,
-			   XVID_DEC_FRAME * frame)
+			   XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats)
 {
 
 	Bitstream bs;
 	uint32_t rounding;
+	uint32_t reduced_resolution;
 	uint32_t quant;
 	uint32_t fcode_forward;
 	uint32_t fcode_backward;
 	uint32_t intra_dc_threshold;
+	VECTOR gmc_mv[5];
 	uint32_t vop_type;
+	int success = 0;
 
 	start_global_timer();
 
@@ -1385,29 +1649,68 @@
 
 	BitstreamInit(&bs, frame->bitstream, frame->length);
 
-	if(BitstreamShowBits(&bs, 8) == 0x7f)
+	// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's
+	if(frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
+	{
+		if (stats)
+			stats->notify = XVID_DEC_VOP;
+		frame->length = 1;
+		image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
+					 frame->image, frame->stride, frame->colorspace, dec->interlacing);
+		emms();
 		return XVID_ERR_OK;
+	}
 
+start:
 	// add by chenm001 <chenm001@163.com>
 	// for support B-frame to reference last 2 frame
 	dec->frames++;
+
+xxx:
 	vop_type =
-		BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward,
-							 &fcode_backward, &intra_dc_threshold);
+		BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, 
+			&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, gmc_mv);
+
+	//DPRINTF(DPRINTF_HEADER, "vop_type=%i", vop_type);
+
+	if (vop_type == -1 && success)
+		goto done;
+
+	if (vop_type == -2 || vop_type == -3)
+	{
+		if (vop_type == -3)
+			decoder_resize(dec);
+			
+		if (stats)
+		{
+			stats->notify = XVID_DEC_VOL;
+			stats->data.vol.general = 0;
+			if (dec->interlacing)
+				stats->data.vol.general |= XVID_INTERLACING;
+			stats->data.vol.width = dec->width;
+			stats->data.vol.height = dec->height;
+			stats->data.vol.aspect_ratio = dec->aspect_ratio;
+			stats->data.vol.par_width = dec->par_width;
+			stats->data.vol.par_height = dec->par_height;
+			frame->length = BitstreamPos(&bs) / 8;
+			return XVID_ERR_OK;
+		}
+		goto xxx;
+	} 
 
 	dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0;	// init pred vector to 0
 
 	switch (vop_type) {
 	case P_VOP:
-		decoder_pframe(dec, &bs, rounding, quant, fcode_forward,
-					   intra_dc_threshold);
+		decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, 
+						fcode_forward, intra_dc_threshold, NULL);
 #ifdef BFRAMES_DEC
 		DEBUG1("P_VOP  Time=", dec->time);
 #endif
 		break;
 
 	case I_VOP:
-		decoder_iframe(dec, &bs, quant, intra_dc_threshold);
+		decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold);
 #ifdef BFRAMES_DEC
 		DEBUG1("I_VOP  Time=", dec->time);
 #endif
@@ -1426,6 +1729,11 @@
 #endif
 		break;
 
+	case S_VOP :
+		decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, 
+						fcode_forward, intra_dc_threshold, gmc_mv);
+		break;
+
 	case N_VOP:				// vop not coded
 		// when low_delay==0, N_VOP's should interpolate between the past and future frames
 		image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
@@ -1435,43 +1743,107 @@
 		break;
 
 	default:
+		if (stats)
+			stats->notify = 0;
+
+		emms();
 		return XVID_ERR_FAIL;
 	}
 
-#ifdef BFRAMES_DEC_DEBUG
-	if (frame->length != BitstreamPos(&bs) / 8){
-		DEBUG2("InLen/UseLen",frame->length, BitstreamPos(&bs) / 8);
+
+	/* reduced resolution deblocking filter */
+
+	if (reduced_resolution)
+	{
+		const int rmb_width = (dec->width + 31) / 32;
+		const int rmb_height = (dec->height + 31) / 32;
+		const int edged_width2 = dec->edged_width /2;
+		int i,j;
+
+		/* horizontal deblocking */
+
+		for (j = 1; j < rmb_height*2; j++)	// luma: j,i in block units
+		for (i = 0; i < rmb_width*2; i++)
+		{
+			if (dec->mbs[(j-1)/2*dec->mb_width + (i/2)].mode != MODE_NOT_CODED ||
+				dec->mbs[(j+0)/2*dec->mb_width + (i/2)].mode != MODE_NOT_CODED)
+			{
+				xvid_HFilter_31_C(dec->cur.y + (j*16 - 1)*dec->edged_width + i*16,
+							      dec->cur.y + (j*16 + 0)*dec->edged_width + i*16, 2);
+			}
+		}
+
+		for (j = 1; j < rmb_height; j++)	// chroma
+		for (i = 0; i < rmb_width; i++)
+		{
+			if (dec->mbs[(j-1)*dec->mb_width + i].mode != MODE_NOT_CODED || 
+				dec->mbs[(j+0)*dec->mb_width + i].mode != MODE_NOT_CODED)
+			{
+				hfilter_31(dec->cur.u + (j*16 - 1)*edged_width2 + i*16,
+								  dec->cur.u + (j*16 + 0)*edged_width2 + i*16, 2);
+				hfilter_31(dec->cur.v + (j*16 - 1)*edged_width2 + i*16,
+								  dec->cur.v + (j*16 + 0)*edged_width2 + i*16, 2);
+			}
+		}
+
+		/* vertical deblocking */
+
+		for (j = 0; j < rmb_height*2; j++)		// luma: i,j in block units
+		for (i = 1; i < rmb_width*2; i++)
+		{
+			if (dec->mbs[(j/2)*dec->mb_width + (i-1)/2].mode != MODE_NOT_CODED ||
+				dec->mbs[(j/2)*dec->mb_width + (i+0)/2].mode != MODE_NOT_CODED)
+			{
+				vfilter_31(dec->cur.y + (j*16)*dec->edged_width + i*16 - 1,
+							      dec->cur.y + (j*16)*dec->edged_width + i*16 + 0,
+								  dec->edged_width, 2);
+			}
+		}
+
+		for (j = 0; j < rmb_height; j++)	// chroma
+		for (i = 1; i < rmb_width; i++)
+		{
+			if (dec->mbs[j*dec->mb_width + i - 1].mode != MODE_NOT_CODED ||
+				dec->mbs[j*dec->mb_width + i + 0].mode != MODE_NOT_CODED) 
+			{
+				vfilter_31(dec->cur.u + (j*16)*edged_width2 + i*16 - 1,
+								  dec->cur.u + (j*16)*edged_width2 + i*16 + 0,
+								  edged_width2, 2);
+				vfilter_31(dec->cur.v + (j*16)*edged_width2 + i*16 - 1,
+								  dec->cur.v + (j*16)*edged_width2 + i*16 + 0,
+								  edged_width2, 2);
+			}
+		}
 	}
-#endif
-	frame->length = BitstreamPos(&bs) / 8;
 
+	BitstreamByteAlign(&bs);
 
 #ifdef BFRAMES_DEC
 	// test if no B_VOP
-        if (dec->low_delay || dec->frames == 0) {
+	if (dec->low_delay || dec->frames == 0 || ((dec->packed_mode) && !(frame->length > BitstreamPos(&bs) / 8))) {
 #endif
-	image_output(&dec->cur, dec->width, dec->height, dec->edged_width,
-					 frame->image, frame->stride, frame->colorspace);
+		image_output(&dec->cur, dec->width, dec->height, dec->edged_width,
+					 frame->image, frame->stride, frame->colorspace, dec->interlacing);
 
 #ifdef BFRAMES_DEC
 	} else {
-                if (dec->frames >= 1) {
+		if (dec->frames >= 1 && !(dec->packed_mode)) {
 			start_timer();
-			if ((vop_type == I_VOP || vop_type == P_VOP)) {
+			if ((vop_type == I_VOP || vop_type == P_VOP || vop_type == S_VOP)) {
 				image_output(&dec->refn[0], dec->width, dec->height,
 							 dec->edged_width, frame->image, frame->stride,
-							 frame->colorspace);
+							 frame->colorspace, dec->interlacing);
 			} else if (vop_type == B_VOP) {
 				image_output(&dec->cur, dec->width, dec->height,
 							 dec->edged_width, frame->image, frame->stride,
-							 frame->colorspace);
+							 frame->colorspace, dec->interlacing);
 			}
 			stop_conv_timer();
 		}
 	}
 #endif
 
-	if (vop_type == I_VOP || vop_type == P_VOP) {
+	if (vop_type == I_VOP || vop_type == P_VOP || vop_type == S_VOP) {
 		image_swap(&dec->refn[0], &dec->refn[1]);
 		image_swap(&dec->cur, &dec->refn[0]);
 
@@ -1487,6 +1859,25 @@
 			mb_swap(&dec->mbs, &dec->last_mbs);
 	}
 
+
+	if (success == 0 && dec->packed_mode)
+	{
+		success = 1;
+	//	if (frame->length > BitstreamPos(&bs) / 8)	// multiple vops packed together
+		goto start;
+	}
+
+done :
+
+	frame->length = BitstreamPos(&bs) / 8;
+
+	if (stats)
+	{
+		stats->notify = XVID_DEC_VOP;
+		stats->data.vop.time_base = (int)dec->time_base;
+		stats->data.vop.time_increment = 0;	//XXX: todo
+	}
+
 	emms();
 
 	stop_global_timer();