--- decoder.c	2002/12/14 09:39:42	1.37.2.22
+++ decoder.c	2002/10/19 12:20:33	1.42
@@ -3,6 +3,9 @@
  *  XVID MPEG-4 VIDEO CODEC
  *  -  Decoder main module  -
  *
+ *  Copyright(C) 2002 MinChen <chenm001@163.com>
+ *               2002 Peter Ross <pross@xvid.org>
+ *
  *  This program is an implementation of a part of one or more MPEG-4
  *  Video tools as specified in ISO/IEC 14496-2 standard.  Those intending
  *  to use this software module in hardware or software products are
@@ -33,7 +36,7 @@
  *  History:
  *
  *  15.07.2002  fix a bug in B-frame decode at DIRECT mode
- *              MinChen <chenm001@163.com>
+
  *  10.07.2002  added BFRAMES_DEC_DEBUG support
  *              Fix a little bug for low_delay flage
  *              MinChen <chenm001@163.com>
@@ -53,9 +56,9 @@
  *  26.03.2002  interlacing support - moved transfers outside decode loop
  *  26.12.2001  decoder_mbinter: dequant/idct moved within if(coded) block
  *  22.12.2001  lock based interpolation
- *  01.12.2001  inital version; (c)2001 peter ross <pross@cs.rmit.edu.au>
+ *  01.12.2001  inital version; (c)2001 peter ross <pross@xvid.org>
  *
- *  $Id: decoder.c,v 1.37.2.22 2002/12/14 09:39:42 suxen_drol Exp $
+ *  $Id: decoder.c,v 1.42 2002/10/19 12:20:33 edgomez Exp $
  *
  *************************************************************************/
 
@@ -79,42 +82,36 @@
 #include "dct/fdct.h"
 #include "utils/mem_transfer.h"
 #include "image/interpolate8x8.h"
-#include "image/reduced.h"
-#include "image/font.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_resize(DECODER * dec)
+decoder_create(XVID_DEC_PARAM * param)
 {
-	/* free existing */
-
-	image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
-	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->tmp, dec->edged_width, dec->edged_height);
-	image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
+	DECODER *dec;
 
-	if (dec->last_mbs) 
-		xvid_free(dec->last_mbs);
-	if (dec->mbs)
-		xvid_free(dec->mbs);
+	dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
+	if (dec == NULL) {
+		return XVID_ERR_MEMORY;
+	}
+	param->handle = dec;
 
-	/* realloc */
+	dec->width = param->width;
+	dec->height = param->height;
 
 	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);
@@ -126,7 +123,6 @@
 		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)) {
@@ -135,7 +131,7 @@
 		xvid_free(dec);
 		return XVID_ERR_MEMORY;
 	}
-	if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) {
+	if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) {
 		image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
 		image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
 		image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
@@ -143,15 +139,6 @@
 		return XVID_ERR_MEMORY;
 	}
 
-	if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) {
-		image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
-		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->tmp, dec->edged_width, dec->edged_height);
-		xvid_free(dec);
-		return XVID_ERR_MEMORY;
-	}
-
 	dec->mbs =
 		xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
 					CACHE_LINE);
@@ -159,11 +146,11 @@
 		image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
 		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->tmp, dec->edged_width, dec->edged_height);
-		image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
+		image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height);
 		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>
@@ -176,58 +163,21 @@
 		image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
 		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->tmp, dec->edged_width, dec->edged_height);
-		image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
+		image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height);
 		xvid_free(dec);
 		return XVID_ERR_MEMORY;
 	}
 
 	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->tmp);
-	image_null(&dec->qtmp);
-
-	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 = 0;
+	dec->frames = -1;
 	dec->time = dec->time_base = dec->last_time_base = 0;
-	dec->low_delay = 0;
-	dec->packed_mode = 0;
-
-	dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
 
-	if (dec->fixed_dimensions)
-		return decoder_resize(dec);
-	else
-		return XVID_ERR_OK;		
+	return XVID_ERR_OK;
 }
 
 
@@ -238,8 +188,7 @@
 	xvid_free(dec->mbs);
 	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->tmp, dec->edged_width, dec->edged_height);
-	image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
+	image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height);
 	image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
 	xvid_free(dec);
 
@@ -268,8 +217,7 @@
 				Bitstream * bs,
 				const uint32_t quant,
 				const uint32_t intra_dc_threshold,
-				const unsigned int bound,
-				const int reduced_resolution)
+				const unsigned int bound)
 {
 
 	DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
@@ -282,15 +230,9 @@
 	uint32_t iQuant = pMB->quant;
 	uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
 
-	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);
-	}
+	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
 
@@ -329,10 +271,8 @@
 		start_timer();
 		if (cbp & (1 << (5 - i)))	// coded
 		{
-			int direction = dec->alternate_vertical_scan ?
-				2 : pMB->acpred_directions[i];
-
-			get_intra_block(bs, &block[i * 64], direction, start_coeff);
+			get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i],
+							start_coeff);
 		}
 		stop_coding_timer();
 
@@ -351,7 +291,6 @@
 		start_timer();
 		idct(&data[i * 64]);
 		stop_idct_timer();
-
 	}
 
 	if (dec->interlacing && pMB->field_dct) {
@@ -360,24 +299,12 @@
 	}
 
 	start_timer();
-
-	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);
-	}
+	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();
 }
 
@@ -387,6 +314,9 @@
 
 #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
 
@@ -399,8 +329,7 @@
 				const uint32_t cbp,
 				Bitstream * bs,
 				const uint32_t quant,
-				const uint32_t rounding,
-				const int reduced_resolution)
+				const uint32_t rounding)
 {
 
 	DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
@@ -408,152 +337,85 @@
 
 	uint32_t stride = dec->edged_width;
 	uint32_t stride2 = stride / 2;
-	uint32_t next_block = stride * (reduced_resolution ? 16 : 8);
+	uint32_t next_block = stride * 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];	/* local copy of mvs */
 
-	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);
-		for (i = 0; i < 4; i++)	{
-			mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x);
-			mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y);
-		}
-	}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);
-		for (i = 0; i < 4; i++)
-			mv[i] = pMB->mvs[i];
-	}
+	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 = mv[0].x;
-		uv_dy = mv[0].y;
+		uv_dx = pMB->mvs[0].x;
+		uv_dy = pMB->mvs[0].y;
 
 		if (dec->quarterpel)
 		{
-			uv_dx /= 2;
-			uv_dy /= 2;
+			uv_dx = (uv_dx >> 1) | (uv_dx & 1);
+			uv_dy = (uv_dy >> 1) | (uv_dy & 1);
 		}
 
-		uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
-		uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
+		uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2;
+		uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2;
+	} else {
+		int sum;
+		sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
 
-		start_timer();
-		if (reduced_resolution)
+		if (dec->quarterpel)
 		{
-			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);
-
+			sum /= 2;
 		}
-		else
-		{
-			if(dec->quarterpel) {
-				interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64,
-	 										dec->qtmp.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);
-		}
-		stop_comp_timer();
-
-	} else {	/* MODE_INTER4V */
-		int sum;
-		
-		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;
+		uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
 
-		uv_dy = (sum >> 3) + roundtab_76[sum & 0xf];
+		sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
 
-		start_timer();
-		if (reduced_resolution)
-		{
-			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)
 		{
-			if(dec->quarterpel) {
-				interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64,
-										  dec->qtmp.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->qtmp.y, dec->qtmp.y + 64,
-										  dec->qtmp.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->qtmp.y, dec->qtmp.y + 64,
-										  dec->qtmp.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->qtmp.y, dec->qtmp.y + 64,
-										  dec->qtmp.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);
-			}
-
-			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);
+			sum /= 2;
 		}
-		stop_comp_timer();
+
+		uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
 	}
 
-	for (i = 0; i < 6; i++) {
-		int direction = dec->alternate_vertical_scan ? 2 : 0;
+	start_timer();
+	if(dec->quarterpel) {
+		DPRINTF(DPRINTF_DEBUG, "QUARTERPEL\n");
+		interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 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, 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, 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, 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);
+	stop_comp_timer();
+
+	for (i = 0; i < 6; i++) {
 		if (cbp & (1 << (5 - i)))	// coded
 		{
 			memset(&block[i * 64], 0, 64 * sizeof(int16_t));	// clear
 
 			start_timer();
-			get_inter_block(bs, &block[i * 64], direction);
+			get_inter_block(bs, &block[i * 64]);
 			stop_coding_timer();
 
 			start_timer();
@@ -576,36 +438,18 @@
 	}
 
 	start_timer();
-	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);
-	}
+	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();
 }
 
@@ -613,25 +457,16 @@
 void
 decoder_iframe(DECODER * dec,
 			   Bitstream * bs,
-			   int reduced_resolution,
 			   int quant,
 			   int intra_dc_threshold)
 {
 	uint32_t bound;
 	uint32_t x, y;
-	uint32_t mb_width = dec->mb_width;
-	uint32_t 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 < mb_height; y++) {
-		for (x = 0; x < mb_width; x++) {
+	for (y = 0; y < dec->mb_height; y++) {
+		for (x = 0; x < dec->mb_width; x++) {
 			MACROBLOCK *mb;
 			uint32_t mcbpc;
 			uint32_t cbpc;
@@ -644,10 +479,9 @@
 
 			if (check_resync_marker(bs, 0))
 			{
-				bound = read_video_packet_header(bs, dec, 0, 
-							&quant, NULL, NULL, &intra_dc_threshold);
-				x = bound % mb_width;
-				y = bound / mb_width;
+				bound = read_video_packet_header(bs, 0, &quant);
+				x = bound % dec->mb_width;
+				y = bound / dec->mb_width;
 			}
 			mb = &dec->mbs[y * dec->mb_width + x];
 
@@ -678,15 +512,15 @@
 
 			if (dec->interlacing) {
 				mb->field_dct = BitstreamGetBit(bs);
-				DPRINTF(DPRINTF_MB,"deci: field_dct: %i", mb->field_dct);
+				DPRINTF(DPRINTF_DEBUG, "deci: field_dct: %d", mb->field_dct);
 			}
 
 			decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
-							intra_dc_threshold, bound, reduced_resolution);
-
+							intra_dc_threshold, bound);
 		}
 		if(dec->out_frm)
-		  output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
+		  output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width);
+
 	}
 
 }
@@ -698,7 +532,7 @@
 				  int x,
 				  int y,
 				  int k,
-				  VECTOR * ret_mv,
+				  VECTOR * mv,
 				  int fcode,
 				  const int bound)
 {
@@ -709,73 +543,48 @@
 	int range = (64 * scale_fac);
 
 	VECTOR pmv;
-	VECTOR mv;
+	int mv_x, mv_y;
 
 	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;
 	}
 
-	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;
+	mv->x = mv_x;
+	mv->y = mv_y;
 
-	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,
-			   VECTOR * gmc_mv)
+			   int intra_dc_threshold)
 {
 
 	uint32_t x, y;
 	uint32_t bound;
 	int cp_mb, st_mb;
-	uint32_t mb_width = dec->mb_width;
-	uint32_t 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,
@@ -784,9 +593,9 @@
 
 	bound = 0;
 
-	for (y = 0; y < mb_height; y++) {
+	for (y = 0; y < dec->mb_height; y++) {
 		cp_mb = st_mb = 0;
-		for (x = 0; x < mb_width; x++) {
+		for (x = 0; x < dec->mb_width; x++) {
 			MACROBLOCK *mb;
 
 			// skip stuffing
@@ -795,10 +604,9 @@
 
 			if (check_resync_marker(bs, fcode - 1))
 			{
-				bound = read_video_packet_header(bs, dec, fcode - 1, 
-					&quant, &fcode, NULL, &intra_dc_threshold);
-				x = bound % mb_width;
-				y = bound / mb_width;
+				bound = read_video_packet_header(bs, fcode - 1, &quant);
+				x = bound % dec->mb_width;
+				y = bound / dec->mb_width;
 			}
 			mb = &dec->mbs[y * dec->mb_width + x];
 
@@ -813,7 +621,6 @@
 				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);
@@ -830,11 +637,6 @@
 					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);
 
@@ -856,30 +658,24 @@
 				if (dec->interlacing) {
 					if (cbp || intra) {
 						mb->field_dct = BitstreamGetBit(bs);
-						DPRINTF(DPRINTF_MB,"decp: field_dct: %i", mb->field_dct);
+						DPRINTF(DPRINTF_DEBUG, "decp: field_dct: %d", mb->field_dct);
 					}
 
 					if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
 						mb->field_pred = BitstreamGetBit(bs);
-						DPRINTF(DPRINTF_MB, "decp: field_pred: %i", mb->field_pred);
+						DPRINTF(DPRINTF_DEBUG, "decp: field_pred: %d", mb->field_pred);
 
 						if (mb->field_pred) {
 							mb->field_for_top = BitstreamGetBit(bs);
-							DPRINTF(DPRINTF_MB,"decp: field_for_top: %i", mb->field_for_top);
+							DPRINTF(DPRINTF_DEBUG, "decp: field_for_top: %d", mb->field_for_top);
 							mb->field_for_bot = BitstreamGetBit(bs);
-							DPRINTF(DPRINTF_MB,"decp: field_for_bot: %i", mb->field_for_bot);
+							DPRINTF(DPRINTF_DEBUG, "decp: field_for_bot: %d", mb->field_for_bot);
 						}
 					}
 				}
 
 				if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
-
-					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) {
+					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],
@@ -905,62 +701,58 @@
 					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, reduced_resolution);
+									intra_dc_threshold, bound);
 					continue;
 				}
 
 				decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant,
-								rounding, reduced_resolution);
-
-			} 
-			else if (gmc_mv)	/* not coded S_VOP macroblock */
+								rounding);
+			} else				// not coded
 			{
-				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;
+				DPRINTF(DPRINTF_DEBUG, "P-frame MB at (X,Y)=(%d,%d)", x, y);
 
+				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();
 
-				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);
-				}
-				
+				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);
 				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;
@@ -1052,49 +844,35 @@
 		uv_dx = pMB->mvs[0].x;
 		uv_dy = pMB->mvs[0].y;
 
-		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];
+		uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2;
+		uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2;
 	} else {
 		int sum;
 
-		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];
+		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));
 	}
 
 	start_timer();
-	if(dec->quarterpel) {
-		interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
- 								    dec->qtmp.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.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,
@@ -1102,14 +880,12 @@
 	stop_comp_timer();
 
 	for (i = 0; i < 6; i++) {
-		int direction = dec->alternate_vertical_scan ? 2 : 0;
-
 		if (cbp & (1 << (5 - i)))	// coded
 		{
 			memset(&block[i * 64], 0, 64 * sizeof(int16_t));	// clear
 
 			start_timer();
-			get_inter_block(bs, &block[i * 64], direction);
+			get_inter_block(bs, &block[i * 64]);
 			stop_coding_timer();
 
 			start_timer();
@@ -1147,6 +923,7 @@
 	stop_transfer_timer();
 }
 
+
 // add by MinChen <chenm001@163.com>
 // decode an B-frame direct &  inter macroblock
 void
@@ -1181,176 +958,100 @@
 		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;
 
-		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];
+		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;
 	} else {
 		int sum;
 
-		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];
+		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));
 	}
 
 
 	start_timer();
-	if(dec->quarterpel) {
-		if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q))
-			interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.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->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.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->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.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->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.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->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.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.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);
 
 
-	if(dec->quarterpel) {
-		if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q))
-			interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
-									    pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
-		else {
-			interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
-									    pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
-			interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
-									    pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
-			interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
-									    pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
-			interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
- 									    dec->qtmp.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->tmp.y, backward.y, 16 * x_pos, 16 * y_pos,
-							  pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
-		interpolate8x8_switch(dec->tmp.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->tmp.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->tmp.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->tmp.u, backward.u, 8 * x_pos, 8 * y_pos,
+	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->tmp.v, backward.v, 8 * x_pos, 8 * y_pos,
+	interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos,
 						  b_uv_dx, b_uv_dy, stride2, 0);
 
-	interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
-						dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
-						dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos,
-						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->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8,
-						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->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
-						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->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
-						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->tmp.u + (8 * y_pos * stride2) + 8 * x_pos,
-						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->tmp.v + (8 * y_pos * stride2) + 8 * x_pos,
-						stride2, 1, 8);
-
+	interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos,
+					 stride);
+	interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos,
+					 stride);
+	interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8,
+					 stride);
+	interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8,
+					 16 * y_pos + 8, stride);
+	interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos,
+					 stride2);
+	interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos,
+					 stride2);
 	stop_comp_timer();
 
 	for (i = 0; i < 6; i++) {
-		int direction = dec->alternate_vertical_scan ? 2 : 0;
-
 		if (cbp & (1 << (5 - i)))	// coded
 		{
 			memset(&block[i * 64], 0, 64 * sizeof(int16_t));	// clear
 
 			start_timer();
-			get_inter_block(bs, &block[i * 64], direction);
+			get_inter_block(bs, &block[i * 64]);
 			stop_coding_timer();
 
 			start_timer();
@@ -1581,7 +1282,7 @@
 				break;
 
 			default:
-				DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type);
+				DPRINTF(DPRINTF_ERROR, "Not support B-frame mb_type = %d", mb->mb_type);
 			}
 
 		}						// end of FOR
@@ -1606,258 +1307,121 @@
 	*mb2 = temp;
 }
 
-
-/* perform post processing if necessary, and output the image */
-void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, 
-					const XVID_DEC_FRAME * frame, int pp_disable)
-{
-
-	if ((frame->general & (XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV)) && !pp_disable)	/* post process */
-	{
-		/* note: image is stored to tmp */
-		image_copy(&dec->tmp, img, dec->edged_width, dec->height);
-		image_deblock_rrv(&dec->tmp, dec->edged_width, 
-						mbs, dec->mb_width, dec->mb_height, dec->mb_width,
-						8, frame->general);
-		img = &dec->tmp;
-	}
-
-	image_output(img, dec->width, dec->height,
-				 dec->edged_width, frame->image, frame->stride,
-				 frame->colorspace, dec->interlacing);
-}
-
-
 int
 decoder_decode(DECODER * dec,
-			   XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats)
+			   XVID_DEC_FRAME * frame)
 {
 
 	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;
-	int output = 0;
-	int seen_something = 0;
 
 	start_global_timer();
 
-	dec->low_delay_default = (frame->general & XVID_DEC_LOWDELAY);
 	dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL;
 
-	if (frame->length < 0)	/* decoder flush */
-	{
-		/* if  not decoding "low_delay/packed", and this isn't low_delay and
-		    we have a reference frame, then outout the reference frame */
-		if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0)
-		{
-			decoder_output(dec, &dec->refn[0], dec->mbs, frame, dec->last_reduced_resolution);
-			output = 1;
-		}
-
-		frame->length = 0;
-		if (stats)
-		{
-			stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING;
-			stats->data.vop.time_base = (int)dec->time_base;
-			stats->data.vop.time_increment = 0;	//XXX: todo
-		}
-
-		emms();
-
-		stop_global_timer();
-		return XVID_ERR_OK;
-	}
-
 	BitstreamInit(&bs, frame->bitstream, frame->length);
 
-	// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's
-	if(dec->low_delay_default && 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;
-	}
-
-repeat:
-
-	vop_type =	BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, 
-			&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, gmc_mv);
-
-	DPRINTF(DPRINTF_HEADER, "vop_type=%i,  packed=%i,  time=%i,  time_pp=%i,  time_bp=%i", 
-							vop_type,	dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
-
-	if (vop_type == - 1)
-	{
-		if (success) goto done;
-		return XVID_ERR_FAIL;
-	}
-
-	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 repeat;
-	} 
+	// add by chenm001 <chenm001@163.com>
+	// for support B-frame to reference last 2 frame
+	dec->frames++;
+	vop_type =
+		BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward,
+							 &fcode_backward, &intra_dc_threshold);
 
 	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);
+#ifdef BFRAMES_DEC
+		DPRINTF(DPRINTF_DEBUG, "P_VOP  Time=%d", dec->time);
+#endif
+		break;
 
-	/* packed_mode: special-N_VOP treament */
-	if (dec->packed_mode && vop_type == N_VOP)
-	{
-		if (dec->low_delay_default && dec->frames > 0)
-		{
-			decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution);
-			output = 1;
-		}
-		/* ignore otherwise */
-	}
-	else if (vop_type != B_VOP)
-	{
-		switch(vop_type)
-		{
-		case I_VOP :
-			decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold);
-			break;
-		case P_VOP :
-			decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, 
-						fcode_forward, intra_dc_threshold, NULL);
-			break;
-		case S_VOP :
-			decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, 
-						fcode_forward, intra_dc_threshold, gmc_mv);
-			break;
-		case N_VOP :
-			image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
-			break;
-		}
-
-		if (reduced_resolution)
-		{
-			image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs,
-				(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width,
-				16, XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV);
-		}
-
-		/* note: for packed_mode, output is performed when the special-N_VOP is decoded */
-		if (!(dec->low_delay_default && dec->packed_mode))
-		{
-			if (dec->low_delay)
-			{
-				decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution);
-				output = 1;
-			}
-			else if (dec->frames > 0)	/* is the reference frame valid? */
-			{
-				/* output the reference frame */
-				decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution);
-				output = 1;
-			}
-		}
-
-		image_swap(&dec->refn[0], &dec->refn[1]);
-		image_swap(&dec->cur, &dec->refn[0]);
-		mb_swap(&dec->mbs, &dec->last_mbs);
-		dec->last_reduced_resolution = reduced_resolution;
-
-		dec->frames++;
-		seen_something = 1;
-
-	}else{	/* B_VOP */
-
-		if (dec->low_delay)
-		{
-			DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream");
-			dec->low_delay = 1;
-		}
+	case I_VOP:
+		decoder_iframe(dec, &bs, quant, intra_dc_threshold);
+#ifdef BFRAMES_DEC
+		DPRINTF(DPRINTF_DEBUG, "I_VOP  Time=%d", dec->time);
+#endif
+		break;
 
-		if (dec->frames < 2)
-		{
-			/* attemping to decode a bvop without atleast 2 reference frames */
-			image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
-						"broken b-frame, mising ref frames");
-		}else if (dec->time_pp <= dec->time_bp) {
-			/* this occurs when dx50_bvop_compatibility==0 sequences are 
-			decoded in vfw. */
-			image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
-						"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
-		}else{
+	case B_VOP:
+#ifdef BFRAMES_DEC
+		if (dec->time_pp > dec->time_bp) {
+			DPRINTF(DPRINTF_DEBUG, "B_VOP  Time=%d", dec->time);
 			decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
+		} else {
+			DPRINTF(DPRINTF_DEBUG, "Broken B_VOP");
 		}
+#else
+		image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
+#endif
+		break;
 
-		decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution);
-		output = 1;
-		dec->frames++;
-	}
+	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);
+		break;
 
-	BitstreamByteAlign(&bs);
+	default:
+		return XVID_ERR_FAIL;
+	}
 
-	/* low_delay_default mode: repeat in packed_mode */
-	if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0)
-	{
-		success = 1;
-		goto repeat;
+#ifdef BFRAMES_DEC_DEBUG
+	if (frame->length != BitstreamPos(&bs) / 8){
+		DPRINTF(DPRINTF_DEBUG, "InLen: %d / UseLen: %d", frame->length, BitstreamPos(&bs) / 8);
 	}
+#endif
+	frame->length = BitstreamPos(&bs) / 8;
 
-done :
 
-	/* low_delay_default mode: if we've gotten here without outputting anything,
-	   then output the recently decoded frame, or print an error message  */
-	if (dec->low_delay_default && output == 0)
-	{
-		if (dec->packed_mode && seen_something)
-		{
-			/* output the recently decoded frame */
-			decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution);
-		}
-		else
-		{
-			image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
-			image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
-				"warning: nothing to output");
-			image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64,
-				"bframe decoder lag");
+#ifdef BFRAMES_DEC
+	// test if no B_VOP
+        if (dec->low_delay || dec->frames == 0) {
+#endif
+	image_output(&dec->cur, dec->width, dec->height, dec->edged_width,
+					 frame->image, frame->stride, frame->colorspace);
 
-			decoder_output(dec, &dec->cur, NULL, frame, 1 /*disable pp*/);	
+#ifdef BFRAMES_DEC
+	} else {
+                if (dec->frames >= 1) {
+			start_timer();
+			if ((vop_type == I_VOP || vop_type == P_VOP)) {
+				image_output(&dec->refn[0], dec->width, dec->height,
+							 dec->edged_width, frame->image, frame->stride,
+							 frame->colorspace);
+			} else if (vop_type == B_VOP) {
+				image_output(&dec->cur, dec->width, dec->height,
+							 dec->edged_width, frame->image, frame->stride,
+							 frame->colorspace);
+			}
+			stop_conv_timer();
 		}
-		output = 1;
 	}
+#endif
 
-	frame->length = BitstreamPos(&bs) / 8;
+	if (vop_type == I_VOP || vop_type == P_VOP) {
+		image_swap(&dec->refn[0], &dec->refn[1]);
+		image_swap(&dec->cur, &dec->refn[0]);
 
-	if (stats)
-	{
-		stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING;
-		stats->data.vop.time_base = (int)dec->time_base;
-		stats->data.vop.time_increment = 0;	//XXX: todo
+		// swap MACROBLOCK
+                // the Divx will not set the low_delay flage some times
+                // so follow code will wrong to not swap at that time
+                // this will broken bitstream! so I'm change it,
+                // But that is not the best way! can anyone tell me how
+                // to do another way?
+                // 18-07-2002   MinChen<chenm001@163.com>
+                //if (!dec->low_delay && vop_type == P_VOP)
+                if (vop_type == P_VOP)
+			mb_swap(&dec->mbs, &dec->last_mbs);
 	}
-	
+
 	emms();
 
 	stop_global_timer();