--- decoder.c	2004/07/10 17:49:31	1.61
+++ decoder.c	2005/03/27 03:59:41	1.69
@@ -20,7 +20,7 @@
  *  along with this program ; if not, write to the Free Software
  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  *
- * $Id: decoder.c,v 1.61 2004/07/10 17:49:31 edgomez Exp $
+ * $Id: decoder.c,v 1.69 2005/03/27 03:59:41 suxen_drol Exp $
  *
  ****************************************************************************/
 
@@ -46,8 +46,8 @@
 #include "dct/fdct.h"
 #include "utils/mem_transfer.h"
 #include "image/interpolate8x8.h"
-#include "image/reduced.h"
 #include "image/font.h"
+#include "image/qpel.h"
 
 #include "bitstream/mbcoding.h"
 #include "prediction/mbprediction.h"
@@ -270,8 +270,7 @@
 				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);
@@ -284,15 +283,9 @@
 	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 */
 
@@ -362,24 +355,12 @@
 	}
 
 	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();
 }
 
@@ -390,14 +371,12 @@
 				uint8_t * pY_Cur,
 				uint8_t * pU_Cur,
 				uint8_t * pV_Cur,
-				const int reduced_resolution,
 				const MACROBLOCK * pMB)
 {
-	DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
+	DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE);
 
 	int stride = dec->edged_width;
-	int next_block = stride * (reduced_resolution ? 16 : 8);
-	const int stride2 = stride/2;
+	int next_block = stride * 8;
 	int i;
 	const uint32_t iQuant = pMB->quant;
 	const int direction = dec->alternate_vertical_scan ? 2 : 0;
@@ -407,63 +386,93 @@
 			int direction,
 			const int quant,
 			const uint16_t *matrix);
+	typedef void (*add_residual_function_t)(
+			uint8_t *predicted_block,
+			const int16_t *residual,
+			int stride);
 
 	const get_inter_block_function_t get_inter_block = (dec->quant_type == 0)
-		? get_inter_block_h263
-		: get_inter_block_mpeg;
+		? (get_inter_block_function_t)get_inter_block_h263
+		: (get_inter_block_function_t)get_inter_block_mpeg;
+
+	uint8_t *dst[6];
+	int strides[6];
 	
-	memset(&data[0], 0, 6*64*sizeof(int16_t));	/* clear */
 
-	for (i = 0; i < 6; i++) {
+	if (dec->interlacing && pMB->field_dct) {
+		next_block = stride;
+		stride *= 2;
+	}
 
-		if (cbp & (1 << (5 - i))) {	/* coded */
+	dst[0] = pY_Cur;
+	dst[2] = pY_Cur + next_block;
+	dst[1] = dst[0] + 8;
+	dst[3] = dst[2] + 8;
+	dst[4] = pU_Cur;
+	dst[5] = pV_Cur;
+	strides[0] = strides[1] = strides[2] = strides[3] = stride;
+	strides[4] = stride/2;
+	strides[5] = stride/2;
+	
+	for (i = 0; i < 6; i++) {
+		/* Process only coded blocks */
+		if (cbp & (1 << (5 - i))) {
 
+			/* Clear the block */
+			memset(&data[0], 0, 64*sizeof(int16_t));
 
 			/* Decode coeffs and dequantize on the fly */
 			start_timer();
-			get_inter_block(bs, &data[i*64], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices));
+			get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices));
 			stop_coding_timer();
 
+			/* iDCT */
 			start_timer();
-			idct(&data[i * 64]);
+			idct(&data[0]);
 			stop_idct_timer();
-		}
-	}
 
-	if (dec->interlacing && pMB->field_dct) {
-		next_block = stride;
-		stride *= 2;
+			/* Add this residual to the predicted block */
+			start_timer();
+			transfer_16to8add(dst[i], &data[0], strides[i]);
+			stop_transfer_timer();
+		}
 	}
+}
 
-	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);
-	}
-	stop_transfer_timer();
+static void __inline
+validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec)
+{
+	/* clip a vector to valid range 
+	   prevents crashes if bitstream is broken 
+	*/
+	int shift = 5 + dec->quarterpel;
+	int xborder_high = (int)(dec->mb_width - x_pos) << shift;
+	int xborder_low = (-(int)x_pos-1) << shift;
+	int yborder_high = (int)(dec->mb_height - y_pos) << shift;
+	int yborder_low = (-(int)y_pos-1) << shift;
+
+#define CHECK_MV(mv) \
+	do { \
+	if ((mv).x > xborder_high) { \
+		DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \
+		(mv).x = xborder_high; \
+	} else if ((mv).x < xborder_low) { \
+		DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \
+		(mv).x = xborder_low; \
+	} \
+	if ((mv).y > yborder_high) { \
+		DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \
+		(mv).y = yborder_high; \
+	} else if ((mv).y < yborder_low) { \
+		DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \
+		(mv).y = yborder_low; \
+	} \
+	} while (0)
+
+	CHECK_MV(mv[0]);
+	CHECK_MV(mv[1]);
+	CHECK_MV(mv[2]);
+	CHECK_MV(mv[3]);
 }
 
 /* decode an inter macroblock */
@@ -475,7 +484,6 @@
 				const uint32_t cbp,
 				Bitstream * bs,
 				const uint32_t rounding,
-				const int reduced_resolution,
 				const int ref)
 {
 	uint32_t stride = dec->edged_width;
@@ -487,48 +495,13 @@
 	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);
+	for (i = 0; i < 4; i++)
+		mv[i] = pMB->mvs[i];
 
-	for (i = 0; i < 4; i++) {
-		/* clip to valid range */
-		int border = (int)(dec->mb_width - x_pos) << (5 + dec->quarterpel);
-		if (mv[i].x > border) {
-			DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", mv[i].x, border, x_pos, y_pos);
-			mv[i].x = border;
-		} else {
-			border = (-(int)x_pos-1) << (5 + dec->quarterpel);
-			if (mv[i].x < border) {
-				DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", mv[i].x, border, x_pos, y_pos);
-				mv[i].x = border;
-			}
-		}
-
-		border = (int)(dec->mb_height - y_pos) << (5 + dec->quarterpel);
-		if (mv[i].y >  border) {
-			DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", mv[i].y, border, x_pos, y_pos);
-			mv[i].y = border;
-		} else {
-			border = (-(int)y_pos-1) << (5 + dec->quarterpel);
-			if (mv[i].y < border) {
-				DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", mv[i].y, border, x_pos, y_pos);
-				mv[i].y = border;
-			}
-		}
-	}
+	validate_vector(mv, x_pos, y_pos, dec);
 
 	start_timer();
 
@@ -543,10 +516,7 @@
 		uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
 		uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
 
-		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);
-		else if (dec->quarterpel)
+		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,
 											mv[0].x, mv[0].y, stride, rounding);
@@ -567,21 +537,7 @@
 		uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
 		uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
 
-		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);
-
-		} 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);
@@ -607,23 +563,15 @@
 	}
 
 	/* chroma */
-	if (reduced_resolution) {
-		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 {
-		interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
-								uv_dx, uv_dy, stride2, rounding);
-		interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
-								uv_dx, uv_dy, stride2, rounding);
-	}
+	interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
+							uv_dx, uv_dy, stride2, rounding);
+	interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
+							uv_dx, uv_dy, stride2, rounding);
 
 	stop_comp_timer();
 
 	if (cbp)
-		decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur,
-							reduced_resolution, pMB);
+		decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
 }
 
 static void
@@ -670,7 +618,7 @@
 	stop_transfer_timer();
 
 	if (cbp)
-		decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
+		decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
 
 }
 
@@ -678,19 +626,13 @@
 static 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;
-	}
+	const uint32_t mb_width = dec->mb_width;
+	const uint32_t mb_height = dec->mb_height;
 
 	bound = 0;
 
@@ -746,7 +688,7 @@
 			}
 
 			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)
@@ -804,7 +746,6 @@
 decoder_pframe(DECODER * dec,
 				Bitstream * bs,
 				int rounding,
-				int reduced_resolution,
 				int quant,
 				int fcode,
 				int intra_dc_threshold,
@@ -813,13 +754,8 @@
 	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;
-	}
+	const uint32_t mb_width = dec->mb_width;
+	const uint32_t mb_height = dec->mb_height;
 
 	if (!dec->is_edged[0]) {
 		start_timer();
@@ -938,12 +874,11 @@
 					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;
 					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, cbp, bs,
-								rounding, reduced_resolution, 0);
+				decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0);
 
 			} else if (gmc_warp) {	/* a not coded S(GMC)-VOP macroblock */
 				mb->mode = MODE_NOT_CODED_GMC;
@@ -963,7 +898,7 @@
 				mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
 
 				decoder_mbinter(dec, mb, x, y, 0, bs,
-								rounding, reduced_resolution, 0);
+								rounding, 0);
 
 				if(dec->out_frm && cp_mb > 0) {
 					output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
@@ -1018,7 +953,7 @@
 decoder_bf_interpolate_mbinter(DECODER * dec,
 								IMAGE forward,
 								IMAGE backward,
-								const MACROBLOCK * pMB,
+								MACROBLOCK * pMB,
 								const uint32_t x_pos,
 								const uint32_t y_pos,
 								Bitstream * bs,
@@ -1035,10 +970,12 @@
 	pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
 	pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
 
+	validate_vector(pMB->mvs, x_pos, y_pos, dec);
+	validate_vector(pMB->b_mvs, x_pos, y_pos, dec);
+
 	if (!direct) {
 		uv_dx = pMB->mvs[0].x;
 		uv_dy = pMB->mvs[0].y;
-
 		b_uv_dx = pMB->b_mvs[0].x;
 		b_uv_dy = pMB->b_mvs[0].y;
 
@@ -1051,21 +988,20 @@
 
 		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 {
-		if(dec->quarterpel) {
-			uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
-			uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
-			b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2);
-			b_uv_dy = (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 {
-			uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
-			uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
-			b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
-			b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
+		uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
+		uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
+		b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
+		b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
+
+		if (dec->quarterpel) {
+			uv_dx /= 2;
+			uv_dy /= 2;
+			b_uv_dx /= 2;
+			b_uv_dy /= 2;
 		}
 
 		uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
@@ -1113,73 +1049,43 @@
 
 	if(dec->quarterpel) {
 		if(!direct) {
-			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);
+			interpolate16x16_add_quarterpel(dec->cur.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);
+			interpolate8x8_add_quarterpel(dec->cur.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_add_quarterpel(dec->cur.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_add_quarterpel(dec->cur.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_add_quarterpel(dec->cur.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,
-						b_uv_dx, b_uv_dy, stride2, 0);
-	interpolate8x8_switch(dec->tmp.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_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos,
+				pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
+		interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
+				16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
+		interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos,
+				16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
+		interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8,
+				16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
+	}
+
+	interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos,
+			b_uv_dx, b_uv_dy, stride2, 0);
+	interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos,
+			b_uv_dx, b_uv_dy, stride2, 0);
 
 	stop_comp_timer();
 
 	if (cbp)
-		decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
+		decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB);
 }
 
 /* for decode B-frame dbquant */
@@ -1274,7 +1180,7 @@
 			if (last_mb->mode == MODE_NOT_CODED) {
 				mb->cbp = 0;
 				mb->mode = MODE_FORWARD;
-				decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
+				decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1);
 				continue;
 			}
 
@@ -1357,14 +1263,14 @@
 				get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y);
 				dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] =	mb->mvs[0];
 
-				decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0);
+				decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0);
 				break;
 
 			case MODE_FORWARD:
 				get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y);
 				dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] =	mb->mvs[0];
 
-				decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
+				decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1);
 				break;
 
 			default:
@@ -1414,7 +1320,6 @@
 	}
 }
 
-
 int
 decoder_decode(DECODER * dec,
 				xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
@@ -1422,7 +1327,6 @@
 
 	Bitstream bs;
 	uint32_t rounding;
-	uint32_t reduced_resolution;
 	uint32_t quant = 2;
 	uint32_t fcode_forward;
 	uint32_t fcode_backward;
@@ -1477,10 +1381,16 @@
 
 repeat:
 
-	coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution,
+	coding_type = BitstreamReadHeaders(&bs, dec, &rounding,
 			&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
 
-	DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i,  packed=%i,  time=%lli,  time_pp=%i,  time_bp=%i\n",
+	DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i,  packed=%i,  time=%"
+#if defined(_MSC_VER)
+    "I64"
+#else
+    "ll"
+#endif
+    "i,  time_pp=%i,  time_bp=%i\n",
 							coding_type,	dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
 
 	if (coding_type == -1) { /* nothing */
@@ -1528,14 +1438,14 @@
 	} else if (coding_type != B_VOP) {
 		switch(coding_type) {
 		case I_VOP :
-			decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold);
+			decoder_iframe(dec, &bs, quant, intra_dc_threshold);
 			break;
 		case P_VOP :
-			decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
+			decoder_pframe(dec, &bs, rounding, quant,
 						fcode_forward, intra_dc_threshold, NULL);
 			break;
 		case S_VOP :
-			decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
+			decoder_pframe(dec, &bs, rounding, quant,
 						fcode_forward, intra_dc_threshold, &gmc_warp);
 			break;
 		case N_VOP :
@@ -1546,12 +1456,6 @@
 			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, 0);
-		}
-
 		/* 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) {
@@ -1569,7 +1473,6 @@
 		image_swap(&dec->cur, &dec->refn[0]);
 		dec->is_edged[0] = 0;
 		SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
-		dec->last_reduced_resolution = reduced_resolution;
 		dec->last_coding_type = coding_type;
 
 		dec->frames++;
@@ -1614,19 +1517,22 @@
 
 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 */
+  /* if we reach here without outputing anything _and_
+     the calling application has specified low_delay_default,
+     we *must* output something.
+     this always occurs on the first call to decode() call
+     when bframes are present in the bitstream. it may also
+     occur if no vops  were seen in the bitstream
+
+     if packed_mode is enabled, then we output the recently
+     decoded frame (the very first ivop). otherwise we have
+     nothing to display, and therefore output a black screen.
+  */
+  if (dec->low_delay_default && output == 0) {
+    if (dec->packed_mode && seen_something) {
 			decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant);
-		} else {
+    } 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");
-
 			decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant);
 			if (stats) stats->type = XVID_TYPE_NOTHING;
 		}