--- decoder.c	2002/11/07 10:28:15	1.37.2.8
+++ decoder.c	2002/11/20 22:28:25	1.37.2.14
@@ -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 1.37.2.8 2002/11/07 10:28:15 suxen_drol Exp $
+ *  $Id: decoder.c,v 1.37.2.14 2002/11/20 22:28:25 Isibaar Exp $
  *
  *************************************************************************/
 
@@ -965,35 +965,49 @@
 		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,
@@ -1080,74 +1094,130 @@
 		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,
@@ -1156,32 +1226,32 @@
 	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();
 
@@ -1463,6 +1533,7 @@
 	uint32_t intra_dc_threshold;
 	VECTOR gmc_mv[5];
 	uint32_t vop_type;
+	int success = 0;
 
 	start_global_timer();
 
@@ -1470,19 +1541,19 @@
 
 	BitstreamInit(&bs, frame->bitstream, frame->length);
 
-#ifdef BFRAMES_DEC
 	// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's
-	if(BitstreamShowBits(&bs, 8) == 0x7f)
+	if(frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
 	{
 		if (stats)
 			stats->notify = XVID_DEC_VOP;
 		frame->length = 1;
-		image_output(&dec->cur, dec->width, dec->height, dec->edged_width,
+		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;
 	}
-#endif
 
+start:
 	// add by chenm001 <chenm001@163.com>
 	// for support B-frame to reference last 2 frame
 	dec->frames++;
@@ -1494,6 +1565,9 @@
 
 	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)
@@ -1563,27 +1637,23 @@
 	default:
 		if (stats)
 			stats->notify = 0;
-		return XVID_ERR_FAIL;
-	}
 
-#ifdef BFRAMES_DEC_DEBUG
-	if (frame->length != BitstreamPos(&bs) / 8){
-		DEBUG2("InLen/UseLen",frame->length, BitstreamPos(&bs) / 8);
+		emms();
+		return XVID_ERR_FAIL;
 	}
-#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, 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 || vop_type == S_VOP)) {
 				image_output(&dec->refn[0], dec->width, dec->height,
@@ -1615,6 +1685,15 @@
 			mb_swap(&dec->mbs, &dec->last_mbs);
 	}
 
+	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;