--- decoder.c	2002/09/26 01:54:54	1.37.2.1
+++ decoder.c	2002/10/11 15:07:32	1.37.2.5
@@ -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 1.37.2.1 2002/09/26 01:54:54 h Exp $
+ *  $Id: decoder.c,v 1.37.2.5 2002/10/11 15:07:32 Isibaar Exp $
  *
  *************************************************************************/
 
@@ -136,6 +136,15 @@
 		return XVID_ERR_MEMORY;
 	}
 
+	if (image_create(&dec->refh, 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->refn[2], 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);
@@ -144,6 +153,7 @@
 		image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
 		image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
 		image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height);
+		image_destroy(&dec->refh, dec->edged_width, dec->edged_height);
 		xvid_free(dec);
 		return XVID_ERR_MEMORY;
 	}
@@ -161,6 +171,7 @@
 		image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
 		image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
 		image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height);
+		image_destroy(&dec->refh, dec->edged_width, dec->edged_height);
 		xvid_free(dec);
 		return XVID_ERR_MEMORY;
 	}
@@ -186,6 +197,7 @@
 	image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
 	image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
 	image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height);
+	image_destroy(&dec->refh, dec->edged_width, dec->edged_height);
 	image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
 	xvid_free(dec);
 
@@ -358,6 +370,30 @@
 
 		uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2;
 		uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2;
+
+		start_timer();
+		if(dec->quarterpel) {
+			interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+	 								    dec->refh.y + 128, 16*x_pos, 16*y_pos,
+									    pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
+		}
+		else {
+			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos,
+								  pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
+			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos,
+							      pMB->mvs[1].x, pMB->mvs[1].y, stride,  rounding);
+			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8,
+								  pMB->mvs[2].x, pMB->mvs[2].y, stride,  rounding);
+			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, 
+								  pMB->mvs[3].x, pMB->mvs[3].y, stride,  rounding);
+		}
+
+		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 {
 		int sum;
 		sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
@@ -377,36 +413,39 @@
 		}
 
 		uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
-	}
 
-	start_timer();
-	if(dec->quarterpel) {
-		DEBUG("QUARTERPEL");
-		interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos,
+		start_timer();
+		if(dec->quarterpel) {
+			interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+									  dec->refh.y + 128, 16*x_pos, 16*y_pos,
+									  pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
+			interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+									  dec->refh.y + 128, 16*x_pos + 8, 16*y_pos,
+									  pMB->mvs[1].x, pMB->mvs[1].y, stride,  rounding);
+			interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+									  dec->refh.y + 128, 16*x_pos, 16*y_pos + 8,
+									  pMB->mvs[2].x, pMB->mvs[2].y, stride,  rounding);
+			interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64,
+									  dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8,
+									  pMB->mvs[3].x, pMB->mvs[3].y, stride,  rounding);
+		}
+		else {
+			interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos,
 								  pMB->mvs[0].x, pMB->mvs[0].y, stride,  rounding);
-		interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos,
+			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_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8,
+			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_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8,
+			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);
-	}
-	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();
+		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++) {
 		int direction = dec->alternate_vertical_scan ? 2 : 0;
@@ -710,7 +749,6 @@
 								rounding);
 			} else				// not coded
 			{
-				DEBUG2("P-frame MB at (X,Y)=",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;
@@ -925,7 +963,6 @@
 	stop_transfer_timer();
 }
 
-
 // add by MinChen <chenm001@163.com>
 // decode an B-frame direct &  inter macroblock
 void
@@ -1033,18 +1070,36 @@
 	interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos,
 						  b_uv_dx, b_uv_dy, stride2, 0);
 
-	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);
+	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);
+
+	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);
+
+	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);
+
+	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);
+
+	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);
+
+	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);
+
 	stop_comp_timer();
 
 	for (i = 0; i < 6; i++) {
@@ -1330,6 +1385,9 @@
 
 	BitstreamInit(&bs, frame->bitstream, frame->length);
 
+	if(BitstreamShowBits(&bs, 8) == 0x7f)
+		return XVID_ERR_OK;
+
 	// add by chenm001 <chenm001@163.com>
 	// for support B-frame to reference last 2 frame
 	dec->frames++;
@@ -1371,6 +1429,9 @@
 	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);
+#ifdef BFRAMES_DEC
+		DEBUG1("N_VOP  Time=", dec->time);
+#endif
 		break;
 
 	default: