--- trunk/xvidcore/src/motion/motion_comp.c	2002/07/24 23:17:19	340
+++ trunk/xvidcore/src/motion/motion_comp.c	2002/07/25 00:43:19	341
@@ -197,11 +197,11 @@
 	const int32_t edged_width = pParam->edged_width;
 	int32_t dx, dy;
 	int32_t b_dx, b_dy;
+	int k,sum;
 	int x = i;
 	int y = j;
 
 
-
 	switch (mb->mode) {
 	case MODE_FORWARD:
 		dx = mb->mvs[0].x;
@@ -287,61 +287,26 @@
 		break;
 
 
-	case MODE_INTERPOLATE:
+	case MODE_INTERPOLATE:		/* _could_ use DIRECT, but would be overkill (no 4MV there) */
+
 		dx = mb->mvs[0].x;
 		dy = mb->mvs[0].y;
+
 		b_dx = mb->b_mvs[0].x;
 		b_dy = mb->b_mvs[0].y;
 
-		transfer_8to16sub2_c(&dct_codes[0 * 64],
-							 cur->y + (i * 16) + (j * 16) * edged_width,
-							 get_ref(f_ref->y, f_refh->y, f_refv->y,
-									 f_refhv->y, 16 * i, 16 * j, 1, dx, dy,
-									 edged_width), get_ref(b_ref->y, b_refh->y,
-														   b_refv->y,
-														   b_refhv->y, 16 * i,
-														   16 * j, 1, b_dx,
-														   b_dy, edged_width),
-							 edged_width);
-
-		transfer_8to16sub2_c(&dct_codes[1 * 64],
-							 cur->y + (i * 16 + 8) + (j * 16) * edged_width,
-							 get_ref(f_ref->y, f_refh->y, f_refv->y,
-									 f_refhv->y, 16 * i + 8, 16 * j, 1, dx, dy,
-									 edged_width), get_ref(b_ref->y, b_refh->y,
-														   b_refv->y,
-														   b_refhv->y,
-														   16 * i + 8, 16 * j,
-														   1, b_dx, b_dy,
-														   edged_width),
-							 edged_width);
-
-		transfer_8to16sub2_c(&dct_codes[2 * 64],
-							 cur->y + (i * 16) + (j * 16 + 8) * edged_width,
+		for (k=0;k<4;k++)
+		{
+			transfer_8to16sub2_c(&dct_codes[k * 64],
+							 cur->y + (i * 16+(k&1)*8) + (j * 16+((k>>1)*8)) * edged_width,
 							 get_ref(f_ref->y, f_refh->y, f_refv->y,
-									 f_refhv->y, 16 * i, 16 * j + 8, 1, dx, dy,
-									 edged_width), get_ref(b_ref->y, b_refh->y,
-														   b_refv->y,
-														   b_refhv->y, 16 * i,
-														   16 * j + 8, 1, b_dx,
-														   b_dy, edged_width),
+									 f_refhv->y, 2*i + (k&1), 2*j + (k>>1), 8, dx, dy,
+									 edged_width), 
+							 get_ref(b_ref->y, b_refh->y, b_refv->y,
+									 b_refhv->y, 2*i + (k&1), 2 * j+(k>>1), 8, b_dx, b_dy, 
+									 edged_width),
 							 edged_width);
-
-		transfer_8to16sub2_c(&dct_codes[3 * 64],
-							 cur->y + (i * 16 + 8) + (j * 16 +
-													  8) * edged_width,
-							 get_ref(f_ref->y, f_refh->y, f_refv->y,
-									 f_refhv->y, 16 * i + 8, 16 * j + 8, 1, dx,
-									 dy, edged_width), get_ref(b_ref->y,
-															   b_refh->y,
-															   b_refv->y,
-															   b_refhv->y,
-															   16 * i + 8,
-															   16 * j + 8, 1,
-															   b_dx, b_dy,
-															   edged_width),
-							 edged_width);
-
+		}
 
 		dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2;
 		dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2;
@@ -352,36 +317,88 @@
 		transfer_8to16sub2_c(&dct_codes[4 * 64],
 							 cur->u + (y * 8) * edged_width / 2 + (x * 8),
 							 get_ref(f_ref->u, f_refh->u, f_refv->u,
-									 f_refhv->u, 8 * i, 8 * j, 1, dx, dy,
-									 edged_width / 2), get_ref(b_ref->u,
-															   b_refh->u,
-															   b_refv->u,
-															   b_refhv->u,
-															   8 * i, 8 * j, 1,
-															   b_dx, b_dy,
-															   edged_width /
-															   2),
+									 f_refhv->u, i, j, 8, dx, dy,
+									 edged_width / 2), 
+							 get_ref(b_ref->u, b_refh->u, b_refv->u,
+									 b_refhv->u, i, j, 8, b_dx, b_dy,
+									 edged_width / 2),
 							 edged_width / 2);
 
 		transfer_8to16sub2_c(&dct_codes[5 * 64],
 							 cur->v + (y * 8) * edged_width / 2 + (x * 8),
 							 get_ref(f_ref->v, f_refh->v, f_refv->v,
 									 f_refhv->v, 8 * i, 8 * j, 1, dx, dy,
-									 edged_width / 2), get_ref(b_ref->v,
-															   b_refh->v,
-															   b_refv->v,
-															   b_refhv->v,
-															   8 * i, 8 * j, 1,
-															   b_dx, b_dy,
-															   edged_width /
-															   2),
+									 edged_width / 2), 
+							 get_ref(b_ref->v, b_refh->v, b_refv->v, 
+							 		 b_refhv->v, 8 * i, 8 * j, 1, b_dx, b_dy,
+									 edged_width / 2),
 							 edged_width / 2);
 
 		break;
-
+	
 	case MODE_DIRECT:
-		// todo
+
+		for (k=0;k<4;k++)
+		{
+			dx = mb->mvs[k].x;
+			dy = mb->mvs[k].y;
+		
+			b_dx = mb->b_mvs[k].x;
+			b_dy = mb->b_mvs[k].y;
+
+//		fprintf(stderr,"Direct Vector %d -- %d:%d    %d:%d\n",k,dx,dy,b_dx,b_dy);
+
+			transfer_8to16sub2_c(&dct_codes[k * 64],
+							 cur->y + (i*16 + (k&1)*8) + (j*16 + (k>>1)*8 ) * edged_width,
+							 get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, 
+							 		 2*i + (k&1), 2*j + (k>>1), 8, dx, dy,
+									 edged_width), 
+							 get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, 
+							 		 2*i + (k&1), 2*j + (k>>1), 8, b_dx, b_dy, 
+									 edged_width),
+							 edged_width);
+		}
+
+		sum = mb->mvs[0].x + mb->mvs[1].x + mb->mvs[2].x + mb->mvs[3].x;
+		dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
+
+		sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y;
+		dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
+
+
+		sum = mb->b_mvs[0].x + mb->b_mvs[1].x + mb->b_mvs[2].x + mb->b_mvs[3].x;
+		b_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
+
+		sum = mb->b_mvs[0].y + mb->b_mvs[1].y + mb->b_mvs[2].y + mb->b_mvs[3].y;
+		b_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2));
+
+/*		// for QPel don't forget to always do
+
+		if (quarterpel)
+			sum /= 2;		
+*/
+
+		transfer_8to16sub2_c(&dct_codes[4 * 64],
+							 cur->u + (y * 8) * edged_width / 2 + (x * 8),
+							 get_ref(f_ref->u, f_refh->u, f_refv->u,
+									 f_refhv->u, i, j, 8, dx, dy,
+									 edged_width / 2), 
+							 get_ref(b_ref->u, b_refh->u, b_refv->u,
+									 b_refhv->u, i, j, 8, b_dx, b_dy,
+									 edged_width / 2),
+							 edged_width / 2);
+
+		transfer_8to16sub2_c(&dct_codes[5 * 64],
+							 cur->v + (y * 8) * edged_width / 2 + (x * 8),
+							 get_ref(f_ref->v, f_refh->v, f_refv->v,
+									 f_refhv->v, i, j, 8, dx, dy,
+									 edged_width / 2), 
+							 get_ref(b_ref->v, b_refh->v, b_refv->v, 
+							 		 b_refhv->v, i, j, 8, b_dx, b_dy,
+									 edged_width / 2),
+							 edged_width / 2);
+
+
 		break;
 	}
-
 }