commit 02632b8

Unrolled the sprite routine a bit more

 1@@ -522,51 +522,63 @@ screen_deo_sprite(void)
 2 	const int ctrl = dev[0x2f];
 3 	const int flipx = ctrl & 0x10, flipy = ctrl & 0x20;
 4 	const int dx = flipx ? -rDY : rDY, dy = flipy ? -rDX : rDX;
 5-	const int row_start = flipx ? 0 : 7, row_delta = flipx ? 1 : -1;
 6-	const int col_start = flipy ? 7 : 0, col_delta = flipy ? -1 : 1;
 7 	const int layer = ctrl & 0x40, layer_mask = layer ? 0x3 : 0xc;
 8-	const int stride = screen_wmar2;
 9 	const int blend = ctrl & 0xf;
10-	const Uint8 opaque_mask = blend % 5;
11+	const int opaque_mask = blend % 5;
12+	const int shift_right = flipx ? 0 : 1;
13+	const int row_step = flipy ? -1 : 1;
14+	const int row_start = flipy ? 7 : 0;
15 	const Uint8 *table = blend_lut[blend][layer >> 6];
16 	if(ctrl & 0x80) {
17 		const int addr_incr = rMA << 2;
18 		for(i = 0; i <= rML; i++, x += dx, y += dy, rA += addr_incr) {
19 			const Uint16 x0 = x + 8, y0 = y + 8;
20-			if(x0 + 8 >= stride || y0 + 8 >= screen_hmar2) continue;
21-			Uint8 *dst = screen_layers + y0 * stride + x0;
22-			const Uint8 *col = &ram[rA + col_start];
23-			for(j = 0; j < 8; j++, dst += stride, col += col_delta) {
24-				Uint8 *d = dst;
25-				const int ch1 = *col, ch2 = col[8];
26-				for(int k = 0, row = row_start; k < 8; k++, d++, row += row_delta) {
27-					const int color = ((ch1 >> row) & 1) | (((ch2 >> row) & 1) << 1);
28-					if(opaque_mask || color)
29-						*d = (*d & layer_mask) | table[color];
30-				}
31+			if(x0 + 8 >= screen_wmar2 || y0 + 8 >= screen_hmar2) continue;
32+			Uint8 *dst = screen_layers + y0 * screen_wmar2 + x0;
33+			const Uint8 *rowptr = &ram[rA + row_start];
34+			for(j = 0; j < 8; j++, dst += screen_wmar2, rowptr += row_step) {
35+				Uint8 *d = dst, bits1 = rowptr[0], bits2 = rowptr[8];
36+				if(shift_right)
37+					for(int k = 0; k < 8; k++, d++, bits1 <<= 1, bits2 <<= 1) {
38+						int color = ((bits1 >> 7) & 1) | ((bits2 >> 6) & 2);
39+						if(opaque_mask || color)
40+							*d = (*d & layer_mask) | table[color];
41+					}
42+				else
43+					for(int k = 0; k < 8; k++, d++, bits1 >>= 1, bits2 >>= 1) {
44+						int color = (bits1 & 1) | ((bits2 & 1) << 1);
45+						if(opaque_mask || color)
46+							*d = (*d & layer_mask) | table[color];
47+					}
48 			}
49 		}
50 	} else {
51 		const int addr_incr = rMA << 1;
52 		for(i = 0; i <= rML; i++, x += dx, y += dy, rA += addr_incr) {
53 			const Uint16 x0 = x + 8, y0 = y + 8;
54-			if(x0 + 8 >= stride || y0 + 8 >= screen_hmar2) continue;
55-			Uint8 *dst = screen_layers + y0 * stride + x0;
56-			const Uint8 *col = &ram[rA + col_start];
57-			for(j = 0; j < 8; j++, dst += stride, col += col_delta) {
58-				Uint8 *d = dst;
59-				const int ch1 = *col;
60-				for(int k = 0, row = row_start; k < 8; k++, d++, row += row_delta) {
61-					const int color = (ch1 >> row) & 1;
62-					if(opaque_mask || color)
63-						*d = (*d & layer_mask) | table[color];
64-				}
65+			if(x0 + 8 >= screen_wmar2 || y0 + 8 >= screen_hmar2) continue;
66+			Uint8 *dst = screen_layers + y0 * screen_wmar2 + x0;
67+			const Uint8 *rowptr = &ram[rA + row_start];
68+			for(j = 0; j < 8; j++, dst += screen_wmar2, rowptr += row_step) {
69+				Uint8 *d = dst, bits = rowptr[0];
70+				if(shift_right)
71+					for(int k = 0; k < 8; k++, d++, bits <<= 1) {
72+						int color = (bits >> 7) & 1;
73+						if(opaque_mask || color)
74+							*d = (*d & layer_mask) | table[color];
75+					}
76+				else
77+					for(int k = 0; k < 8; k++, d++, bits >>= 1) {
78+						int color = bits & 1;
79+						if(opaque_mask || color)
80+							*d = (*d & layer_mask) | table[color];
81+					}
82 			}
83 		}
84 	}
85 	if(!screen_reqdraw) {
86-		int x1 = flipx ? x : rX, x2 = flipx ? rX : x;
87-		int y1 = flipy ? y : rY, y2 = flipy ? rY : y;
88+		int x1 = flipx ? x : rX, y1 = flipy ? y : rY;
89+		int x2 = flipx ? rX : x, y2 = flipy ? rY : y;
90 		screen_change(x1 - 8, y1 - 8, x2 + 8, y2 + 8);
91 	}
92 	if(rMX) rX += flipx ? -rDX : rDX;