Espozo wrote:
Oops. I really don't know why, but I thought it was 6. If anything, It's closer to 8.
Probably because of early Macs (if I recall correctly), those also had a 68000 running at 7.67MHz, but because of the video hardware hogging the memory bus in practice it was more like 6MHz.
Espozo wrote:
(About the 68000 vs. the 65816, is the Genesis's CPU truly about twice as fast then, or is it less efficient? I have a bit of a hard time believing that it's twice as fast...)
The 68000 is running at double the speed but also takes up double the amount of cycles for equivalent instructions, which evens it out. Still the 68000 instruction set allows it to do more stuff in less instructions (less need to move things around that you'll use only once, and usually you won't need to touch RAM for variables in subroutines), as well as having 32-bit instructions (which are slightly slower, but still much faster than doing it with two 16-bit operations), so it still has an edge there. How much it helps depends on how complex your calculations need to be.
This is comparable to Z80 vs 6502, really.
Espozo wrote:
Were you the one that made the demo? I'm just curious.
lolno (I know how gasega68k's demo works, though)
Espozo wrote:
Another thing I've always wondered is what's so bad about the SNES's graphics format?
Each pixel is spread over four bytes, and each byte has data for eight pixels. When just storing graphics (as nearly every game does) it's a non-problem, the problem is when software rendering kicks in.
Note, it still may be helpful. I know that one trick on the Amiga is to reserve one or two bitplanes for lighting and use the rest for coloring the polygons =P Then just use an appropriate palette. (it does reduce how many colors you can use, but still...)
psycopathicteen wrote:
Does anybody know how much faster the 68000 is than the 65816 at running C with the most efficient compilers around? Shiru's DK Classic game runs at a solid 60fps.
No idea, but I would guess 68000 may be faster because it has more registers (no idea how much of an issue is using zero page on the 65816 though). Also I imagine that 65816 would be using 16-bit ints, while 68000 uses 32-bit ints.
That said, could be anything really, since the C ABI on the 68000 is pretty wasteful... (only two integers and two pointers can be passed through registers, and everything is stored as 32-bit unless it's an array, even chars *shudders*)