Re: A86: a pretty fast multiply method...
[Prev][Next][Index][Thread]
Re: A86: a pretty fast multiply method...
Well, our routine is 115 cycles. The one on that page is between 101 and 124
cycles. It doesn't load the address of the table inside the routine (that's
11 cycles), and it outputs in AC (a pretty useless way to have a 16-bit
result). So I'm gonna have to prefer ours because it's faster and also fixed
t-state =)
But we might find some good tips on this page... Thanks for pointing it out.
Also, on the pages before, there were some good things about line clipping
and polygons, looks like I'll have fun. Thanks =)
> Well, I only got a TI calculator and started programming in Z80 a week
> and a half ago - I didn't know you guys were looking for information like
> this :)
>
> But yeah, it's great to try yourself, then look at other people's ideas
> after - you may come up with something better that way.
>
> > Anyway, between Jimmy's optimizations and a few tweaks, looks like it's
> > gonna come in at about 115 cycles (including ret) with a 1280 byte
table.
>
> Looks like there's similarly efficient Z80 version at:
> http://www.ffd2.com/fridge/speccy/Spec3D a bit more than half-way down.
>
> Steve Judd managed to do it in 43 cycles on the 6502, and only 25 cycles
> if only y changes for further iterations - which can be very common in
> 3d gfx. Of course, all this wonderful reference material was generated
> by a flamewar - and who says nothing good comes from (flame)war?
> --
> Robin Harbron macbeth@psw.nu
> http://www.tbaytel.net/macbeth
> http://www.psw.nu/
References: