Oh, I idiotically didn't close SIM Coupe before throwing the DSK file
at my FTP program. Sorry!
Re: code, I really don't think that cycle pinching is the main problem
— I think it's algorithmic. I'm going to need to think about that a
bit more. At the minute I'm a bit indoctrinated as to how I think a 3d
pipeline should go, and I may need to rethink that. Also to be clear
though, I'm using 8.8 fixed point, so divides and multiplies need to
work in that space. Which, at a most naive level, means doing a 16x16 -
> 24 multiply and a 24x16 -> 16 divide. Most of the time I'm only
doing 2.8 x 2.8 multiplies, i.e. 10x10 -> 16, and those go through my
table multiplier which I really, really, don't think could be
improved. It does a multiply in between 268 and 368 cycles (depending
on raster position), including a call/ret pair. So I can probably
shave 10% off just by inlining, but then I'm probably pretty much done
unless there's some sort of special tricks for 16bit table lookups
that I haven't thought of yet.
On 17 May 2008, at 20:39, Dan Dooré wrote:
Thomas Harte wrote:
It's originally on Fred 50, but not really in the sense that you
have to expand it to another floppy disk and then boot it off that.
But, here, I'll save you the effort of all that:
http://members.allegro.cc/ThomasHarte/temp/DWC.DSK
This appears to be a blank disc image with SamDOS on it :-(
You can download a copy from WoSam: http://www.worldofsam.org/node/617
Dan.
