[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: SANE on GS

khorster@attila.stevens-tech.edu (Karl Horster) wrote:

>Soenke Behrens (sbehrens@contech.demon.co.uk) wrote:
>
>: SANE's kind of cool ... it's slow, but very accurate. If you can give
>: up a few points of precision on trig routines, a NC card (68881/2) is
>: a definite plus --- much faster. Heavy floating point processing
>: suddenly actually gets doable :)
>
>	Couldn't the SANE be patched to give up some precision for 
>speed?  Like in some applications where you might not need such high 
>precision?

Yes, one could do that. You would need to essentially rewrite the
parts you patch, of course. So, you would need to get an IEEE-754
implementation (or part thereof), port the bits you want to the GS
(maybe initially just mult/div/add/sub), make sure you got things
right (which is the hard part ... losing a few bits of precision is
OK, getting an outright wrong result is definitely not), and then
patch SANE to substitute your new code. Possible. Maybe even feasible.
And still slower than a NumberCruncher :).

Soenke
--
I don't know what it is he's doing, it must be art