hplus at mindcontrol.org
Wed Dec 4 17:30:00 EST 2002
Denormals and other special case values may slow you down a factor
10 - 1000 though, depending on architecture. It's still a neat
observation, though :-)
> -----Original Message-----
> From: music-dsp-admin at aulos.calarts.edu
> [mailto:music-dsp-admin at aulos.calarts.edu]On Behalf Of Jay Stelly
> Sent: Wednesday, December 04, 2002 4:36 PM
> To: 'music-dsp at shoko.calarts.edu'
> Subject: RE: [music-dsp] 0!=0?
> One cool thing to remember is that computers are actually fast
> enough now to
> iterate the entire 32-bit space in a short period of time. So if
> I wanted a
> definitive answer to this question I would simply write a loop
> that iterated
> all 32 bit patterns, stuffed them into a float and then multiplied them by
> zero - outputting the ones that were not zero. My guess is that
> you've got
> some NaN or INF value in there, but it's nice to be able to know the
> behavior definitively with a simple bit of test code. You can actually
> write unit tests for simple 32-bit float or int functions that take the
> _entire_ domain of the function into account now. That's pretty cool.
> > -----Original Message-----
> > From: Russell Borogove [mailto:kaleja at estarcion.com]
> > Sent: Wednesday, December 04, 2002 4:01 PM
> > To: music-dsp at shoko.calarts.edu
> > Subject: Re: [music-dsp] 0!=0?
> > > should it be that x*0.0f (where x is some float, maybe very
> > large) in
> > > c++ does not evaluate to zero?
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book
> reviews, dsp links http://shoko.calarts.edu/musicdsp/
More information about the music-dsp