Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!samsung!dali.cs.montana.edu!uakari.primate.wisc.edu!zaphod.mps.ohio-state.edu!think.com!linus!linus!mwunix.mitre.org!twegner From: twegner@mwunix.mitre.org (Timothy Wegner) Newsgroups: comp.graphics Subject: Re: int math vs. float math Message-ID: Date: 7 Jan 91 20:58:23 GMT References: <26189@uflorida.cis.ufl.EDU> Sender: usenet@linus.mitre.org Distribution: usa Lines: 17 Nntp-Posting-Host: mwunix.mitre.org jdb@reef.cis.ufl.edu (Brian K. W. Hook) writes: >I have heard a lot about using integer math instead of floating point >since it is much faster (e.g. Fractint supposedly uses 32-bit int math >on a 386 instead of floats and Autodesk theoretically only uses 10% float >math)....however, I have not been able to find a good reference on this >subject? Any help? I don't know about a reference, but I would like to point out that Fractint does either integer or floating point math, so that you can easily benchmark the benefits. The downside of integer math is that fixed point has a limited dynamic range. But if the nature of the computation is that the numbers stay within known limits (as in a Mandelbrot calculation), fixed point math using 32 bit registers is much faster. Incredibly faster if your machine does not have a math coprocessor. For the technically minded, the Fractint source code is available so that you can see how it is done.