I discovered a -mieee gcc switch that solves this problem. Mea culpa. (it's described only in the info docs, not in man pages )
On Sun, 8 Apr 2001, Dmitrii Pasechnik wrote: >Date: Sun, 8 Apr 2001 14:34:46 +0200 (MET DST) >From: Dmitrii Pasechnik <[EMAIL PROTECTED]> >To: [email protected] >Subject: an FPU problem - nextafter too small? > >I'm running the latest stable Debian 2.2 on Alpha XL266. >(alphaev56 processor) > >There appears to be a problem with the values returned by >the function nextafter (see <math.h>). >Namely, this value (nextafter(0.,1.) returns 4.9407e-324) >is too small for doing arithmetic operations on it, >that is, for instance, 2.*nextafter(0.,1.) produces a floating point >exception. >This is with gcc (version 2.95.2 20000220 (Debian GNU/Linux)) >and no extra options (except -lm). > >The smallest possible value x that still allows for 2.*x with my setup >is about 1.e-307. > >Is this a feature or a bug (in gcc or in glibc ?) ? >Thanks in advance, >Dmitrii. > -- Dmitrii Pasechnik e-mail: [EMAIL PROTECTED] http://ssor.twi.tudelft.nl/~dima/
