Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!asuvax!ncar!elroy.jpl.nasa.gov!jato!vsnyder
From: vsnyder@jato.jpl.nasa.gov (Van Snyder)
Newsgroups: comp.lang.fortran
Subject: Re: Bug in Sun numerics library
Message-ID: <1991Apr17.234253.3444@jato.jpl.nasa.gov>
Date: 17 Apr 91 23:42:53 GMT
Reply-To: vsnyder@jato.Jpl.Nasa.Gov (Van Snyder)
Organization: Jet Propulsion Laboratory, Pasadena, CA
Lines: 14

One of our engineers discovered the Sun fortran compilers introduce lots
more error in computing 10.0d0**(-abs(j)), than do Univac 1100, Vax or
PC compilers.  The error goes up as the logarithm of j.  The folks I've
talked to at Sun say this is because the routine "pow_di" computes
(1/x)**abs(j) when asked to compute x**(-abs(j)).  It should compute
1/(x**abs(j)) instead.

Has anybody else noticed this problem, and if so, have you developed a
new pow_di?

-- 
vsnyder@jato.Jpl.Nasa.Gov
ames!elroy!jato!vsnyder
vsnyder@jato.uucp