Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!caen!uflorida!mlb.semi.harris.com!sloth.mlb.semi.harris.com!jdr
From: jdr@sloth.mlb.semi.harris.com (Jim Ray)
Newsgroups: comp.lang.fortran
Subject: validating fortran compiler's numerical results
Keywords: fortran, validating
Message-ID: <1991Apr9.023737.18140@mlb.semi.harris.com>
Date: 9 Apr 91 02:37:37 GMT
Sender: news@mlb.semi.harris.com
Organization: Harris Semiconductor, Melbourne FL
Lines: 22
Nntp-Posting-Host: sloth.mlb.semi.harris.com

As the topic states, I am looking for some fortran program that one
could use to compare the numeric (floating point) results with some
"known" values?  Specifically, I am looking for such a program that
would exercise (extensively and repeditively ) most of the
transcendental functions plus the normal run of the mill ones.  This
"program" would checking for roundoff and complience with the IEEE
standard floating point specs.

The reason I am looking for such a program, is the wide divergence of
numerical results we have seen between several releases of fortran
compilers on several workstations including SUN. Problem is that I
can't figgure which one(s) are "correct".

Any comments are welcome....

Thanks,


-- 
Jim Ray                                Harris Semiconductor
Internet:  jdr@semi.harris.com         PO Box 883   MS 62B-022
Phone:     (407) 729-5059              Melbourne, FL  32901