Path: utzoo!attcan!uunet!lll-winken!ames!mailrus!cwjcc!gatech!hubcap!Eugene
From: eugene@eos.arc.nasa.gov (Eugene Miya)
Newsgroups: comp.parallel
Subject: Another NBSLIB problem/solution
Message-ID: <4177@hubcap.UUCP>
Date: 24 Jan 89 14:42:37 GMT
Sender: fpst@hubcap.UUCP
Lines: 10
Approved: parallel@hubcap.clemson.edu

It appears we are getting different results for some of the NBSLIB
tests (e.g., Mendez/Euler1) when we compile and run the same code
on the same compiler with different compiler options/switches.
It would be handy to have a "correct" run on the machine which the
code was written for (in this case, and IBM or clone).  Then a
program like spiff(1) could be used for at least an approximate comparison,
because we can't completely evaluate the accuracy of our results
here on the ported code.

--eugene