Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!spool.mu.edu!olivea!tardis!tymix!uunet!seas.gwu.edu!gritz
From: gritz@seas.gwu.edu (Larry Gritz)
Newsgroups: comp.sys.hp
Subject: HP9000/425 floating point: fixed
Message-ID: <3305@sparko.gwu.edu>
Date: 14 Jun 91 15:06:50 GMT
Article-I.D.: sparko.3305
Reply-To: lg@galileo.usno.navy.mil (Larry Gritz)
Organization: The U.S. Naval Observatory, Washington D.C.
Lines: 25

Last week I posted an example program showing how certain floating point
operations work more slowly on a 425 (68040) than on the 400t (68030),
in particular the floating point --> ascii conversion used by the *printf
family of routines.  I also related our initial difficulty in finding HP
representatives who were aware of the problem and/or knew of a solution.

I thank those in netland for their help in narrowing down the problem.
I am still answering inquiries and requests for the demo programs.

We got a patch for the problem HAND DELIVERED by HP yesterday.  It replaces
three small object files in the /lib/libc.a library.  This fixes the problem
entirely.  HP representatives were most helpful, once we got hold of some
who were actually aware of the problem.  I have been informed that (1) the
fix will be a part of HP-UX 8.0, which is scheduled for release soon, and
(2) the patch to HP-UX 7.05 will be generally released in about two weeks
through the usual customer support channels.

Thanks to all,
     -- Larry Gritz, US Naval Observatory


-- 
Larry Gritz                                   lg@galileo.usno.navy.mil
US Naval Observatory                          phone: 202-653-1034
Washington, DC 20392-5100                     also: gritz@seas.gwu.edu