Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!mips!pacbell.com!att!ucbvax!PAN.SSEC.HONEYWELL.COM!erstad
From: erstad@PAN.SSEC.HONEYWELL.COM (Dave Erstad)
Newsgroups: comp.sys.apollo
Subject: re: Problems with FTN 10.[78] / 425t
Message-ID: <9106232226.AA04236@pan.ssec.honeywell.com>
Date: 23 Jun 91 22:26:58 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 31


      
>We are having some strange results from some simulations running on a 400t and 
>a 425t.  I was wondering if there are known differences with the math done on 
>a 68882 -vs- 68040.  The results that follow are using FTN 10.7 / SR10.2.4 
>(10.8 on 10.3.4 provided similar results)...
      
Yes, we've seen numerous changes in arithmetic results.  Nothing as large
as the example cited, but that might be application dependent, depending
on the stability of the calculations.  This is with C and Pascal 
applications.

We did receive a patch OS from HP/Apollo which appears to address many 
of the inconsistencies:

        **** Node 29D5E.2E747 ****   "//isildur"
   Domain/OS kernel(11), revision 10.3.4.3, test  May 16, 1991  8:41:55 am
                                            ^ Never had an OS say that before!

       
>Any ideas? (hmmm.... nick here says that JT from SSEC.honeywell.com will be
>the first to reply.   The stop watch is running....)

Well, I haven't seen him today so I might have him beat.  He is engaged,
you know   :-)

But I suspect he'll have some more concrete info to add.

Dave Erstad
Honeywell SSEC
erstad@pan.ssec.honeywell.com