*BSD News Article 11043


Return to BSD News archive

Received: by minnie.vk1xwt.ampr.org with NNTP
	id AA1191 ; Tue, 23 Feb 93 14:29:33 EST
Path: sserve!manuel.anu.edu.au!munnari.oz.au!spool.mu.edu!howland.reston.ans.net!usc!cs.utexas.edu!not-for-mail
From: vazquez@iqm.unicamp.br (Pedro A.M. Vazquez)
Newsgroups: comp.unix.bsd
Subject: [386bsd] npx-0.1 causes lockup in npxprobe() - now what? (
Date: 12 Feb 1993 08:47:00 -0600
Organization: UTexas Mail-to-News Gateway
Lines: 52
Sender: daemon@cs.utexas.edu
Message-ID: <9302121437.AA09426@styx.iqm.unicamp.br>
NNTP-Posting-Host: cs.utexas.edu

> >Newsgroups: comp.unix.bsd
> >Path: ccsun.unicamp.br!dcc.unicamp.br!agate!spool.mu.edu!decwrl!pacbell.com!UB.com!quack!dfox
> >From: dfox@quack.sac.ca.us (David Fox)
> >Subject: [386bsd] npx-0.1 causes lockup in npxprobe() - now what?
> >Message-ID: <fZaiZXG@quack.sac.ca.us>
> >Keywords: npx cyrix npxprobe ndp
> >Organization: The Duck Pond public unix: +1 408 249 9630, log in as 'guest'.
> >Date: 11 Feb 1993 08:38:16 UTC
> >Lines: 60
> 
> 
> 
> Subsequent postings (which I replied to by email) suggested I could
> comment out the fp_divideby0() and/or the npxprobe() routine.  I tried
> the first, didn't want to do the second, but did comment out a small
> section where it was looking for irq16 (this applies only for 486s which
> I don't have, it seems).
> 
I have a similar problem: I'm unable to run two (or more) fpu intensive
applications because a panic:npxintr crashes the system.

This ocurred before applying patch-kit0.2 and less frequently after 
applying it. But a new problem came with patch-kit0.2: I can't display
images with xv, it allways occur a fp overflow and the program aborts.
The same kernel runs xv without this problem in another machine without
a fpu unit.

I applied npx-0.3-diff with the same results of David Fox, the system
hangs on boot. Then I comented out the fp_divideby0() as sugested and 
the new kernel booted ok but issued the follow message:

npx0<Error reporting broken (no IRQ13)> at 0xf0 on isa

With this kernel I started my torture test: I put an molecular eletronic
structure abinitio program to run in background, then started a rayshade
and a raypaint of the same scene description file and checked if  the 
clinpack and  flops20 benchmarks would give diferent results from the
previous kernel.

No problems! Wonderfull! :-)

Then I called xv to display a gif image and the system hang :-(

After rebooting I tried to run paranoia on the console (no X) and the
system hanged again when the program was looking for the overflow limit.


Many resets after I confirmed the problem has to do with fp overflows,
I hope this does not implies an incompatibility of IIT fpus with 386bsd
as occurred with the ULSI chips (I was one of those ULSI unhappy owners ). 

Pedro