Return to BSD News archive
Newsgroups: comp.os.386bsd.bugs Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!elroy.jpl.nasa.gov!swrinde!cs.utexas.edu!uunet!quack!dfox From: dfox@quack.kfu.com (David Fox) Subject: Re: Floating exceptions? Message-ID: <f0366W2@quack.kfu.com> Organization: The Duck Pond public unix: +1 408 249 9630, log in as 'guest'. References: <f0XUP76@quack.kfu.com> <cproto.733032990@marsh> Date: 25 Mar 1993 22:47:44 UTC Lines: 43 In article <cproto.733032990@marsh> cproto@cs.curtin.edu.au (Computer Protocol) writes: >mrapple@quack.kfu.com (Nick Sayer) writes: > >>0.2.2 running on a 486-50 with 16M RAM. I compiled xv_calctool >>(patchlevel 12). If I try and find ln(10000)/ln(10), it crashes with >>a floating point exception. The same code on a sun does not. > >>Has anyone seen this behavior before? Might there be some delay between >>the occurance of the problem and the exception or something? > >>-- >>Nick Sayer <mrapple@quack.kfu.com> | "Do you know how lucky you are we >I'm having the same problem with Ghostscript 2.5.2. See my article in >comp.os.386bsd.apps. > >Regards - Tibor Sashegyi (cproto@cs.curtin.edu.au) I responded to Tibor by email to his article in comp.os.386bsd.apps with the belief that much of this is caused by inadequate numerics handling in the kernel without Bruce Evans' npx patches. I had somewhat similar problems when doing heavy FP in one vterm (rayshade) and intermittent heavy FP in another vterm (a few gs runs). The kernel panicked, but it panicked long after, when I wasn't doing any FP. After I applied Bruce Evans' npx patches, this problem hasn't resurfaced. Charles Hannum said that the only way to guarantee that a SIGFPE is reported when the FP instruction executes is to insert FWAITs before every FP instruction. *Please* don't. It would severely cripple numerics performance on 386BSD. -- ------------------------------------------------------------------------ in real life: david fox email: dfox@quack.kfu.com 5479 Castle Manor Drive, San Jose, CA 95129 "Thanks for letting me change some USA magnetic patterns on your HD" -me 408-253-7992 "You've just won an absolutely FREE trip to the Isles of Langerhans" -me ------------------------------------------------------------------------