*BSD News Article 2254


Return to BSD News archive

Path: sserve!manuel!munnari.oz.au!mips!mips!sdd.hp.com!usc!news
From: merlin@neuro.usc.edu (merlin)
Newsgroups: comp.unix.bsd
Subject: Jolitz 386BSD-0.1 -- floating point performance -- need gcc flags?
Date: 22 Jul 1992 02:55:45 -0700
Organization: University of Southern California, Los Angeles, CA
Lines: 39
Sender: merlin@neuro.usc.edu (merlin)
Message-ID: <l6qc51INN1gu@neuro.usc.edu>
References: <1992Jul21.215716.267@ugle.unit.no>
NNTP-Posting-Host: neuro.usc.edu

I have most of the US Army BRLCAD three dimensional CSG modeling and
distributed ray tracing system ported to the Jolitz 386BSD-0.1.  But,
I am getting only about one fifth of the floating point performance
previously measured using AT&T pcc and GNU gcc 1.4x on ATT UNIX SYSV.

Does the compiler default to '387 emulation?  Is there some flag which
needs to be set to actually use the coprocessor?  Or are there reasons
386BSD-0.1 would exhibit relatively poor floating point performance?

This is certainly not a complaint.  Porting BRLCAD has been extremely
straighforward.  However the performance gap between naive use of the
cc/gcc compilers on 386BSD-0.1 vs ATT SYSV 386 UNIX seems interesting.
I am hoping there is some simple flag or system configuration option
which I just happened to have missed in getting our installation up &
running with 386BSD-0.1, BRLCAD, and X11R5.

All in all I am very impressed with this system as a cheap development 
platform.  I have to move an occasional file to /BAD to get rid of bad
sectors.  But, apart from overloading the system with too many xterm's
or too many simultaneous memory hogging ray tracing tasks, the system
has neither crashed or exhibited any other pathological behaviors.  It
has also exhibited remarkably good citizenship as a USC ethernet node.  

Thanks, AJ

p.s.  Apart from it's intrinsic value as a three dimensional modeling
package, developers of 386BSD may find BRLCAD a very usefull tool for
stressing process scheduling, memory management, network links, and
filesystem performance of their various implementation alternatives.

------------------------------------------------------------------------------
Alexander-James Annala
Principal Investigator
Neuroscience Image Analysis Network
HEDCO Neuroscience Building, Fifth Floor
University of Southern California
University Park
Los Angeles, CA 90089-2520
------------------------------------------------------------------------------