Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!lll-lcc!ames!ucbcad!ucbvax!jade!eris!mwm
From: mwm@eris.BERKELEY.EDU (Mike (My watch has windows) Meyer)
Newsgroups: comp.graphics,comp.sys.amiga
Subject: Re: DBW-Tracer on UNIX
Message-ID: <3422@jade.BERKELEY.EDU>
Date: Sat, 2-May-87 21:22:20 EDT
Article-I.D.: jade.3422
Posted: Sat May  2 21:22:20 1987
Date-Received: Sun, 3-May-87 19:36:33 EDT
References: <213@ndmath.UUCP> <214@brandx.klinzhai.RUTGERS.EDU>
Sender: usenet@jade.BERKELEY.EDU
Reply-To: mwm@eris.BERKELEY.EDU (Mike (My watch has windows) Meyer)
Organization: Missionaria Phonibalonica
Lines: 55
Xref: mnetor comp.graphics:597 comp.sys.amiga:4426

In article <214@brandx.klinzhai.RUTGERS.EDU> webber@klinzhai.RUTGERS.EDU (Webber) writes:
<In article <213@ndmath.UUCP>, milo@ndmath.UUCP (Greg Corson) writes:
<> Recently I was messing around with the DBW (AMIGA) ray tracing program on a
<> SUN workstation here at Notre Dame.  The main program (RAY) compiles and runs
<> just fine here, however there are some things that seem puzzling to me.  The
<> documents to DBW tracer state the demo program "glass" runs for about 11 hours
<> on an Amiga.  On the Sun with SKY floating point board it ran 14 hours...the
<> only reason I can figure is that the Sun version might have automatically 
<> calculated the picture to greater precision.  If anyone out there has actually
<
<Well, I think you are close to the problem, but got it backwards.  On
<most unix systems (including sun) it is faster to work in double
<precision than single precision.  This is because all the libraries
<are double precision, so anytime you pass a float variable or capture
<a result in a float variable, a conversion has to be done.  You can
<probably speed things up by changing all your `float' declarations
<to be `double' declarations.  (Even when you pass a varible of type
<float to a procedure whose formal parameter is declared to be of type
<float, it still usually gets converted to double.)


To quote: No, NO, NO!!!

Anybody who's serious about supporting C fixed their compiler long ago
to have a flag that makes things happen in single precision instead of
double. For the Sun, under SunOS 3.0, that flags is -fsingle. I'd try
that instead. But I'm working on the assumption that DBW will have
declared things as double if he felt they needed double precision.
It's not clear what to check on to verify it, but I suspect that
things will work right with -fsingle. In any case, this is easier than
converting all your floats to doubles, and will make things go even
faster.

Also, for some ungodly reason, the default in that same environment is
to do "software floating-point calls." Not sure what this means, but I
suspect that its software emulation. It looks like you want -fsky if
you can use it, otherwise -fswitch with the environment set right. All
this is for SunOS 3.0 (3.2? not sure where the man pages are coming
from with rman), so check your man pages.

HH: How about working for Apple?
Me: Will they pay me to write Amiga applications?
HH: No.
Me: No.
HH: Well, then, how about Sun?
Me: Will they expect me to support their software?
HH: Yes.
Me: No.

	<mike
--
Take a magic carpet to the olden days			Mike Meyer
To a mythical land where everybody lays			ucbvax!mwm
Around in the clouds in a happy daze			mwm@berkeley.edu
In Kizmiaz ... Kizmiaz					mwm@ucbjade.BITNET