Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!wuarchive!uunet!orca!undies!pmartz
From: pmartz@undies.dsd.es.com (Paul Martz)
Newsgroups: comp.graphics
Subject: Re: Rendering performance
Message-ID: <1991May8.175257.11847@dsd.es.com>
Date: 8 May 91 17:52:57 GMT
References: <1991Apr29.162712.1905@canon.co.uk> <1991May3.190648.13574@ucunix.san.uc.edu> <1991May3.224900.12807@dsd.es.com> <1991May6.114747.12681@canon.co.uk> <1991May7.190509.23004@fido.wpd.sgi.com> <1991May8.125546.1722@canon.co.uk>
Sender: usenet@dsd.es.com
Reply-To: pmartz@undies.dsd.es.com (Paul Martz)
Organization: Evans & Sutherland Computer Corp., Salt Lake City, UT
Lines: 41
Nntp-Posting-Host: 130.187.85.56

I think everyone has the basic idea now that peak performance numbers
quoted by marketing departments are pretty mich useless in calculatin
ghow fast applications X Y and Z will run on a workstation. An
alternate benchmarking method is needed.

Robert Skinner suggested in a previous article, "...use *your
application* as the benchmark on the the machines you consider
purchasing." This is probably the best way to judge your performance
for that applcation, but is often impractical for large applications
that would require substantial porting to some vendor's proprietary
interface. (Now if you've coded to PHIGS or some other standard, this
problem is reduced. But anyway...) So something better is needed.

I think Rich Thomson mentioned GPC but I don't remember him explaining
it very thoroughly. GPC is a benchmarking standard developed by the
Graphics Performance Characterization Committee. Basically it allows
anyone to code up a simple test file (in a format called BIF --
Benchmark Interchange Format). Some sample BIF files I've seen
typically put up several frames, pan and zoom around the object/image,
contain lines, local transform matrices, etc. You can make it as
complicated as you want. Or if you don't have the resources to write
up your own BIF file, there are already several available, and chances
are one of them might resemble sample output from the application
you'd like to run.

The point is, GPC benchmarks can take into account whatever factors
you're likely to experience in your application, such as swapping
buffers or clearing areas between frames. Many vendors are members of
GPC, and are ready to run whatever BIF file a potential customer would
like to run.

Further information concerning GPC is available through NCGA:

	NCGA
	PO Box 3412
	McLean, VA 22103-9832

-- 

   -paul	pmartz@dsd.es.com
		Evans & Sutherland