Re: OpenGVS Benchmark Appli

New Message Reply Date view Thread view Subject view Author view

Steve Baker (steve++at++mred.bgm.link.com)
Tue, 30 Apr 96 21:49:14 -0500

> First, this is not a contest between OpenGVS and Performer;
> Second, Gemini has *not* changed to a new position as you have suggested.

What exactly *are* you proposing? It's beginning to sound a lot like you are only
really producing an OpenGL performance suite. If you are really aiming at the latter,
then may I suggest that you pool your efforts with those already attacking this problem:

Check out the OpenGL Performance Characterization (OPC) Project Proposal at

             http://sunsite.unc.edu/gpc/opc.html

...and

the Viewperf OpenGL benchmark code at

             http://www.sgi.com/Technology/openGL/OPC.html

I'm sure that most simulation people will tell you that raw OpenGL thoughput
doesn't tell you much about the overall performance of a system in normal use.

                                    ---oooOOOooo---

This discussion is getting nowhere fast. I'm getting the impression that no two
people agree on what they think they will get from this so-called benchmark suite.

Would John Archdeacon please answer the following questions as simply as possible
so that we can all understand what he is trying to achieve:-

                                    ---oooOOOooo---

1) What are you attempting to compare using your benchmark suite?

    a) All IG's?

    b) Only IG's that currently run OpenGL?

    c) Only IG's that currently run OpenGVS?

    d) Only IG's that currently run OpenGVS on top of OpenGL?

2) For a benchmark to be meaningful, all the systems under test must
   presumably produce essentially the same end result. What end results are
   required in order to be considered as having sucessfully run the
   benchmark? There must be some kind of standard or else I can get
   really good times just by halving all the LOD transition ranges,
   or ignoring the textures or not bothering to Gouraud shade anything.

3) Am I allowed to restructure the database in order to get it to run faster?
   If so, to what degree? I have often managed to get a 4:1 performance
   improvement by restructuring the database to better suite the IG's inner
   workings.

4) Is the only test software that we are allowed to use OpenGVS? If so, what
   benchmark figure does a single-pipe, two-CPU RE2 and a single-pipe, four-CPU
   RE2 achieve? If the answer is anything close to 1.0 then I think we can all
   end this discussion right now since the ability to make use of multiple CPUs
   in the image generation process is a really critical issue in IG performance.

Thanks John.

                                    ---oooOOOooo---

BTW: I know that Gemini are in favour of standards and portability - it's a shame
     that their web page: http://www.gemtech.com/rwbench/rwbv1.pdf is formatted in
     a format that I can't read on anything except a PC or a Mac.
     Ghostscript choked and died when I fed it the postscript version.

     How about an HTML version guys!

  Steve Baker 817-323-1361 (Vox-Lab)
  Hughes Training Inc. 817-695-8776 (Vox-Office/vMail)
  2200 Arlington Downs Road 817-695-4028 (Fax)
  Arlington, Texas. TX 76005-6171 steve++at++mred.bgm.link.com (eMail)

New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:52:49 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.