home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!haven.umd.edu!darwin.sura.net!jvnc.net!yale.edu!qt.cs.utexas.edu!cs.utexas.edu!sun-barr!olivea!sgigate!odin!fido!babar.asd.sgi.com!mtj
- From: mtj@babar.asd.sgi.com (Michael Jones)
- Newsgroups: comp.graphics
- Subject: Re: NCGA Picture Level Benchmark (was: Time to put up or shut up)
- Keywords: benchmarks NCGA GPC
- Message-ID: <pak809s@fido.asd.sgi.com>
- Date: 1 Sep 92 21:07:11 GMT
- References: <1992Aug20.172200.17418@dsd.es.com> <ortnla8@fido.asd.sgi.com> <1992Sep1.194132.2601@dsd.es.com>
- Sender: news@fido.asd.sgi.com (Usenet News Admin)
- Organization: Silicon Graphics, Inc.
- Lines: 47
-
- In article <...>, rthomson@mesa.dsd.es.com (Rich Thomson) writes:
-
- |> SGI is very eager to publish its SPECmarks for the processors on its
- |> machines. The SPEC benchmark suite is similar to the PLB benchmark
- |> suite in conception. The only conclusion I can draw from this is that
- |> SGI must not perform very well on the benchmark and doesn't want to
- |> publish "bad" numbers.
-
- The PLB issue seems the perfect example of the Akin/Thomson dialogues
- on graphics programming interfaces. My understanding is that the PLB
- suite spends quite a bit of effort (thus time) "editing" the databases
- it is drawing. It does this *precisely because applications layered on
- API's like PHIGS and PEXlib* tend to spend quite a bit of their time
- convincing the graphics layer of their desires. This structural, non-
- graphic burden is a major factor in many applications, such as visual
- editors, CAD systems, and such.
-
- The direct interface provided by GL and OpenGL avoids this, leaving it
- to the application to decide the structural issues. Often, the cost of
- editing the structure is zero -- once the desired changes are "known"
- to the application, it is then *immediately* able to issue the new
- graphics commands.
-
- I feel that an SGI PLB suite benchmark would be hard to interpret. It
- would show, in many cases, what the cost (time) would be to do the
- very things that SGI has helped its customers avoid needing to do. On
- the other hand, it might make a good story ;-)
-
- (customer) "But why is this PLB application so much slower than
- all the similar applications I see on SGI machines?"
-
- (salesman) "Because we're using a PHIGS/PEXlib style API."
-
- (customer) "Oh. I guess it must be the extra data movement and
- redundant PHIGS structure editing."
-
- (salesman) "Exactly. Now let me show you the OpenGL version."
-
- (customer) "Wow!. No wonder they stopped making workstations."
-
- Just my personal opinion. I have no doubt that the PLB effort is a
- valid one, and I freely admit that meaningful graphics benchmarks are
- very hard to develop.
-
- -- Michael Jones mtj@sgi.com 415.390.1455 M/S 7U-550
- Silicon Graphics, Advanced Systems Division
- 2011 N. Shoreline Blvd., Mtn. View, CA 94039-7311
-