home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pipex!bnr.co.uk!uknet!mcsun!news.funet.fi!ajk.tele.fi!funic!sauna.cs.hut.fi!news.cs.hut.fi!apm
- From: apm@kikka.hut.fi (Antti Miettinen)
- Newsgroups: comp.graphics.opengl
- Subject: Re: OpenGL and GL tmesh
- Message-ID: <APM.92Dec31005004@kikka.hut.fi>
- Date: 31 Dec 92 00:50:04 GMT
- References: <APM.92Dec19175627@kikka.hut.fi> <1h4vojINN9s5@fido.asd.sgi.com>
- Distribution: inet
- Organization: Helsinki University of Technology, Finland
- Lines: 40
- NNTP-Posting-Host: kikka.hut.fi
- In-reply-to: kurt@cashew.asd.sgi.com's message of 21 Dec 1992 17:43:47 GMT
-
- In article <1h4vojINN9s5@fido.asd.sgi.com> kurt@cashew.asd.sgi.com (Kurt Akeley) writes:
- >In article <APM.92Dec19175627@kikka.hut.fi>, apm@kikka.hut.fi (Antti Miettinen) writes:
- >|> Please say it's not true.
- >
- >Sorry, it's true.
-
- I guess I'll have to acceept that.
-
- >Not necessarily. Our experience with implementing swaptmesh() on our
- >own machines is that it is easy and straighforward on pipeline machines,
- >(such as the GT for which it was designed) but difficult and complex on
- >the more parallel machines that we have designed more recently (VGX,
- >Elan, RealityEngine).
-
- If you think this is of general interest, could you elaborate? Or is
- there some publicly available document that would explain things
- related to the issue?
-
- >We agree that the savings aren't very great. We decided, when we were
- >designing the OpenGL, that the savings in vertexes wouldn't be worth the
- >cost in both implementation complexity and the likely overall reduction
- >in implementation performance.
-
- Some elaboration on this also? I think the traffic in this group isn't
- excessive. I'm sure these things are evident to the persons who
- designed VGX, Elan and RE but to me the reduction in performance is a
- mystery.
-
- Do you think that some applications will take a performance hit when
- ported from GL to OpenGL (from the fact that there is no swaptmesh()
- equivalent)?
-
- I may seem insistent on swaptmesh() but basically I would just like to
- know about the issues involved. The applications I have in mind are
- two very different ones. One would draw a lot of triangles and try to
- draw them in as large trimeshes as possible. The other would callobj()
- a lot of objects which could be drawn as single trimeshes if
- swaptmesh() were available. Without swaptmesh() the objects would have
- to be broken into several trimeshes. Can you estimate whether the
- performance hit in these applications would be significant?
-