home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.sgi
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!mips!odin!deadhead.asd.sgi.com!mogal
- From: mogal@deadhead.asd.sgi.com (Joshua Mogal)
- Subject: Re: Marketing Triangle performance (how do I acheive it)
- Message-ID: <1992Aug19.181506.26531@odin.corp.sgi.com>
- Sender: news@odin.corp.sgi.com (Net News)
- Nntp-Posting-Host: deadhead.asd.sgi.com
- Organization: Silicon Graphics, Inc.
- References: <14580@borg.cs.unc.edu>
- Distribution: usa
- Date: Wed, 19 Aug 1992 18:15:06 GMT
- Lines: 35
-
- In article <14580@borg.cs.unc.edu>, cullip@pooh.cs.unc.edu (Timothy Cullip) writes:
- |> I'm sure this has been covered in the past (probably several times), but
- |> I am looking for example code that gives the 1 Mega tris/sec for a VGX
- |> that the marketing pamplets all talk about.
- |>
- |> We have a PowerSeries 4D/340 VGX with 128 Megs of memory.
- |>
- |> I've written a program that opens a 500x500 pixel window in RGBmode,
- |> double buffered with zbuffer off.
- |> It uses triangle meshes (200 meshes of 400 triangles each).
- |> Each triangle typically covers about a 3x3 pixel area (very small).
- |> It makes no calls to n3f (just v3f for the vertices) and uses shademodel(FLAT).
- |> It does no lmdef or lmbinds to define material or ligh properties.
- |>
- |> It only gets 450 K tris/sec.
- |>
- |>
- |> Interestingly, if I use an lmdef and lmbind to define a material and light,
- |> (use phong lighting with gouraud shading) and use n3f() and v3f() at each
- |> vertex and turn the zbuffer on I still get about 400 K tris/sec.
- |>
- |> I'm not surprised that with all the bells and whistles turned on I get
- |> 400 K tris/sec, but I am surprised that with them turned off I only
- |> get 450 K tris/sec.
- |>
- |> Have I missed something?
-
-
- Have you made sure that you set SUBPIXEL(TRUE) ? If this is not set you will get approx 50% of the peak machine performance...
-
-
-
- Joshua Mogal
- Product Manager
- Advanced Graphics Division
-