Re: Occulting problems with polygon subfaces in Performer

New Message Reply Date view Thread view Subject view Author view

Michael T. Jones (mtj++at++babar.asd.sgi.com)
Mon, 21 Oct 1996 12:26:25 -0700


On Oct 21, 10:48am, Andrew Walker wrote:
> Subject: Re: Occulting problems with polygon subfaces in Performer

> Some things to think about when using subfaces.
>
> 1. They are very slow on the SGI. I would recommend acvoiding them.

I'm not defensive, but I should clarify this. Subfaces per se are
neither fast nor slow. Subface rendering methods (of which there
are several) *do* range from fast to slow as a function of style,
situation, and system.

Here goes:

cut-out-subfaces
   +less fill (less than *any* of the other methods)
   -more polys

don't-cut-out-subfaces
   a) use stencil
       1. on RE, RE2 - 50% fill penalty for base and layer geom
       2. on iR, only a 5% fill penalty
      either way, though, the draw order *must* be base, then each
      layer from bottom to top, which reduces our chances to sort
      the database by texture and lighting model (probably not a
      big deal for the typical decals)
   b) use displace ("glPolygonOffset")
       1. in IRIS GL, simplistic implementation
       2. in OpenGL, fancy/robust implementation
      either way, this should be fast.
   c) use reference plane
       -only on iR.
       +full speed
       +our idea, so we're proud if you use it. ;-)
   d) static manual offset ("stipes hover off runway")
       -no "proper" place if eye-to-layer distance can vary greatly.
   e) dynamic manual offset ("displace a DCS as a function of range")
       -never better than opengl-style polygon offset
       -DCS prevents geoset bbox culling :-(
       -just say no.

Quality:

   Perfect: cut-out-subfaces, and layer styles a and c
   Seemingly Perfect (can be made indetectable from Perfect): b, e
   Expensive and not-perfect: d

So, my vote is:
   1. cut-outs if you can afford the extra polygons (you can on iR)
   2. reference plane (only available on iR)
   3. displace (OpenGL method only)
   4. stencil (fast on iR only)

Michael
=======================================================================
List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
            Submissions: info-performer++at++sgi.com
        Admin. requests: info-performer-request++at++sgi.com


New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:53:47 PDT

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