Re: Fast texturemapping on Indys?

New Message Reply Date view Thread view Subject view Author view

Rob Jenkins (robj++at++quid)
Wed, 2 Oct 1996 08:24:10 -0700

On Oct 2, 2:29pm, Morten Eriksen wrote:
> Subject: Fast texturemapping on Indys?
> A few months ago there was some talk (from someone in the Performer
> team, I believe) about the possibility of making faster software routines
> for texturemapping on the SGI machines which lack hardware texturemapping
> support. Any news on this?
>
> Morten
>
> =======================================================================
> List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
> Submissions: info-performer++at++sgi.com
> Admin. requests: info-performer-request++at++sgi.com
>-- End of excerpt from Morten Eriksen

This is what was posted ( from George Kyriazis ), I believe it was in Irix 6.2
( note at the bottom it mentions platforms, R5K being best ), gl_dev will have
more details on new features for OpenGL in Irix 6.2:

"
Ok, I got asked to be a bit more specific about the upcoming Indy
fast texturing.

It actually has two aspects, which, when combined provide the desired
result. Performance figures are included below.

1. Certain "cheap" texturing paths are optimized. This is basically
   the following combinations:

        2-d textures
        no texture borders
        min/mag filters: GL_NEAREST / GL_NEAREST (no mip-mapping)
        repeat/clamp modes: GL_REPEAT (both s&t directions)
        glTexEnv modes: GL_REPLACE_EXT is fastest (no lighting).
                                GL_MODULATE is fast, but not top speed.
        GL_PERSPECTIVE_CORRECTION_HINT should NOT be set to GL_NICEST
                (This way a divide is avoided per pixel).
        GL_RGB and GL_RGBA textures are accelerated
                                (blending supported for RGBA textures)
        depth buffering is supported.
        no stencil.
        no scissoring when depth buffer is enabled.

Just this optimization gives you upto 2Mpxls/sec depth-buffered throughput,
which is about 5-10x performance increase, but it's not enough..

2. An OpenGL SGIX (machine dependent) extension is added that allows
   rasterization to be performed at a level courser than 1-pixel. This
   way, the (software) rasterizer performs less work. This "zoom"
   factor can be altered on a frame-by-frame basis.

   Traditional (non-textured) rendering suffers a hit using this mode
   (because of the hardware support existing for flat/gouraud polygons).
   Only when the zoom factor is >= 3 you get performance higher than
   traditional rasterization.

   For textured polygons, using the above settings, you can get pixel
   throughputs which can reach about 20Mpxls/sec, with a decreased
   visual quality, of course.

For simple scenes, you CAN get fully-textured real-time 30Hz.

The above also applies for the just-announced R5000 Indy. The R5Ks
provides the highest texture performance, followed by R4400s, and then
by R4600s.

  --george
"

Cheers
Rob

-- 
________________________________________________________________
Rob Jenkins
Silicon Graphics
robj++at++csd.sgi.com
=======================================================================
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 - 17:53:42 PDT

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