Re: Non-MultiSample Anti-aliasing

New Message Reply Date view Thread view Subject view Author view

Orad Hi-Tec Systems (orad_u++at++netvision.net.il)
Mon, 03 Feb 1997 09:14:13 +0200

Timothy Moore wrote:
>
> From: "Angus Dorbie" <dorbie++at++bitch.reading.sgi.com>
> Date: Fri, 31 Jan 1997 19:27:21 +0000
>
> In general anti-aliasing isn't that big a problem for IMPACT for
> several reasons:
>
> 1) trilinear MIP mapping and point sampled polygon edges mean you
> only need it where texture ID changes or you get a shilouette.
> 2) You can use a wireline pass very effectively to get high quality
> blended edges in a scene and can control the load. The fill performance
> hit is manageable because it's wireline.
>
> Cool idea.
>
> 3) Soft edge blending alpha textures like tree lines mean you you never
> need it for blended alpha textures.
>
> Don't you then need to depth sort these blended edge polygons? Also,
> I thought that blending modes are pretty expensive.
>
> Tim

If we are willing to go to depth-sorted rendering, why is it necessary
to do a wireline pass? The old "framebuffer alpha saturation blending"
can be used. This is assuming the hw supports GL_POLYGON_SMOOTH mode,
and alpha bitplanes in the framebuffer. I think both are true for the
impact?

>From man glBlendFunc:

     Polygon antialiasing is optimized using blend function
     (GL_SRC_ALPHA_SATURATE, GL_ONE) with polygons sorted from nearest
to
     farthest. (See the glEnable, glDisable reference page and the
     GL_POLYGON_SMOOTH argument for information on polygon
antialiasing.)
     Destination alpha bitplanes, which must be present for this blend
     function to operate correctly, store the accumulated coverage.

Moshe Nissim
Orad Hi-Tec Systems
=======================================================================
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 - 17:54:33 PDT

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