Re: Layer anomalies

New Message Reply Date view Thread view Subject view Author view

John Rohlf (jrohlf++at++tubes)
Wed, 22 Mar 95 10:32:10 PST


>
>
> Ran Yakir, Gary Hall, Dr. Goh Eng Lim, Michael Jones, Allan Schaffer,
> Angus Dorbie, and all other co-planar polygon, z-buffer, pfLayer, pfDecal,
> etc experts:
>
> We are still struggling with co-planar planes and z-buffering in our
> application. In our database the roads are co-planar with the terrain.
> These causes the roads to be "Z fighting" or "flimmering" like the man
> pages describe them.
>
> Ran, we have tried to setup stencils using the stensize() command with no
> difference in the behavior of our application. Maybe we are not inserting
> the calls in the right place? Can you more specific as to were this command
> should be placed within the program.
>
> Allan, we are staying away from "stitching" the road polygons because it
> will have to be done manually and there is a lot of roads, rail roads,
> rivers and other networks. Maybe somebody out there has an algorithm that
> will implement this "stitching" function. The algorithm will replace the
> terrain polygons with road polygons, however this is not a trivial task so
> we are also staying away from developing such algorithm.
>
> Angus, we are also staying away from modifying the z-buffer before and after
> drawing the road polygons because we are running on an Indigo Extreme and
> it will kill our frame rate (15 Hz).
>
> The best solution so far is being to label the roads as subfaces. This
> has taken care of the "shimmering" polygons but it has created another
> rendering anomalies. Some times depending on the angle and view point
> location roads can be seen through hills and buildings. The only changes
> here is to the database, and the anomalies can be seen using Perfly. We
> have tried stensize() calls but they do not seam to make any difference.
> Can anybody explain what Performer is doing when it sees this sub-face
> polygons and why we are seen this rendering anomalies.
>
> We have placed a copy of the database in our anonymous ftp site for all of
> those that will like to take a look at these anomalies and maybe suggest a
> fix to this problem. The database can be viewed with Perfly or any other
> program the will load a Flight 14.X format database.
>
> The ftp site is "ftp.vsl.ist.ucf.edu" and the files are in the following
> directory:
>
> /pub/vtb/graf_tdb
>
> There are two files in there:
>
> terrain.tar.Z and
> terrain.uu
>
> The terrain.tar.Z is a UNIX LZW compressed tar file of the database
> and terrain.uu is a uuencoded version of file terrain.tar.Z.
> Please feel free to ftp the database or let me know if you will like
> me to e-mail it to you.
>
>
> thanks
> |
> |\
> | \ /
> |/ /
> / .
> .... ... ...
> ... )\ )\
> / . \ / \ / \
> ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
> Benito Graniela University of Central Florida
> Visual Systems Lab Institute for Simulation & Training
> (407) 658-5075 3280 Progress Drive
> bgraniela++at++vsl.ist.ucf.edu Orlando, FL 32806-0544
> ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~

        Performer uses displacepolygon() for fast rendering
and stencil() for not-so-fast rendering of coplanar geometry.
Unfortunately, IrisGL's displacepolygon() is inadequate in
certain situations - you can see poke through when layers
have large z-slopes and flimmering when they have small z-slopes.

        Note that if you specify PFDECAL_BASE_HIGH_QUALITY when
building your Performer database (See pfLayerMode) you will always get
accurate coplanar rendering. You need not call stensize() nor stencil().

        I am currently working on a fix for the
displacepolygon() problem that will be available in 2.0
and will experiment with your database.
         


New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:51:06 PDT

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