[RC5] film rendering

Eric Gindrup gindrup at okway.okstate.edu
Mon Jan 25 09:37:41 EST 1999


     Alternatively, if the rendering method is hemi-cube based rather 
     than raycasting based, the problem is almost always reducible to 
     some single-model calculations followed by *significant* effort to 
     invert a matrix (or with equal difficulty, triangularize it).
     
     As a funny aside, MPEGged hemi-cube output has been described as 
     performing frequent linear interpolations on the time-dependent 
     version of the "cross-illumination" matrix.  This estimate is not 
     entirely accurate, but it does point at an interesting variation...
     
     Recursively decompose the time domain for the rendering.  For (as 
     yet) unrendered portions, linearize the time-dependencies for a 
     frame generated by a client.  This gives a decent approximation for 
     the 'tween frames.  If the 'tweens don't differ for the two 
     approximations of the same time slice, don't subdivide any more.  
     (This reduces the total work for normal cinematography 
     significantly.  However, since each frame now requires computer 
     algebra (CA) to calculate, the per-frame cost will be higher.)
     
     There exist parallel decomposition methods for inverting matrices.  
     Several distributed efforts are essentially large matrix inverters.
            -- Eric Gindrup ! gindrup at okway.okstate.edu


______________________________ Reply Separator _________________________________
Subject: Re: [RC5] film rendering 
Author:  <rc5 at lists.distributed.net> at SMTP
Date:    1/22/99 1:21 PM


At 06:14 AM 1/22/1999 , you wrote:
>However, rendering _does_ require more network bandwidth than plain ol' 
>calculations like OGR or RC5 - you have to dl the 3D model, and then 
>upload the finished BIIIIIIG (40Mb+ for 32mm) pictures.  Plus of course 
>you'd need LOTS more memory and hard disk space.  Modem users need not 
>apply ;-)
     
What if it wasn't done in whole frames per user, but bunches of pixels? 
You would fetch the 3D model only once per frame, and then the 
"pixelserver" would give you a range of pixels to crunch out.  Multiple
people could all be working on the same frame at once.  You'd only download 
a new model once your frame was finished.
     
[snip]
-----------------------
Scott I. Remick                        mailto:scott at computeralt.com 
Network and Information Systems        (802)388-7545  FAX:(802)388-3697 
Computer Alternatives, Inc.                http://www.computeralt.com

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list