[Chromium-dev] Explaination of TileSort

[<t-l...@no...>]

Hello, 

     Sorry to be asking such a broad question, but I've been working with the 
tilesort code for a couple of weeks now, and I'm still trying to surmount 
obstacles to a CAVE implementation.  I stripped out the sorting function in the 
tilesort_bucket.c file, and wrote my own configuration scripts, so I am able to 
get a working system (I can broadcast the entire geometry and sort geometry 
based on a server's viewing frustum).  My trouble now is being able to supply 
each server with its separate viewing parameters - specifically its own 
viewport and matrix stack.  Even though the geometry can be sorted, it still 
needs the appropriate viewing transforms applied to it. 
     The tilesort SPU appears to only be using one giant viewport and one 
matrix stack to run its servers (there seems to be one primary context that is 
accessed in the bucket sort algorithm to retrieve the transformation 
information).  I want each server to have its own, and this doesn't seem 
possible with my understanding of the tilesort SPU architecture.  Is there any 
documentation of the tilesort code?  The comments haven't provided me with much 
help.  I'm specifically interested in the relationship of the "contexts" to the 
servers and the spu, and how threading plays a role. 

     Somewhat unrelated, I would like to be able to override the call to 
gluLookAt if possible.  Since this isn't in the opengl library, would it be 
difficult to support it? 

If anyone could provide some help or advice, I'd be much obliged. 

tom