Saarcor at SIGGRAPH 2005

[Titanio]

Apparently the same guys ported their implementation to Cell with IBM Germany, and will be presenting the results of that at Siggraph also.

http://www.uni-saarland.de/de/medien/2005/07/1122474975

Doesn't really say how it compares, except that scenes run in realtime at "full resolution". Of course, realtime may mean a few frames per second where this is concerned

 

[MfA]

Oh ... internal competition.

 

[Simon F]

RT complexity is logarthmic if an implementation make use of some spatial subdivision schemes as octrees, kdtree...

The setup time for the subdivision structure is going to be (at least) linear so, at some point, the log time must be overtaken by the linear