Originally Posted by darkened_fetus
...why would they go to all the trouble...
Interesting point !
However
(assuming that this story is accurate), then your earlier pinpoint reference may well hold the key
All of the ultra-clever-uber-bods whose names appear on the original paper are listed as being 'Stanford Super Smarties'
With 'drive' from guys like Mike Houston, Stanford seems to have led the way with a lot of the hardcore GPGPU stuff that is beginning to revolutionise the world
Having heard him speak - I would not pretend to understand the nature of the work that they do...
...but he did say
(Sep 2005) a lot of things that sounded like "ATI's latest technology means that cross compilation of massively parallel tasks is easier than it has ever been" etc
If any of you ever feel alert/superior - then you can give yourself a swift dose of soporiphic inferiority with a visit down these threads...
http://forum.beyond3d.com/forumdisplay.php?f=42
I must confess that I regularly lose sleep worrying about bringing "...together researchers and practitioners working on feedback-directed optimization and back-end compilation techniques..."
Bottom line is that replicating 'previously ~impossible tasks that had to be rendered off line' in an online, realtime way would be
exactly the kind of demo you would want to run if you had something clever
That's why you'd go to the trouble...