[Algorithms] Multi core/cell processing and game algorithms...

[Yordan G. <yg...@gy...>]

With all the fuss around the next-gen consoles it would be interesting to 
know what people are doing... (quietly in their corner thinking that their 
wheel is going to be more round that the other's ;)

First its obvious that concurrency is going to be "the thing"... we all 
already do that but may be not on that scale. I was going through some scea 
presentation and saw "rich set synchronization primitives" mentioned... and 
after that mutexes, semaphores and stuff... but what about lock free 
programming? Anyone doing lock-free stuff or just sticking with the old lock 
based approach?

Next we want to apply this divide (into small tasks) and conquer (on 
different cores) approach to as many fields as we can. Of the top of my head 
physics can probably scale well here... but I struggle to image collision 
detection being the best candidate. I have done some cool animation stuff on 
PS2 that will scale wonderfully on multiple cores but what about other 
fields of application? Sound maybe?

Next one is more subtle. It seems to me that just when we all start to think 
that the time has come for game play and AI to move one step ahead the 
next-gen answer is to provide more floating point units to produce better 
graphics and better physics. So, without going into rant mode does anyone 
have any sensible  plan of laying out good gameplay and AI architectures 
that will scale well in the "new era"?

Anyway it will be nice to see what people think...

Thanks,
Yordan