AMD looks to Folding@home language for gaming physics

Posted at: 4:17am 30th April 2008 by Ben Hardwidge

Stanford University’s Brook Plus language could be used for calculating physics in games

Nvidia may be working towards making its newly acquired PhysX technology the gaming physics standard, but AMD reckons that there are plenty of other alternatives out there, including using Stanford University’s and highly parallel Brook language for gaming physics.

AMD’s technical director of sales and marketing, Giusseppe Amato, told Custom PC that Brook ‘allows you to use the parallel SIMD (single instruction multiple data) array that we have in every stream processor engine…so basically you can make any type of 3D pictorial mathematics calculation. Can it also be used for the physics of the game? The answer is yes.’

Brook is currently used by Stanford University in the GPU Folding@home client, and the director of the Folding@home project, Vijay Pande, told us that there’s no reason why it couldn’t work on Nvidia GPUs too. However, Amato says that while the kernel of Brook was used for the Folding@home project, AMD has developed an extended version called Brook Plus, which could be used for gaming physics, as well as many other tasks.

‘We took Brook from Stanford,’ explained Amato, ‘and we extended it. It’s an open language, and the libraries are now starting to be built, so this is all very promising.’ According to Amato, Brook Plus is still hardware independent, so it could be adopted by other hardware manufacturers too.

Now that Intel has bought Havok, AMD is looking at Brook as one of several ways of calculating gaming physics on a GPU. Other methods include using a future version of DirectX, which Amato says is moving more to ‘a general purpose language.’ As well as this, Amato didn’t rule out supporting Nvidia’s newly acquired PhysX API, which Nvidia recently said would be available to third party GPU vendors.

The future of gaming physics, according to Amato, is up to the games industry. ‘Shifting computational power from a CPU to a GPU is something that requires standardisation,’ says Amato, ‘and acceptance from the application program interface, and from the ISVs – the software writers. We as AMD have opened our hardware layer so that the ISV can have full access to the instruction library, and the architecture of our GPU, but we’re also working with the game industry to figure out how they will see the evolution of physics calculations in their games.’

How do you think game developers will want to develop physics in the future? Let us know your thoughts.