NVIDIA announces CUDA-x86

[Apoptosis]

NVIDIA announced this morning that they have entered a partnership with The Portland Group and that the company is bringing the CUDA parallel computing framework for x86 processors. This means that CUDA-x86 will compile CUDA applications in order for them to work with x86 processors. This is great news for NVIDIA as their CUDA applications will now run in theory run on "any computer, or any server in the world." There will be some exceptions, but this is good news.

cudax86.jpg (14.16 KiB) Viewed 3165 times

[DarkStarr]

This has been around for a bit, take a look at something called Ocelot. http://code.google.com/p/gpuocelot/

[hark]

Pretty cool that they're expanding the platforms that CUDA code can run on. Too bad they're being stubborn about keeping GPU CUDA exclusive to Nvidia GPUs. OpenCL still seems a more likely standard because of how open it is and how it can work on a lot more platforms.

That Ocelot thing sounds interesting. If they manage to get CUDA applications to run on ATI GPUs, would that mean that PhysX could be accelerated on ATI GPUs?