Many familiar names dot the list of founding members, including chip companies such as AMD, Nvidia, and Intel, mobile industry representatives such as ARM, Motorola, Samsung, and TI, and Apple.
GPUs are very good at taking specific tasks, breaking them down into pieces, and solving them at an extremely high rate of speed using multiple processor cores. But they aren’t good at handling the random assortment of software that we all have on our PCs or Macs, which in turn hasn’t been programmed to take advantage of multiple processor cores, for the most part.
To that end, companies like Nvidia, AMD, and Intel have all been working on ways to make it easier for software developers outside of the scientific computing industry to take advantage of the unique characteristics of the GPU.
The Khronos Group, an industry consortium that already administers well-known standards like OpenGL, announced the creation of a Compute Working Group on Monday to develop an industry standard for allowing software developers to tap into the performance offered by graphics processors, or GPUs.
GPUs are perhaps best-known for rendering realistic mayhem in the never-ending sequence of PC shooter games, but they are taking on new roles. Newer operating systems like Vista are placing more graphical demands on the PC, and programmers in the scientific community are also interested in using the power of GPUs to process certain types of applications.
That’s because there’s a notable company missing from the founding members of this group: Microsoft. If Microsoft chooses to go down a different path with the next implementations of Windows and Windows Mobile, it will be hard for the chip companies to resist following suit.
When Apple unveiled Mac OS X Snow Leopard last week during its Worldwide Developers Conference, the company noted that the operating system would feature a technology called OpenCL to make it easier for software developers to access graphics processors. The Khronos Group will evaluate OpenCL as a proposed standard, but there’s no guarantee that all companies will eventually head down that path.
In typical fashion, however, they all came up with different implementations. Nvidia has CUDA, AMD has Stream Computing, and Intel has its Larrabee project, which actually hasn’t been released.
The PC and mobile-computing industries are getting together to propose a standard for computing on graphics processors, and they are going to start their evaluation with Apple’s OpenCL technology.