We’re looking forward to the first handheld supercomputer, thanks to an Apple-developed technology the company is now working to see accepted as an open standard.
Apple announced Snow Leopard at WWDC last week, equipped with the technology – Open Computing Language (OpenCL) – that will allow the OS to use the processing power of graphics processors for non-graphics apps.
OpenCL is based on the C programming language and has been proposed as an open standard. As part of the standard-setting process, Apple has joined a consortium of companies (the Khronos Group) to develop OpenCL as a standard.
Apple CEO Steve Jobs last week told the New York Times a little about OpenCL. “Basically it lets you use graphics processors to do computation,” he said. “It’s way beyond what Nvidia or anyone else has, and it’s really simple.” Interestingly, both Nvidia and ATI are part of the standards-setting consortium.
The standards setting procedure is taking place under the aegis of the Khronos Group, which has set up the Compute Working Group to investigate how best to use the increasingly powerful capabilities of GPUs to share computational tasks with the host processor.
Khronos Group president, Neil Trevett, said, "The Compute Working Group potentially will be one of the most significant standardisation efforts at Khronos. Highly-accelerated parallel computation across GPUs and CPUs is essential to many emerging rich consumer applications that will transform the computing experience of diverse users."
Oh but the web we weave – Apple’s secrets continue to unravel, with Trevett’s next revelation: "Significantly, this initiative is aimed at both desktop and embedded devices – the day when you will be able to hold a supercomputer in the palm of your hand is perhaps not so far away."
Other members of the working group include AMD, ARM, IBM, Intel, and Nokia.
How is this beyond what NVidia has when they have CUDA for just this purpose?
It’s likely in the implementation. Just because they have another idea to accomplish the same purpose doesn’t mean it’s just as effective or efficient or easy to use, etc. If such is the case, you may see NVidia using CUDA only in the interim until OpenCL is finished and standardized.
My understanding is that OpenCL may be applied to many different CPU/GPU configurations whereas NVIDIA and ATI likely have tools for their specific products. So, if I’m a movie producer and for some reason (post purchase add on) my computer has both an ATI and a NVIDIA card, OpenCL should allow me to utilize the GPUs on both of those cards. NVIDIAs CUDA implementation wouldn’t use the ATI GPU.