OpenCL: the next GPGPU standard?

Yesterday, in addition to the presentation of the FireStream 9250, AMD also announced that it was joining the Khronos Compute Working Group. This working group aims to develop standards for parallel calculation and to offer specifications for OpenCL (Open Computer Language), as language for GPGPUs based on C was created with by an initiative by Apple. The group already includes 3Dlabs, AMD, ARM, Codeplay, Ericsson, Freescale, Graphic Remedy, IBM, Imagination Technologies, Intel, Nokia, Nvidia, Motorola, QNX, Qualcomm, Samsung, Seaweed, TI as well as the University of Umeå.

While with CUDA, a programming language (also close to C), which enables using the core of the GeForce supporting DirectX 10, Nvidia seems to be well on its way AMD’s FireStream SDK solution has been offered for several months now and a beta version (v1.01.0) for Linux was recently put online.

Efforts from AMD’s graphic division seem to be bearing their fruit because Rogue Wave Software (specialized in C++ components) and RapidMind (Multi-Core Development Platform, for simplifying parallel calculations) have just announced their support for the Canadian company’s solution. In both cases, these are solutions capable of using the combined power of ET CPUs and GPUs, suggesting some attractive perspectives. Other companies such as CAPS entreprise, Mercury Computer Systems and VizExperts are developing and working with AMD on FireStream.

Surely, this isn’t that significant compared to the amount of attention CUDA is currently receiving, but it does confirm that besides their traditional functions GPUs sold by AMD can and are starting to be used for something other than folding molecules with Folding@Home. And with at least two or three companies (by counting Intel with its Larrabee) initially in the race, the launch of OpenCL comes at the right time.