KEEPING THE ENDORSEMENTS coming for the OpenCL platform, AMD has also announced it will be supporting the newly ratified OpenCL 1.0 specification.
The firm says it plans to release a developer version of its ATI Stream SDK with support for OpenCL 1.0 for content developers by the first half of 2009. Apparently, AMDs engineers have been slaving away on OpenCL since its early spec days and have already started running code on its initial implementation.
Probably much to Nvidia's chagrin, AMD boasts it was the first company to even launch a GPGPU strategy, unveiling its Stream Computing vision in 2006, also gushing in a press release it had been the pioneer in launching GPGPU software development tools.
The Green Goblin also announced its love and support for the Open Computing Language spec drawn up by the multi vendor Khronus group (shurely crony group?) at Siggraph Asia the other day, despite remaining deeply passionate about its own Cuda.
It will also be the first time OpenCL, a compute API which developers can use for faster, GPU powered, parallel computing, will be able to run on Nvidia's own cards, providing an alternative to Cuda.
"We don't think one size fits all", Nvidia's general manager of Cuda, Manju Hegde, told the INQUIRER, adding "developers are too diverse a group to fall into that mantra". Hegde also noted how important it had been to the firm when Apple had helped efforts to push OpenCL forward, noting "Their recognition that the GPU will now play an essential role in consumer applications is a significant milestone in the history of computing."
Senior Veep and GM of AMD's Graphics Products Group, Rick Bergman, also noted "The potential benefits of having applications run on both the CPU and GPU within a system are enormous," adding that up until now programmers could only choose proprietary programming languages that limited their ability to write vendor-neutral, cross-platform applications. "With today's ratification of OpenCL 1.0, I'm happy to say those days are over. Developers now have a better, truly open choice," Bergman said.
Nvidia seems to agree with the sentiment, claiming anything that will generally push GPU computing forward and make computers faster is a good thing. Especially if it makes people want to rush out and buy a better GPU to use for gaming, video transcoding, picture editing, scientific modeling, oil and gas exploration or any other similarly entertaining hobby.
But just because NV's gone all free love, doesn't mean it's forgetting its own, much plugged Cuda. Cuda, or C is for Cuda as the Nvidia sesame street gang likes to call it, is also a massively parallel computing ISA and hardware architecture, designed to natively support all parallel computing interfaces. It provides for industry standard languages including C, Java, Fortran and Python and will now run Open CL for programmers who want that little bit more flexibility when writing code. µ