AMD Helps Advance Parallel Computing with OpenCL University Kit

[btarunr]

AMD today announced the introduction of the OpenCL University Kit, a set of materials that can be leveraged by any university to assist them in teaching a semester course in OpenCL programming. This effort underscores AMD's commitment to the educational community, which currently includes a number of strategic research initiatives, to enable the next generation of software developers and programmers with the knowledge needed to lead the era of heterogeneous computing. OpenCL, the only non-proprietary industry standard available today for true heterogeneous computing, helps developers to harness the full compute power of both the CPU and GPU to create innovative applications for vivid computing experiences.

"As a former professor at Washington University in St. Louis, I firmly believe that the university setting is a vital environment to cultivate the best and brightest minds and set them on a path to succeed," said Manju Hegde, corporate vice president, AMD Fusion Experience Program. "By ensuring that an industry standard like OpenCL is a central element of the education process, we are helping to put the PC application ecosystem in good hands to take full advantage of a heterogeneous computing future."

"Teaching students to effectively leverage the OpenCL standard involves all the intricacies of parallel programming plus support for a new class of heterogeneous computing devices built on a variety of hardware technologies," said David Kaeli, professor and associate dean of undergraduate programs, Northeastern University College of Engineering. "The OpenCL University Kit introduced by AMD is an easy tool to enable educators to quickly introduce OpenCL learning into their curriculum, helping them strike a balance between teaching syntax and higher level architectural issues."

Included in the University Kit is a 13 lecture series, equipped with instructor and speaker notes, as well as code examples where necessary. An advanced understanding of OpenCL is not needed to understand the course materials; students only require a basic knowledge of C/C++ programming. A C/C++ compiler and an OpenCL implementation (such as the AMD APP SDK) are needed to complete the exercises.

OpenCL is gaining popularity in academia, with a number of universities already offering similar OpenCL courses. For more information on the university courses currently offered, please visit the OpenCL Zone here. If you offer a course that would like to add to the list, you can also do that here.

For students and developers who are interested in presenting technical papers on heterogeneous computing, AMD will be holding its first AMD Fusion Developer Summit from June 13-16 in Seattle, Washington. Proposals can be submitted until February 25. Summit participants will engage in interactive sessions and hands-on labs to deepen their knowledge of advanced CPU and GPU programmability, and gain a better understanding of how software applications can take full advantage of the parallel processing power of APUs, bringing supercomputer-like performance to everyday computing tasks.

View at TechPowerUp Main Site

 

[Mussels]

this was badly needed, openCL needs a kickstart

 

[LAN_deRf_HA]

Is openCL what was supposed to bring in the era of standardized physics in direct X11? Somebody needs to sponsor a game with that already, preferably a neutral party but even AMD would do. Would certainly be better than being sponsored by nvidia for physx implementation as then your profits are limited to owners of one brand.

 

[Mussels]

Is openCL what was supposed to bring in the era of standardized physics in direct X11? Somebody needs to sponsor a game with that already, preferably a neutral party but even AMD would do. Would certainly be better than being sponsored by nvidia for physx implementation as then your profits are limited to owners of one brand.

yes, it was.


openCL is the open 'free for everyone to use' coding language for GPU acceleration.


every time i see a program coded for CUDA, i get sad and think it shoulda been done in openCL

 

[Swamp Monster]

every time i see a program coded for CUDA, i get sad and think it shoulda been done in openCL

+1
Witcher2 will use Havok, and I hope it runs on GPU too via OpenCL.
Good move from AMD, because young people will start to use it more, and I hope game devs too.

 

[theubersmurf]

+1
Witcher2 will use Havok, and I hope it runs on GPU too via OpenCL.
Good move from AMD, because young people will start to use it more, and I hope game devs too.

Havok is owned by intel, somehow I don't get the feeling intel will want openCL implementations of Havoks API.

 

[Swamp Monster]

^ They have already showed demo "Havok Cloth" which uses OpenCL. I hope it develops further. Intel would benefit too, because OpenCL can run on CPU too.

 

[Mussels]

Havok is owned by intel, somehow I don't get the feeling intel will want openCL implementations of Havoks API.

openCL is the 'runs on everyones hardware' standard - it runs on CPU too.


its a good move for intel to support it, especially if they add in support for it to their next gen IGP's...

 

[theubersmurf]

openCL is the 'runs on everyones hardware' standard - it runs on CPU too.


its a good move for intel to support it, especially if they add in support for it to their next gen IGP's...

Even with that being the case, if Add-in gpus outperform their own cpu implementation, I get the feeling they'll scuttle the openCL implementation, or hold off on releasing a version until their own optimizations for their cpu implementations make Add-in gpus useless. Intel has done a lot to downplay gpu compute tools and push for CPU implementations of many of those tools, since they seem to want to make add in gpus valueless.

 

[wahdangun]

its not gpu valueless that intel chase but to make cpu more important for gamer, let face it, even core i7 or amd x6 was useless for games right now, but shame intel seems didn't push havock hard enough to use all potential cpu power to do physic

 

[Imsochobo]

YAAAY for AMD!

AMD have done good for the open source community lately!

 

[Cheeseball]

AMD has already been doing this in some Uni's here in Asia (albeit on-hand training, not a kit), so it's great to hear that they've compiled their resources into an easy rollout kit.

Now that they can gather publicity, all AMD has to do is properly implement OpenCL 1.1 specifications.

every time i see a program coded for CUDA, i get sad and think it shoulda been done in openCL

The reason why developers use CUDA is because:
a.) Unfortunately NVIDIA is sponsoring the developers.
b.) OpenCL 1.1 alone is still very lacking.

What the Khronos group needs to do is implement CUDA into OpenCL because from what I've seen, converting code from CUDA to OpenCL is very fuckin' (I curse here because it's that good) easy to do.

The other issue is AMD's implementation which is still buggy as of December 2010. Their implementation still doesn't support Crossfire (I hate this), proper implementation of BLAS, FFT, and reductions libraries (this is why Adobe uses CUDA now for post-processing in their apps), lacking documentation, the dwindling lack of *nix support, and greater 1D texture support (like CUDA).