The Khronos Group announce Heterogeneous Computing Initiative

Tuesday 17th June 2008, 11:15:00 PM, written by Rys

The Khronos Group, an organised collection of interested parties that collaborate to push and develop open standards for certain classes of computing, have announced what they call the Heterogeneous Computing Initiative.

The hand-waving impetus behind the initiative is to create a set of open standards behind the idea of task and data parallel computing, via what they're calling the Compute Working Group.  The heterogenous part comes from the execution of the same codes on both CPUs and GPUs.

The Compute Working Group has initial membership from the likes of Intel, AMD, NVIDIA, ImgTec, Nokia, Apple, 3DLabs, ARM, Freescale, TI and Qualcomm (we highlighted the discrete and embedded GPU guys on purpose, sorry to those on the list in the PR that we've not namechecked here).  OpenCL is laid out as the focal point of the group for the time being, as the industry gets behind it, and representatives of the companies mentioned will be at SIGGRAPH to talk about the Compute Working Group, OpenCL and related topics.

The announcement is a big deal, and we'll cover it in a bit more depth in the near future.  Until then, you can read the press release, and the forums are getting warmed up with discussion as you read this.

Discuss on the forums

Tagging

b3d ± khronos, group, compute, working, opencl, opengl, heterogenous, gpu, cpu


Latest Thread Comments (334 total)
Posted by Dade on Saturday, 02-Jan-10 17:48:56 UTC
Quoting Jawed
http://sa09.idav.ucdavis.edu/
Thanks, very useful and interesting :!:

Posted by Arnold Beckenbauer on Thursday, 14-Jan-10 19:44:55 UTC
Some more details on the GPU3 core, regarding OpenCL (http://folding.typepad.com/news/2010/01/some-more-details-on-the-gpu3-core-regarding-opencl.html)

Quote
This means that the core will only roll out for NVIDIA first. *ATI has depreciated Brook, but does not have a fully-functioning OpenCL implementation,* so we are stuck in between support on the ATI side. Once the OpenCL implementation matures (on both ATI and NVIDIA), we will be able to finish and optimize our OpenCL code (we can't reliably optimize code until the implementations are more finalized), and then the OpenCL portions will go into QA
How is this possible?

Posted by mhouston on Thursday, 14-Jan-10 20:12:58 UTC
Sigh, the *OpenMM OpenCL port is not fully functional*, the AMD implementation is fully functional and many other software packages and projects are already using it. For example, we showed a bunch of computational chemistry/bio/physics apps running on AMD's implementation (both CPU and GPU) at Supercomputing. You can also see around these forums that our OpenCL implementation is pretty good, although not without bugs. Vijay was not sufficiently clear in that post. GPU3 is running CUDA on Nvidia, not OpenCL. The OpenCL port is not working completely anywhere (although I just fixed a bug preventing both Nvidia's and AMD's latest implementations from working at all...) and the base port was just completed at the end of last year. Now it is entering testing and tuning.And, Brook runs just fine still, but it is no longer in active development by AMD, it has transitioned back to a fully opensource project as we said we would do.

Posted by entity279 on Thursday, 14-Jan-10 21:44:17 UTC
Quoting mhouston
Vijay was not sufficiently clear in that post..
Yeah but unclear is one thing, and saying AMD's implementation is not fully functional is another. It is plain false.

Or do they have some specific feature in mind when they say it?

Posted by mhouston on Thursday, 14-Jan-10 22:09:28 UTC
It's true that OpenMM on AMD's OpenCL is not currently fully functional, i.e. there is a bug found on compiling 2 kernels making our GPU compiler unhappy, but we are seeing some other issues that haven't been tracked yet.

Posted by Arnold Beckenbauer on Thursday, 14-Jan-10 22:58:10 UTC
The HD4000s do not support OpenCL's Local Memory: Is it a big problem for the GPU3 core?
(Ok, assuming HD4000s can run the GPU3 core)

Posted by mhouston on Thursday, 14-Jan-10 22:59:57 UTC
We don't know yet what the performance impact will be. It will be kernel dependent for sure and have a lot to do with ALU/Bandwidth ratio of the code.

Posted by Arnold Beckenbauer on Friday, 15-Jan-10 00:01:33 UTC
Ok. But why don't you port the old ATi physics demos to OpenCL? :lol:

Posted by Jawed Geeking Out with AMD OpenCL on Thursday, 21-Jan-10 23:01:05 UTC
http://www.hardocp.com/article/2010/01/18/geeking_out_amd_opencl Jawed

Posted by Dade on Friday, 22-Jan-10 17:36:35 UTC
Quoting Arnold Beckenbauer
Ok. But why don't you port the old ATi physics demos to OpenCL? :lol:
Few OpenCL/Physic related links I have seen in the latest days:- http://code.google.com/p/bullet/issues/detail?id=336 (http://bulletphysics.org/)- http://www.youtube.com/watch?v=33rU1axSKhQ


Add your comment in the forums

Related b3d News

Really Techy Worldly Hot Par 2010 coverage
Carmack Rage: either console could look superior
Sub $100 graphics at Tech Report
Broadcom purchase AMD's DTV business
Lucid Hydra 100 multi-GPU scaling demonstrated at IDF
3DMark, the Game
PC's Assassin Creed Official System Requirements: 2GB of RAM
Beyond3D's 2007 'Bricks & Bouquets'
Illuminate Labs adds its lighting baking solutions to UE3
Are foundries trying to inflate wafer prices?