Sign up for our newsletter and get the latest HPC news and analysis.

RapidMind acquired by Intel, likely good news for the community

Remember way back when PeakStream was acquired by Google? That acquisition was characterized by total silence, and then PeakStream disappeared into the Googleplex to become a competitive advantage for Google’s own code development effort.

Intel logoThe newest multicore development tool acquisition is at least starting differently. A reader pointed us to a blog post and thence to RapidMind’s web site today where we find this

We are now part of Intel Corporation, the leader in software development tools for software parallelism and performance. This change is a great development for our customers and for easing the effort in parallel programming. With Intel, our shared vision for enabling parallelism will lead to strengthening our software developer tools for multicore and manycore processors and accelerators. Later this year, we will provide more details about the integration of the RapidMind platform and Intel software products and technologies, including the Intel Ct technology for data parallelism.

We continue to sell the RapidMind platform and we encourage you to contact us to explore your interest in solutions for software parallelism. We respect and strive to preserve and expand the investments our customers have made in our products. You are encouraged to visit the Intel Developer Forum in September 2009 to hear more about Intel software developer tools.

And James Reinders, Intel’s software acquisition bullhorn (he announced the Cilk Arts buy a couple weeks ago too), has this to say on Intel’s software blog

Data Parallelism need not seem like parallelism when writing software. The advantages in avoiding “feeling like you have to learn, or deal with, parallelism” are substantial. Imagine programming, feeling familiar and intuitive, which is scalable for more and more processor cores and has safety by default (safety from data races and deadlock). That’s been our vision for our product using Intel Ct technology. Now we’ve joined up with Rapidmind, people who share our vision.

…We’ve been talking with customers, who are using Rapidmind products, about our plans together and the feedback has been very positive. We have exciting things in store for the future. Our product plans for Intel Ct technology are on track for beta before the end of the year, and the integration with Rapidmind products will come in phases after the first beta is available. This will enable us to offer the best solution our joint team can create.

James also confirms that Intel will continue to sell and support RapidMind as a separate product. So, unlike the very bad omen I felt the PeakStream acquisition was for the community, I think Intel’s move here is on balance positive, and will likely result in more resources and a bigger platform being given to this technology. Congrats to the RapidMind team, by the way.

Comments

  1. It’s great to hear McCool and the boys over at RapidMind are doing great. I do hope that the RapidMind IP does stick around. That deal with PeakStream and google was a bit of a kick in the teeth for those who were working with them. I know on our side, we were hoping to start getting some software converted to PeakStream and then they went dark. With RapidMind though, I am very curious as to what the direction and status of their current lineup of software will be. They had essentially gotten their start in GPU and Cell based solutions before they moved to multi-core.

    Also, with OpenCL beginning to take shape in the computing world, where does this end up leaving technologies like RapidMind in the foreseeable future? A middleware solution for OpenCL integration or a competing technology to it?

  2. Chris – are you using Rapidmind now? If so, I’m confident you’ll like the future as it unfolds.

    Rapidmind is not disappearing. Combining with our product for Ct technology is like a marriage made in heaven.

    Rapidmind has expanded from Cell, GPUs to multicore as they were able to get to a vision we share: harnessing data parallelism, without requiring deep parallel programming knowledge or reasoning, with satefy by default (in my opinion, one of three key barriers to being able to parallel programming widely adopted). This vision really makes the most sense when it is cross-platform – not tied to one processor/GPU/accelerator nor a single OS.

    I’m proud of the success we’ve had with Intel Threading Building Blocks (TBB). It is so successful in no small part because it delivers on the cross-plaftform support. Actions speak louder than words. I’m confident believing in Intel as a tools provider is a good bet. Personally, I think it is the best bet you can make in parallel programming. As an Intel employee, you might double check my enthusiasm – but I’ll stand behind it.

    Nothing happens overnight, including OpenCL. OpenCL holds promise for tools vendors as a target. Rapidmind currently targets CUDA. It would be a bonus for us, and our future tools, if the promise of a non-proprietary CUDA (as some have describe OpenCL) emerges as an alternate target with more reach. We’re seeing a lot of interest in providing abstract targets for tools – like Microsoft’s Concurrency Runtime (for Windows, in beta now), and Apple’s Grand Central Dispatch (for Mac OS X – in Snow Leopard, in beta now). Microsoft is targetting tool writers, Apple is much more aimed at users, and OpenCL is in between. Each is interesting, and definitely no overlapping. Offering a lot of “opportunity.”

    The future of wide spread parallel programming is in higher level abstract and portable methods. OpenMP, Intel Threading Building Blocks, and Ct/Rapidmind are amoung the answers.

    James Reinders, Intel

  3. Hi James, thanks for the comment. These solutions I would like to understand a bit more about. I had done work with Rapidmind on a couple of fronts. Testing with Cell and GPU. This was probably nearly 2 years ago, but I do like to keep an eye and see how RapidMind is doing. I have been more focused on developing hardware solutions around CUDA as of late simply because when I had started developing platforms, we had multi GPU in mind, and CUDA not only had a solution that operated multi GPU, from our findings, it supported N number of GPUs. In our case, it was 6 Tesla C870 GPUs. On top of that, there was actually an application in development that supported CUDA multi GPU processing (VMD) that even scaled to 6 GPUs.

    If you have the chance, I would like to discuss this more with you directly, either by phone or e-mail.

  4. Being acquired by a larger company – the fate of all successful small companies.

Trackbacks

  1. […] 20, 2009 by bearcrossings This morning I was reading John West’s article about Intel’s acquisition of RapidMind.  It’s the latest example of the High Performance Computing (HPC) industry recognizing the […]

Resource Links: