共用方式為

Concurrency, HPC, the Bay Area Office, and the Right People

  • 發行項

Concurrency and HPC

For as long as I can remember parallel computing has been the "next big thing". Yet it always seemed to be just around the corner. Like the young child on a trip with his parents to Disneyland, “Are we there yet?”, was the question asked every 10 minutes. For the first time in my life I really think we've arrived, I can see Mickey.

 

Looking out at the landscape you can see things are different today as every major hardware manufacturer has moved to building multicore processors. Concurrency will be the key to performance (not faster clocks), see Herb Sutter’s article on the “Free Lunch”. Us working on developers tools land have an interesting problem in front of us. Not only do we need to be able to take advantage of concurrency going forward (you do want faster compilers and IDEs don't you?), but we have to enable the developers that use our tools to do the same.

 

I'm sure you noticed that Microsoft has moved into HPC in a big way, with the announcement of Compute Cluster Server 2003, due out later this year. This is an exciting entrance into an area where Microsoft has been absent throughout the years. Additionally, we actually did some work in Visual Studio 2005 to help enable the HPC developer (OpenMP and parallel debugging support). And Project Phoenix is underway now, which will lead us to the next generation of tools and compilers that target concurrency and HPC. So we've started, but we realize that we need to do more. But to do more we need the right people…

 

The Bay Area Office and the Right People

We are well on our way to addressing all the various aspects of HPC and concurrency with a few very notable hires as of late.

 

Let’s start with Carol Eidt, who comes from Hewlett-Packard and was an HP Fellow. She will be the group manager of the Bay Area Office. That's right, we're actually opening a Bay Area Office where we will be actively looking to hire the best in parallelism, HPC, and compiler, and developer tools technology (if you thought Phoenix was already enticing, this raises the stakes). So those of you who did not want to move to Redmond, but could live in the Bay Area, dust off those resumes -- we have some incredible opportunities ready for you.

 

Another great find was David Callahan from Cray. If you've done any work in compilers then you know his work. He also led the development of the Tera parallelizing compiler, which to this day is probably my favorite parallelizing compiler.

 

Across the company we've also had some high profile hires including Burton Smith, the founder of Tera and ex-Chief Scientist of Cray. He's one of those rare true industry luminaries who lives up to the hype -- and exceeds them.

 

Certainly that would be enough, but we also had recently hired Tony Hey, Fellow of the Royal Academy of Engineering, and ex-Director of the UK e-Science program. He did incredible work in the HPC community, plus he worked with Richard Feynman. :-)

 

So as you can see we are off to a great start, but we still need your help. And whether you want to work in Redmond or the Bay Area, we can make it work.

 

Well that's just an introduction. I should have more details as they arrive. Signing off, and looking forward to hearing from you.

Comments

  • Anonymous
    January 15, 2006
    As someone who used OpenMP on Linux and Visual Studio, I can tell you how easy you folks made it for us to develop in VS 2005.

    The MPI debugging still needs a lot of work, and we found TotalView to be more useful :)

    I hope C# gets OMP support too, especially since Java already have JOMP
  • Anonymous
    January 18, 2006
    This is excellent news. We at Digipede are looking forward to a Bay Area presence for these tools and technologies.
    Cheers,
    Robert
  • Anonymous
    January 25, 2006
    Will Microsoft once again offer a Fortran compiler? They dropped that from VS, how many years ago now? (I'm currently using the Lahey/Toshiba Fortran 95 compiler with Fortran 2000 extensions that targets .NET and runs in the Visual Studio 2003 IDE.) I think a parallelizing Fortran compiler (along with C++) will be important if MS intends to lure away users from Beowulf clusters.
  • Anonymous
    January 29, 2006
    No plans at the moment to do a Fortran compiler. We have some great partners doing Fortran compilers.

    How do you like the Lahey/Toshiba compiler?