Another Voice for Re-Thinking OSes in the Age of Multi-Core

I am disappointed in the fall off in coverage for efforts to improve programming of parallel systems. Maybe it is a bias in my own personal sources but it seems like the energy of a few years ago has abated somewhat.

Maybe this Network World story to which Slashdot links will help re-invigorate discussion. It covers a presentation by Dave Probert, a kernel architect at Microsoft. As the article notes, not even Dave’s cohorts at the Redmond software giant necessarily agree with his views.

I don’t think what he’s suggesting, from improving CPU scheduling to adopting a more hypervisor like approach, hasn’t been covered by other researchers. I am curious whether he’s looked at the field of parallel programming as being pursued by competitors and academics alike.

Regardless, maybe the question coming from within such a high profile company will re-ignite interest, even if it arises from the midst of a bit of friction at Microsoft’s expense.

3 Replies to “Another Voice for Re-Thinking OSes in the Age of Multi-Core”

    1. OK, a quick read of Wikipedia’s entry on GCD reveals it is similar to threading approaches I have used before. Probert no doubt would lump it in with existing approaches he feels have failed for reasons of complexity but GCD actually seems to nicely streamline the idea of queue into a pool, the actually processors, units of work. I especially like that it can cut out the busy-wait non-sense specifically for event handling.

      I tried to find anything relating to Probert’s presentation, even an executive summary or abstract, to see what it is more specifically he is suggesting but have yet to turn anything up.

  1. That’s an excellent question, one that had occurred to me. I haven’t read enough about the technology itself to really say. This is probably as good an excuse as any to dig in a bit more.

Leave a Reply

Your email address will not be published. Required fields are marked *