DON CLARK at the Wall Street Journal (sub required) seems to have got a press release about parallel computing earlier than anyone else today.
He writes that His Voleness and La Intella will announce a major investment to promote programming for multicore chips.
This will be led by boffins at Berkeley and there’s probably going to be a lot more money put in than the cash prizes of $250 AMD said it was offering a week or two back.
Intel – like AMD – is really hoist by its own petard. After running out of places to go in the megahurts wars, attention was turned to multicore chips, and no doubt we’ll probably see Intel “Atom” MIDs soon with multiple cores. But the big big problem is how to write software that will take advantage of these hardware capabilities.
And it’s not a new big big problem. Software boffins have struggled with the concept for years and years. The Journal quotes William Dally, a Stanford professor, as saying that while the chip makers are hurtling pell mell towards multicores, no one has a clue on how to program for them.
That no doubt includes Microsoft, which couldn’t even be bothered to program for Intel’s marketing scheme called HT – that’s hyperthreading, not hypertension – in the glory [surely gory, Ed.] days of Chipzilla’s Pentium 4.♣
The reason nobody has been programming for them is because there has been little incentive to do so (outside of academia and specialized programming houses who preferred to keep such knowledge in house).
Now that multi-core chips are becoming mainstream (or are they mainstream already), I think people will start taking advantage of them.
Mind you, none of this new-fangled innovation will come from existing mainstream companies who always play conservatively and never want to take risks (at least if they are staffed with the kinds of managers I have experience with).
Thanks for the topic
شكرا علي الموضوع
http://www.freeprogam.com