I am working rigth now to sync with max speed multiple chunks of code/data on 8cores... and from my experience compiler wouldnt do magic trick for lot of ppl ...

you need to understand application data structure+dependency and statistics to make huge jump in application speed on multicores.

BTW there would be half day event about it on THE GRID soon about it from Intel: http://www.life20.net/intel1reg.php



Just send John McClane then. When it comes to fighting, McClane always wins out in the end. He'll be able to wring out the parallelism in anything !

Wow this RapidMind has such a good marketing team.

They manage to push their bull**** t product pr all across the interweb.

A product that has some success in the HPC market which is parallel by nature and will do nothing to more generic software.

If programmers care about parallelism (and 98.7 of them don’t) they should learn to use locks and events and stop looking for a magic solution.

It's called distributed computing, and cluster computing. People have been using it for years. 

Same as what I posted in reply to a similar story here yesterday - what is required is a common assembley language that works on any hardware, and the distributed & cluster middleware uses the hardware calls directly - that'll speed up the instructs being executed.

I notice you didn't allow that comment to be put online though. Bet you thought you could magic it away by pretending I never wrote it.