On Mar 14, 2007, at 22:11 UTC, Daniel Stenning wrote: > "A company called Codeplay showed up at this week's GDC to talk about > their new auto-parallelizing compiler, called Sieve. Sieve takes in > single-threaded C/C++ code, examines it for dependencies and > parallelization opportunities, and turns it into multithreaded code > for use on multicore processors" > > Might this be something that RS could also implement for RB ? Since > we are now in a "multiprocessor age" , the ability to easily create > MP exes from RB by such means could be a good selling/marketing point > for RS.
Well, the RB framework already took advantage of some parallel processing by means of the Altivec unit (back in the PowerPC days, of course). Intel chips don't (or at least, didn't) have as much vector-processing muscle, but I think RB's framework may make use of what it can of that, too. But multi-core processing is a whole different kettle of fish of another color, so to speak. My wife researches this sort of thing for a living, and from what I've picked up, I'm very skeptical about how much benefit Codeplay's tricks would produce on real code. The same would apply to RB code. Most code is simply not very parallelisable (to use their made-up word) without a lot of serious refactoring. A lot more bang for the buck would be had by implementing more common optimizations. (Don't believe those who say that RB is not an optimizing compiler -- it certainly DOES do optimizations, but it could probably stand to do a lot more.) Best, - Joe -- Joe Strout -- [EMAIL PROTECTED] Verified Express, LLC "Making the Internet a Better Place" http://www.verex.com/ _______________________________________________ Unsubscribe or switch delivery mode: <http://www.realsoftware.com/support/listmanager/> Search the archives: <http://support.realsoftware.com/listarchives/lists.html>
