Quote Originally Posted by RealNC View Post
Didn't people in some previous thread say that the LLVM dudes do not want to provide OpenMP at all because it doesn't scale and they instead want to work on an alternative? Turns out this was just a bunch of nonsense.
One of the reasons LLVM has taken off like wildfire is that it easy to modify and easy to use as the basis for research projects, and there are plenty of such projects, both in companies and at universities.

Point is: to the extent that there are "LLVM dudes" who have their own opinions about the best way to handle parallelism, that does not prevent anyone else from doing what they like to add OpenMP to the code.
And, based on previous history (and in spite of whatever "Apple is the spawn of satan" theories people may claim), once the patches appear to be stable and unproblematic, they'll probably be rolled into the mainline.

I imagine at some point, once more immediately pressing matters have been resolved, we'll see Apple's future vision for parallelism (which probably consists of syntactic changes to Objective-C, rather than littering the code with #pragmas) rolled into LLVM, and that's where Apple's going to concentrate the time of its people.
But if Intel want to go ahead with OpenMP, why not? It's no different from IBM adding SystemZ support, or some PhD student adding support for a fancy new optimization he's invented.