On Fri, 2015-04-03 at 15:34 +0000, John Colvin via Digitalmars-d wrote: […] > > I've had a look at Chapel and I don't get what the big deal is.
It's PGAS programming, so you control the whole parallel computation in a single program. Though for clustering you may have to suffer MPI and hence an element of SPMD nastiness. Also there can be problems distinguishing cores, processors, and computers – the hardware has three levels of parallelism, PGAS only two. > There's some nice syntax and good thinking about parallelism in > there*, but I don't see what's exciting after that... Maybe D has > spoiled me for seeing power in a language. I believe it would be most useful to unpack this to try and decide what D has right, what Chapel has right, what D has wrong and what Chapel has wrong. My thinking here is that Chapel seems to have everything needed for the practicing (not practising, though maybe that as well) HPC programmer. D has lots but the core parallelism story is classic single bus and so cannot handle the core/processor/computer split that is easy with Chapel. > I guess what I'm saying is I can see that they've put a lot of > thought in to good abstractions for parallelism in HPC, we should > steal a bunch of it because D is eminently capable of supporting > similar abstractions, while being a much more rounded language in > other regards. But what does D have that people using Chapel should be demanding? This is not to tell Brad et al. what they need to look at, but that as well, more to say, why would these people consider D if the parallelism message of D was stronger. -- Russel. ============================================================================= Dr Russel Winder t: +44 20 7585 2200 voip: sip:russel.win...@ekiga.net 41 Buckmaster Road m: +44 7770 465 077 xmpp: rus...@winder.org.uk London SW11 1EN, UK w: www.russel.org.uk skype: russel_winder
signature.asc
Description: This is a digitally signed message part