Path: csiph.com!newsfeed.hal-mli.net!feeder3.hal-mli.net!newsfeed.hal-mli.net!feeder1.hal-mli.net!border3.nntp.dca.giganews.com!border1.nntp.dca.giganews.com!border4.nntp.dca.giganews.com!border2.nntp.dca.giganews.com!nntp.giganews.com!news.iecc.com!nerds-end From: Marco van de Voort Newsgroups: comp.compilers Subject: Re: HPC and parallel programming, was Have we reached the asymptotic Date: Wed, 21 Mar 2012 10:28:27 +0000 (UTC) Organization: Stack Usenet News Service Lines: 23 Sender: news@iecc.com Approved: comp.compilers@iecc.com Message-ID: <12-03-048@comp.compilers> References: <12-03-012@comp.compilers> <12-03-013@comp.compilers> <12-03-047@comp.compilers> NNTP-Posting-Host: news.iecc.com X-Trace: leila.iecc.com 1332441541 37698 64.57.183.58 (22 Mar 2012 18:39:01 GMT) X-Complaints-To: abuse@iecc.com NNTP-Posting-Date: Thu, 22 Mar 2012 18:39:01 +0000 (UTC) Keywords: design, parallel Posted-Date: 22 Mar 2012 14:39:01 EDT X-submission-address: compilers@iecc.com X-moderator-address: compilers-request@iecc.com X-FAQ-and-archives: http://compilers.iecc.com Xref: csiph.com comp.compilers:515 On 2012-03-19, Victor Eijkhout wrote: >> The biggest thing that I think needs innovation is parallel programming: >> we're still waiting for the innovation like OOP that makes large >> commercial parallel programs (leaving aside HPC as wanting something >> different) much easier to program. > > Can you characterize non-HPC parallelism? Productivity before performance. > Why is it different? HPC is more performance before productivity. > What are the problems to be addressed? Most programming is sequential. There is no general method to change a sequential program into a parallel one, and even the methods for specific problems domains that are there require a high programmer skill/experience level, and is then still laboursome and risky. This makes workers more expensive and takes longer training, with potential retraining needed for a different problem domain.