From: m_b_metcalf on 13 Jul 2010 08:36 Some of you might find this reassuring: http://cacm.acm.org/magazines/2010/7/95060-the-ideal-hpc-programming-language/fulltext Regards, Mike Metcalf
From: nmm1 on 13 Jul 2010 08:57 In article <1f91c3eb-6fe4-4997-9043-0c948de13e72(a)k39g2000yqb.googlegroups.com>, m_b_metcalf <michaelmetcalf(a)compuserve.com> wrote: >Some of you might find this reassuring: > >http://cacm.acm.org/magazines/2010/7/95060-the-ideal-hpc-programming-language/fulltext Well, he did choose codes that are fairly 'natural' for Fortran. When one is using rectangular matrices and/or complex numbers, modern Fortran is more-or-less a verbose and inelegant direct mapping of the mathematics. I use Cholesky as an example of that in my course, and it really is obvious. Try some others, like regular expression manipulation, and things become a lot less clear. His paper stands, but the preferred language would not be Fortran-like - perhaps Python-like. Regards, Nick Maclaren.
From: viper-2 on 13 Jul 2010 10:15 On Jul 13, 8:57 am, n...(a)cam.ac.uk wrote: > In article <1f91c3eb-6fe4-4997-9043-0c948de13...(a)k39g2000yqb.googlegroups..com>, > > m_b_metcalf <michaelmetc...(a)compuserve.com> wrote: > >Some of you might find this reassuring: > > >http://cacm.acm.org/magazines/2010/7/95060-the-ideal-hpc-programming-... > > <snip> His paper stands, but the preferred > language would not be Fortran-like - perhaps Python-like. For those of us who have chosen to invest the time to learn modern Fortran before other languages it is, indeed, reassuring.:-) I'm not sure that an interpreted language like Python could compete with Fortran in HPC. Python's ease of use is a great selling point, but I suspect therein lie a few drawbacks. Being so easy to learn, Python appears to be growing in popularity. I've seen, for example, where certain Lispers migrated to Python claiming to have become disillusioned with Lisp. Time will tell. agt -- Freedom - no pane, all gaiGN! Code Art Now http://codeartnow.com Email: agt(a)codeartnow.com
From: e p chandler on 13 Jul 2010 11:12 "m_b_metcalf" <michaelmetcalf(a)compuserve.com> wrote in message news:1f91c3eb-6fe4-4997-9043-0c948de13e72(a)k39g2000yqb.googlegroups.com... > Some of you might find this reassuring: > > http://cacm.acm.org/magazines/2010/7/95060-the-ideal-hpc-programming-language/fulltext > > Regards, > > Mike Metcalf Very interesting. The article by Seeley linked to is also quite good. I only write "toy" Fortran programs, but I think the article's message is really two fold: A good programming language is easy to write and easy to read. I think that modern Fortran is a vast improvement over FORTRAN. Even with F2003 partly available, it's still possible to write real nasty programs that are difficult to read and almost impossible to maintain. I will point out two long programs recently posted in this newsgroup. One was posted as a problem with output disappearing because unit 6 was redefined. The other was posed as a problem in un-initialized variables. Both could have been snipped, but the long listings are still somewhat instructive. The first program was difficult to read. As I mentioned, a large number of optimizations were possible even without knowing about the actual problem. Since it was homework, I did not bother to post a condensed program. A shorter program which uses Fortran 95+ gives a much better sense of what is going on. The first program is a mix of numerical integration and differentiation. It uses two series of recurrence relations. The second program was a complete beast. A printed listing would run to 20 pages. What does it do? I don't know. It is obvious that there is a lot of parallel structure that could have been exploited by its author. Maybe re-casting with user functions and subroutines would have helped (at the expense of some performance). A modern language helps, but the programmer also needs some aptitude and experience. ---- e
From: viper-2 on 13 Jul 2010 11:26 On Jul 13, 10:12 am, "e p chandler" <e...(a)juno.com> wrote: > "m_b_metcalf" <michaelmetc...(a)compuserve.com> wrote in message > > news:1f91c3eb-6fe4-4997-9043-0c948de13e72(a)k39g2000yqb.googlegroups.com... > > > I only write "toy" Fortran programs, What exactly do you mean by "toy" programs? agt -- Freedom - no pane, all gaiGN! Code Art Now http://codeartnow.com Email: agt(a)codeartnow.com
|
Next
|
Last
Pages: 1 2 Prev: Cheap Wholesale A&F Long Sleeve Next: Fortran 2003 Dependency Issues (and beyond) in Babel |