[Beowulf] Scientific =?iso-8859-1?Q?computing=E2=80=99s_future:_Can_any_coding_language_top_a_?=1950s behemoth ?

Douglas Eadline deadline at eadline.org
Tue May 13 12:23:04 PDT 2014


> But I'd ask (extending the analogy) if I build an addition to my parents'
> house, should I use pex or copper? Should I use copper because the old
> part
> of the house uses copper?

So I'm not a plumber and I don't want to push the analogy too far,
and either is easy enough to use either depending on your needs. Now,
if you used copper, then it might be easier/cheaper to find someone
to do the work than with pex, or maybe pex is the easiest way
to run a long line to the new project thus reducing the overall
cost. My point being, don't use pex because it is new and cool,
use what is right for the job.


> How important is it for us to develop in the language of the legacy code
> base? Vs, how important is it for us leverage the training the new young
> guys got in ...I don't know, Java?

Like everything it all depends. And, it runs a little deeper.
So who should/will be writing the next generation code?
The domain specialist or the numerical programming specialist?
Both are hard to find and the best HPC programmers tend to be
somewhere in the middle (in my experience).

My opinion is a programming language that
keeps the domain specialist closer to their domain is
often the better choice. As the article stated
being able to enter y = (a + b)/c (or use complex numbers,
multi-dimensional arrays etc) is much more natural
for the chemist or physicist. There is also the
availability of well tested libraries that keep
programmers from having to re-invent the
"computational wheel."

As much as I like the idea of Functional programming
I'm not sure learning to use implicit loops with recursion
(keep doing this recursively until you reach a terminal
condition) vs writing an explicit loop (Do this a number
of times and stop) is the best way to go ?

To be clear, I'm not advocating one way or another,
without more context such arguments are fruitless.
I do however, believe Julia can become the "HPC BASIC"
we need in technical computing.

  http://clustermonkey.net/Parallel-Programming/julia-hpc-basic.html


--
Doug

> Peter
>
>
> On Tue, May 13, 2014 at 9:39 AM, Douglas Eadline
> <deadline at eadline.org>wrote:
>
>>
>> > Young pretenders to FORTRAN's throne.
>> >
>> >
>> http://arstechnica.com/science/2014/05/scientific-computings-future-can-any-coding-language-top-a-1950s-behemoth/?kw=100k_pvs&search=100k_pvs
>> >
>>
>> Somewhat myopic view of the "possibilities." A pretty good
>> description of the languages it covers (Fortran, Haskell,
>> Clojure, and Julia) Really misses on C/C++ and some other
>> things like OpenMP, MPI to name a few. I like Julia BTW,
>> seems to be the right mix or pragmatism and performance.
>>
>> I also find this reoccurring notion that HPC needs to run out
>> and replace Fortran with something else kind of silly.
>> Fortran is doing just fine, thank you. And if you
>> can make a case or have a requirement for some other
>> language, then use what works for your requirements.
>>
>> When young programmers make snide remarks about
>> Fortran and rewriting codes in modern languages,
>> I ask them if they also considered replacing
>> the old copper plumbing in their house (or parents house)
>> because there are modern products like Pex. I don't
>> even wait for an answer. I just tell them to get off
>> my lawn.
>>
>> --
>> Doug
>>
>> > --
>> > Mailscanner: Clean
>> >
>> > _______________________________________________
>> > Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin
>> Computing
>> > To change your subscription (digest mode or unsubscribe) visit
>> > http://www.beowulf.org/mailman/listinfo/beowulf
>> >
>>
>>
>> --
>> Doug
>>
>> --
>> Mailscanner: Clean
>>
>> _______________________________________________
>> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
>> To change your subscription (digest mode or unsubscribe) visit
>> http://www.beowulf.org/mailman/listinfo/beowulf
>>
>
> --
> Mailscanner: Clean
>
>


--
Doug

-- 
Mailscanner: Clean




More information about the Beowulf mailing list