Re: Mathematica 7 is now available
- To: mathgroup at smc.vnet.net
- Subject: [mg93748] Re: Mathematica 7 is now available
- From: Mark Westwood <markc.westwood at gmail.com>
- Date: Mon, 24 Nov 2008 06:45:22 -0500 (EST)
- References: <email@example.com>
There are 2 possible justifications for parallel computing:
1) Just because it's there, and it's an interesting set of problems
to tackle. But it's mainly 'there' because:
2) Serial execution on a single CPU is too slow (or, as a variant,
your problem requires so much memory that you need to use the address
space of more than one process(or))
When is serial execution on a single CPU too slow ? Well, that's the
question you have to answer for your problems and practices. Do you
want to wait 3 days for a serial answer, or spend 3 weeks
parallelising your program to wait 1 hour ? Depends on how often you
are going to run the program I suspect.
The questions you raise about parallel Mathematica and parallel
Fortran or C are interesting ones too. I do a fair bit of Mathematica
programming but my day job is really Fortran programming for solving
problems in computational electromagnetics. This is one of those
fields where parallel computing is firmly entrenched; solving large
inverse problems just needs lots of CPUs and (usually) lots of memory.
For me, the availability of parallel Mathematica, and a desktop with 2
quad core processors, changes some of the variables in the equations
which determine, for me, whether or not a program should be written in
Mathematica or in Fortran. However, I have a strong preference for
slow programs easily-written in Mathematica, easily integrated with
graphics and all the rest.
I expect you'll get a lot of other views on this one.
On Nov 21, 10:33 am, Alexei Boulbitch <Alexei.Boulbi... at iee.lu> wrote:
> Dear MatheGroup members,
> I would like to put a trivial though an important question to you. For
> some time I try to find problems to be best solved*
> using Mathematica, and where parallel computing would be necessary. So
> far I did not succeed in finding them. This tells
> me that I do not understand something important. I would be grateful, if
> you could give few examples. Very frankly,
> this my mail is not aimed to state a uselessness of parallelization. In
> contrast, I need few realistic examples to make my mind and
> begin understanding, if I may need using a parallelization myself.
> * Just to explain the point: molecular dynamics simulation for instance,
> often requires parallelization, but here Fortran
> of C would be probably methods of choice, rather than Mathematica. In
> analytical calculations in contrast, Mathematica
> would be the method of choice, but in this case I do not see how
> parallelization may be benefited from.
> It is evident how this my question is related to the M7 release.
> Best regards, and thank you in advance, Alexei
> Alexei Boulbitch, Dr., Habil.
> Senior Scientist
> IEE S.A.
> ZAE Weiergewan
> 11, rue Edmond Reuter
> L-5326 Contern
> Phone: +352 2454 2566
> Fax: +352 2454 3566
> This e-mail may contain trade secrets or privileged, undisclosed or other=
wise confidential information. If you are not the intended recipient and ha=
ve received this e-mail in error, you are hereby notified that any review, =
copying or distribution of it is strictly prohibited. Please inform us imme=
diately and destroy the original transmittal from your system. Thank you fo=
r your co-operation.
Prev by Date:
Re: ContourPlot and ColorFunction Opacity
Next by Date:
Re: How to animate a (scrolling) ListPlot in a procedural Mathematica
Previous by thread:
Re: Mathematica 7 is now available
Next by thread:
Cost of Composition vs Pure Function