On 6/9/2012 10:45 AM, Richard Hennessy wrote:
> Does Fateman speak for the entire Maxima development team? I am not a
> member of the team, so I don't know who is. Do you all agree with him?
>
> Richard Hennessy
I'm not sure there is any "team" and if so, I certainly don't speak on
behalf of anyone else!
Many people (including me) have worked on projects involving
parallelism, programming languages,
and computer algebra algorithms (separately and together). If you have
innovative, well-thought-out ideas that you are willing and able to
implement to make all this work better, that would be great.
To know if something is innovative, you need to be informed on past work.
See for example the PASCO series of conferences
http://pasco2010.imag.fr/
or some of the other 221,000 google hits for {"computer algebra"
parallel}.
If there is something specific that works for Maple or Mathematica,
something that
they have designed nicely, and that improves performance in some
application of
interest, tell us about it.
I suspect that all they do is launch multiple kernels and have them
communicate via sockets or something similar.
This is not a good way of doing things generally. It has one advantage:
it is obvious.
I tried Mathematica with ParallelTable and found it slower (I have 4
processors) on one problem.
I tried a larger problem, and it jammed up my computer totally so I had
to kill the process.
I suspect that having MORE processors would be even worse.
Just because multiple CPUs is new to you does not mean people have not
thought about
these issues.
RJF
>
>
> -----Original Message----- From: Richard Hennessy
> Sent: Saturday, June 09, 2012 1:29 PM
> To: Richard Fateman
> Cc: maxima at math.utexas.edu
> Subject: Re: [Maxima] parallel cell executing and wxmaxima
>
> Resistance to this idea is very strong, I sense. I am glad I have
> Mathematica and Maple too.
>
> Rich
>
>
>
> -----Original Message----- From: Richard Fateman
> Sent: Saturday, June 09, 2012 1:21 PM
> To: Richard Hennessy
> Cc: maxima at math.utexas.edu
> Subject: Re: [Maxima] parallel cell executing and wxmaxima
>
> On 6/9/2012 10:15 AM, Richard Hennessy wrote:
>> So in other words, you are not going to help. I have to do it all
>> myself. If that is what you are waiting for then you should just give
>> up. I am not going to do it myself.
> I didn't think you would. However, it sometimes pays to see what other
> people have done
> before speculating on what could be done. Parallelism in computer
> algebra is not a new
> idea. Languages that support parallelism is not new either.
>
> In fact I supervised a PhD student (Carl Ponder) working on this
> topic. about 15 years ago.
> The easy way to have parallelism in Maxima is to reduce an algebraic
> problem to a
> numerical problem which can be solved by a parallel subroutine in some
> library.
> If this has not been done, it could be done quite easily in most lisps
> with foreign function
> interfaces.
>
> For other ideas, look up Ponder's thesis or some of the conference
> proceedings on this
> topic.
>
> I'm not worried about Mathematica or Maple taking over the world because
> of parallelism.
>
>>
>>
>> -----Original Message----- From: Richard Fateman
>> Sent: Saturday, June 09, 2012 1:08 PM
>> To: Richard Hennessy
>> Cc: maxima at math.utexas.edu
>> Subject: Re: [Maxima] parallel cell executing and wxmaxima
>>
>> You are of course welcome to define and implement a new language that
>> is some kind of superset of
>> the Maxima language. If you can implement it on all systems that
>> already run Maxima, that would be impressive.
>>
>
> _______________________________________________
> Maxima mailing list
> Maxima at math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima