Some comments from a user:
This may be a longshot, but I would like maxima to be a free program
with most of the "bells and whistles" of programs like Maple and Mathematica.
The advantages for education and research are obvious. Also, the
availability of the source code in a convenient place could help
budding programmers and developers in writing new code.
If it is possible to simply call up the LAPACK routines in maxima when
one is needed, then please tell me how to do it. An example would be
most welcome. For instance, how about an example which finds the
top 10 eigenvalues (assuming all are non-zero) of a 50x50
non-symmetric sparse matrix?
In solving certain research problems, LAPACK is very useful. If one
simply says that this or that implementation of maxima does not have
LAPACK, then
either
(a) that sends a (potentially large user group) to other
venues, or
(b) one is forced to develop one's own code for this or that
problem. This has to be done in many cases anyway, but it
should not have to be done for standardized code such as the LAPACK
routines. This is re-inventing the wheel, and a huge waste of
time. Also, the code, even if contributed, would most likely
have the optimization of that produced by experienced
programmers.
For those who have resources and experience with many types of
languages (maybe even make a living in that) it is no problem to pull
up this or that appropriate tool for this or that problem.
But, most users don't get to that situation. They have specific
problems and want solutions in a small set of packages.
I think the modular aspect of maxima is good enough so that it does
not have to become bloated. Packages can be loaded as needed. So, I
put my vote in for the inclusion of LAPACK in the source, either as
generic lisp code or as maxima packages.
--Just my two-cents (or maybe half a cent).
-sen
On Sat, 23 Dec 2006, Richard Fateman wrote:
> I kind of disagree here.
>
> Instead of inserting LAPACK source into the Maxima source, how about
> pointing to a few binary forms, one for each plausible host environment,
> with a suitable interface. Perhaps we could even point to matlisp, which
> apparently solves the problem of the foreign function interface for cmucl
> and allegro.
> So what is needed is CLISP and GCL. Or just say that LAPACK is not available
> for certain implementations. (The set of features of Maxima in my version
> of Allegro CL is missing other pieces, like plotting; on the other hand, it
> has voice output as an option. LAPACK need not be universal.)
>
> If someone wants LAPACk for GCL, it would require coming up with suitable
> interface code and suitable binary (.dll, .so, .lib ??). It seems to me that
> this is a GCL project, suitable for someone to do who knows almost nothing
> about maxima. Perhaps it should be a CFFI implementation for GCL.
>
> There are many other programs in netlib that could potentially be of
> interest to maxima users, and for someone on this list to speculate on what
> is most useful and spend time converting or interfacing such a program --
> without an application in mind -- seems to have the potential for a major
> distraction. If the same person could, instead, figure out a way of
> improving maxima in a way that is more nearly unique -- ie provides a
> symbolic mathematical facility that is not available in netlib or any other
> programming framework... or perhaps is only available in proprietary form in
> Mathematica or Maple... that seems to me to be more productive.
>
> If people are lacking in ideas of what symbolic facility might be possible,
> there are items in this mailing list that have come up. Or alternatively,
> just browse through the (Free, online) mathematica or maple documentation
> and find something neat that maxima does not yet do. And either mimick the
> design, or perhaps improve on it.
>
> Or look at the potential project list for a class in algebraic manipulation
> are posted near the end of this document...
> http://www.cs.berkeley.edu/~fateman/282/hand1.pdf
>
> My suggestion though: if you have no more math background than a year of
> freshman calculus, your chance of making a thoughtful contribution will be
> considerably lower than if you have a more sophisticated background in math.
> This background can be obtained through studying some of the literature or
> "the competition". While enthusiasm and energetic programming effort are
> necessary, they are rarely sufficient to overcome a naive design of symbolic
> math facilities.
>
> RJF
>
>
>
>> -----Original Message-----
>> From: maxima-bounces at math.utexas.edu
>> [mailto:maxima-bounces at math.utexas.edu] On Behalf Of Leo
>> Sent: Saturday, December 23, 2006 7:04 AM
>> To: maxima at math.utexas.edu
>> Subject: Re: [Maxima] LAPACK with maxima
>>
>> * Robert Dodier (2006-12-22 23:21 -0700) said:
>> ^^^^^^^^^^^^^
>>> On 12/22/06, Raymond Toy <raymond.toy at ericsson.com> wrote:
>>>
>>>> It's just very big, that's all. I haven't and don't plan on
>>>> creating interfaces to all the functions. Eigenvalue/eigenvector
>>>> and SVD seem useful, so I've done those.
>>>>
>>>> Perhaps it can be a share or contrib package, but I'm not even sure
>>>> it should be checked into maxima at all.
>>>
>>> If you mean some or all of LAPACK -- yes, I believe it should be
>>> committed to cvs.
>>
>> Is this similar to what http://matlisp.sourceforge.net/ does?
>>
>>>
>>> I believe it is important and widely used, so it should in a
>>> subdirectory of src (not share). We can work on creating interfaces
>>> apart from just getting the f2cl output loaded into Maxima
>>>
>>> Thanks for working on this problem,
>>>
>>> Robert
>>
>> --
>> Leo <sdl.web AT gmail.com> (GPG Key: 9283AA3F)
>> _______________________________________________
>> Maxima mailing list
>> Maxima at math.utexas.edu
>> http://www.math.utexas.edu/mailman/listinfo/maxima
>>
>
> _______________________________________________
> Maxima mailing list
> Maxima at math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima
>
--
---------------------------------------------------------------------------
| Sheldon E. Newhouse | e-mail: sen1 at math.msu.edu |
| Mathematics Department | |
| Michigan State University | telephone: 517-355-9684 |
| E. Lansing, MI 48824-1027 USA | FAX: 517-432-1562 |
---------------------------------------------------------------------------