GSL, blas and Mathematica

GSL, blas and Mathematica

[Alberto Simões/EPL]

Hi.

I'm developing a mathematica alike program for linux (GSL) named 
numexp (numexp.sf.net). At the moment, it is only a numeric processor
(not algebraic functionalities at all (almost)).

We are using GSL to implement a bunch of functions. I've read something
about blas, but I don't have certain I understood it... is it realted
to algebraic processment? If not, where can I get good documentation
on how to implement this hard stuff?

Thanks
Alberto
-- 
Departamento de Informatica - Universidade do Minho

"The C Programming Language -- A language which combines the
flexibility of assembly language with the power of assembly language."

Re: GSL, blas and Mathematica

[Robert G. Brown]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: TEXT/PLAIN; charset=X-UNKNOWN, Size: 2908 bytes --]

On Fri, 18 Oct 2002, Alberto [iso-8859-1] Simões/EPL wrote:

> 
> Hi.
> 
> I'm developing a mathematica alike program for linux (GSL) named 
> numexp (numexp.sf.net). At the moment, it is only a numeric processor
> (not algebraic functionalities at all (almost)).
> 
> We are using GSL to implement a bunch of functions. I've read something
> about blas, but I don't have certain I understood it... is it realted
> to algebraic processment? If not, where can I get good documentation
> on how to implement this hard stuff?

BLAS stands for Basic Linear Algebra System -- it is an API for
low-level routines that perform "atomic" linear algebra functions, from
which more complex linear algebra operations can be build.  In order to
proceed optimally efficiently, a lot of those operations need to be
tuned (both blocksize and algorithm) to the operational hardware on the
system so that things like cache hits and fetch times can be optimized.
This is the basis for the ATLAS (automatically tuned linear algebra
system) which basically builds a special BLAS library that can be used
in LAPACK (a standard library for higher order linear algebra calls) or
GSL (which already lets you specify a blas library IIRC) and other
things.

A good place to learn about this sort of stuff online is at the oak
ridge netlib repository.  Jack Dongarra (onetime author of PVM) is also
one of the co-parents of ATLAS (written mostly by his student/colleague
Clint Whaley), and there is a lot of documentation there on BLAS and
linear algebra libraries in general.  This group is Very Bright and has
been for some time.

Netlib (netlib.org, follow link to browse the repository) has a blas
link with all sorts of stuff -- references, fortran sources, and more.
Ditto atlas.  Ditto, in fact, a ton of stuff -- netlib has been around
long enough that I've accessed it with ftp, gopher, its own custom
interface, and the web.  An unfortunate bias towards fortran libraries,
perhaps, but still a lovely place to look for certain things.

I for one would love to see a clone of mathematica good enough to
completely replace it, built on top of the GSL.  Good luck.  I think a
lot of mathematica (and formac, maple, and the various algebraic
manipulation languages over the ages) comes down to token parsing and
lexical analysis, though, more than application of numerical algorithms
-- they are in a sense the "easy" part.

   rgb

> 
> Thanks
> Alberto
> -- 
> Departamento de Informatica - Universidade do Minho
> 
> "The C Programming Language -- A language which combines the
> flexibility of assembly language with the power of assembly language."
> 

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb@phy.duke.edu

Re: GSL, blas and Mathematica

[James Amundson]

On Fri, 2002-10-18 at 12:43, Robert G. Brown wrote:
> I for one would love to see a clone of mathematica good enough to
> completely replace it, built on top of the GSL.  Good luck.

Impolite though it may be to mention another project in response to this
query, please let me point out that Maxima, <http://maxima.sf.net>, is a
mature GPL'ed mathematics program. We don't utilize GSL yet, but there
are plans to do so in the future.

--Jim Amundson

Evolution .. Survival of the fittest ...

[Manoj Warrier]

Hi ppl ...

First of all a BIG thanks to the GSL team (I use it in my research).

I used to wonder why there exist so many programs that do allmost
similar things? For example (octave, Scilab, RLAB, etc .... (MATLAB
like). Then U have MuPad, Maxima, YACAS, etc doing a fair amount of
symbolic computation (which I naively refer to as mathematica like).
A more comprehensible list is available at scilinux.sf.net/mathpack.html
Possible reasons are:
1) They start off covering an area not addressed by the other package,
   (but then spread to cover the areas covered by other software).
2) License dissatisfactions,
3) Coding styles,
4) Programming satisfaction
..... N) etc.

Initially it seems that it is better if all combine and develop
a better product together, but on reflection it seems that an evolutionary
model might be better and we might get a decent software which will
be better than the best ...

I also wonder how the GSL interface to PYTHON, and also the numeric
PYTHON extensions has postured PYTHON as a good alternative to these
various MATLAB like mathematical packages. Of course it is also a far
cry from doing symbolic computations like Mathematica or MuPAD does.
I have been promising myself since a long time to check out Maxima.

Why is all this relevant to this list? Well, it shows how a nice numerical
library (GSL) can be the heart of many mathematical packages.

Manoj Warrier (manoj.warrier@ipp.mpg.de)

Stellaratortheorie, Max-Planck Institut Fur Plasmaphysik
TeilInstitut Greifswald Wendelsteinstrasse 1
D-17491 Greifswald Germany Tel: +49-3834-882434

--------- History of Computing 10-11-3003 ---------------
Then there used to be this great user friendly OS which
overwrote your MBR whenever you installed it.
---------------------------------------------------------

On 18 Oct 2002, James Amundson wrote:

> On Fri, 2002-10-18 at 12:43, Robert G. Brown wrote:
> > I for one would love to see a clone of mathematica good enough to
> > completely replace it, built on top of the GSL.  Good luck.
>
> Impolite though it may be to mention another project in response to this
> query, please let me point out that Maxima, <http://maxima.sf.net>, is a
> mature GPL'ed mathematics program. We don't utilize GSL yet, but there
> are plans to do so in the future.
>
> --Jim Amundson

Re: Evolution .. Survival of the fittest ...

[Christos Siopis]

[ Apologies for the long posting; I hope it is relevant to this list ]

Manoj,

Thank you for posting this!

I would also like to express my gratitude to the coders of GSL, Python,
Numeric Python and other open source/free numeric software. I have not
written a single line to contribute to these projects, so some might say i
am not entitled to say what i am about to say, but i nevertheless will to
emphasize the user's perspective, and out of interest for the community.

I would like to "complain" about the large degree of "degeneracy" in open
source numeric projects: many projects overlap substantially, not (alas!)  
in terms of code base but in terms of intent. As an example, the Python
interface to GSL comes to mind, which (as a goal) overlaps with the
Numeric Python project. There are several other cases, of course.

The end result is that we have LOTS of programs doing all sorts of things,
but not a SINGLE (or a few) well-done application that can do ALL three
main things the numeric researcher needs today: "array"/vector
calculations, "professional"-grade graphics, and library support for a
large number of standard problems. Commercial programs such as IDL and
Matlab come to mind (which doesn't mean they are perfect or that we cannot
do better). Symbolic support would probably be nice, but i wouldn't touch
this before the other three components are done!

You may think (as Manoj put it) that GSL could function as the basis for
what Numeric Python tries to do. Well, i am not so sure! For instance: how
well could GSL work with Numeric array objects? How many or the array-
oriented operations and ufuncs of Numeric could be implemented in terms of
GSL functions? I suspect that GSL was "made for" serial computations, and
adapting it to the vector environment of Numeric would not be efficient,
except perhaps for some high-level operations.

Does this mean that GSL is redundant? Not at all. It intends to fill in a
free software gap in the niches where NAG, IMSL or the numerical recipes
would otherwise be used. But one can't help wondering if GSL could have
started in a way that other projects, such as Numeric, could benefit from
all the effort that went into it. I believe that one of the nice things
about GSL, compared to using some other free (beer) alternative like
Netlib, is the uniformity in the interface and, perhaps, the expectation
that a similar level of attention went to most functions, and that it is
better maintained than a Netlib program which was "dumped" by its author
some time ago without anyone taking care of it or its bugs any more. In
other words, there is a community behind GSL, which will do a better job
maintaining GSL as long as it does not split its attention among many
GSL-like projects!

Now take a look at Numeric Python: they also want to add much of the
functionality that GSL has or intends to have. They do it in a way which,
in my very humble opinion, is more "anarchic" than GSL, with people
writing their code from scratch or "adapting" code from Netlib and other
projects. There is less consistency in the API of these functions, not to
mention things like error reporting and propagation. Would it not be nice
to be able to use GSL as the foundation for Numeric Python?

Then we have the PyGSL project. Again, i thank the person who started it,
and i do think that there are some cases where it may be useful. But why
not use a full-fledged language like Python to take care of the syntactic
stuff and, instead, re-invent the wheel? Why not put the effort currently 
going into PyGSL to some better (for the numerical community) use?

Of course, there is no way, and probably no sense, in stopping people from
starting their own projects whenever they feel like it. In fact many do it 
not so much for the sake of the project as for the sake of e.g. learning a 
programming language or environment. Fine.

But we should also understand that the "amateur" numerical community's
resources are limited, perhaps more limited than the resources of the
software engineering community that created and maintains GNU/linux. This
may explain why, while that community has managed to create a full-fledged
operating system with all sorts of utilities, the scientific community,
after 7+ years of efforts, still does not have a "standard" platform that
includes production-quality vector computing, graphics and library
support.

I think it is important for our community to decide (somehow) on a minimum
number of "core projects" to work on; e.g., a science library like GSL, a
core plotting/graphics language (possibly based on one of many such
libraries floating around), and a hierarchy of array objects. Once the
basic stuff is there, different communities can then modularly build on
top of it, based on their respective needs (and again, Python is probably
the perfect language for that).

This would not just be a free (as in beer and as in speech) infrastructure
for the numerical scientist to use; it would also offer a much greater
level of transparency (no black boxes!) and consistency: when enough
people use e.g. its FFT transform module, the module will not only
eventually be bug-free, but its overall behaviour would be known and
understood. So whether i link to it through my C code or i use it through
Python's integrated environment, i will expect the same behavior. I will
not wonder, with every new package i use, whether its FFT transforms work
well, because they will all use that same FFT module. The more useful this
environment becomes, the more people will use it, the more people will
contribute to it, and the more useful it will become. This would be a
situation almost competely reverse of the current, fragmented landscape.

My programming skills are not good enough to contribute to the foundations
of this project. In fact, i do not know if the abilities of most numerical
scientists are sufficient. This may be an area where some government
agency should lay the seeds by hiring some professional team to get it
started. But once the overall foundation is there, there are things that i
(and most other numerical scientists) could contribute.

So i would like to ask: is there ANY chance at all that such a project
might one day become reality? Or will we have to forever use the IDL's and
Matlab's of this world?

***************************************************************
/  Christos Siopis              |                             \
/  Postdoctoral Research Fellow |    ''CARPE NOCTUM''         \
/  Department of Astronomy      |_____________________________\
/  University of Michigan       |                             \
/  Ann Arbor, MI 48109-1090     | E-mail : siopis@umich.edu   \
/  U.S.A.             __________|                             \
/                    / http://www.astro.lsa.umich.edu/~siopis \
***************************************************************


On Sat, 19 Oct 2002, Manoj Warrier wrote:

> Hi ppl ...
> 
> First of all a BIG thanks to the GSL team (I use it in my research).
> 
> I used to wonder why there exist so many programs that do allmost
> similar things? For example (octave, Scilab, RLAB, etc .... (MATLAB
> like). Then U have MuPad, Maxima, YACAS, etc doing a fair amount of
> symbolic computation (which I naively refer to as mathematica like).
> A more comprehensible list is available at scilinux.sf.net/mathpack.html
> Possible reasons are:
> 1) They start off covering an area not addressed by the other package,
>    (but then spread to cover the areas covered by other software).
> 2) License dissatisfactions,
> 3) Coding styles,
> 4) Programming satisfaction
> ..... N) etc.
> 
> Initially it seems that it is better if all combine and develop
> a better product together, but on reflection it seems that an evolutionary
> model might be better and we might get a decent software which will
> be better than the best ...
> 
> I also wonder how the GSL interface to PYTHON, and also the numeric
> PYTHON extensions has postured PYTHON as a good alternative to these
> various MATLAB like mathematical packages. Of course it is also a far
> cry from doing symbolic computations like Mathematica or MuPAD does.
> I have been promising myself since a long time to check out Maxima.
> 
> Why is all this relevant to this list? Well, it shows how a nice numerical
> library (GSL) can be the heart of many mathematical packages.
> 
> Manoj Warrier (manoj.warrier@ipp.mpg.de)
> 
> Stellaratortheorie, Max-Planck Institut Fur Plasmaphysik
> TeilInstitut Greifswald Wendelsteinstrasse 1
> D-17491 Greifswald Germany Tel: +49-3834-882434
> 
> --------- History of Computing 10-11-3003 ---------------
> Then there used to be this great user friendly OS which
> overwrote your MBR whenever you installed it.
> ---------------------------------------------------------
> 
> On 18 Oct 2002, James Amundson wrote:
> 
> > On Fri, 2002-10-18 at 12:43, Robert G. Brown wrote:
> > > I for one would love to see a clone of mathematica good enough to
> > > completely replace it, built on top of the GSL.  Good luck.
> >
> > Impolite though it may be to mention another project in response to this
> > query, please let me point out that Maxima, <http://maxima.sf.net>, is a
> > mature GPL'ed mathematics program. We don't utilize GSL yet, but there
> > are plans to do so in the future.
> >
> > --Jim Amundson

Re: GSL, blas and Mathematica

[Brian Gough]

James Amundson writes:
 > On Fri, 2002-10-18 at 12:43, Robert G. Brown wrote:
 > > I for one would love to see a clone of mathematica good enough to
 > > completely replace it, built on top of the GSL.  Good luck.
 > Impolite though it may be to mention another project in response to this
 > query, please let me point out that Maxima, <http://maxima.sf.net>, is a
 > mature GPL'ed mathematics program. We don't utilize GSL yet, but there
 > are plans to do so in the future.

I encourage everyone to use Maxima. It is a very powerful program.
For people who want a free symbolic algebra system there is no longer
any reason to use proprietary software.

Brian

Re: GSL, blas and Mathematica

[Alberto Simões/EPL]

On Sun, Oct 20, 2002 at 01:55:33PM +0100, Brian Gough wrote:
| James Amundson writes:
|  > On Fri, 2002-10-18 at 12:43, Robert G. Brown wrote:
|  > > I for one would love to see a clone of mathematica good enough to
|  > > completely replace it, built on top of the GSL.  Good luck.
|  > Impolite though it may be to mention another project in response to this
|  > query, please let me point out that Maxima, <http://maxima.sf.net>, is a
|  > mature GPL'ed mathematics program. We don't utilize GSL yet, but there
|  > are plans to do so in the future.
| 
| I encourage everyone to use Maxima. It is a very powerful program.
| For people who want a free symbolic algebra system there is no longer
| any reason to use proprietary software.

Buaaa! And I was thinking my project could be usefull to someone ;(

Goos work...
Alberto Simões

-- 
Departamento de Informatica - Universidade do Minho

"The C Programming Language -- A language which combines the
flexibility of assembly language with the power of assembly language."

Re: GSL, blas and Mathematica

[Brian Gough]

Alberto Simões/EPL writes:
 > Buaaa! And I was thinking my project could be usefull to someone ;(
 > 
 > Goos work...
 > Alberto Simões
 > 

Maybe you can link to maxima for any symbolic algebra.