Re: a Guile linear-algebra package, RFC

[Lars Arvestad <arve@nada.kth.se>]

  CL> 1) Is anyone interested in this package (if no one is
  CL>    interested, maybe I should not bother).

Yes! I had planned to sit down this afternoon and put together my own
little package for simple matrix operations when I read your
message. Now I'll probably wait a little...

  CL> 2) This code has been used to support a couple of my own
  CL>    personal projects, but is not yet widely used or fully
  CL>    tested. The underlying FORTRAN libraries are reliable, but it
  CL>    is possible that there are bugs in the code which calls those
  CL>    libraries which could result in incorrect results.  Should I
  CL>    let people play with it at their own risk and help me
  CL>    flush-out bugs bazzare-style, or is this simply too risky for
  CL>    numerical code.  Maybe I should wait until I can test it more
  CL>    fully (a month or two) to reduce the risks?  At minimum, I
  CL>    will mark the routines which I haven't used much so that they
  CL>    can be inspected before they are tried and trusted.

I vote for publishing it with warnings.

  CL> 4) Should I assume that people know what BLAS and LAPACK are?
  CL>    Can I let people find these themselves or should I try to
  CL>    write-up some instructions for finding/compiling/installing
  CL>    them (I'm not an expert on this, but I've done a little).
  CL>    CL> 5) Does anyone have serious problems with column-major
  CL>    matrix representation?  I chose to be compatible with the
  CL>    FORTRAN convention because it simplified the project as a
  CL>    whole.

I've heard of LAPACK, but not BLAS. I don't think you can assume that
people know these. I think you should include pointers to where to get 
them. I would write instructions like: "Get BLAS (URL) and LAPACK
(URL) and install them. Then install my lib."

  CL> 6) Should I include only a subset of the above libraries?

Is it easy to break up in parts? It would be cool to be able to choose 
the components one needs.

  CL> 7) Would people be willing to port this to their own systems and
  CL>    send me patches to make the code more universal?  I've set-up
  CL>    the autoconf scripts so that it works on my Irix 6.3 and
  CL>    Debian Linux 2.0 machines, but the way FORTRAN libraries are
  CL>    used on these systems is very different.  Tweaking
  CL>    configure.in for other systems may take a few tries.

Yes. 

  CL> 6) I don't make much use of LAPACK yet.  LAPACK-related code be
  CL>    in a separate library so that people without LAPACK can use
  CL>    the BLAS-only stuff?  (Same with FFTPACK?)

That would be great. 

Lars

--
Lars Arvestad
Dept. of Numerical Analysis and Computing Science
Royal Institute of Technology (KTH), Stockholm, Sweden