Re: C++ Matrix & Linear Algebra library

Dear Evgenii,

I think I'm still on the oonumerics maillist, but as you say, there is
no more traffic. On this list used to be a reasonable effort in
gathering all projects related to Object Oriented Numerics (not only
linear algebra). But this has quickly died out. If I'm not mistaken, it
was the author of Blitz who set up this list/webpage, Todd Veldhuizen.

It seems that, even several years after the introduction of C++ in
scientific computing, people have moved back to "old style" Fortran
because "the software is already there and it works" or they go for
"full option packages" like Matlab, Mathematica, Maple,...

Fortran90 and beyond try to be a bit OO, but I'm not convinced. When I
wrote my Master thesis (about 10 years ago), I had to code a numerical
algorithm. My professor gave me the choice between C++ and Fortran90 and
it was up to me to find the necessary linear algebra libraries. At that
time, the choice was quickly made: there was no decent (stable,
tested,...) linear algebra package in C++ that I could use, so I went
for Fortran90 and LAPACK. Worst six months of my (coding) life... Three
compilers (IBM, Sun, SGI IRIX) often gave me three different compiler
outputs (some warnings, some errors, but not consistent among the three
compilers)...

I've also tried SciPy (I'm a big fan of Python), but I don't have the
time right now to look into it in more detail.

I wouldn't count on the umbrella pages anymore. When OON went frozen, I
didn't see anybody pick it up. I used to visit E. Baum's site often
because he gathered an index of all free scientific software available.
But that went down also some time ago (years already maybe). Google and
this newsgroup seem to be the only repository of information...

bye,
gert

On 08/21/2008 08:24 PM, Evgenii Rudnyi wrote:
On Aug 21, 3:09 am, Rob McDonald <rob.a.mcdon...@xxxxxxxxx> wrote:
Gert & Evgenii,

Thanks for the great links. It seems that there are a lot of projects
out there that are difficult to find.

If we look at the list

http://www.oonumerics.org/mailman/listinfo.cgi/oon-list/

there is almost no traffic there. It seems that people lost interest
for something like this.

Well, the mainstream development goes through the use of interpreters
like Matlab, Mathematica and others. Recently I have tried SciPy - it
looks really nice. From this viewpoint it is not that clear the role
of the library in C++.

Hopefully some of the umbrella
pages which keep track of this sort of thing will try to get back up-
to-date.

Recently I have started

http://MatrixProgramming.com

but I have overestimated my spare time and it is not that much there.

Evgenii
.

Relevant Pages

  • Re: Python arrays and sting formatting options
    ... If fortran90 you would just do: ... How would the python equivalent go? ... You would drag yourself out of the 1960s, install numpy, and ... (Google scipy), which gives you arrays and matrices. ...
    (comp.lang.python)
  • Re: Books for efficient Fortran90 programming
    ... sunnyabc wrote: ... I have been looking for book to improve the speed of my Fortran90 ... An interesting question is which practices can be expected to be useful across a range of the better compilers, and whether a range of hardware architectures is to be addressed. ... This is a moving target, as the quality of optimizations in those compilers improves, and usage of current syntax in important benchmarks increases. ...
    (comp.lang.fortran)