[boost] Re: matrix classes in boost

From: Gabriel Dos Reis (Gabriel.Dos-Reis_at_[hidden])
Date: 2000-01-04 19:29:23

Martin Schürch <mschuerch_at_[hidden]> writes:

| Hi
| 
| I'm following silently the boost mailing list since a couple of days and I  like the
| idea off code as free as boost code.
| 
| Since a couple of months I'm playing with the idea to write a free simple matrix
| library. I know that there are
| many around already, but I didn't find one which fulfills the following
| characteristics:
| 
| 1) free also for commercial projects (e.g. MTL I think does not fulfill that)
| 2) available also for VC (e.g. not for blitz)
| 3) easy to use
| 4) fast but does not have to be as fast as e.g. MTL or Blitz

  0) General enought to be accepted by the computing community 
     (harder to achieve IMHO)

| My ideas:
| 1) based on valarray,... and slice, gslice -> std-implementors have to write the
| optimized low-level code
| 2) RowMajor and ColumnMajor mode chooseable at compile time via a trait class

Why just 'at compile time?'

| 3) LU and SVD decomposition classes
| 4) a set of relating functions e.g det, trace, product, solve
| 5) perhaps using some BLAS routines
| 
| -> similar to Rogue Waves math++ but with concret classes
| 
| The user in mind would be people who need a simple and stable matrix class for
| average
| numerical tasks. I think there are a lot of them.
| 
| Are there any ultimate libraries I missed?

Did you browse OON?

| Is this outside the direction of boost?

I think such a project needs vaste involvement from the numerical
computing community, not just the 'private' boost subscribers :-).

-- Gaby