|
Boost : |
From: David Abrahams (david.abrahams_at_[hidden])
Date: 2001-08-20 11:45:21
----- Original Message -----
From: "Dave Steffen" <steffend_at_[hidden]>
> helmut.zeisel_at_[hidden] writes:
> > --- In boost_at_y..., Hayati Ayguen <Ayguen_at_b...> wrote:
> > Exactly.
> > However, if we wait for a matrix library that covers
> > all these issues, we will never have a standard C++ matrix interface.
>
> Right, and amen. Plus, the sooner we have _something_, the sooner we
> can all start improving it. ;-)
>
> > What is the actual state of MTL 3? If it is far enough, it might
> > be a good idea to use the interface of MTL 3 as a basic matrix
> > interface.
>
> This is my understanding: MTL 2.whatever (the currently available
> one) has been around for a while; someone out there (Andrew
> Lumsdaine?) is going to round up a bunch of patches and update the
> current release. I _think_ that we then proceed on, over here in
> Boost, to release 3
>
> Jeremy Siek is(AFAIK) the MTL expert 'round these parts. Jeremy, any
> comments or clarifications?
Jeremy has checked a prototype version of MTL 3 into CVS on the
matrix_development branch of boost/numeric. It is a significant departure
from MTL 2. There is also the ublas work of Joerg Walter and Mathias Koch
rooted at the same location in the CVS tree. The plan is to select the best
parts of each of these and merge them into a single library. I have been
planning to work on this, too. Incidentally, both efforts support expression
templates. The ublas stuff works with MSVC6. IMO, expression templates are
important; it's too easy to create unneccessary temporaries and too
difficult to write efficient generic matrix code without them.
-Dave
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk