From: Paul C. Leopardi (leopardi_at_[hidden])
Date: 2006-02-16 06:32:13
Since the GluCat classes ( http://glucat.sf.net ) were designed to be drop-in
replacements for float, double, complex<>, and since it is easy to create
multivectors which do not commute, it may be possible to adapt an existing
uBLAS test suite for use with GluCat, and see whether the assumption of
commutatitity is used in practice in uBLAS. Is there a uBLAS test suite which
could be adapted in this way?
On Thursday 16 February 2006 21:01, Gunter Winkler wrote:
> Vardan Akopian wrote:
> >A quick note and a question.
> >For block matrices to be possible with ublas, we also need to make
> >sure that there are no implicit assumptions of the element (scalar)
> >product being commutative, i.e. a*b = b*a.
> >Can anyone confirm or deny the presence for these kind of assumptions?
> Yes and no.
> First, ublas always assumes that the product of two scalars is commuative.
> Second, most algorithms does not use the fact. Over the last few years
> this issue was discussed several times, but we always decided to require
> products of scalars to be commutative because dropping this assumption
> would force us to check every algorithm whether it fails or not. This
> would be a lot of work and any volunteer is welcome ...
> Personally, I would love to have a native ublas block sparse matrix -
> but I do not have the time to do the implementation. I think that GLAS
> will provide such classes, but I am not involved in this developement.