
Ublas : 
Subject: Re: [ublas] [bindings][lapack] Initial highlevel solve
From: Jesse Perla (jesseperla_at_[hidden])
Date: 20090324 13:14:04
Looks great. One thing that would be nice is if the B could support vectors
as well as matrices. Quite often you want to solve AX=b with b a ublas
vector and don't want to convert between the two types. I am not sure if
there is the ability for the traits to treat a vector as an Nx1 matrix. I
assume this could use some kind of matrix adaptor or decoration to say
whether it is an Nx1 vs. 1xN for conformability.
I assume that it will be based on some sort of traits similar to
http://lists.boost.org/MailArchives/ublas/2009/03/3384.php
<http://lists.boost.org/MailArchives/ublas/2009/03/3384.php>It would be
nice if there was an equivalent decoration to these traits for calling this
function so the user isn't forced to use a bunch of adaptors with every
call.
I look forward to testing!
On Sun, Mar 22, 2009 at 7:08 AM, Rutger ter Borg <rutger_at_[hidden]> wrote:
>
> Guys,
>
> I've created an initial prototype for a highlevel interface to solve. It
> uses Boost.Parameter to allow a user to pass parameters by keyword;
> lowlevel routines create a temporary pivot vector if not provided and
> needed by that routine.
>
> See below for the syntax, all expressions are valid.
>
> #include <boost/numeric/bindings/lapack/solve.hpp>
>
> using namespace lapack::keywords;
>
> // basic usage
> lapack::solve( A, B );
> lapack::solve( A, B, _pivot = ipiv );
>
> // this is also valid
> lapack::solve( _B = B, _A = A );
> lapack::solve( A, _pivot = ipiv, _B = B );
>
> What we could use next is a good set of keywords/type decorators for
> addressing all solve routines and expert driver options. Suggestions are
> very welcome.
>
> Cheers,
>
> Rutger
>
>
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
>