dolfin team mailing list archive

["Garth N. Wells" <g.n.wells@xxxxxxxxxx>]

On Thu, 2006-04-06 at 10:34 -0500, Anders Logg wrote:
> On Thu, Apr 06, 2006 at 05:20:00PM +0200, Garth N. Wells wrote:
> > I've created a NewMatrix template to allow a matrix to be created which
> > is either dense or sparse. Some benchmarks are in src/test/main.cpp for 
> > 
> >   NewMatrix<DenseMatrix> A;
> > 
> > There is no overhead when operating on A with respect to directly
> > creating a uBlas matrix and operating on it.
> 
> Very good!
> 
> The only drawback as I see it compared to the envelope-letter design
> is that we can't change the storage dynamically, but we can probably
> live without that.
> 
> It should be enough if we provide functionality for converting from
> one type to the other.
> 
> > I suggest putting the necessary functions which are common to dense and
> > sparse matrices into NewMatrix, and changing the name of Matrix to
> > SparseMatrix, and creating a similar template for vectors. Any
> > objections or suggestions?
> > 
> > Garth
> 
> No objections. This sounds like a very good solution.
> 
> One of the functions we need to have is Matrix::add() for assembly.
> 
> Is it possible to set a default type? So one would do just
> 
>     Matrix A;
> 
> to get a sparse matrix and would need to do something different to get
> a dense one. I'd prefer something else than Matrix<DenseMatrix>,
> perhaps
> 
>     Matrix<dense> A;
> 
> if possible.
> 

This is possible using typedef.

Garth

> Or we could rename the DenseMatrix and SparseMatrix classes to
> something like
> 
>     DenseMatrixImplementation / SparseMatrixImplementation 
>     DenseMatrixStorage / SparseMatrixStorage
>     DenseMatrixReal / SparseMatrixReal
>     other options?
> 
> and then do
> 
>    typedef MatrixInterface<DenseMatrixImplementation> DenseMatrix;
>    typedef MatrixInterface<SparseMatrixImplementation> SparseMatrix;
>    typedef MatrixInterface<SparseMatrixImplementation> Matrix
> 
> where MatrixInterface is the name of the template NewMatrix after we
> rename it.
> 
> This would make it easy for the user to create matrices, but it would
> make life harder for us when we write functions that take a general
> matrix (sparse or dense) as an argument, but we can probably find a
> nice solution for that too...
> 
> /Anders
> 
> 
> > 
> > On Thu, 2006-04-06 at 16:50 +0200, DOLFIN wrote:
> > > One or more new changesets pushed to the primary DOLFIN repository.
> > > A short summary of the last three changesets is included below.
> > > 
> > > changeset:   1845:4c75b74c95a56c23f1bb6d073c7e8261515a2ffd
> > > tag:         tip
> > > user:        "Garth N. Wells <g.n.wells@xxxxxxxxxx>"
> > > date:        Thu Apr  6 16:49:50 2006 +0200
> > > files:       configure configure.ac src/kernel/la/DenseMatrix.cpp src/kernel/la/Makefile.am src/kernel/la/Makefile.in src/kernel/la/dolfin/DenseMatrix.h src/kernel/la/dolfin/Makefile.am src/kernel/la/dolfin/Makefile.in src/kernel/la/dolfin/dolfin_la.h src/test/main.cpp
> > > description:
> > > Create template class NewMatrix.
> > > 
> > > 
> > > changeset:   1844:6b24c4aa352caa578d97cf8eee398793f9a65499
> > > user:        "Anders Logg <logg@xxxxxxxxx>"
> > > date:        Wed Apr  5 16:37:57 2006 -0500
> > > files:       configure configure.ac
> > > description:
> > > Remove -Werror when compiling with --enable-optimization
> > > 
> > > 
> > > changeset:   1843:a665e3cd78c7ec75ef59cf6f61f26c24630a1f22
> > > parent:      1842:05da1868689e5ec15c64172ce8f8d56ab9a65ad0
> > > parent:      1841:80c6859e328196bae99a9f15d475a3fae5b60f83
> > > user:        "Garth N. Wells <g.n.wells@xxxxxxxxxx>"
> > > date:        Wed Apr  5 14:48:52 2006 +0200
> > > files:       
> > > description:
> > > merge
> > > 
> > > 
> > > -------------------------------------------------------
> > > For more details, visit http://www.fenics.org/hg/dolfin
> > > 
> > > _______________________________________________
> > > DOLFIN-dev mailing list
> > > DOLFIN-dev@xxxxxxxxxx
> > > http://www.fenics.org/cgi-bin/mailman/listinfo/dolfin-dev
> > 
> > 
> > _______________________________________________
> > DOLFIN-dev mailing list
> > DOLFIN-dev@xxxxxxxxxx
> > http://www.fenics.org/cgi-bin/mailman/listinfo/dolfin-dev
> > 
> 
> -- 
> Anders Logg
> Research Assistant Professor
> Toyota Technological Institute at Chicago
> http://www.tti-c.org/logg/
> 
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/cgi-bin/mailman/listinfo/dolfin-dev