Re: [ublas] vector * Matrix is slow compared to hand written code

From: Gunter Winkler (guwi17_at_[hidden])
Date: 2007-06-29 15:19:10

Am Freitag, 29. Juni 2007 12:53 schrieb Markus Weimer:
> Hi,
>
> we just did some experiments on the following case, which occurs
> often in our code:
>
> v = prod(u,M)

use axpy_prod(A,x,y,flag) from operation.hpp. This is optimized for 
maximal speed

simple version:
axpy_prod( trans(M), u, v, false ); // v = M^T * u

advanced version:

compressed_matrix<double, row_major> M;

v.clear();
axpy_prod( M, u, v, colum_major_tag );
// yes, column_major_tag for a row_major matrix is intentional!

mfg
Gunter

PS: Check my homepage
http://www.guwi17.de/ublas/index.html