Subject: Re: [ublas] Testing the new bindings
From: Rutger ter Borg (rutger_at_[hidden])
Date: 2009-03-08 12:35:22
Jesse Perla wrote:
> Also, should I play with the higher level "solve" interface yet? And if
> there is discussion on the next high level interface, I put in my vote for
> the linear least squares and the SVD routines.
I think we'll have to deal with the auto-initializers in the low-level
bindings first - it is next on my todo list. I'm thinking of an "auto_init"
type, which can be passed to any low-level binding. E.g., the expression
gesv( a, auto_init(), b );
will cause ipiv to be initialized. If this is supported, the high-level
solve interface can start supporting optional parameter passing,
lapack::solve( a, b ); // calls XXsv( a, auto_init(), b );
lapack::solve( a, b, pivots(i) ); // calls XXsv( a, i, b );
and this could then be extended for all the other optional parameters
defined in the expert/divide-and-conquer/etc. drivers).