Re: [ublas] Testing getri

Subject: Re: [ublas] Testing getri
From: Rutger ter Borg (rutger_at_[hidden])
Date: 2009-03-22 12:49:46

Jesse Perla wrote:
> 
> Please tell me if I am missing out on the triangular traits classes and
> when I should test these two functions otherwise
> Also:  Do you expect the interfaces for the trtri and trtrp to change
> much?
>  For example, should I expect the 'N' vs. 'D' parameter to disappear as we
> detect the ublas::lower vs. ublas::unit_lower?
> 
> Thanks guys,
> Jesse

Dear Jesse,

As far as I can tell, there is no support (yet) for triangular matrices in
the traits system. 

W.r.t. the interfaces, I think we could move the uplo-tags detection to the
free functions, enabling us to offer both interfaces, with and without the
uplo/diag arguments. I think we could end up with low-level drivers that
support both type-decorated matrices and true-triangular matrices. I.e.,
something like 

some_func( _upper(A), ... );
some_func( U, ... );

where A is not a triangular type, but U is, "_upper" is a keyword which
type-decorates A. _upper(A) is not a reference, so we can use this to
distinct these forms:

some_func( matrix& A, .. ) {
 // use traits tags for uplo, diag
}

some_func( matrix const& A, ... ) {
 // use type decoration tags for uplo, diag
}

Cheers, 

Rutger