|
|
Ublas : |
Subject: Re: [ublas] [bindings] lapack::lange: where is the computed norm?
From: Karl Meerbergen (Karl.Meerbergen_at_[hidden])
Date: 2010-07-23 04:21:59
Hi,
A more fundamental question is, when we would like to use LAPACK norm
functions. Are they more efficient than c++ codes?
For BLAS-1 functions, e.g., there is no advantage (at least in my
experience), and the c++ (ublas, mtl, eigen, glas,...) are easier to use.
Thanks
Karl
Rutger ter Borg wrote:
> Marco Guazzone wrote:
>
>
>> Hi,
>>
>> In auxiliary, there are a family of ?LAN?? functions
>> http://www.netlib.org/lapack/lug/node144.html
>>
>> used for computing the norm of a matrix.
>> These functions differ from the type of matrix they work with.
>> For instance, ?LANGE works with general matrices, while ?LANGB work
>> with general banded matrices and ?LANHB works with Hermitian banded
>> matrices.
>> It appears that all these functions are FORTRAN FUNCTION returning some
>> value. In bindings, it seems they are implemented as FORTRAN SUBROUTINE
>> (i.e., returns a std::ptrdiff_t instead of the value computed by the
>> function).
>>
>>
>
>
> It requires adjusting the LAPACK code generator a bit. Most of the required
> functionality is already present in the BLAS code generator, so hopefully it
> will not be a huge undertaking to incorporate support for these functions.
>
> Cheers,
>
> Rutger
>
>
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
> Sent to: karl.meerbergen_at_[hidden]
>
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm