|
|
Ublas : |
Subject: Re: [ublas] [bindings] matrix traits
From: Karl Meerbergen (karl.meerbergen_at_[hidden])
Date: 2009-01-08 06:01:34
Dear all,
I have committed the idea of the workspace that we discussed last year.
I also removed some bugs. some functions still need furth work to use
workspace3
Best regards,
Karl
Karl Meerbergen wrote:
> Dear all,
>
> I modified (my local version) so that the workspace1,2,3 classes allow
> to do the following:
>
> template <typename A, typename W, typename WC, typename WR>
> int heev (char jobz, char uplo, A& a, W& w,
> detail::workspace2<WC,WR> workspace ) {
> typedef typename A::value_type
> value_type ;
> typedef typename traits::type_traits<value_type>::real_type
> real_type ;
>
> return detail::heev( jobz, uplo, a, w, workspace.select(
> value_type() ), workspace.select( real_type() ) );
> }
>
> You can call:
>
> lapack::heev( 'N', UPLO, a2, e2, workspace(wr,wc) ) ;
> lapack::heev( 'N', UPLO, a2, e2, workspace(wc,wr) ) ;
>
> Karl
>
>
>
> Karl Meerbergen wrote:
>
>> Hi Rutger,
>>
>> I do not think this is possible. In some cases a row_major matrix is
>> interpreted as a transpose column_major matrix.
>> Why would you need this for LAPACK bindings?
>>
>> Best,
>>
>> Karl
>>
>>
>> Rutger ter Borg wrote:
>>
>>
>>> Dear all,
>>>
>>> looking into what might be expected by lapack and what is available through
>>> the traits system, I was wondering whether it is possible to detect a
>>> transposed type through the matrix_traits? Or is this never used? Or does
>>> this work differently?
>>>
>>> Thanks,
>>>
>>> Rutger
>>>
>>>
>>> _______________________________________________
>>> ublas mailing list
>>> ublas_at_[hidden]
>>> http://lists.boost.org/mailman/listinfo.cgi/ublas
>>>
>>>
>>>
>>
>> _______________________________________________
>> ublas mailing list
>> ublas_at_[hidden]
>> http://lists.boost.org/mailman/listinfo.cgi/ublas
>>
>>
>
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
>