|
Ublas : |
Subject: Re: [ublas] bindings for matrix_vector_range, matrix_vector_slice, bounded_matrix, bounded_vector
From: Karl Meerbergen (Karl.Meerbergen_at_[hidden])
Date: 2008-10-22 03:17:03
Dear Thomas, Markus,
stride1 and stride2 are indeed a ublas issue since they are not used in
any bindings code (BLAS or LAPACK e.g.). That is the reason why we took
it out of the description of the bindings.
Best,
Karl
Thomas Klimpel wrote:
> Hi,
>
>
>> I've attached a selection of code to test these bindings. I hope this helps.
>>
>
> Thank you. I added it to the test. I also added
>
> //
> // Copyright Markus Rickert 2008
> //
> // Distributed under the Boost Software License, Version 1.0.
> // (See accompanying file LICENSE_1_0.txt or copy at
> // http://www.boost.org/LICENSE_1_0.txt)
> //
>
> at the top of the file. Is this OK?
>
>
>
>>> The report at
>>> http://www.cs.kuleuven.be/publicaties/rapporten/tw/TW506.abs.html
>>> doesn't mention stride1/stride2 functionality. I think the stride1/stride2 functionality
>>> is an implementation detail of ublas_matrix.hpp used to implement the "storage" functionality.
>>> Since ublas_vector2.hpp doesn't seem to depend on ublas_matrix.hpp,
>>> I don't think that it should have stride1/stride2 functionality.
>>>
>> I saw it listed in matrix_traits.hpp and found it rather useful while
>> creating bindings for the Intel IPP libraries (e.g. ippmMul_mt_64f()
>> with two vectors as input).
>>
>
> I added your patch, but commented out the stride1/stride2 functionality in ublas_vector2.hpp.
> In case the stride1/stride2 functionality is important, free accessor functions dense_matrix_stride1 and dense_matrix_stride2 should be created instead, where matrix_traits<M>::leading_dimension(m) / 1 would be returned, dependent on matrix_traits<M>::ordering_type.
>
> Regards,
> Thomas
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
>