|
|
Ublas : |
Subject: Re: [ublas] question about a binding of ublas
From: Kaveh Kohan (kaveh.kohan_at_[hidden])
Date: 2010-11-03 12:52:07
>Kaveh Kohan wrote:
>> I think the version of boost I am using is quite old (1.35.0) and it
>> seems that there have been significant changes with respect to the current
>>version (is it 1.6?).
>
>If I remember correctly, the current bindings require a more recent Boost.MPL
>than 1.35.0. The current boost release is 1.44.0. The numeric_bindings don't
>>have version numbers, and the current numeric_bindings is significantly
>different from numeric_bindings-v1 in many ways. The numeric_bindings-v1 should
>>work with both older and more recent Boost versions, but I would suggest
>upgrading Boost instead (if Boost.MPL should really turn out to be too old).
>
>To use atlas, you have to link against the corresponding libraries, and define
>
>BOOST_NUMERIC_BINDINGS_BLAS_CBLAS
>BOOST_NUMERIC_BINDINGS_LAPACK_CLAPACK
>
>to select ATLAS as backend.
First I have to thank you all for your replies. It is such a responsive group (I
hope my question was not too provocative though :) ).
I am trying to re-implement a project which uses a lot of linear algebra with
Boost, so please bear with me if my questions are very elementary. Here is my
curious question about the design of the numerical binding:
I thought that one design could be to provide functionalities of BLAS through
uBlas but only by turning ON/OFF a compiler flag. Let me explain what I mean by
an example: we have a command like prod in uBlas which is a uBlas implementation
of matrix-matrix product but if one wants to use BLAS, she has to use gemm
command. Why didn't you design it such that prod does both of them and a
compiler flag can decide which implementation (uBlas or BLAS) to be compiled?
Again, I am sorry if it is such naive question, I am newbie to boost.
Regards,
Kayhan