Subject: Re: [boost] performance of a linear algebra/matrix library
From: vicente.botet (vicente.botet_at_[hidden])
Date: 2010-05-06 14:12:46
----- Original Message -----
From: "DE" <satan66613_at_[hidden]>
To: "vicente.botet" <boost_at_[hidden]>
Sent: Thursday, May 06, 2010 8:03 PM
Subject: Re: [boost] performance of a linear algebra/matrix library
>
> on 06.05.2010 at 21:44
> vicente.botet wrote :
>> Why you don't use the same pointer arithmetic in C++?
> it's not easy because the data is hidden behind an abstraction
> that is when you access an element like 'm(i, j)' it may be retrieved
> from some kind of storage or even computed on the fly
>
>> I though you wanted performances :)
> performance is an important thing but i will not go C anyway
I was referring to Boost.LA. But if you need configurable sizes for your matrixes
>> have you compared with uBlas?
> i haven't yet
> uBLAS docs claim there is no abstraction penalty for relativly large
> matrices but i don't know with what kind of C code they compared,
> maybe it is inefficient by the same factor as ublas routines giving
> apparently no abstraction penalty
The best thing to do is check it yourself. It should be not too complex.
Vicente