getfem-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Getfem-users] questions about some warnings in GMM++


From: Yves Renard
Subject: Re: [Getfem-users] questions about some warnings in GMM++
Date: Fri, 13 Mar 2009 13:32:10 +0100
User-agent: KMail/1.9.5

On Tuesday 10 March 2009 23:02, Kayhan Batmanghelich wrote:
> Dear GetFEM users,
>
> I am newbie to your community. I have a question about GMM++ rather than
> GetFEM and I hope that this is the correct forum to ask such question from.
>
> I wrote a program using gmm++ for an algorithm which involves iterative
> large scale dense and sparse matrix multiplication, thus performance is
> cruciual for me. I received the following warnings at the run time.
> Firstly, I don't know what these means and where to get some descriptions
> of these and secondly  which lines inside of my program rise such issues
> and lastly how to resolve them:
>
> the first warning I got:
>  Level 3 Warning in /usr/include/gmm/gmm_blas.h, line 1973: Inefficient row
> matrix - col matrix mult for sparse matrices, using temporary

In your program, you multiply a sparse row matrix by a sparse column matrix. 
This multiplication is inefficient, so a copy of the row matrix is first 
performed into a column matrix. If it is possible, it could be better to 
change the storage of one of your matrix to avoid this copy.

Unfortunately, there is no simple mean to detect in which line of your program 
it occurs except to insert some output messages in your program.

>
> the second warning I got:
>  Level 2 Warning in /usr/include/gmm/gmm_blas.h, line 905: Warning : a
> conflict is possible in copy
>

This message appear when you copy a part of a vector/matrix in itself. This is 
a potentialy dangerous operation, especially for sparse vector/matrices.


Best regards,

Yves.


-- 

  Yves Renard (address@hidden)       tel : (33) 04.72.43.87.08
  Pole de Mathematiques, INSA-Lyon             fax : (33) 04.72.43.85.29
  20, rue Albert Einstein
  69621 Villeurbanne Cedex, FRANCE
  http://math.univ-lyon1.fr/~renard

---------



reply via email to

[Prev in Thread] Current Thread [Next in Thread]