dolfin team mailing list archive

[Andy Terrel <aterrel@xxxxxxxxxxxx>]

Alessio Quaglino wrote:
> Yes, I link dolfin and my code with -pg but not dolfin and UMFPACK, hence
> I guess I must use tic() and toc() in order to time the LU factorization.
>
> There is another problem with UMFPACK, it seems that with all the default
> values the call:
>
> umfpack_dl_numeric( (const long int*) Rp, (const long int*) Ri, Rx,
> Symbolic, &Numeric, control, info);
> control[UMFPACK_PRL] = 2;
> umfpack_dl_report_info(control, info);
>
> returns, in particular, the following:
>
> BLAS library used:  none.  UMFPACK will be slow.
>   
This is because your UMFPACK was built without BLAS. In order to make 
UMFPACK more portable this is the default, you have to edit the 
UMFPACKv4.4/UMFPACK/Make/Make.<arch> and 
UMFPACKv4.4/AMD/Make/Make.<arch> files to compile with a BLAS library.  
Dolfin can link to UMFPACK like this just fine, it just has timing 
implications.

Andy

> so I suspect that my UMFPACK and dolfin are not linked properly. I'm using
> UMFPACK 4.4 and uBlas. Has anyone experienced something similar?
>
> Alessio
>
>
>   
> > I don't have alot of experiance with gprof, nor am I familiar with
> > UMFPACK.
> > But code that isn't compiled AND linked with -pg are ignored by the
> > profiler.
> > Maybe the functions you mention that are ignored aren't compiled and
> > linked with
> > -pg?
> >
> >     
> > > I'm trying to profile my code using gprof adding -gp to the flags when
> > > compiling. However it seems that the functions called by UMFPACK are
> > > completely ignored. Is there a solution to include them or should I use
> > > another profiling tool? Thanks.
> > >
> > > Regards,
> > > Alessio Quaglino
> > >
> > > _______________________________________________
> > > DOLFIN-dev mailing list
> > > DOLFIN-dev@xxxxxxxxxx
> > > http://www.fenics.org/mailman/listinfo/dolfin-dev
> > >       
> > _______________________________________________
> > DOLFIN-dev mailing list
> > DOLFIN-dev@xxxxxxxxxx
> > http://www.fenics.org/mailman/listinfo/dolfin-dev
> >
> >     
>
>
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev
>   


--

====================
Andy Terrel
Computer Science Dept
University of Chicago
aterrel@xxxxxxxxxxxx
---------------------

Men are equal; 
it is not birth but virtue that makes the difference.
                                           -Voltaire