[Trilinos-Users] Epetra_FECrsGraph question
Williams, Alan B
william at sandia.gov
Wed Aug 13 14:11:31 MDT 2008
Ok, I see what you mean.
I'll do some profiling of this area of epetra and let you know what I can do about optimizing it or recommending other usage.
> -----Original Message-----
> From: trilinos-users-bounces at software.sandia.gov
> [mailto:trilinos-users-bounces at software.sandia.gov] On Behalf
> Of kent-and at simula.no
> Sent: Wednesday, August 13, 2008 10:29 AM
> To: Williams, Alan B
> Cc: Kent-Andre Mardal; trilinos-users at software.sandia.gov
> Subject: Re: [Trilinos-Users] Epetra_FECrsGraph question
> Dear Alan,
> The :InsertGlobalIndices call is performed one time element
> so it should
> be the same number of calls in all examples (except that the 2D and 3D
> mesh are different).
> I have attached the Dolfin code for creating the sparsity pattern.
> >> -----Original Message-----
> >> From: trilinos-users-bounces at software.sandia.gov
> >> [mailto:trilinos-users-bounces at software.sandia.gov] On Behalf
> >> Of Kent-Andre Mardal
> >> Hello,
> >> But what I don't understand is that there is such a great
> >> difference in
> >> the case where the finite element method contains vector
> type elements
> >> in 3D. While in the scalar case, Epetra is comparable to the other
> >> backends.
> > I see what you mean, the timings are substantially
> different for those
> > last two cases.
> > The pseudo-code you sent indicates you're calling
> > CrsGraph::InsertGlobalIndices, which takes a single row. If you have
> > vector fields it just means that you call
> InsertGlobalIndices more times.
> > Is it possible that you could send more details? Perhaps
> more of your
> > code, or some profiling data.
> > Alan
More information about the Trilinos-Users