I'm currently using FEFieldFunction to transfer my finite element solution between different grid levels and it works great if I use the VectorTools::interpolate function. However, what I really need is the L^2 projection as the interpolation isn't equal to the numerical method that I have so is giving bad results when my time steps are small. Now, using VectorTools::project I get weird results. Firstly, the meshes are coarsened and refined (potentially) at each time step so I use a large number of quadrature points (QGauss(20)) to account for this.
Unless you refine/coarsen the mesh by more than one level, you shouldn't need this many quadrature points. If you refine by only one level, it should be sufficient to iterate the QGauss(degree+1) formula once using the QIterated class.
> Now, the
odd thing is, my program can work perfectly one minute (run through all time steps, all meshes) without any problems whatsoever, but other times I get "segmentation faults", not only that but they appear completely at random, sometimes near the first time step, other times near the final time step, do you think there could be a bug in either FEFieldFunction of in VectorTools::project? (Since interpolate works fine and perfectly...)
Yes, there was a bug in FEFieldFunction when multiple threads were used. What version of deal.II are you using? I believe that this was fixed for 7.1.0.
Best W. ------------------------------------------------------------------------ Wolfgang Bangerth email: [email protected] www: http://www.math.tamu.edu/~bangerth/ _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
