On Tue, 28 Feb 2017, Alex Lindsay wrote: > I have a pretty off-the-wall question. In plasma physics we often deal > with up to six dimensions of phase space in solving the Boltzmann > equation: x, y, z, vx, vy, vz. How difficult would it be for libmesh to > support solution of a problem in which the "mesh" dimension is greater > than the standard geometric three? Easy? Impossible? In an ideal world, > one would also be able to perform adaptive mesh refinement in the > velocity space as well as the configuration/geometric space.
Right now? Short-term impossible, due to the lack of dim>3 elements, and long-term undesireable, due to the lack of anisotropic refinement. We do have users solving 7-dimensional problems (space, radiation angle, radiation energy, time), but they do so by adding *lots* of spatially-dependent variables. (and by paying me to fix the worst of the performance breakage when it turned out that libMesh was not originally designed to cope with tens of thousands of variables...) You can still do adaptive refinement in velocity space this way; e.g. by choosing a hierarchic discretization there and then using per-subdomain variables to get coarser or finer resolution in different areas in space. --- Roy ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
