> a) Has deal.ii been used to solve massively parallel problems, say, 10^9 > unknowns? I am not sure if this is in actuality a naive question since I > have not been following the FE literature for a while. My understanding > a few years ago was that massively parallel problems were still > intractable with FE.
We have solved problems with almost 2 billion unknowns recently but the code is not yet publicly available because we use a library that isn't currently available publicly yet. We have some hope that this happens sometime soon. 2 billion was actually not a big problem, I think with a large enough cluster it should easily be possible to solve something 10 times bigger. The main obstacle is that 32 bit integers can't address more than 2B unknowns :-) > b) This is probably more appropriate for a CFD list but was hoping > someone here could point me in the correct direction. Is FE the way to > go for multiphase flow? Again, the little literature that I have come > across seems to favour lattice-Boltzmann for large problems. The answer probably depends on whom you ask. I would venture the guess that only few will pick lattice-Boltzmann methods, though. I would imagine that finite volume techniques would be mentioned more often. W. ------------------------------------------------------------------------- Wolfgang Bangerth email: [email protected] www: http://www.math.tamu.edu/~bangerth/ _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
