[aspect-devel] Duplicate Vertices

Eric Heien emheien at ucdavis.edu
Mon Dec 10 15:49:11 PST 2012


Hello all,

While working on visualizing some large datasets resulting from Aspect runs, I noticed something odd in the output files.  Large numbers of the vertices in the output mesh are duplicates and it's not clear to me why this is.  The temperature, pressure and velocity are also all identical for equivalent vertices, though the error indicator is not.

With the box.prm parameter file, the first output step has 1648 vertices recorded, though only 481 vertices are unique (1167 are duplicates).  Since there are 412 cells, I'm guessing the output stores the vertices for each cell (4*412=1648) in order regardless of whether it's been recorded already.  In the larger dataset I'm working with, 90% of the vertices are duplicates.  This means each 2.2GB file I'm working with has almost 2GB of apparently redundant data, which has a massive impact on performance for large simulations.

Although I'm working with HDF5 output, I believe this also applies to the other output formats as well.  Can someone explain why duplicate vertices (and associated values) exist in Aspect?  Is this an artifact of how the mesh is constructed or is there another reason?  Can we merge these in the output and/or simulation to improve performance or are they fundamentally required?

Thanks,

-Eric



More information about the Aspect-devel mailing list