XDMF is a file format which is designed for very large simulation datasets. The main file is XML, but there is provision for the "heavy data" to be stored in HDF5 using MPI-IO in parallel. Datasets may be hundreds of GB in size. In FEniCS we have used XDMF to produce visualisations, but it is also highly desirable to use for checkpointing, i.e. saving the current state of a simulation and reading it back in later. The data structures in FEniCS do not map directly onto the data structures expected by the current implementation of XDMF. We would like to be able to save FEniCS data structures directly in XDMF, and still be able to visualise the data.

Organization

Student

Michal Habera

Mentors

  • Christopher Richardson
  • Jack S. Hale
  • David DeMarle
  • Jan Blechta
close

2017