search

CEED / FMS

The CEED high-order Field and Mesh Specification
http://ceed.exascaleproject.org/fms
BSD 2-Clause "Simplified" License
14 stars 1 forks source link

Existing visualization tools? DG support? #1

Open ibaned opened 5 years ago

ibaned commented 5 years ago

Hello, this project looks interesting! I have a couple of questions:

  1. What visualization tools read this format or are actively being enhanced to read this format?
  2. Is there or will there be support for Discontinuous Galerkin fields?

Thanks!

v-dobrev commented 5 years ago

Discontinuous fields are definitely supported: https://github.com/CEED/FMS/blob/bdccccbbe7e19a456b9cdf9cf630f0bdb4326570/src/fms.h#L316-L322

Regarding visualization: FMS is still in the early prototype stage, so there is no support for it in any visualization tool. However, one of the main goals of FMS is to promote (and serve as the basis for) broader adoption and support for high-order meshes and fields in visualization tools.

We (the CEED ECP project) will be talking to the ECP visualization projects (which include members of the ParaView and VisIt teams) to get their feedback and discuss a path forward for adding support for FMS in VisIt and ParaView.

Currently, FMS defines only a data exchange interface and the idea is to implement file I/O format(s) on top of this interface.

For postprocessing and visualization, one will probably need a finite element library to interpret the basis function and degree-of-freedom specifications used by FMS and provide a richer mesh and field query interface.

tzanio commented 5 years ago

To add to @v-dobrev's answer: as shown in the PUMI/MFEM example, we can read the FMS data in MFEM, convert to MFEM's mesh and visualize with GLVis, so that's one option that could be used currently.

ibaned commented 5 years ago

@tzanio so for PUMI/MFEM is FMS the middleman, e.g. PUMI -> FMS -> MFEM? I'll look into GLVis, thanks!

tzanio commented 5 years ago

@ibaned: yes, see https://github.com/CEED/FMS/blob/master/examples/README.md

bangerth commented 5 years ago

Interesting. We'd be interested to write the corresponding interfaces in deal.II as well.

VTK in newer versions has support for higher order polynomial fields defined on individual cells. This seems to work reasonably well in Paraview already, and we can write these from deal.II. Since essentially every finite element is a subset of a (complete) polynomial space of sufficiently high degree, this allows exact visualization. deal.II implemented this in https://github.com/dealii/dealii/pull/6994 and some follow-up patches.

v-dobrev commented 5 years ago

Hi @bangerth,

We're glad you are interested!

We are trying to engage as many FE libraries/applications as possible, so we can create a rich, arbitrary order FE-focused mesh and data exchange interface and file format (or suite of file formats).

The FMS data exchange interface is a first beta (v0.1), so any feedback on that is welcome.

Thanks, Veselin