[ishan_ae]

Hi,

I want to build a project using C for myself which will be like a numerical playground for 2D unstructured FVM. Nothing fancy at the moment; only diffusion equation term but Navier-Stokes equations in the future.

I want to know what kind mesh file formats are you using to read the data and also in what data format are you storing the calculations. I would like to read the solutions into ParaView. What can you recommend?

Based on my background information finding CGSN looks good. It is open source and has C libraries.

[LuckyTran]

Paraview plays really nicely with vtk and tecplot formats.

[praveen]

Gmsh seems to be best freely available grid generator for unstructured grids, so having ability to read in gmsh format is good. It is easy to add other formats to read in, you will have to write your own code to perform some geometry calculations. You can also try out DMPlex from Petsc for handling unstructured grids. For saving solutions vtu or xdmf/hdf formats can be used. I mainly use VisIt but Paraview is another alternative.

[ishan_ae]

Thanks Lucky and Praveen for your inputs.

Praveen, are you writing from Gmsh a MSH file in ASCii format and then parsing it for nodal/face information?

[naffrancois]

Hello,

The basic information gmsh gives you is:
- a list of nodes coordinates
- a list of elements tagged by physical groups and connected to the list of nodes

You can output a simple mesh as ASCII v2 to understand the format.

Then it entirely depends on how you intend to manage data in your solver. For example, compressible unstructured cell-centered finite volume solvers usually work with elements and faces lists. You compute fluxes on the face and add it to the two neighbouring elements this face belongs to. In other words from the basic information given by gmsh you need to construct lists of connectivities. This is relatively easy work as long as performance is not essential (e.g. few elements). Building these lists from that basic information efficiently is quite more complicated though. In any case this is something you need to do prior to the solver, it should be in a preprocessing stage, separate program.

[ishan_ae]

Thank you @naffrancois for the information. I am going to look into this format now.