Slow and crashing when using tetgen for mesh initialization

[chaseshyu]

Currently, DES can run for model scale up to 5 million elements. However, the mesh initiation might crash sometimes or take forever. So, I am trying to use a coarser Tetgen mesh + MMG to initialize an irregular mesh. The new code shows only 1/8 of the duration time for mesh initialization, from 84s to 11s. The mesh numbers are similar, but MMG's mesh has a smoother change between different resolutions. Also, the same MMG Lib generated initiated mesh results in a smoother mesh change between before and after MMG remeshing. Please feel free to provide comments and suggestions. Thanks!

tetgen:

MMG:

tetgen:

MMG:

[tan2]

This is looking good. Do you have to modify the code to achieve this? Also, how well does this technique work with a more complicated geometry?

[chaseshyu]

This changes are implemented via adding an USEMMG branch in the function points_to_mesh(). So, every mesh initialization using this function can be benefited by this change. I will push the feature branch later.

This is looking good. Do you have to modify the code to achieve this? Also, how well does this technique work with a more complicated geometry?

[echoi]

@tan2 I think this two-step approach might still be slow for a complex geometry because tetgen will have to tackle it first anyway.
It's certainly a feature worth pulling, though.