Add dry run capability

[mbkuhn]

Summary

Had a request for a dry run capability so the setup could be viewed without running a simulation. This is a sort-of automated way of doing the least amount of work while still populating arrays and outputting a plotfile. I also made sure that a plotfile will still be written when doing a dry run for a restart.

Pull request type

• Bugfix
• Feature
• Code style update (formatting, renaming)
• Refactoring (no functional changes, no api changes)
• Build related changes
• Documentation content changes
• Other (please describe):

Checklist

The following is included:

• new unit-test(s)
• new regression test(s)
• documentation for new capability

This PR was tested by running:

• the unit tests
  • on GPU
  • on CPU
• the regression tests
  • on GPU
  • on CPU

Additional background

Before merging, I'll need to add this to the documentation. I could incorporate it into it's own reg test, but I'm not sure it's worth the hassle.

[mbkuhn]

This approach to a dry run would still require similar memory demands to the actual simulation, but it would require basically no time and very little computing power. Not sure how to avoid memory needs when the goal is to actually initialize the arrays.

[mbkuhn]

It would be good to chat with @hgopalan, @rthedin, and @MaxPal to see if this would be useful for their workflows.

[hgopalan]

I think if it is for the complex terrain support, we need to read the TerrainDrag module which initializes the terrain_blank field and provides information on the terrain mapping before running simulations. I register the velocity field to create amrex::MFIter mfi(velocity); mfi.isValid(); ++mfi since I did not know how to do it otherwise with all new variables. If there is a way to avoid calling the velocity field, I can update the TerrainDrag and we would need only to register the new variables and plot them.

[marchdf]

I don't think we need a regtest for this. But documentation would be good ;)

[mbkuhn]

I think if it is for the complex terrain support, we need to read the TerrainDrag module which initializes the terrain_blank field and provides information on the terrain mapping before running simulations. I register the velocity field to create amrex::MFIter mfi(velocity); mfi.isValid(); ++mfi since I did not know how to do it otherwise with all new variables. If there is a way to avoid calling the velocity field, I can update the TerrainDrag and we would need only to register the new variables and plot them.

Thanks for sharing. For the sake of the terrain, there probably could be a standalone tool devised to address that need, which would be nimbler in terms of memory needs. This approach is more general and should help others who would appreciate the capability with different problems. One nice thing with this approach is it uses the actual code path of AMR-Wind, so it should run into the same problems that a real run would (e.g., if something is not set up properly in the input).