Smarter domain decomposition

[gomezzz]

Description

• Added altitude-based MPI decomposition (each rank is responsible for some altitude interval)
• More logging
• Refactoring of MPI cooms
• Fixed bug in getRank of RegularGrid

TODO:

• [x] Fix conjunction detection (possibly on immigrants) - [ ] Add logger for altitude-based (will be done when required)
• [x] Convert some cout to logger
• [x] Add tests
• [x] Automatically optimize altitude bins according to distribution of particles

Related Pull Requests

• 138 (based on it)

Resolved Issues

• [x] fixes #141

How Has This Been Tested?

• Running initial population
• TBD

[gomezzz]

@FG-TUM any suggestions for what tests we should add for this?

[FG-TUM]

@FG-TUM any suggestions for what tests we should add for this?

Add a few MPI tests. Make sure

• [x] the partitioning works (do we catch the case if there are more ranks than particles? Do we care? :grin: ),
• [x] the migration works
• [ ] collisions during migration are caught.

FYI: The MPI tests are always executed with 8 ranks.

[gomezzz]

@FG-TUM any suggestions for what tests we should add for this?

Add a few MPI tests. Make sure

* [x]  the partitioning works (do we catch the case if there are more ranks than particles? Do we care? 😁 ),

* [x]  the migration works

* [ ]  collisions during migration are caught.

FYI: The MPI tests are always executed with 8 ranks.

Added most tests. Collisions not atm as that is pretty complex actually (basically need to run a whole iteration) and we still have to fix #145. I'd suggest to do it there so we can wrap this one up first. Also, because this check also should be done for regularGrid, I reckon?

[FG-TUM]

Also, because this check also should be done for regularGrid, I reckon?

Yes this should be done independently of the underlying decomposition.

I'd suggest to do it there so we can wrap this one up first

Ok