Migrating multiple ParticleContainers to use runtime components

[BenWibking]

I am trying to convert Quokka from having many ParticleContainers with ParticleType = Particle<N,M> to instead all use Particle<0,0> with a variable number of runtime components. We would then have many ParticleContainers with the same ParticleType.

However, it appears that NextID() is static and re-used for all ParticleContainers of the same underlying type. Is this the intended design? Is this safe in general?

Would it work better if we used our own internal counter per-container? It looks like the particle checkpoints depend on the NextID() value. We currently have the option to split particles into multiple particles on restart, and that would break without additional changes under a shared NextID() paradigm. Also, I would like to avoid breaking particle restarts before/after the switch to runtime components, if that's possible.

[atmyers]

That is the intended design. However, you're allowed to set the particle ids however you like, including having them be non-unique if that's you want. Checkpoints store the NextID() so that it will always return a unique value across all ranks, but you don't have to use that function in Quokka. The only thing that matters from an AMReX perspective is whether the ids register as valid or not.

[BenWibking]

Ok, great. That's exactly what I wanted to know. Thanks!