reneSchm

I have a few open items that may need a bit of discussion: - Naming: I moved several methods around, without changing their names. None of them are particularly bad,...

On performance: There are some small improvements, but unsurprisingly nothing big. Here is single threaded (OMP_NUM_THREADS=1) performance of the current main branch as reference: ``` Running ../build/bin/abm_benchmark Run on (56...

> We could omit the parameter LocationType in a lot of places where it is only necessary to identify the location, as the Id is unique for a location. My...

> In the benchmark get_subpopulation_combined is very slow, this maybe needs to be investigated. This is partially intentional, as get_subpopulation needs to iterate over all persons instead of all persons...

Recompilation could be circumvented by reusing the skbuild directory produced by setup.py. That is, instead of creating a temporary build directory and calling cmake on it from within the test...

I suggest clamping to zero exactly, and only use tolerances on the upper bound. Then you don't need a relative tolerance, at least I would've just subtracted `tol`. You can...

> my seir2v model is not yet in main yet so if I branch from main that model will be missing I think? Depends. If you have nothing commited, then...

I am not sure. If the problem still persists, it would be possible to catch these errors in the simulation, as the SDE models use their own simulations anyways.

Here I will try to list problems of or around as well as implicit requirements on `ScalarType` or `FP`. Solutions or alternatives listed are to be understood as ideas, and...

We use the somewhat complicated type `Eigen::Matrix` regularly, even for e.g. `get_flows`, which has to be implemented by users. Should we alias it to make it easier to use? I...