Vlad Neverov

I discovered this when I tried to run a new SOLPS demo for generomak cherab/solps#39 by @jacklovell after a system upgrade I made today. So, it is not related to...

> This is a general problem with Python's multiprocessing module and Numpy: we see it at Culham with codes completely unrelated to Cherab and Raysect too, when codes using multiprocessing...

For some reason the threadpoolctl solution is not working for me. It fails to introspect numpy to get thread-pools description and also fails to limit the number of threads at...

I will also be happy to help with the new interpolators and caching functions. By the way, the current implementation of caching uses general `find_index` routine ignoring the fact that...

> The demos produce the correct output at the moment and performance will be sufficient for most use cases. The performance will depend on the number of parallel threads per...

Hi @Mateasek, This will be a useful option. I also agree that this should be implemented at the atomic data level. Each atomic rate must have a dedicated fallback value,...

Just realised that, for example, a user can have different impact excitation PECs in the same simulation, and if the fallback value is given at the `AtomicData` level, it applies...

> Applying fallback at the atomic data level is imo fine, because you can repeat the observation with different fallback values. I'm not against the fallback values, but to provide...

In #336 @jacklovell just suggested that these tools might not be needed at all. Well, indeed, maybe demo scripts will do better.

Thanks for the tips with `FloatToVector3DFunction2D`, @jacklovell. I suspected that something like that must exist. Now I think that the demos would do better. These function do not cover all...