Threads

[jtravs]

Basic threads attempt. It works, and gives a disappointing 24% speedup for 4 threads for radial free space sims.

I implemented this now because I was getting fed up of very slow modal simulations with vector plasma, but haven;t yet implemented modal threads.

• Radial free space threads
• Full free space threads
• Get this working for modal sims
• Benchmark xy sims too
• Check tests pass with varying number of threads

Fixes #40

[chrisbrahms]

That you get such little improvement is probably at least partly because the Hankel transform itself takes up a significant chunk of the runtime (btw what version of Hankel are you using? I made it up to 8x faster from 0.2 to 0.2.1). Multithreading that is on my to-do list but not very high up right now.

[jtravs]

Yes, it should be much better in the modal case, and also when the nonlinear part is much larger (e.g. Raman). But in fnfep even in the modal case the gains were not so huge. I just tried this now because, as you can see, the changes required were trivial.

[jtravs]

Modal sims at least 2x faster with 4 threads. Still not great but on par with fnfep.

[jtravs]

Why was this closed? It was working and still the correct approach I think, and was decently fast. It did need updating.

[chrisbrahms]

Ehh, no idea, didn't even realise that happened.