Option to add num_threads options to PstFrom.add_parameters?

[kbefus]

I'm having a sporadic issue where I get hung up on "starting interp point loop for x points", and it looks like the unstructured mesh I'm using is defaulting to num_threads=10 and the utils.geostats.OrdinaryKrige._calc_factors_mp call instead of the single thread org option. It looks like my unstructured (DISV) grid is catching a True here:

pyemu/pyemu/utils/pst_from.py

Line 2652 in b301698

if structured:

and then num_threads=10 is getting stuck in my system. In my local copy of pyemu, I put in a 1 here (and added verbose=True) and then the hang up went away using the org routine instead of the mp one. I don't think I can replicate this with a simple example, but seems like a multiprocessing problem on my machine. Would be great to have a num_threads kwarg or similar that could be passed through add_parameters that would trickle to the underlying calls.

[jtwhite79]

Yeah it's a problem. We've just about got pypestutils baked in so that if it's available, it will be used to solve the kriging equations (which is heaps faster bc it's compiled fortran). But we probably should sneak in a num_threads option too...

[langevin-usgs]

+1 for Fortran!

[kbefus]

Sweet.. Thanks :). It's also me breaking things on my first time through with my own model after watching heaps of your videos and goofing around with notebooks. I haven't gone down the route of figuring out why my disv grid is switching the structured flag to true either.