search

CovertLab / arrow

Stochastic simulations in python
MIT License
3 stars 1 forks source link

Suspend alternative propensity calculation 'form's #27

Closed jmason42 closed 5 years ago

jmason42 commented 5 years ago

Playing with Numba - I don't think it likes passing around function handles. I get performance hits when trying to wrap propensity and choose with @numba.jit, but if I pull out the form option, I get roughly a four-fold speed-up. There's probably a way around this but I think it's going to be easier to step back before moving forward - we also don't have any non-default test cases for the form feature. @prismofeverything thoughts?

(Numba FAQs suggest that this is an active area for improvement.)

prismofeverything commented 5 years ago

Yeah I'm happy to walk back the form feature if it is holding us back, feel free! It was for a feature Heejo needed, but it turns out she doesn't need it anymore

jmason42 commented 5 years ago

Addressed via #28.