John Stachurski

I agree with the need to profile. Question: Can we automate parallelization of the max step in application of T or computation of greedy policies and get significant gains through...

Yes, good point. In my understanding this is still experimental, whereas the `parallel` flag on `@vectorize` is already standard.

As far as I know this is still an open issue and we'd love to see it pushed forward. Thanks for your interest @zhoujianberkeley . @oyamad / @mmcky, do you...

@davidrpugh Many thanks. Actually, I see now that my Numba knowledge is out of date: We can just pass jitted functions to jitted functions: http://numba.pydata.org/numba-doc/dev/user/jit.html#calling-and-inlining-other-functions This is probably the easiest...

@oyamad Great! I ran your code with the lastest from Numba and got similar output. It's a shame the `__call__` method doesn't work. I hope they fix that. Other than...

Awesome! Until someone else provides these, it would be nice to have some basic jitted scalar optimization and root finding routines in QE.py...

I didn't fully understand the second comment but certainly it should be possible to call `stationary_distributions` after `lss = LSS(A, C, G)` . In fact I would have thought we...

Thanks for your feedback @jacobadenbaum and @SeongDeokKo . I'm hoping @oyamad or @sglyon will have time to respond. (Also CC @jlperla.)

Thanks for the report @duncanhobbs. These are big, complex models that we have in mind to re-write. Would you mind to give us some information on your execution environment?

Roger that. Thanks for pointing it out. @mmcky , it looks like a problem in the linear optimization routine. Have we seen this one before?