A question: are you getting good performance from Julia? I've seen posts that say so, but my code that does Monte Carlo simulation (both vectorized and devectorized) seem to be very very slow compared to Matlab or Numpy.
I've been testing Julia on and off for the past few month (rewriting various bits of real world numpy code I've written for work) and I've seen everything from 2 times slowdown to 20 times speedup when compared to numpy. Most of the time I'm not seeing much more than 2 times speedup.
We implemented a "serious" version of simplex algorithm for linear programming (sparse-matrix / sparse-vector operations that can't be handled by standard libs), and we got performance no worse than half of the equivalent code in C++ with bounds checking, slides here: http://www.mit.edu/~mlubin/informs2013.pdf
giving an example of something that can go wrong would be a good starting point to understand why