If the restricted kind of code that Numba/Pythran can optimize is fine for your application, then lucky for you! We use Julia because we want more flexibility and composability than that, not because Julia is magically faster at simple a[i] = 2*a[i] + 1 style loops over arrays of Float64 scalars (it isn’t). There was a long thread about this that we don’t want to repeat: Julia motivation: why weren't Numpy, Scipy, Numba, good enough?
5 Likes