Hi,
over the last couple of months we have ported significant chunks of our code to Numba and are mostly very happy with the decision. Two features we are missing though and which would be great, are:
1.) Stack based fixed size arrays. Even for small arrays we currently need to create Numpy objects on the heap, which is eating up a lot of performance. The ability to create fixed size arrays on the stack would be amazing.
2.) Explicit Simd constructs. Auto-vectorizers are getting better all the time. Yet, having explicit simd types often provides a more reliable pathway to vectorization in complex codes.
Have there been any discussions or thoughts about these features?