I have nopython @jit function that is working very well with np.arrays.
Now I want to create new np.array in the nopython @jit function and return new array back to python. Could I do such things? Is it OK from memory management and performance point of view?
Hi @alkalinin,
Yes, and most probably, yes and yes! You can create NumPy arrays in compiled functions (np.empty
, np.ones
, np.zeros
are all supported, along with some variants of np.array
, see docs on supported NumPy features). If you have a specific use case in relation to performance or memory that is demonstrably not doing something expected, perhaps share an example to discuss?
Hope this helps?