Peter Würtz

This might be related to the issue I'm seeing in #9712, where Numba-`np.empty` or temporary array allocations are significantly slower than arrays allocated with numpy. To complicate things further, this...

Update: When moving all memory allocations from Numba to Numpy, the total runtime (Numpy allocations + Numba call without allocations) is on-par with Numpy again. ```python @nb.njit(fastmath=True, nogil=True) def func_without_alloc(z1,...

Thanks for your insights! I'm still planning on following up on your `arena` size idea and see if timings correlate the same way on my end (unfortunately I won't be...

I think I can confirm the correlation between fast/slow runs with specific changes in memory arena. This is the set up for calling `mallinfo2` from `libc`: ```python from cffi import...

Issue is still relevant

I can re-run the tests on numba `0.61` when it's released, but the assumption is that the issue is still relevant.

> * Is NumPy using Python's allocator? I would guess it is, but would need to confirm that. Presumably doing that has better performance for typical Python use cases. [Memory...

I've also noticed that for some reason, `np.sum` within Numba isn't able to compete with Numpy at all. However, if you literally type out what sum does, in the most...