Does Numpy allocate temporary arrays in expressions like x += 2 * y?

Question

When evaluating expressions like

x += 2 * y

does Numpy first allocate a new temporary array to hold 2*y , add it to x and then delete it, or can it perform this whole operation in-place?

Answer 1

Yup, that makes a temporary array.

If you find yourself needing to mitigate NumPy's love of giant scratch arrays, additional libraries like Numexpr can help quite a bit, but make sure you're attributing performance problems to the right causes. Naive attempts to save allocations usually cause massive slowdowns instead of performance improvement.