How does NumPy handle memory allocation and deallocation for ndarrays?

NumPy allocates raw memory buffers in C and relies on reference counting for automatic deallocation.

NumPy arrays allocate memory using low-level C allocators (like malloc). Each ndarray holds a reference to a contiguous memory buffer. Python's reference counting system tracks how many objects point to that buffer. When the reference count drops to zero, memory is released. Views do not allocate new memory but share the same buffer, increasing lifetime of the underlying allocation.