Mutex lock in njit function

[hbina]

AFAIK numba on CPU does not have atomics support, the CUDA one does.
I would write the locking mechanism you need in C and use it via FFI.

From this thread, you can use object mode to get your locks, if you so desire.

How to use locks in nopython mode?

Which lock implementation can be used when using @njit(nogil=True)? It refuses to compile with threading.Lock() - import threading from numba import njit lock = threading.Lock() @njit(nogil=True) def child(): lock.acquire() child() Untyped global name 'lock': cannot determine Numba type of <class '_thread.lock'> File "xxx.py", line 67: def child(): lock.acquire() ^