test_backend.py

from __future__ import print_function
import threading
from joblib import Parallel, delayed
import Queue

# Fix print 
_print = print
_rlock = threading.RLock()
def print(*args, **kwargs):
    with _rlock:
        _print(*args, **kwargs)


N_GPU = 4
q = Queue.Queue(maxsize=N_GPU)
for i in range(N_GPU):
    q.put(i)


def runner(x):
    gpu = q.get()
    print(gpu)
    q.put(gpu)


Parallel(n_jobs=N_GPU, backend="threading")(
    delayed(runner)(i) for i in range(8))

# prints:
# 0
# 1
# 2
# 3
# 0
# 1
# 2
# 3

# with backend="multiprocessing"
# 0
# 0
# 0
# 1
# 0
# 1
# 1
# 2