from IPython import parallel
rc = parallel.Client()
ref = parallel.Reference

%precision 2

rc.ids
dview = rc[:]
dview

dview.scatter('rank', rc.ids, flatten=True)

%%px
import numpy as np
A = np.random.random((rank+1,100))
B = np.random.random((100,rank+1))

lview = rc.load_balanced_view()
ref = parallel.Reference

C_async = lview.apply(ref('A.dot'), ref('B'))
C_async

C_async.get()

from IPython.display import display

async_dots = [ lview.apply(ref('A.dot'), ref('B')) for id in rc.ids ]
for ar in async_dots:
    C = ar.get()
    print "A.dot(B) on engine %s:" % ar.engine_id
    print C.shape
    print C

async_dots = [ lview.apply_async(lambda : A.dot(B)) for id in rc.ids ]
for ar in async_dots:
    C = ar.get()
    print "A.dot(B) on engine %s:" % ar.engine_id
    print C.shape
    print C