|
|
"""Example of how to use pylab to plot parallel data.
|
|
|
|
|
|
The idea here is to run matplotlib is the same IPython session
|
|
|
as an ipython parallel Client. That way matplotlib
|
|
|
can be used to plot parallel data that is gathered using
|
|
|
a DirectView.
|
|
|
|
|
|
To run this example, first start the IPython controller and 4
|
|
|
engines::
|
|
|
|
|
|
ipcluster -n 4
|
|
|
|
|
|
Then start ipython in pylab mode::
|
|
|
|
|
|
ipython -pylab
|
|
|
|
|
|
Then a simple "run parallel_pylab.ipy" in IPython will run the
|
|
|
example.
|
|
|
"""
|
|
|
|
|
|
import numpy as N
|
|
|
from pylab import *
|
|
|
from IPython.parallel import Client
|
|
|
|
|
|
# load the parallel magic
|
|
|
%load_ext parallelmagic
|
|
|
|
|
|
# Get an IPython Client
|
|
|
rc = Client()
|
|
|
v = rc[:]
|
|
|
v.activate()
|
|
|
|
|
|
# Create random arrays on the engines
|
|
|
# This is to simulate arrays that you have calculated in parallel
|
|
|
# on the engines.
|
|
|
# Anymore that length 10000 arrays, matplotlib starts to be slow
|
|
|
%px import numpy as N
|
|
|
%px x = N.random.standard_normal(10000)
|
|
|
%px y = N.random.standard_normal(10000)
|
|
|
|
|
|
print v.apply_async(lambda : x[0:10]).get_dict()
|
|
|
print v.apply_async(lambda : y[0:10]).get_dict()
|
|
|
|
|
|
# Bring back the data
|
|
|
x_local = v.gather('x', block=True)
|
|
|
y_local = v.gather('y', block=True)
|
|
|
|
|
|
# Make a scatter plot of the gathered data
|
|
|
plot(x_local, y_local,'ro')
|
|
|
|
