Example: monte_carlo

Monte Carlo simulation example of pi estimation. This example shows how to use the Playdoh interface to execute loosely coupled parallel tasks.

from playdoh import *
import numpy as np


# Any task class must derive from the ParallelTask
class PiMonteCarlo(ParallelTask):
    def initialize(self, n):
        # Specify the number of samples on this node
        self.n = n

    def start(self):
        # Draw n points uniformly in [0,1]^2
        samples = np.random.rand(2, self.n)
        # Count the number of points inside the quarter unit circle
        self.count = np.sum(samples[0, :] ** 2 + samples[1, :] ** 2 < 1)

    def get_result(self):
        # Return the result
        return self.count


def pi_montecarlo(samples, nodes):
    # Calculate the number of samples for each node
    split_samples = [samples / nodes] * nodes
    # Launch the task on the local CPUs
    task = start_task(PiMonteCarlo,  # name of the task class
                      cpu=nodes,  # use <nodes> CPUs on the local machine
                      args=(split_samples,))  # arguments of
                                              # MonteCarlo.initialize
                                              # as a list,
                                              # node #i receives
                                              # split_samples[i]
                                              # as argument
    # Retrieve the result, as a list with one element returned
    # by MonteCarlo.get_result per node
    result = task.get_result()

    # Return the estimation of Pi
    return sum(result) * 4.0 / samples


if __name__ == '__main__':
    # Evaluate Pi with 10,000 samples and 2 CPUs
    print pi_montecarlo(1000000, 2)