Demonstration of CPython vs PyPy performance on a simple task. From the docstring: "This demonstrates that the expectation of the number of random values on the interval 0 <= r < 1 which sum to greater than 1 is the constant 'e'."

pypy_experiment.py

from __future__ import division

import random


def experiment(num_trials):
    """This demonstrates that the expectation of the number of random values
    on the interval 0 <= r < 1 which sum to greater than 1 is the constant 'e'.
    """
    total_num_values = 0

    for _ in xrange(num_trials):
        value_total = 0.0
        num_values = 0

        while value_total <= 1:
            value_total += random.random()
            num_values += 1

        total_num_values += num_values

    return total_num_values / num_trials


print experiment(10000000)

results.txt

System:
    Mac OS X 10.7.5
    Intel Core i7 2.2 GHz
    8 GB 1333 MHz DDR 3

### Python 2.7.1

$ time python pypy_experiment.py 
2.7182563

real    0m11.451s
user    0m11.434s
sys 0m0.015s

$ time python pypy_experiment.py 
2.7180425

real    0m11.321s
user    0m11.300s
sys 0m0.019s

$ time python pypy_experiment.py 
2.7183524

real    0m10.844s
user    0m10.829s
sys 0m0.013s

### PyPy 1.9.0 (Python 2.7.2)

$ time pypy pypy_experiment.py 
2.7182537

real    0m1.415s
user    0m1.381s
sys 0m0.016s

$ time pypy pypy_experiment.py 
2.7187915

real    0m1.396s
user    0m1.379s
sys 0m0.015s

$ time pypy pypy_experiment.py 
2.7184903

real    0m1.392s
user    0m1.377s
sys 0m0.014s