Simplified timeit

[Connelly Barnes]

I've been avoiding the timeit module because it seems
overly complicated.  I finally coded my own function,
which has the following benefits:

 1. Simple
 2. Figures out how many times to evaluate the
    callable passed to it.

Here's the code:

def pytime(f, *args, **kwargs):
  """Calls f many times to determine the average time
     to execute f.  Usually accurate to within 3%."""
  from time import clock
  count = 1
  while True:
    start = clock()
    for i in range(count): f(*args, **kwargs)
    end = clock()
    T = end - start
    if T >= 0.5: break
    count *= 2
  return T / count

Example:

from math import *

def f(x): return x * x - cos(x)

t = pytime(f, 1.0)
print 'Time to execute f: ', t
print 'Calls to f per second: ', 1.0/t

PS: Try repeating this benchmark with psyco.  I get a
70x increase in speed.

Connelly Barnes