Test which function is faster

Question

I have 3 functions that do the same thing but with a different implementation and I want to check which of them is faster.

I know that there is a timer library, but my concern is that if one of them is linear and the other one is exponential I have to make a few tests with different inputs to find the answer. For instance, if the input is a list, maybe the function A will be faster for list with length 20, but function B for length 1000.

So, I was wondering if there is any way to do this automatically. I could create a script to do this and maybe with a few plots with the time in base of the inputs length, but I don't want to invent the wheel again, if there is a library that can do this for me.

Sorry, my crystal ball is broken - could you at least mention the programming language? — Frank Schmitt, Feb 09 '14 at 11:59
Probably I need a coffee. I thought I added it on tags. I use Python. — Sfinos, Feb 09 '14 at 12:20
I think it's a relevant question: http://stackoverflow.com/questions/1593019/is-there-any-simple-way-to-benchmark-python-script — Dmytro Sirenko, Feb 09 '14 at 13:12

Justin O Barber · Answer 1 · 2014-02-09T13:16:29.847

If the functions are simple, you can do something like this (with the timeit module):

>>> import timeit
>>> a = ('A()', 'def A():\n    x = [i for i in range({list_length})]\n    x * 3')
>>> b = ('B()', 'def B():\n    x = []\n    for i in range({list_length}) * 3:\n        x.append(i)')
>>> c = ('C()', 'def C():\n    x = []\n    for i in range({list_length}):\n        x.append(i)\n    x * 3')
>>> for num in (100, 1000, 10000):
    for func in (a, b, c):
        print func[0], '=>', num,
        timeit.timeit(func[0], setup=func[1].format(list_length=num), number=10000)

The result:

A() => 100
0.06415607679355162
B() => 100
0.29436154161112427
C() => 100
0.11945218201799435
A() => 1000
0.4792028142351228
B() => 1000
2.701182674568372
C() => 1000
1.0364000362492334
A() => 10000
4.432893686294392
B() => 10000
26.445674297814094
C() => 10000
10.198394832984377

If the functions are more complex, I would suggest using a Python Profiler.

score 1 · Accepted Answer · edited May 23 '17 at 12:20

1

Use timeit:

import timeit

example = 'map(lambda x: x + 1, range(100000)'
t = timeit.Timer(example)
t.timeit()

If you want something more featureful, this question may be relevant.

edited May 23 '17 at 12:20

Community

1
1

answered Feb 09 '14 at 13:13

Dmytro Sirenko

5,003
21
26

Test which function is faster

2 Answers2