Why is `if` so much faster when checked before a statement than after a statement?

Question

Here's an example of what I mean:

s = """
if x > 10:
    x -= 10
else:
    x = 0
"""
import timeit
print(timeit.timeit(s, setup="x=5", number=99999999))

Outputs approximately 3 seconds on my computer, regardless of the setup (x=5 vs x=15, no difference)

---

If I were to use much shorter code, one that first decreases x -= 10 and only then checks if x < 0, I will get much worse results:

s = """
x -= 10
if x < 0:
    x = 0
"""
import timeit
print(timeit.timeit(s, setup="x=5", number=99999999))

It outputs around 6 seconds, again regardless whether the initial value of x was 5 or 15.

---

I understand that it would be slower when x < 10 since we'd first call x -= 10 and then set x = 0 instead of simply setting x once.

The thing is, 99% of the time the x's initial value in my program is set to a number much higher than 10, so I thought I'd use the shorter version since most of the time I should see no difference in performance.

However, there's a huge difference in performance even when x > 10, why's this?

Answer 1

Your premise is wrong. setup only gets run once for the entire timeit. If you make sure that x stays above 10 then the symptoms disappear:

>>> s1 = """
... if x > 10:
...     x -= 10
... else:
...     x = 0
... """
>>> s2 = """
... x -= 10
... if x < 0:
...     x = 0
... """
>>> import timeit
>>> print(timeit.timeit(s1, setup="x=1000000000", number=99999999))
8.934118068675566
>>> print(timeit.timeit(s2, setup="x=1000000000", number=99999999))
8.744505329313448