time.perf_counter in Python 2?

[Florian Lindner]

Hello,

I wrote a script that does some time measurements. It uses
time.perf_counter() from Python 3 which works very well. Now I need to
backport it to python 2.

Docs say that time.clock() is way to go:

time.clock()
On Unix, return the current processor time as a floating point number
expressed in seconds. The precision, and in fact the very definition of the
meaning of “processor time”, depends on that of the C function of the same
name, but in any case, this is the function to use for benchmarking Python
or timing algorithms.

On Windows, this function returns wall-clock seconds elapsed since the first
call to this function, as a floating point number, based on the Win32
function QueryPerformanceCounter(). The resolution is typically better than
one microsecond.

But for me it always returns the almost same number, nothing time like:

Python 2.7.3 (default, Feb 27 2014, 19:58:35)
[GCC 4.6.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import time
>>> time.clock()
0.03
>>> time.clock()
0.03
>>> time.clock()
0.04
>>> time.clock()
0.04


What's wrong there?

Thanks,
Florian

[Chris Angelico]

On Mon, Oct 20, 2014 at 6:47 PM, Florian Lindner <mailin...@xgm.de> wrote:
> time.clock()
> On Unix, return the current processor time as a floating point number
> expressed in seconds. The precision, and in fact the very definition of the
> meaning of “processor time”, depends on that of the C function of the same
> name, but in any case, this is the function to use for benchmarking Python
> or timing algorithms.
>

> But for me it always returns the almost same number, nothing time like:

You're doing practically no work in there. You're using almost no
processor time - the interpreter's waiting for the user, not actually
busy. If you create a busyloop, you'll see time.clock() go up roughly
linearly:

rosuav@sikorsky:~$ python
Python 2.7.3 (default, Mar 13 2014, 11:03:55)
[GCC 4.7.2] on linux2

Type "help", "copyright", "credits" or "license" for more information.
>>> import time

>>> def killtime():
... start=time.time()
... end=time.clock()+1
... while time.clock()<end: pass
... return time.time()-start
...
>>> killtime()
1.0048420429229736
>>> killtime()
1.004641056060791
>>> killtime()
1.0081689357757568
>>> killtime()
1.0019969940185547
>>> killtime()
1.0044770240783691

This function will do its best to waste one second of CPU time. As you
see, it's taking pretty much exactly one second of wall time to do so
(as measured by time.time()). At the end of a few of these runs,
time.clock() will return a value that's fairly close to 1.0 times the
number of executions of killtime(), as the other work Python's doing
is insignificant by comparison.

ChrisA

[Peter Otten]

Florian Lindner wrote:

> Hello,
>
> I wrote a script that does some time measurements. It uses
> time.perf_counter() from Python 3 which works very well. Now I need to
> backport it to python 2.
>
> Docs say that time.clock() is way to go:
>

> time.clock()
> On Unix, return the current processor time as a floating point number
> expressed in seconds. The precision, and in fact the very definition of
> the meaning of “processor time”, depends on that of the C function of the
> same name, but in any case, this is the function to use for benchmarking
> Python or timing algorithms.
>

> On Windows, this function returns wall-clock seconds elapsed since the
> first call to this function, as a floating point number, based on the
> Win32 function QueryPerformanceCounter(). The resolution is typically
> better than one microsecond.
>

> But for me it always returns the almost same number, nothing time like:
>

> Python 2.7.3 (default, Feb 27 2014, 19:58:35)

> [GCC 4.6.3] on linux2

> Type "help", "copyright", "credits" or "license" for more information.
>>>> import time

>>>> time.clock()
> 0.03
>>>> time.clock()
> 0.03
>>>> time.clock()
> 0.04
>>>> time.clock()
> 0.04
>
>
> What's wrong there?

While clock is not the recommended timer on Linux the granularity is still
higher over here:

>>> time.clock()
0.06874
>>> time.clock()
0.069221
>>> time.clock()
0.069782
>>> time.clock()
0.070309

That or you can type faster ;)

Here's an excerpt from the 2.7 timeit.py:

if sys.platform == "win32":
# On Windows, the best timer is time.clock()
default_timer = time.clock
else:
# On most other platforms the best timer is time.time()
default_timer = time.time

In Python3.4 this has become

default_timer = time.perf_counter

So I'd go with timeit.default_timer().