How accurate is python's time.sleep()?

From the documentation:

On the other hand, the precision of
time() and sleep() is better than
their Unix equivalents: times are
expressed as floating point numbers,
time() returns the most accurate time
available (using Unix gettimeofday
where available), and sleep() will
accept a time with a nonzero fraction
(Unix select is used to implement
this, where available).

And more specifically w.r.t. sleep():

Suspend execution for the given number
of seconds. The argument may be a
floating point number to indicate a
more precise sleep time. The actual
suspension time may be less than that
requested because any caught signal
will terminate the sleep() following
execution of that signal’s catching
routine. Also, the suspension time may
be longer than requested by an
arbitrary amount because of the
scheduling of other activity in the
system.

You can’t really guarantee anything about sleep(), except that it will at least make a best effort to sleep as long as you told it (signals can kill your sleep before the time is up, and lots more things can make it run long).

For sure the minimum you can get on a standard desktop operating system is going to be around 16ms (timer granularity plus time to context switch), but chances are that the % deviation from the provided argument is going to be significant when you’re trying to sleep for 10s of milliseconds.

Signals, other threads holding the GIL, kernel scheduling fun, processor speed stepping, etc. can all play havoc with the duration your thread/process actually sleeps.

Why don’t you find out:

from datetime import datetime
import time

def check_sleep(amount):
    start = datetime.now()
    time.sleep(amount)
    end = datetime.now()
    delta = end-start
    return delta.seconds + delta.microseconds/1000000.

error = sum(abs(check_sleep(0.050)-0.050) for i in xrange(100))*10
print "Average error is %0.2fms" % error

For the record, I get around 0.1ms error on my HTPC and 2ms on my laptop, both linux machines.

People are quite right about the differences between operating systems and kernels, but I do not see any granularity in Ubuntu and I see a 1 ms granularity in MS7. Suggesting a different implementation of time.sleep, not just a different tick rate. Closer inspection suggests a 1μs granularity in Ubuntu by the way, but that is due to the time.time function that I use for measuring the accuracy.

Here’s my follow-up to Wilbert’s answer: the same for Mac OS X Yosemite, since it’s not been mentioned much yet.

Looks like a lot of the time it sleeps about 1.25 times the time that you request and sometimes sleeps between 1 and 1.25 times the time you request. It almost never (~twice out of 1000 samples) sleeps significantly more than 1.25 times the time you request.

Also (not shown explicitly) the 1.25 relationship seems to hold pretty well until you get below about 0.2 ms, after which it starts get a little fuzzy. Additionally, the actual time seems to settle to about 5 ms longer than you request after the amount of time requested gets above 20 ms.

Again, it appears to be a completely different implementation of sleep() in OS X than in Windows or whichever Linux kernal Wilbert was using.

A small correction, several people mention that sleep can be ended early by a signal. In the 3.6 docs it says,

Changed in version 3.5: The function now sleeps at least secs even if
the sleep is interrupted by a signal, except if the signal handler
raises an exception (see PEP 475 for the rationale).

Tested this recently on Python 3.7 on Windows 10. Precision was around 1ms.

def start(self):
    sec_arg = 10.0
    cptr = 0
    time_start = time.time()
    time_init = time.time()
    while True:
        cptr += 1
        time_start = time.time()
        time.sleep(((time_init + (sec_arg * cptr)) - time_start ))

        # AND YOUR CODE .......
        t00 = threading.Thread(name='thread_request', target=self.send_request, args=([]))
        t00.start()

Do not use a variable to pass the argument of sleep (), you must insert the calculation directly into sleep ()

---

And the return of my terminal

1 ───── 17:20:16.891 ───────────────────

2 ───── 17:20:18.891 ───────────────────

3 ───── 17:20:20.891 ───────────────────

4 ───── 17:20:22.891 ───────────────────

5 ───── 17:20:24.891 ───────────────────

….

689 ─── 17:43:12.891 ────────────────────

690 ─── 17:43:14.890 ────────────────────

691 ─── 17:43:16.891 ────────────────────

692 ─── 17:43:18.890 ────────────────────

693 ─── 17:43:20.891 ────────────────────

…

727 ─── 17:44:28.891 ────────────────────

728 ─── 17:44:30.891 ────────────────────

729 ─── 17:44:32.891 ────────────────────

730 ─── 17:44:34.890 ────────────────────

731 ─── 17:44:36.891 ────────────────────

if you need more precision or lower sleep times, consider making your own:

import time

def sleep(duration, get_now=time.perf_counter):
    now = get_now()
    end = now + duration
    while now < end:
        now = get_now()

def test():
    then = time.time()  # get time at the moment
    x = 0
    while time.time() <= then+1:  # stop looping after 1 second
        x += 1
        time.sleep(0.001)  # sleep for 1 ms
    print(x)

On windows 7 / Python 3.8 returned 1000 for me, even if i set the sleep value to 0.0005

so a perfect 1ms

The time.sleep method has been heavily refactored in the upcoming release of Python (3.11). Now similar accuracy can be expected on both Windows and Unix platform, and the highest accuracy is always used by default. Here is the relevant part of the new documentation:

On Windows, if secs is zero, the thread relinquishes the remainder of its time slice to any other thread that is ready to run. If there are no other threads ready to run, the function returns immediately, and the thread continues execution. On Windows 8.1 and newer the implementation uses a high-resolution timer which provides resolution of 100 nanoseconds. If secs is zero, Sleep(0) is used.

Unix implementation:

• Use clock_nanosleep() if available (resolution: 1 nanosecond);
• Or use nanosleep() if available (resolution: 1 nanosecond);
• Or use select() (resolution: 1 microsecond).

So just calling time.sleep will be fine on most platforms starting from python 3.11, which is a great news ! It would be nice to do a cross-platform benchmark of this new implementation similar to the @wilbert ‘s one.