Python perf_counter()

The perf_counter() function returns a floating-point number representing a high-resolution timer value (in seconds) at the moment of the call, and is commonly used to benchmark code by measuring the time taken to execute a piece of code.

Syntax

time.perf_counter()

Parameters:

The function takes no arguments.

Return value:

Returns a float representing the current value of a high-resolution performance counter in seconds.

Example 1

This example shows how to call perf_counter() to get the current high-resolution timer value:

import time

print(time.perf_counter())

Copy to clipboard
Copy to clipboard

A possible output of this code is:

5003.666789873

Copy to clipboard
Copy to clipboard

Importing perf_counter() from time:

from time import perf_counter

print(perf_counter())

Copy to clipboard
Copy to clipboard

A possible outcome of this code is:

1780743.368374651

Copy to clipboard
Copy to clipboard

Example 2

Here, two perf_counter() calls are placed around sleep() which adds a 5-second delay in the execution right after the first method call:

from time import perf_counter, sleep

# record current time before delay
start = perf_counter()

print('Time before sleep():', start)

sleep(5) # pause for 5 seconds

# record current time after delay
end = perf_counter()

print('Time after sleep():', end)

# calculate elapsed time between both calls
elapsed_time = end - start

print('Elapsed time in seconds:', elapsed_time)

Copy to clipboard
Copy to clipboard

The output of this code is:

Time before sleep(): 1781952.36727265
Time after sleep(): 1781957.367474534
Elapsed time in seconds: 5.000201884191483

Copy to clipboard
Copy to clipboard

Codebyte Example

The following program shows two ways to search for an element in a sorted integer list. Both solutions are written in separate functions, and pairs of perf_counter() calls are used to track the time taken to execute each function:

Visit us
Visit us
Hide code
Code
Output
Hide output
Loading...
Copy to your clipboard
Copy Codebyte
Run

Note: The actual numbers will vary each time you run the code, since they represent the current internal timer value, not absolute wall-clock time.

Ideally, with a large nums list the binary search function should take half as much time as the linear search function.