When computing the time of execution of a task using this code:
start_time = time.time()
task()
end_time = time.time()
print "task time=", end_time - start_time
I got random values of time at each re-execution of the code!!!!! Even the task does not change, but every time when I execute the program I got a value which is not the same!!!! Sometimes even an important difference !!!
Now, when testing if parallel processes (multiprocessing package) will reduce the time compared to serial ones this will be very hard to check because really the time varies at every new execution of the program
!!!!!
Can anyone propose or explain that??? and how to check the time of execution of the task or of a process??
Dear Kahloul,
The operating system tends to save files in memory, and also the CPU keeps inquired data in caches. As a result, accessing memory in a cache is faster than using other memory; this could reduce execution time. Instead, you can use profiler or some predefined functions in python to compute time byline or function. Please check this tuto: https://jakevdp.github.io/PythonDataScienceHandbook/01.07-timing-and-profiling.html
Regards.
Dear Khaled
Thanks for your answer. I will see the tutor. I have just to inform you that I have tried timeit class too, and even with this class the values returned as execution time varies at each execution. The same when using time.clock() or time.time(). However, when using Linux results are more reasonable.
I will check the tuto too. Thanks again.
This is because new OS does not execute programs in real-time, but there is a priority queue that changes from time to time. Another thing you should note is that python is fundamentally an interpreted language and this means that the code requires a lot of stochastic steps (in particular if you run your code on a notebook) before reaching the processor. In addition, at runtime, the code is optimized in various ways at each iteration (for example to save memory or energy). A general rule of thumb is to apply some statistics to timing, repeating the code over and over again (storing values in a list for example) until you reach a plateau, then you can calculate your performance metrics (e.g. mean and std). If you have strict timing requirements for your program you need to optimize your code, for example, pre-allocating variables and avoiding loops as plague. Here is a good point to start writing really optimized code:
https://wiki.python.org/moin/PythonSpeed/PerformanceTips
Hi Laid, hope you are fine.
I remembered that we used these links when we tried to optimize our code.
https://www.geeksforgeeks.org/optimization-tips-python-code/
https://www.techbeamers.com/python-code-optimization-tips-tricks/
hope it will be helpful for you
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python