Performance is important for games, luckily there are many Python profiling tools. Profiling is about building a profile of a software program in order to collect information about memory usage or time complexity.
cProfile is a C extension introduced in Python 2.5. It can be used for deterministic profiling. Deterministic profiling means that the time measurements are precise and no sampling is used. Contrast this with statistical profiling, where measurements come from random samples.
The following steps will help you profile your code:
- Creating a profile file: We will profile the collision demo code and store the profile output in a file as follows:
python -m cProfile -o collision_demo.profile collision_demo.py
- The pstats browser: After creating the file, we can view and sort the data in a special command-line browser:
python -m pstats collision_demo.profile Welcome to the profile statistics browser.
- Getting help: Being able to get help is always a good thing, just type the following commands at the command line:
collision_demo.profile% help Documented commands (type help <topic>): ======================================== EOF add callees callers help quit read reverse sort stats strip
- Sorting: We can sort with the following
sortcommand:collision_demo.profile% sort Valid sort keys (unique prefixes are accepted): stdname -- standard name nfl -- name/file/line pcalls -- call count file -- file name calls -- call count time -- internal time line -- line number cumulative -- cumulative time module -- file name name -- function name
- Top 3 called functions: We can get the top 3 called functions by sorting and calling
stats:collision_demo.profile% sort calls collision_demo.profile% stats 3 380943 function calls (380200 primitive calls) in 18.056 seconds Ordered by: call count List reduced from 801 to 3 due to restriction <3> ncalls tottime percall cumtime percall filename:lineno(function) 52156 0.013 0.000 0.013 0.000 {method 'endswith' of 'str' objects} 31505/31368 0.003 0.000 0.003 0.000 {len} 27573 0.022 0.000 0.022 0.000 {method 'lower' of 'str' objects}
We profiled the collision demo. The following table summarizes the profiler output:
|
Column |
Description |
|---|---|
|
|
Number of calls |
|
|
Total time spent in a function |
|
|
Time per call, calculated by dividing the total time by the calls count |
|
|
Cumulative time spent in function and functions called by the function, including recursive calls |