cProfile

TLDR

Profile a Python script

$ python -m cProfile [script.py]

Profile and sort by cumulative time

$ python -m cProfile -s cumtime [script.py]

Profile and save to file

$ python -m cProfile -o [profile.prof] [script.py]

Sort by total time in function

$ python -m cProfile -s tottime [script.py]

Profile specific function in code

$ import cProfile; cProfile.run('[function_call()]')

View saved profile with pstats

$ python -c "import pstats; p = pstats.Stats('[profile.prof]'); p.sort_stats('cumtime').print_stats(20)"

Visualize with snakeviz

$ snakeviz [profile.prof]

SYNOPSIS

python -m cProfile [-o output] [-s sort] script.py [args]

cProfile is Python's built-in deterministic profiler that measures how much time is spent in each function. It's implemented in C for low overhead, making it suitable for profiling production code.
The profiler tracks every function call and return, recording the number of calls and time spent. Output shows each function with call count, total time, time per call, cumulative time, and cumulative time per call.
Profile data can be saved for later analysis with the pstats module or visualized with tools like snakeviz, pyprof2calltree, or gprof2dot. This enables detailed investigation of performance bottlenecks.

PARAMETERS

-o file

Save profile statistics to file for later analysis.

-s sort

Sort output by specified column.

-m module

Profile a module instead of script.

SORT OPTIONS

calls: Number of calls
cumtime: Cumulative time (including subcalls)
cumulative: Alias for cumtime
tottime: Total time in function (excluding subcalls)
time: Alias for tottime
ncalls: Number of calls
name: Function name
filename: File name

PROGRAMMATIC USAGE

$ import cProfile
import pstats

profiler = cProfile.Profile()
profiler.enable()
# Code to profile
profiler.disable()

stats = pstats.Stats(profiler)
stats.sort_stats('cumtime')
stats.print_stats(10)

CAVEATS

Profiling adds overhead that can affect measurements, especially for fast functions. The profiler is deterministic, not statistical, so it captures every call but may impact performance. Use line_profiler for line-by-line profiling or py-spy for sampling-based profiling with lower overhead.

HISTORY

cProfile was added to Python in version 2.5 (2006) as a faster alternative to the pure-Python profile module. It was based on lsprof contributed by Armin Rigo. The C implementation provides roughly 10x less overhead than the original profiler while maintaining the same interface.