Python profiling

python-profiling.md

• line_profiler https://github.com/rkern/line_profiler

  • Requeres a decorator usage
  • does not uses sys.settrace or sys.profile, evailutes the code himself lin by line
  • at the end of the program it prints how much time is line reached, and how much time spent in it

• pyinstrument https://pypi.python.org/pypi/pyinstrument/0.12

  • draws a call tree with timing information at the end

• https://pypi.python.org/pypi/yep

  • requeres google-profile tools, masures time spent in native modules

• https://pypi.python.org/pypi/RunSnakeRun cProfile (pstat) file visualizer

• https://sourceware.org/systemtap/wiki/PythonMarkers stap pyfuntop.stp -D STP_NO_OVERLOAD=1 -D MAXMAPENTRIES=60000 --skip-badvars The overhead is high, the sample rate might be decreased

• eventlet_profiler https://github.com/colinhowe/eventlet_profiler

• requres a decorator

• interactive console browser

• run time significanlt increased

• plop https://pypi.python.org/pypi/plop/

• sampler (low overhead)

• profile can viewed as ball chart via a (built-in) web server

• yappi https://code.google.com/p/yappi/

  • no simple option for creating a pstat file
  • expected to handle multiple threads automatically

• https://fedorahosted.org/gdb-heap/wiki post mortem memory ? (needs to be tried)

https://github.com/modaresimr/auto_profiler
It is also a good tools