Profiling¶

Get the profile stats with `cProfile`¶

import cProfile
prof = cProfile.Profile()
prof.runcall(func, *args, **kwargs)

Print out the stats:

prof.print_stats()

Dump the profile stats for GUI consumption:

prof.dump_stats('stats.prof')

Install RunSnakeRun for GUI view:

$ pip install RunSnakeRun

It needs wxPython:

$ conda install wxpython

Invoke RunSnakeRun:

$ runsnake stats.prof

It is useful to look at Local column. This gives the time spent on a function itself, excluding the time spent calling other functions during the function call. Hence it gives the net time spent.
On the other hand, Cum shows the cumulative time spent on that function, including the time spent for other function calls made during the function call.

Install KCachegrind and pyprof2calltree:

$ zypper in kcachegrind
$ sudo pip install pyprof2calltree

Invoke KCachegrind through pyprof2calltree:

$ pyprof2calltree -i stats.prof -k