Useful Development Tools

Memory Profiling

GObject Instance Count Leak Check

Requires a development (only available in debug mode) version of glib. Jhbuild recommended.

jhbuild shell
GOBJECT_DEBUG=instance-count GTK_DEBUG=interactive ./quodlibet.py

• In the GTK+ Inspector switch to the “Statistics” tab

• Sort by “Cumulative” and press the “Next” multimedia key to quickly switch songs.

• If something in the “Cumulative” column steadily increases we have a leak.

Performance Profiling

cProfile

• https://docs.python.org/2/library/profile.html

• bundled with python

python -m cProfile -s [sort_order] quodlibet.py > cprof.txt

where sort_order can one of the following: calls, cumulative, file, line, module, name, nfl, pcalls, stdname, time

Example output:

         885311 function calls (866204 primitive calls) in 12.110 seconds

   Ordered by: cumulative time

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    0.002    0.002   12.112   12.112 quodlibet.py:11(<module>)
        1    0.007    0.007   12.026   12.026 quodlibet.py:25(main)
19392/13067    0.151    0.000    4.342    0.000 __init__.py:639(__get__)
        1    0.003    0.003    4.232    4.232 quodlibetwindow.py:121(__init__)
        1    0.000    0.000    4.029    4.029 quodlibetwindow.py:549(select_browser)
        1    0.002    0.002    4.022    4.022 albums.py:346(__init__)
        ...
        ...

SnakeViz

• https://jiffyclub.github.io/snakeviz/

• pip install snakeviz

python -m cProfile -o prof.out quodlibet.py
snakeviz prof.out

Run Snake

• http://www.vrplumber.com/programming/runsnakerun/

• package runsnakerun in debian Wheezy

python -m cProfile -o prof.out quodlibet.py
runsnake  prof.out

Example: https://www.google.at/search?q=runsnakerun&tbm=isch

Gprof2Dot

• https://github.com/jrfonseca/gprof2dot

python -m cProfile -o output.pstats ./quodlibet.py
gprof2dot.py -f pstats output.pstats | dot -Tpng -o output.png

Line Profiler

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

# wrap all functions of interest with the @profile decorator
./kernprof.py  -l ./quodlibet.py # creates quodlibet.py.lprof
python line_profiler.py quodlibet.py.lprof > prof.txt

Example output:

Timer unit: 1e-06 s

File: test.py
Function: a at line 2
Total time: 0.001528 s

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
     2                                           @profile
     3                                           def a():
     4         1          134    134.0      8.8      print "hello"
     5         1           12     12.0      0.8      b = []
     6       101          628      6.2     41.1      for i in xrange(100):
     7       100          696      7.0     45.5          b.append(i)
     8         1           58     58.0      3.8      print "world"

strace

• https://linux.die.net/man/1/strace

strace -c ./quodlibet.py

Example output:

 time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
 81.64    0.013274           9      1444       178 read
  7.04    0.001144           0      4041      3259 open
  3.81    0.000619          11        54           getdents64
  2.80    0.000456           0      1004           fstat64
  1.84    0.000299           0      2221      1688 stat64
...
...

IOProfiler

• https://code.google.com/archive/p/ioapps/wikis/ioprofiler.wiki

• strace GUI

• Shows read/write to files (how many reads/writes per file, which parts of the files were affected)

strace -q -a1 -s0 -f -tttT -oOUT_FILE -e trace=file,desc,process,socket ./quodlibet.py
ioreplay -c -f OUT_FILE -o OUT_FILE.bin
ioprofiler.py
# open OUT_FILE.bin

Example: https://code.google.com/archive/p/ioapps/wikis/IOProfilerScreenshots.wiki