[Octave-bug-tracker] [bug #60005] Flame graphs for profiler output

[Rik]

Follow-up Comment #1, bug #60005 (project octave):

I collected some profile data, and then ran the profiler on flamegraph.m
itself.  It looks like the the annotation calls consume most of the time.


Top
  flamegraph: 1 calls, 1.725 total, 0.000 self
    dissect: 1 calls, 1.610 total, 0.000 self
      dissect: 2 calls, 1.453 total, 0.002 self
        1) annotation: 18 calls, 1.181 total, 0.004 self
        2) dissect: 18 calls, 0.270 total, 0.001 self
        3) cell2mat: 4 calls, 0.000 total, 0.000 self
        4) strrep: 18 calls, 0.000 total, 0.000 self
        5) sort: 2 calls, 0.000 total, 0.000 self
        6) binary *: 18 calls, 0.000 total, 0.000 self
        7) binary +: 36 calls, 0.000 total, 0.000 self
        8) binary /: 36 calls, 0.000 total, 0.000 self
        9) sum: 2 calls, 0.000 total, 0.000 self
        10) binary -: 36 calls, 0.000 total, 0.000 self


At this level, 81% of the 1.453 total run time is because of annotation.

If you wanted to speed this part up you could probably replace annotation()
with either rectangle() or patch() combined with a text() call for labeling.

Although, since profiling code is used infrequently it doesn't need to be
especially high performance.  I can wait 5 seconds for a plot to render if I
only need to do it every few months.

    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?60005>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/