Bennett Helm a écrit :
On May 19, 2006, at 11:34 AM, Abdelrazak Younes wrote:
No, I think it would be better if you pass
--disable-debug --disable-stdlib-debug
instead of:
--enable-debug --enable-stdlib-debug
Without --enable-debug, at least, I can't get function names in the
profile.
Yes, I forgot about that, --enable-debug is fine as it should not impact
performance with gcc. But --enable-stdlib-debug do so please use at
least --disable-stdlib-debug.
You use case should be useful. Here is some other use case with a big
document (UserGuide or Extended):
- hold the arrow down key pressed
- hold the page down key pressed
- scroll by dragging the scrollbar slider
- scroll by holding click on the down arrow button of the scrollbar.
I'll wait on doing these until I'm sure you like the format.
Fine.
I am sure the tree view is useful because this line for example
> QWorkArea::keyPressEvent(QKeyEvent*) (lyx)
> | | | | | | | | | | | | | | | | | | | | | | | + 67.0%
indicate that 67% of the time is spent on this method and its
children. But real time is most probably close to zero. Absolute
figures are more informative if you manage to produce that.
OK ... how's this (produced from the same profile, again arbitrarily cut
off near the top)?
Sorry Bennet, I didn't mean "absolute figure" in the sense of "number of
calls" versus "percentage". I mean that we need the actual _time_ spend
in each function. What you are providing seems to be the number of
calls. That said, I got maybe all that wrong... Anybody use to analize
those kind of profile (JMArc)??
Abdel.
