Neil Hodgson wrote:
Robert Roessler:
I have always felt that the Scintilla widget (GTK2 on Windows XP with
Pango font rendering) running a lexer I wrote was "slow", but I just
now watched CPU time while holding down the Page Down key... wow.
I see 100%, with about 70% of that "kernel" time - as an [unfair]
comparison, a VS6 window showing syntax colored C++ under the same
conditions... shows essentially NO CPU time!
I see peaks of around 35-45% for VS .NET 2005 and 50-60% for SciTE.
For you, is SciTE much better than Scintilla/GTK+?
Actually, I just sort of "knew" this was fast (like I knew the widget
case was slow) - so I hadn't done this same comparison... with a
freshly loaded copy of essentially the same file as before, I see
10-15% CPU (all kernel), with the same going from back to front.
Not that there's anything wrong with using 100% CPU in software: if
there's something to do then do it.
True - but that raises the question of whether it is "starved" for CPU
- if is below 100%, then you know that it is getting all of the CPU it
can use... and it is either performing as designed OR is being
throttled by some other resource contention/shortage (presumably not
the case here).
3) is this because of "excess" / unbuffered syntax coloring?
After you've scrolled through the file once its all styled. Does it
still use the same resources?
*YES* - the Scintilla/GTK+ widget case never exhibits any apparent
memory of having been all styled previously... at least the CPU stays
the same. Of course, it does not change in the "good" SciTE case
either - but that CPU usage level is not troubling.
Robert Roessler
[EMAIL PROTECTED]
http://www.rftp.com
_______________________________________________
Scintilla-interest mailing list
[email protected]
http://mailman.lyra.org/mailman/listinfo/scintilla-interest
