John Ehresman wrote:
Robert Roessler wrote:
editor#set_margin_width_n 0
(if get_prop_int p_t "line.numbers" > 0 then
(editor#text_width sci_STYLE_LINENUMBER "_9999") else 0)
... FURTHER refining of the test shows (as you probably already know)
that ALL of the [measurable] time is going to the text_width call!
Are you profiling the first instance created? You might be measuring
the initial pango (the gtk text rendering engine) initialization.
Subsequent calls to pango will hopefully be faster.
Thanks, but unfortunately (that would be too easy), this call
stabilizes at 0.08 - 0.10 secs (I did not post this level of detail
because it was already getting too wordy)... so while the Pango
rendering machinery seems to be on the heavyweight side, I wonder why
this particular call (SCI_TEXTWIDTH) is so expensive?
In case anyone is about to ask, I have left the "buffered draw" and
"two phase draw" settings alone, and they are documented as defaulting
to true.
The good news is that I am sure that things will get faster as GTK
transitions to Pango on the Cairo backend... ;)
For those just joining our program, I was majorly bummed when
SeaMonkey switched from GdiPlus to Cairo for the Windows builds - *I*
think rendering speed took a serious hit, but maybe that's just me.
Robert Roessler
[EMAIL PROTECTED]
http://www.rftp.com
_______________________________________________
Scintilla-interest mailing list
[email protected]
http://mailman.lyra.org/mailman/listinfo/scintilla-interest
