Re: Possible Pango 1.4 ideas
- From: Joaquín Cuenca Abela <cuenca pacaterie u-psud fr>
- To: Owen Taylor <otaylor redhat com>
- Cc: Damon Chaplin <damon karuna uklinux net>, Havoc Pennington <hp redhat com>, gtk-i18n-list gnome org
- Subject: Re: Possible Pango 1.4 ideas
- Date: 24 Jan 2003 19:52:45 +0100
On Thu, 2003-01-23 at 20:17, Owen Taylor wrote:
>
> Joaquín Cuenca Abela <cuenca pacaterie u-psud fr> writes:
>
> > On Wed, 2003-01-22 at 23:51, Owen Taylor wrote:
> > >
> > > Joaquín Cuenca Abela <cuenca pacaterie u-psud fr> writes:
> > >
> > > > [*] ie., I've not measured it, just seen GtkTextView in action... but I
> > > > don't how representative is GtkTextView
> > >
> > > Quick timing is that GtkTextView is about 14 times as fast as
> > > AbiWord 0.99.4 at laying out text. (Yeah, it's an old copy
> > > of AbiWord, but what I had available to try on the machine
> > > where I was doing the timing.)
> >
> > You're right. My apologies. The text files that I was opening with
> > GtkTextView were much bigger than I first thought (they didn't seem to
> > be *so* big until I opened them with abiword and I saw the number of
> > pages), and I got the impression that GtkTextView was lagging.
>
> One thing to realize is that text editors like Emacs or vi are
> _really fast_. They generally don't actually lay out the text,
> and instead simply count lines, or (for Emacs) characters, when
> setting up the scrollbars.
That's really embarrassing... I got so misguided in my appreciation
because I created the text file starting with 1 line, and with a series
of:
cat test.txt >> test2.txt && cat test2.txt >> test.txt
...
and I miserably missed the exponential grown rate...
When I opened it with emacs, it was instantaneous, and with GtkTextView
I had to wait several seconds for the scrollbars to stop growing.
In fact, the file was over 500 pages once I opened it in AbiWord.
At least, I know where the poor performance comes from... several months
ago we identified a little piece of code in AbiWord that raises the
layout complexity to O(n^2) (with n the number of lines), but it was not
yet showing in the top of the profilers, so we (mainly Tomas) started
fixing other stuff (we got a x4 improvement on loading). I hope to be
able to find a bit of free time to fix these issues.
We're doing less complex things than pango most of the time, so I hope
that we will be able to bring abiword up to pango speed.
Again, sorry for the FUD.
--
Joaquín Cuenca Abela
cuenca pacaterie u-psud fr
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]