On Jan 4, 2012, at 4:44 AM, Stephan Witt wrote:
>
> And this is the result:
> I've opened the users guide and started to page down unto the end once.
>
> * 2.0.0, 4.6.3 Carbon, 22 Sekunden
> * 2.0.0, 4.6.3 Cocoa, 23 Sekunden
> * 2.0.2, 4.7.2 Cocoa, 20 Sekunden
> * 2.1svn, 4.8.0 Cocoa, 23 Sekunden
> * 2.1svn, 4.8.0 Carbon, 23 Sekunden
> * 2.1svn, 4.6.3 Carbon, 22 Sekunden
>
> All tests I've made with \force_paint_single_char true and false.
> No real difference.
>
> You can see it yourself - there is no performance gain or loss on my system
> (Macbook Pro OS X 10.6.8).
>
> Jerry, may I ask you to provide some numbers for your use case and for the
> users guide?
>
> Stephan
All tests conducted on a MacBook Pro running 10.6.8, with the LyX window taking
half of the width and the entire height of the screen. Toolbars displayed are
Standard, View/Update, and Extra. Scrolling was done by clicking and holding in
the elevator(?) area of the scrolling control area, not on the scroll arrows.
Scrolling results by Function-downarrow or Function-arrow are the same. All LyX
were binaries downloaded from lyx.org as of today. As I noted in an earlier
post to this thread, my attempt to compile 2.1 from SVN failed. Under Energy
Saver Preferences, the Graphics hardware choice was set to "Better battery
life."
Preference line "\force_paint_single_char false" out
2.0.2
No cursor misplacement
Scroll User's Guide: 20 seconds
2.0.1
No cursor misplacement
Scroll User's Guide: 10 seconds
2.0.0
No cursor misplacement
Scroll User's Guide: 10 seconds
Preference line "\force_paint_single_char false" in
2.0.2
Cursor misplaced
Scroll User's Guide: 10 seconds
2.0.1
No cursor misplacement
Scroll User's Guide: 10 seconds
2.0.0
No cursor misplacement
Scroll User's Guide: 10 seconds
In all cases where I have indicated UG scrolling in 10 seconds, subsequent
scrollings of the User's Guide were about 10% faster, presumably because of
certain objects already being rendered on the first scrolling. However, in the
20-second UG case, I did not notice any such speed-up.
If the 20-second UG case sounds OK, it isn't—scrolling under thoses conditions,
whether in UG or a short document, are still excruciatingly slow especially in
text-heavy parts. (It gets noticeably better in areas of less text.)
In 2.0.2 with the preference line out (the mode related to my complaint), under
situations of even moderately vigorous two-finger-swipe trackpad scrolling,
large pieces are skipped in order to make the scrolling happen in a
"reasonable" time. This makes trying to spot a feature as it speeds by useless,
as the feature that you are looking for might not be displayed at all. On a
small file that prints out to a little over two (rather dense—IEEEtrans) pages,
it is not hard to make the screen jump from the beginning to the end without an
intervening screenful. (The window width set to 1/2 the screen width, requiring
3-4 screenfuls to display with normal clicking in the elevator bar.)
Even in the best cases above, scrolling is still laggy relative to native
applications (but not so bad as to be useless). There is a small but noticeable
lag from the time two-finger scrolling is initiated and when the screen content
moves, and another lag between when scrolling is stopped and when the content
stops moving. Also, try two-finger-scrolling on the trackpad in small rapid
back-and-forth movements and watch as the screen updates in nearly 180 degree
opposition to your finger movements, then repeat with e.g. TextEdit or Preview
or Skim and notice how the content movement matches your finger movement.
However, I don't want to make that a standard that LyX should meet; the
behavior under the "10-second" cases above is good enough to be useful.
Jerry
