Le 07/07/2016 23:58, Jean-Marc Lasgouttes a écrit :
Le 07/07/2016 19:06, Pavel Sanda a écrit :
Can you try if you can see difference between 2.1 and master for this
scenario:
1. set high speed rate for your keyboard: xset r rate 300 180
2. load user guide, fullscreen
3. Hold right arrow until you reach the end of the first page and count
the time
(even approximately, the difference is quite noticeable here).
I remember trying that some time ago and being very frustrated with
having
a keyboard that I could not use anymore :) I'll try again. For now I am
Use xset r rate 300 20 to get back.
OK, I did some testing with the down arrow from the top to the bottom of
the User Guide, because the right arrow really requires too much time
for some reason. [Rereading your message I notice only now that you
meant first page only. Oh well.]
Test run on non-debug builds:
Qt4.8.7 builds:
2.1.5: 3:57min
2.2.1dev: 5:14min
2.3.0dev (with my new secret patch): 1:16min
Qt5.5.1 builds (with build-type=dev, I'm lazy)
2.3.0dev (with only the breakAt patch): 1:32min
2.3.0dev (with the new secret patch): 1:25min
OK, I recompiled with Qt5 in profiler mode with the complete patch (and
with nothing). The results are better:
Qt5.5.1 builds (with build-type=prof)
2.3.0dev (with the complete patch): 1:00min
2.3.0dev (plain master): 1:05min
So it seems that all this caching is not very useful since Qt5 already
does it internally. Therefore I should probably make the caching
Qt4-only for the sake of simplicity and of memory.
Thoughts?
JMarc
