[josm-dev] JOSM Mappaint - major improvements

Rolf Bode-Meyer robome at gmail.com
Tue Jan 13 11:58:38 GMT 2009


2009/1/13 Ulf Lamping <ulf.lamping at googlemail.com>:

> After a week of intensive work, I was able to *significantly* improve
> the performance of JOSM mappaint (the standard way JOSM paints the OSM
> data).

Yes, saw and tested this version yesterday. It definitively is a big
improvement. Thank you very much for your effort.

What I don't understand - with your change and before - is that paint
speed seems to depend on the amount of data in the layer even if it's
outside the view.
I've downloaded an area with scale of 40m (64 KB), JOSM fullscreen (on
1600x1200) and downloaded a huge area (14 MB) containing the smaller
one to another layer.
Moving a way on the small area takes about 62ms, moving the same way
with the same data in view but having much more outside the view it
takes about 140ms.

Though I'm very happy about not being linear to the amount of data on
the layer, I still want to understand the effect.

The complete data, Mappaint Profiler (cache=true, fillareas=true,
fillalpha=50%):
Prepare  :    0ms
Relations:   15ms, n=  141
Areas    :   63ms, n= 1504, visible=4
Ways     :   15ms, n=10510, visible=17
Selected :    0ms, n=    2
Nodes    :   16ms, n=60646, visible=116
Virtual  :   16ms, n=12014
All      :  140ms

Prepare  :    0ms
Relations:   16ms, n=    2
Areas    :   31ms, n=    5, visible=4
Ways     :    0ms, n=   14, visible=14
Selected :    0ms, n=    2
Nodes    :    0ms, n=  316, visible=116
Virtual  :    0ms, n=   19
All      :   62ms

And another point I already posted about a month ago. Having the
LVA-Oberpfalz layer visible in background makes painting about 4 times
slower. Interestingly this degration isn't shown by the profiler but
visible. And it only happens with this image layer, no other like
Yahoo or Landsat imagery. Any ideas?

Rolf




More information about the josm-dev mailing list