Rene Wunderlich <[EMAIL PROTECTED]> writes:
> Now 2 new patches for _unstable ;)
>
> 0001_removing_unused_gd_images.patch.gz
This is applied (and a bit more). Thanks.
> 0002_optimizePngTasks.patch.gz
>
> add a new version from the threaded optimizePng
> in lib/optimizePngTasks.pm
>
> WARNING: it's experimentell DO NOT USE IT!!!!!!
>
> it only work with Fork=0 and the new option cores=1 or more ;)
> on my 4 cores is this version 2-3 times faster as the forked version
>
> i have it only teset with the xy command and not in loop modus
>
> the problem is the GDlib is not thread save...
I have not tried that one because my knowledge about threading is very
basic. So what do you do about GD?
I think the whole forking/threading subject needs rethinking.
Currently, the forking is done by zoom levels (one fork per zoom level
maximum) and one layer is processed after the other. This means that
for the captionless layer there is only 1 fork (not that there would
be any work to split up anyway), for maplint there are max. 5 forks
and for the tile layer there are max. 6 forks. On top of that the
workload across the zoom levels es very uneven (z12 will render much
quicker than z17).
>
> sorry for my bad englich i hope my code is better ;)
^
Yeah, this English is really bad ;-)
Matthias
_______________________________________________
Tilesathome mailing list
[email protected]
http://lists.openstreetmap.org/listinfo/tilesathome
