Hi Jan,
this would be a nice and welcome addition, yes. See also:
https://github.com/graphhopper/graphhopper/issues/479
It should be safe and faster but requires a multiple amount of RAM than
the normal preparation already takes, so the default should still be
thread_count=1.
Also you should not use MMAP which is not thread safe (only MMAP_SYNC or
RAM_STORE). Additionally one would have to make sure that
ghStorage.freeze();
is called synchronized.
Kind Regards,
Peter
On 08.09.2015 17:08, Jan Torben Heuer wrote:
> Hi everybody,
>
> Now that we can have multiple CHs for one base graph: Could I create CHs in
> parallel? I have sufficient idle cores. But is it possible, save and
> performant?
>
> I currently use a simple loop:
>
> for(Weighting w : weightings) {
> PrepareContractionHierarchies pch = // create CH
> pch.doWork();
> pch.close();
> }
>
> Jan
> _______________________________________________
> GraphHopper mailing list
> [email protected]
> https://lists.openstreetmap.org/listinfo/graphhopper
_______________________________________________
GraphHopper mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/graphhopper
