On Sat, Oct 11, 2008 at 10:42, Marcus Wolschon <[EMAIL PROTECTED]> wrote: >> I'm working on testing an idea that works backwards from the >> destination and forwards from the origin at the same time and stops >> when they intersect each other. That sounds like it would be a >> solution to your problem. > I think this will be slower then to go from target to start > because it cripples your caching. In your case you have > half your memory filled with cached roads from the target > and half from the start, both with reduced hit-rates compared > to a single cache storing local roads around the one area > you are currently examining.
The real core of what I'm working on is using previous routes as partial solutions to new requests (which I'm sure has been done before, but I'm the sort of person who likes to write things for himself,) but I'm hoping that the speed of being able to follow something that's already calculated will make up for any slowdown caused by running two route finders in parallel initially. And it's currently more a proof of concept for the algorithm, so I'm not too worried about the speed yet. -- David J. Lynch [EMAIL PROTECTED] _______________________________________________ Routing mailing list [email protected] http://lists.openstreetmap.org/listinfo/routing
