In article ,
Julian Scarfe wrote:
Repeat for other departure airports of interest. Processing time for one
departure airport for my network was about 30s on a fairly typical
desktop
machine. YMMV, literally. ;-)

"Ben Jackson" wrote in message
news:rTzbd.249118$D%.142632@attbi_s51...

I think you can do better than that by ordering your edges better. You
know more than just edge costs, you also have coordinates for each node.
You can choose to explore edges that move you closer to the destination
first.

IIRC, the key to making Dijkstra fast is to find a solution as early as
possible. That establishes a baseline cost that allows massive pruning
of the search space.

That may be the case. I should have said that what I was trying to do was
create a server-based system for large numbers of users with different
dep-dest pairs. In that case, I'm looking at all potential destinations in
parallel. The strategy for a client-based system starting from scratch for
every single flight might be rather different.

Julian