This discussion makes not much sense when you compare both algorhithms. Basically Dijkstra and A* are the same. The only difference is one estimation function that you have to add to Dijkstra and you will automatically get A*. So it will not calculate each possible route on a node, it will estimate with a function where to go to find the shortest route to the goal node. Often they simply use a length-function (length between current node and goal) for that.

So if you have Dijkstra, than you almost have A*. And if you have A*, then Dijkstra is already implemented. You only have to remove this estimation function.

But when you want to calculate all routes in a tree (for each node to each possible node), then you have to use Dijkstra. This makes sense, when you pre-calculate a matrix. With such a matrix you dont have to calculate routes in real-time. You just look them up from this pre-calculated matrix. I thought this is the way Intense-X pathfinding works.

You can find more details on this here:
http://www.amazon.com/Programming-Game-Example-Mat-Buckland/dp/1556220782

Never said that i wnat to workaround the graph with A* o.O. Did A* while ago, i know what it is, the major question was just about the speed(or have i spoken so unclearly? dont know^^, sometimes my english isnt very good).

Oh and nice idea todo a complete precalculation. Will try it(maybe) later. Depends on my current progress, and the current performance.


But thanks for your suggestions^^.

EDIT: ah okay foun the point where you misunderstood me. I used a Graph and an open(closed list system to see whether a node is used or not, so i used a graph in my A* version, too

Greets
Rackscha

Of course that depends how often the "fast enough" code gets run. If paths are being searched for a very large number of objects, even comparatively fast operations can cause a significant slowdown.

where in this case it would not be fast enough or the wrong method..

What I stated was, that if the code is fast enough for the purpose it
is used in, there are other things to work on, than
wasting time on optimization.

--

I have a pathfinding for RTS units on a mobile phone game (Java phone)
that can find a path in realtiime for hundreds of objects with no
remarkable slowdown.
It is then just using a more appropriate metod than
a classic A* pathfinding.
Where it would not be as fast for single units.

The initial choice of the method is more important than the optimization of it.