An all-nodes (or single-source) search is a procedure in which all reachable nodes are searched from a common source vertex. Once an all-nodes search is complete, shortest grid paths can be efficiency sampled between the source vertex and any reachable sample vertex. All five search methods in the previous section have a corresponding all-nodes method, implemented in Central64 using mostly the same code.
The classic search algorithm by Dijkstra (1959) is the all-nodes version of A* Search, and the foundation of all of the path planning methods in the Central64 library. In contrast to A*, nodes in Dijkstra Search are expanded in order of increasing g-cost rather than f-cost.
Canonical Dijkstra, the all-nodes version of Jump Point Search, was introduced by Sturtevant & Rabin (2016).
Bounded Canonical Dijkstra is Canonical Dijkstra with a finite jump cost. It is the all-nodes version of Bounded Jump Point Search. The method is unconventional, since there is no clear benefit to bounding the jump cost if all reachable nodes must be searched anyway.
Mixed Dijkstra is the all-nodes version of Mixed A*. Blocks in Mixed Dijkstra are processed in order of increasing g-cost instead of f-cost.
Mixed Canonical Dijkstra is the all-nodes version of Mixed Jump Point Search. Blocks in Mixed Canonical Dijkstra are processed in order of increasing g-cost instead of f-cost.