Parallel Astar search on message-passing architectures

The authors compare three parallel Astar search algorithms: the shared-list algorithm where a search space is shared among processors; the static-distribution algorithm, where a search space is distributed once to all processors; and the continuous-diffusion algorithm, where a search space is continuously redistributed. The continuous-diffusion algorithms outperform the other two algorithms on message-passing architectures. The grid-flow technique developed for continuous diffusion is of general importance for nearest-neighbor algorithms that need to share some global information.<
>