Abstract
A new parallel search algorithm running on a large computer cluster solves a popular board game by efficiently computing the best moves from all reachable positions. As such, the algorithm uses the main memories for frequently and randomly accessed data and stores terbytes of less frequently acessed intermediate results on disks. All processors repeatably inform each other about positions' intermediate values, generating more than a petabit of interprocessor communication as well as terabytes of disk I/O.
Original language | English |
---|---|
Pages (from-to) | 26-33 |
Journal | Computer |
Volume | 38 |
Issue number | 10 |
DOIs | |
Publication status | Published - 2003 |