Jean-Christophe Weill. The ABDADA distributed minimax search algorithm. In Proceedings 1996. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
Transposition Table Driven Scheduling for Two-Player Games - Kishimoto (2002) (Correct)
....can be generalized so that all the promising branches must be searched before the rest of the branches at a node are searched. This generalization is applied to Feldmann s YBWC algorithm [15] Although there are many variants of YBWC, the only di erences are in 27 the implementation details [25, 53, 21]. The idea of YBWC originates from PV Split [28] which applies the above strategy only at PV nodes. YBWC can be seen as a generalized version of PV Split because the strategy of YBWC is applied to all nodes. YBWC has the following bene ts: First, the minimal game trees can be parallelized with ....
....which received the stolen work. Using a partitioned transposition table causes a communication latency in looking up the transposition entry for the node. Replicated tables also have extra communication overhead when broadcasting the result of searching the node. ABDADA is a variant of YBWC [53]. However, ABDADA does not use work stealing. Instead, ABDADA uses a shared transposition table to control the parallel search. All the processors start searching the root node simultaneously. Each transposition table entry has a eld for the number of processors entering a node, which is used to ....
[Article contains additional citation context not shown here]
J.-C. Weill. The ABDADA Distributed Minimax-Search Algorithm. International Computer Chess Association Journal, 19(1):3-16, 1996.
A Taxonomy Of Parallel Game-Tree Search Algorithms - Brockington (1996) (1 citation) (Correct)
....fffi Transputers Chess NegaScout distributed 6.5 (n=8 8TT) David, 1993) messages CABP Sequent Artificial fffi shared 4. 6 (n=9) Cung, 1994) Balance Trees memory Jamboree CM 5 Chess NegaScout distributed 50 (n=512) Kuszmaul, 1994) messages ABDADA CM 5 Chess NegaScout distributed 15.85 (n=32) (Weill, 1996) ( Othello) messages Dynamic Multiple PV Split AP 1000 Artificial PVS none 32 (n=64) Marsland and Gao, 1995) Trees APHID Sparc 2 Chess NegaScout local 6.04 (n=16) Brockington and Schaeffer, 1995) Network Table 2: Comparison of Parallel fffi based Game Tree Search Implementations A Taxonomy ....
....non evaluated children in the tree. Although both fffi and YBWC use the same decision method for allowing or denying parallelism, fffi uses a shared transposition table to keep the processors working on different parts of the tree, while YBWC uses master slave relationships. In a later paper (Weill, 1996), the combined method was called Alpha Beta Distribut e avec Droit d A inesse, or ABDADA. He showed that ABDADA yields greater speedups than YBWC on a CM 5 when studying chess trees. ABDADA also yields similar speedups to YBWC when studying fffi trees generated by an Othello program. Dynamic ....
Weill, J.-C. (1996). The ABDADA Distributed Minimax-Search Algorithm. ICCA Journal, Vol. 19, No. 1, pp. 3--16.
Asynchronous Parallel Game-Tree Search - Brockington (1998) (7 citations) (Correct)
....3.32 (n=16) 57] TC2000 memory fffi Transputers Chess NegaScout distributed 6.5 (n=8 8TT) 25] messages CABP Sequent Artificial fffi shared 4. 6 (n=9) 24] Balance Trees memory Jamboree CM 5 Chess NegaScout distributed 50 (n=512) 53] messages ABDADA CM 5 Chess NegaScout distributed 15.85 (n=32) [96] ( Othello) messages Dynamic Multiple PV Split AP 1000 Artificial PVS none 32 (n=64) 62] Trees Table 3.2: Comparison of Parallel fffi based Game Tree Search Implementations 50 The second column describes the underlying hardware used to host the selected implementation. A software simulation ....
....other non evaluated children in the tree. Although both fffi and YBWC use the same heuristic for allowing or denying parallelism, fffi uses a shared transposition table to keep the processors working on different parts of the tree, while YBWC uses master slave relationships. In a later paper [96], the combined method was called Alpha Beta Distribut e avec Droit d A inesse, or ABDADA. Weill showed that ABDADA yields greater speedups than YBWC on a CM 5 when studying chess trees. ABDADA also yields similar 60 speedups to YBWC when studying fffi trees generated by an Othello program. ....
[Article contains additional citation context not shown here]
J.-C. Weill. The ABDADA Distributed Minimax-Search Algorithm. ICCA Journal, 19(1):3--16, 1996. (49,59,76)
Multithreaded Pruned Tree Search - In Distributed Systems (Correct)
No context found.
Jean-Christophe Weill. The ABDADA distributed minimax search algorithm. In Proceedings 1996.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC