Abstract: "Load-dispatching strategies are proposed for a parallel inference machine prototype, and their performance are evaluated by the simulation based on the loosely-coupled cluster model, using 6-queens benchmark. Sender-initiate concept is appropriate to the bunch layer of a parallel inf...

Ausführliche Beschreibung

Bibliographische Detailangaben

Format:	Buch
Sprache:	English
Veröffentlicht:	Tokyo, Japan 1988
Schriftenreihe:	Shin-Sedai-Konpyūta-Gijutsu-Kaihatsu-Kikō <Tōkyō>: ICOT technical report 371
Schlagworte:	Logic programming Machine theory Parallel processing (Electronic computers)
Zusammenfassung:	Abstract: "Load-dispatching strategies are proposed for a parallel inference machine prototype, and their performance are evaluated by the simulation based on the loosely-coupled cluster model, using 6-queens benchmark. Sender-initiate concept is appropriate to the bunch layer of a parallel inference machine prototype. The strategy in which the cluster with maximum ready goals dispatches a goal to the cluster with minimum ready goals brings the lowest load-dispatching rate limit, but it is not expected to realize stable performance at real program execution, since it covers too narrow load-dispatching rate region The strategy in which the goal dispatch target cluster is determined at random and then this goal dispatch is aborted on the condition that the dispatch target cluster has more ready goals than the dispatching cluster or on the condition that the dispatching cluster has fewer ready goals than the threshold brings the second lowest load-displacing rate limit and is expected to realize high performance stably at real program execution. More than 70% averaged utilization is achieved in the region that the load-displacing rate is higher than 5%. It is confirmed that 0.54 of the maximum performance of a parallel inference machine prototype can be achieved by applying this load-dispatching strategy.
Beschreibung:	15 S. graph. Darst.