並列LU分解を用いる系統過渡安定度計算に適したマルチプロセッサアーキテクチャ

書誌事項

タイトル別名
  • A Multiprocessor System Architecture Suited for Power System Transient Calculation
  • ヘイレツ LU ブンカイ オ モチイル ケイトウ カト アンテイド ケイサン

この論文をさがす

抄録

In the field of power system calculation, much study has been made in the area of applying parallel computers in order to increase computation speed. However, unlike examples in other application areas, such as circuit simulation or computational fluid dynamics, the performance gain is small. Also, the performance gain begins to saturate just after small number of processors are utilized. This is due to the essential feature of power system calculation: mainly (1) The size of the problem is usually much smaller than the other application fields. (2) The sparsity of the problem is high. These two features results in the fact that effect of interprocessor communication overhead is severely reflected.<br>In this study, the authors tried to determine the most suitable parallel architecture for Power System Calculation, restricting the number of processors to 16 or so. The architectural feature we wanted to detemine are Memory architecture and Inter processor connection method. The candidates we chose were (a) Distributed memory architecture connected by an inter-processor communication network, (b) Distributed memory architecture connected by a bus, (c) Shared memory architecture connected by a bus, and (d) Composite distributed/shared memory architecture connected by a bus. We evaluated each architecture using the data for transient stability analysis from a single PE excution, and evaluating each memory operation, necessary interprocessor communication and computation. The results showed that while it was very hard to attain much speed up using a distributed shard memory architecture connected by an interconnection network, it was possible to attain a speedup of 6.5 using 16 PEs on a Composite distributed/shared memory architecture connected by a bus.

収録刊行物

被引用文献 (1)*注記

もっと見る

参考文献 (19)*注記

もっと見る

詳細情報 詳細情報について

問題の指摘

ページトップへ