Aperiodic mapping system using power-of-two stride access to interleaved devices

作者: Scott P. Wakefield , Gregoty F. Pfister , Keven P. McAuliffe , Vern A. Norton , Evelyn A. Melton

DOI:

关键词:

摘要: An aperiodic mapping procedure for the of logical to physical addresses is defined as a permutation function generating optimized stride accesses in an interleaved multiple device system such large, parallel processing shared memory wherein comprises bit-matrix multiplication presented first (logical) address with predetermined matrix produce second (physical) address. The maps from space improved performance system. Assuming that has n bits and 2d separately accessible devices (where d≦n) utilizes n-d offset within referenced node. includes performing bit between successive roows said d

参考文章(26)
William Jalby, Jean Marc Frailong, Jacques Lenfant, XOR-Schemes: A Flexible Data Organization in Parallel Memories. international conference on parallel processing. pp. 276- 283 ,(1985)
Akira Hattori, Takamitsu Tsuchimoto, Data processing system having hierarchical memories ,(1980)
Fabrizio Saltini, Vittorio Eccettuato, Electronic computer with cyclic program memory ,(1970)
Terry A. Welch, Martin Cohn, Address sequencer for overwrite avoidance ,(1982)
Henri Verdier, Jacques Bienvenu, Claude Carre, Duc Luu, Process synchronization utilizing semaphores ,(1981)
Peter M. Kogge, Skewed matrix address generator ,(1981)