Distributed array processor

The CRAY-1 vector processing computer at the Science Research Council s (S.E.R.C) Daresbury Laboratory, is at the centre of a network providing large scale computational facilities for Universities in the United Kingdom. This is the only supercomputer available at present to Quantum Chemists in the U.K., and this article will therefore be restricted to experience gained on the CRAY-1, although this experience will undoubtedly be relevant to future applications on machines such as the ICL Distributed Array Processor (DAP) (see reference (2) for a detailed description) and the CDC Cyber 203/205. 

Agglomerative Clustering Using the ICE Distributed Array Processor. 

M. P. Allen, Theor. Chim. Acta, 84, 399 (1993). Simulation of Condensed Phases Using the Distributed Array Processor. 

D. J. Wild and P, Willett,/. Chem. Inf. Comput. Sci., 34,224 (1994). Similarity Searching in Files of Three-Dimensional Chemical Structures Implementation of Atom Mapping on the Distributed Array Processor DAP-610, the MasPar MP-1104 and the Connection Machine CM-200. 

Rasmussen, E.M. Willett, P. Wilson, T. Chemical Structure Handling Using th< Distributed Array Processor . In these Proceedings. 

CHEMICAL STRUCTURE HANDLING USING THE DISTRIBUTED ARRAY PROCESSOR... 

Gostick, R.W. Software and Algorithms for the Distributed Array Processor . ICL Technical Journal 1979,2, 116-135. 

Parallel Database Processing. Text Retrieval and Cluster Analysis Using the Distributed Array Processor Willett, P. Rasmussen, E.M. Pitman London, 1990. 

Glinski and Roe, 1994] Glinski, S. and Roe, D. (1994). Spoken Language Recognition on a DSP Array Processor. IEEE Trans, on Parallel and Distributed Systems, 5(7) 697-703. 

E. Kyriakis-Bitzaros and C. Goutis. An efficient decomposition technique for mapping nested loops with constant dependencies onto regular processor array processors. Journal of Parallel and Distributed Computing, 16, pages 258-264, 1992. 

The eigenvector and potential vector arrays are row-wise distributed in the same fashion. The program only requires an index function indx(i, j, k) which returns the number of the processor in which the grid point (i, j, k) resides. 

In collaboration with our IRISA partner, we have succeeded in applying the dependence mapping strategy to retrieve the folded array of section 4 in a systematic way [2]. Automated synthesis techniques that include extensions to the basic linear scheduling and allocation techniques, and that can handle the APP demonstrator, are also presented in chapters 4, 5, and 6. Moreover, if several different processor elements with different modes have to be dealt with, they need to be controlled by local controllers. For this purpose, control signals have to be distributed, as discussed in chapter 4. One particular extension that is very useful for decreasing the execution time even further is discussed in the next subsection. 

DM-MIMD MPP machines are undoubtedly the fastest-growing class in the family of supercomputers, although this type of machine is more difficult to deal with than shared-memory machines and processor-array machines. For shared-memory systems the data distribution is completely transparent to the user. This is quite different for DM-MIMD systems, where the user has to distribute the data over the processors, and also the data exchange between processors has to be performed explicitly. The initial reluctance to use DM-MIMD machines has decreased lately. This is partly due to the now-existing standards for communication software such as MPI (message passing interface) and PVM (parallel virtual machine) and is partly because, at least theoretically, this class of systems is able to outperform all other types of machines. 

---