PEPE: A Trace-Driven Simulator to Evaluate Reconfigurable Multicomputer Architectures

Recent research on parallel systems with distributed memory has shown that the most difficult problem for system designers and users is related with the interconnection network. In this paper, we describe a programming and evaluating tool for multicomputers, named PEPE. It allows the execution of parallel programs and the evaluation of the network architecture. PEPE takes a parallel program as input and generates a communication trace obtained from this program. Next, PEPE simulates and evaluates the behaviour of a multicomputer architecture for this trace. The most important parameters of the multicomputer can be adjusted by the user. PEPE generates performance estimates and quality measures for the interconnection network. Another important feature of this tool is that it allows us to evaluate networks whose topology is reconfigurable. Reconfigurable networks are good alternatives to the classical approach. However, only recently this idea became the focus of much interest, due to technological developments (optical interconnection) that made it more viable. A reconfigurable network yields a variety of possible topologies for the network and enables the program to exploit this topological variety to speed up the computation.

[1]  Larry Rudolph,et al.  PIE: A Programming and Instrumentation Environment for Parallel Processing , 1985, IEEE Software.

[2]  William J. Dally,et al.  Deadlock-Free Message Routing in Multiprocessor Interconnection Networks , 1987, IEEE Transactions on Computers.

[3]  Alan H. Karp,et al.  Programming for Parallelism , 1987, Computer.

[4]  Kathleen M. Nichols,et al.  Modeling multicomputer systems with PARET , 1988, Computer.

[5]  Deborah Hix,et al.  Human-computer interface development: concepts and systems for its management , 1989, CSUR.

[6]  John L. Hennessy,et al.  Multiprocessor Simulation and Tracing Using Tango , 1991, ICPP.

[7]  David Peleg,et al.  The Power of Reconfiguration , 1991, J. Parallel Distributed Comput..

[8]  Eric A. Brewer,et al.  PROTEUS: a high-performance parallel-architecture simulator , 1992, SIGMETRICS '92/PERFORMANCE '92.

[9]  José M. García Carrasco A new language for multicomputer programming , 1992 .

[10]  C. Bonello,et al.  Programming environment for phase-reconfigurable parallel programming on SuperNode , 1992, [1992] Proceedings of the Fourth IEEE Symposium on Parallel and Distributed Processing.

[11]  José M. García A new language for multicomputer programming , 1992, SIGP.

[12]  José Duato,et al.  A New Theory of Deadlock-Free Adaptive Routing in Wormhole Networks , 1993, IEEE Trans. Parallel Distributed Syst..

[13]  José Duato,et al.  Dynamic reconfiguration of multicomputer networks: limitations and tradeoffs , 1993, 1993 Euromicro Workshop on Parallel and Distributed Processing.

[14]  Jean-Marc Adamo,et al.  Programming Environment for Phase-Reconfigurable Parallel Programming on SuperNode , 1994, J. Parallel Distributed Comput..

[15]  D. Peleg,et al.  The Power of Recon guration , 1998 .