Parallel Algorithm Fundamentals and Analysis

. This session explores, through the use of formal methods, the "intuition" used in creating a parallel algorithm design and realizing this design on distributed memory hardware. The algorithm class NC and the LSTM machine are used to show why some algorithms realize their promise of speedup better than others and the algorithm class NP is used to show why other algorithms will never be good for parallelization. The realities of algorithm design are presented through partitioning and mapping issues and models. Finally, correctness through cooperative axiomatic reasoning provides an additional basis for understanding parallel algorithm design and specification and is used for run-time assurance of distributed computing systems through operational evaluation. Key Words: Algorithm Design, Embeddings, Speedup, Class NC, Reasoning, Operational Evaluation This paper appears, in its entirety, in the Proceedings of the International Summer Institute on Parallel Computer Architectures, Language..