Parallel algorithm development workbench

The authors discuss the need for rapid prototyping during algorithm and system development. They propose a workbench environment to support an analysis effort to optimize algorithms, communications schemes, and CPU designs for parallel processing applications. This environment allows these facets of a parallel system to be investigated and optimized, either singly or jointly. The authors describe how an algorithm can be evaluated at various levels of abstraction and how the environment can support the design decomposition down to real code. The environment should allow algorithm and system designers with minimal modeling experience to experiment and optimize applications expending minimal effort