Comparing Alternative Evaluation Strategies for Stream-Based Parallel Functional Languages

摘要

In parallel functional languages, like Eden, lazy and strict evaluation are commonly mixed. Thus, the parallel performance of these languages depends on the strategy used to fix the degrees of laziness/strictness. By using an implementation of Eden's operational semantics, we analyze the influence of alternative evaluation models on Eden skeletons performance. In particular, we assess the performance of different implementations of a skeleton that uses stream-based communications.