Runtime support for execution of fine grain parallel code on coarse grain multiprocessors

The goal of this research is to provide systems support that allows fine grain, data parallel code to execute efficiently on much coarser grain multiprocessors. The task of writing parallel applications is simplified by allowing the programmer to assume a number of processors convenient to the algorithm being implemented. This paper describes and evaluates a runtime approach that efficiently manages thousands of "virtual" processors per actual processor. Tight integration and specialization of scheduling, communication, and context switching is used to significantly reduce the overhead of running fine grain parallel code. A Paragon prototype of this runtime approach is evaluated by comparing implementations of two numerical problems. The overhead due to scheduling, communication, and context switching is analyzed. The implementation and analysis show that fine grain code can be efficiently executed in a coarse grain multiprocessor using a runtime approach.<
>