Parallelism extraction in acyclic code

摘要

We describe a framework for parallelism extraction andpartitioning in acyclic code regions. This framework is an extension ofM. Girkar's (1981) work on functional parallelism, using a Petri netmodel to represent parallel code, and applying modified optimizationtechniques to minimize the overheads of explicit synchronization. Themodifications introduced are directed towards the generation ofefficient multi threaded code. We also describe a simple partitioningtechnique that can be used to artificially increase the granularity ofthe extracted parallelism up to a desired level