Performing data flow analysis in parallel

The authors have designed a family of parallel dataflow analysis algorithms for execution on a message-passing MIMD (multiple instruction multiple data) architecture, based on general purpose, hybrid dataflow analysis algorithms. They have exploited the natural task partitioning of the hybrid algorithms and have explored a static mapping-dynamic scheduling strategy. Alternative mapping-scheduling choices and refinements of the flow graph condensation utilized are discussed. This parallel hybrid algorithm family is illustrated on the reaching definitions problem, although parallel algorithms also exist for many interprocedural (e.g., aliasing) and intraprocedural (e.g., available expressions) problems