I.Rogers, Lazy interprocedural analysis for dynamic loop parallelization, proc

Abstract. Dynamic compilation is becoming a dominant compilation technique. Runtime compilation has to avoid slow compile times by tar-geting optimizations to areas where it has a performance impact. For parallelization optimizations this can lead to not exposing opportuni-ties for parallelization. To enable fuller optimization we present a sim-ple interprocedural analysis. Our analysis and parallelization phases are performed as part of the Jikes RVM. Our approach succeeds in finding coarser grain loops and increased performance in a number of benchmark kernels on a research chip multi-processor architecture.