A Novel Probabilistic Data Flow Framework

Classical data ow analysis determines whether a data ow fact may hold or does not hold at some program point. Probabilistic data ow systems compute a range, i.e. a probability, with which a data ow fact will hold at some program point. In this paper we develop a novel, practicable framework for probabilistic data ow problems. In contrast to other approaches, we utilize execution history for calculating the probabilities of data ow facts. In this way we achieve signiicantly better results. EEectiveness and eeciency of our approach are shown by compiling and running the SPECint95 benchmark suite.