Section: New Results
Semantic Array Dataflow Analysis
Participants : Laure Gonnord, Paul Iannetta.
Together with Lionel Morel (Insa/CEA) and Tomofumi Yuki (Inria, Rennes), we revisited the polyhedral model's key analysis, dependency analysis. The semantic formulation we propose allows a new definition of the notion of dependency and the computation of the dependency set. As a side effect, we propose a general algorithm to compute an over-approximation of the dependency set of general imperative programs.
We argue that this new formalization will later allow for a new vision of the polyhedral model in terms of semantics, which will help us fully characterize its expressivity and applicability. We also believe that abstract semantics will be the key for designing an approximate abstract model in order to enhance the applicability of the polyhedral model.
The results is published in a research report .