May-Happen-in-Parallel Analysis for Priority-Based Scheduling

A may-happen-in-parallel (MHP) analysis infers the sets of pairs of program points that may execute in parallel along a program’s execution. This is an essential piece of information to detect data races, and also to infer more complex properties of concurrent programs, e.g., deadlock freeness, termination and resource consumption analyses can greatly benefit from the MHP relations to increase their accuracy. Previous MHP analyses have assumed a worst case scenario by adopting a simplistic (non-deterministic) task scheduler which can select any available task. While the results of the analysis for a non-deterministic scheduler are obviously sound, they can lead to an overly pessimistic result. We present an MHP analysis for an asynchronous language with prioritized tasks buffers. Priority-based scheduling is arguably the most common scheduling strategy adopted in the implementation of concurrent languages. The challenge is to be able to take task priorities into account at static analysis time in order to filter out unfeasible MHP pairs.