A Cross-Language Framework for Verifying Compiler Optimizations

Most compiler correctness efforts, whether based on validation or once-and-for-all verification, are tightly tied to the particular language(s) under consideration. Proof techniques may be replicated for other targets, and parts of the compiler chain may be shared for new input or output languages, but the extent to which common elements can be generalized across multiple targets has not been fully explored. In this paper, we lay out a general approach to specifying and verifying optimizations and transformations on low-level intermediate languages. By generalizing across elements such as concurrent memory models and single-thread operational semantics, we can build a library of facts that can be reused in verifying optimizations for dramatically different target languages, such as stack-machine and register-machine languages.