Sound and Complete Refinement for SSA with Substructural Effects
This program is tentative and subject to change.
In this talk, we will show how to give a denotational model for static single assignment (SSA), the dominant intermediate representation for modern production compilers, and show how it enables proving the correctness of a variety of compiler optimizations. Because SSA is an effectful programming language, with features like undefined behaviour, state, and concurrency, we need to equip our calculus with a directed notion of program refinement (an inequational theory) rather than the equational theory as with most purely functional languages. Furthermore, which refinements are sound depends on which effects are used, and whether effects are duplicated, moved, or dropped. This gives rise to a natural notion of substructural effect system for our language.
This program is tentative and subject to change.
Sun 12 OctDisplayed time zone: Perth change
14:00 - 15:30 | |||
14:00 30mTalk | Compositions for Free: Reasoning about Effectful Lenses HOPE | ||
14:30 30mTalk | Sound and Complete Refinement for SSA with Substructural Effects HOPE | ||
15:00 30mTalk | Substructural Weakest Preconditions in Fibrations HOPE John Li Northeastern University, Pedro H. Azevedo de Amorim University of Oxford, Ryan Doenges Northeastern University |