The content discusses the development of realizability and realization logic for semantics-guided synthesis. It introduces a new algorithm for synthesizing solutions to tasks, emphasizing the importance of compositionality and abstraction. The approach aims to reduce backtracking and improve scalability in search-based synthesis.
The syntax-guided synthesis initiative has been influential in program synthesis technology, focusing on standardized input formats independent of solvers. Semantics-guided synthesis offers domain-independent solutions by allowing users to define program semantics. Realizability logic analyzes sets of programs eagerly, ensuring efficient synthesis without backtracking.
The paper introduces realization logic to derive rewriting steps between proof outlines in realizability logic. The approach eliminates unsuitable alternatives iteratively until a suitable program is found. The algorithms developed aim to automate the process and provide guarantees of success without backtracking.
Experiments with a prototype tool in memory reclamation for lock-free data structures demonstrate the effectiveness of the proposed approach. The content highlights the importance of user guidance, loop invariants, and abstract domains in achieving successful program synthesis.
Vers une autre langue
à partir du contenu source
arxiv.org
Questions plus approfondies