Scalable Linear Invariant Generation with Farkas' Lemma

Scalable Linear Invariant Generation with Farkas' Lemma

Explore a conference talk on generating tight linear invariants for affine programs using Farkas' Lemma. Delve into the challenges of scalability in existing approaches and discover two novel improvements that significantly enhance performance. Learn about the innovative technique of generating invariants at individual program locations, enabling parallel processing and reducing combinatorial explosion. Examine the segmented subsumption testing method in CNF-to-DNF expansion for identifying local subsumptions. Understand how these advancements maintain accuracy while dramatically improving runtime efficiency, marking a major breakthrough in Farkas' Lemma-based approaches for linear invariant generation in nearly two decades.