Stack-Liberated Abstract Garbage Collection скачать в хорошем качестве

Stack-Liberated Abstract Garbage Collection

Properly devised, stack-precise control-flow analyses—those that model control using a pushdown system—exhibit the natural and appealing property of context irrelevance under which the result of an evaluation does not depend on its continuation. The addition of abstract garbage collection to such analyses destroys context irrelevance as the precise abstract transitions taken during the evaluation of the expression depend on (the root set of reachability contained in) the continuation. This paper reconciles context irrelevance and abstract garbage collection by narrowing the scope of the store to just the local environment. Within this narrowed scope, garbage collection is trivially straightforward. However, with the store no longer acting as a global repository of effects, we require an explicit channel between evaluations to communicate effects. We define an effect log for each evaluation to act as this channel, which we can apply to explicitly propagate effects. After the dust settles, we’re left with a fully-compositional analysis that exhibits the benefits of perfect stack precision, garbage collection, and context irrelevance. Presenter: Kimball Germane Authors: Kimball Germane and Michael Adams Paper: https://thomas.gilray.org/scheme-2019... Presented at the Scheme workshop 2019: https://icfp19.sigplan.org/home/schem..., part of ICFP'19: icfp19.sigplan.org