Moito-n Runtime Slowdown Conjecture

Determine whether the Moito variant that uses Grammar Flow Analysis to compute precise hole abstractions (Moito-n) is sometimes slower than the variant that uses default top abstractions (Moito-h), specifically due to the cost of propagating non-top interval semantics when the added precision does not increase pruning, and ascertain the conditions under which this slowdown occurs.

Background

The paper introduces Moito, a SemGuS-based synthesis tool with two configurations: Moito-h, which uses default hole abstractions set to the top interval, and Moito-n, which computes more precise hole abstractions via Grammar Flow Analysis. While Moito-n can prune more candidates in many benchmarks, the authors note that carrying and propagating precise intervals may introduce overhead.

In the evaluation of precise hole abstractions, the authors explicitly conjecture a runtime tradeoff: that Moito-n can be slower than Moito-h when the increased precision fails to yield additional pruning, due to the cost of propagating non-top intervals through the abstract semantics.