Stack-Liberated Abstract Garbage Collection
summary
Paper claims abstract garbage collection can keep context irrelevance by scoping the store to a single local environment and passing effects explicitly. It implements this in an abstracting definitional-interpreter / pushdown-CFA setting, gives each evaluation an effect log that is replayed on the caller’s store, and recovers perfect stack-precise pushdown CFA while still collecting garbage. Result is a fully compositional analysis that preserves control precision, enables stronger memoization, and empirically restores precision lost by global stores.
AI-written summary · about