Optimistic Stack Allocation and Dynamic Heapification for Managed Runtimes

Explore a 20-minute conference talk from PLDI 2024 that introduces a novel approach to optimizing memory allocation in managed object-oriented languages. Delve into the proposed static+dynamic scheme that leverages precise static escape analysis for stack allocation while addressing soundness and efficiency concerns in runtime environments. Learn about three key ideas: optimistic stack allocation during JIT compilation, dynamic heapification to handle optimism invalidation, and stack ordering to reduce overhead. Discover how this approach, implemented in the Soot optimization framework and Eclipse OpenJ9 VM, significantly reduces garbage collection cycles and improves performance, particularly in constrained-memory environments. Gain insights into the evaluation results comparing this scheme with existing escape analysis techniques.