Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Watch this 14-minute conference presentation from OOPSLA 2025 that introduces CoSSJIT, a groundbreaking approach that combines static analysis with speculative optimization in just-in-time (JIT) compilers. Learn how researchers from IIT Bombay and IBM developed the first-of-its-kind method to bridge the gap between the precision limitations of JIT compilers and the conservative nature of ahead-of-time static compilers. Discover the three key contributions of CoSSJIT: identifying scenarios where static analysis makes conservative assumptions but JIT compilation can deliver precision through runtime speculation, understanding the concept of 'speculative conditions' integrated into static interprocedural dataflow analysis for stack allocation optimization, and exploring how production JIT compilers can be extended to read and enrich static analysis results with resolved speculative condition values. Explore the practical implementation on a production Java Virtual Machine that achieved a 5.7x improvement in stack allocation while maintaining functional correctness during JIT compilation. Gain insights into escape analysis, stack allocation techniques, and speculative optimization strategies that demonstrate how combining the strengths of both static and dynamic compilation approaches can significantly enhance performance in modern virtual machine environments.
