Exploration of the Relationship Between Just-in-Time Compilation Policy and Number of Cores

摘要

Just-in-Time (JIT) compilation is a key technique for programs written in managed languages, such as Java and JavaScript. Traditionally, a conservative JIT compilation policy is used without impacting application threads too much on single-core machines. Nowadays, modern machines provide more and more processor cores, which are abundant computing resources. Modern virtual machines also have the ability to use an aggressive compilation policy, such as spawning multiple concurrent compiler threads, which is suitable to multicore situation. However, the suitable JIT compilation policy varies with the number of microprocessor cores. The goal of this work is to explore the relationship between the number of microprocessor cores on modern machines and the suitable JIT compilation policies that can enable existing as well as future VMs to realize better program performance. In this work, we design novel experiments and implement new VM configurations to effectively control the compiler aggressiveness in the industry-standard Oracle HotSpot Java VM to achieve the goal. We notice that when single core is used, traditional foreground compilation with single compiler thread has better performance. As the number of cores increases, which makes more abundant computing resources available, background compilation with more compiler threads and smaller compilation threshold reaches better performance. After comparison between the foreground compilation and the background compilation, we propose a novel compilation policy, the throttling compilation, which stops interpretation when the compilation queue is too long. This policy combines the advantages of the foreground compilation and the background compilation, which gains better performance.