Eclipse OMR garbage collection for tracing JIT-based virtual machines.

The RPython framework provides a toolchain with reusable tracing Just-In-Time (JIT) compilation, and garbage collection (GC) that Virtual Machines (VMs) written in it inherit. This toolchain has provided an efficient way over the years for implementing dynamic programming languages. However, it is unclear if garbage collection modularity through a framework like Eclipse OMR has any advantages to offer languages that are RPython-based. In this paper, we present the design, implementation, and experimental evaluation of OMR-based garbage collection for tracing-JIT-based virtual machines that use the RPython translation toolchain. Our findings agree with existing work that using a GC framework results in a GC that is simple, easier to maintain, and has fewer bugs. The experimental results show that GC modularity through the Eclipse OMR GC framework can still achieve high performance for components like garbage collection for dynamic languages. We evaluate our implementation by applying it to two RPython-based interpreters, namely, PyPy, a Python implementation, and PyHP, a PHP implementation.