Speeding Up the Plush Garbage Collector
Summary
The article is a technical blog post detailing performance tuning of a copying garbage collector in the Plush language. It explains the shift from a hash-map-based forwarding approach to a Cheney-like algorithm with an undo-list, achieving substantial speedups (down to 7 ms for a million objects) and memory optimizations via mmap-backed dynamic message allocators. It also discusses trade-offs, profiling, and potential future improvements, with references to the Plush project on GitHub.