Plush Language Optimizes Garbage Collector for Creative Coding

Maxime, the creator of the Plush programming language, has implemented a rewritten garbage collector (GC) to enhance performance and memory efficiency. The update aims to make the language viable for real-time applications, such as games with high allocation rates.

According to Pointers Gone Wild, the new GC implementation removes a previous reliance on a bulky hash map during the collection process, streamlining how the language handles memory. Additionally, Maxime utilized an "mmap trick" to reduce the baseline Resident Set Size (RSS), effectively lowering the amount of physical memory the process occupies. These optimizations are integrated into Plush's existing architecture, which employs actor-based parallelism. In this model, each actor maintains its own private allocator alongside a separate message allocator used specifically for receiving communications.

The Evolution of Plush

Plush is a toy, dynamically-typed, Lox-like language specifically designed for creative programming, sound, and graphics. Developed as a personal project by Maxime, the language serves as an experimental platform for exploring graphics and sound programming, drawing inspiration from a variety of established languages including Rust, Python, Lua, JavaScript, and Lox. The project has undergone a series of iterative optimizations to improve its utility for performance-heavy tasks, including the implementation of a copying GC and various interpreter enhancements tailored for workloads like raytracing.

Impact on Creative Coding

For experimental or toy languages, garbage collection overhead is frequently the primary bottleneck preventing their use in production-like environments. By removing hash map overhead and leveraging low-level memory management techniques like mmap, this update demonstrates how a high-level dynamic language can be optimized to handle high allocation rates without significant memory bloat or disruptive pauses. This is particularly critical for creative coding, where consistent frame rates and low latency are required for interactive graphics and audio synthesis.

Future Outlook

While the technical improvements are concrete, the practical application remains an area of observation. Maxime notes that the rewritten GC is presumably sufficient for a real-time game characterized by frequent allocations. Future development will likely focus on whether these memory management gains translate into stable performance across more complex, large-scale creative projects.