Coexistence of PPE/VGC/Yield Memory with JIT, AOT, SIMD, and GPU acceleration while preserving determinism and bounded memory

Investigate integration strategies by which just-in-time compilation, ahead-of-time compilation, SIMD acceleration, and GPU offloading can coexist with the PPE runtime architecture, VGC, and Yield Memory while maintaining determinism and strictly bounded memory behavior.

Background

The paper deliberately excludes JIT, AOT, SIMD, and GPU techniques to focus on architectural feasibility, determinism, and bounded memory. However, practical high-performance systems often rely on these technologies.

The authors explicitly flag exploring coexistence with these techniques as a future direction, emphasizing the requirement to preserve determinism and memory bounds—key properties of the proposed architecture.

References

Several directions remain open for future exploration. Finally, while this work intentionally excludes just-in-time compilation, ahead-of-time compilation, SIMD acceleration, and GPU offloading, future research may explore how such techniques could coexist with PPE, VGC, and Yield Memory without compromising determinism or bounded memory behavior.

— VGC: A High-Performance Zone-Based Garbage Collector Architecture for Python with Partitioning and Parallel Execution  (2512.23768 - M, 29 Dec 2025) in Section Future Goals