Spegion: Implicit and Non-Lexical Regions with Sized Allocations

2025-06-02

AI Summary

Existing region-based memory management systems either rely on lexical scoping with explicit annotations or require a substructural type system to ensure safety for non-lexical regions, thus struggling to balance simplicity and safety. This paper introduces Spegion, a novel language featuring the first implicit non-lexical region management mechanism. It combines a splittable region model with size-aware memory allocation to achieve static memory safety—without subtyping constraints or explicit region annotations. Its effect-based type system enables fine-grained heap memory control; we formalize its small-step operational semantics and rigorously prove type soundness. By transcending the expressiveness limitations of traditional stack- or region-based memory management, Spegion establishes a new paradigm for secure, efficient, and low-overhead dynamic memory management.

Technology Category

• Machine Learning: Hardware-aware ML
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Virtualization and resource management in Web systems and infrastructures
• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Region based memory management is a powerful tool designed with the goal of ensuring memory safety statically. The region calculus of Tofte and Talpin is a well known example of a region based system, which uses regions to manage memory in a stack-like fashion. However, the region calculus is lexically scoped and requires explicit annotation of memory regions, which can be cumbersome for the programmer. Other systems have addressed non-lexical regions, but these approaches typically require the use of a substructural type system to track the lifetimes of regions. We present Spegion, a language with implicit non-lexical regions, which provides these same memory safety guarantees for programs that go beyond using memory allocation in a stack-like manner. We are able to achieve this with a concise syntax, and without the use of substructural types, relying instead on an effect system to enforce constraints on region allocation and deallocation. These regions may be divided into sub-regions, i.e., Splittable rEgions, allowing fine grained control over memory allocation. Furthermore, Spegion permits sized allocations, where each value has an associated size which is used to ensure that regions are not over-allocated into. We present a type system for Spegion and prove it is type safe with respect to a small-step operational semantics.

Problem

• Ensuring memory safety without lexical scoping
• Managing implicit non-lexical regions without substructural types
• Enabling fine-grained memory control via splittable sized regions

Innovation

• Implicit non-lexical regions for memory safety
• Effect system enforces region allocation constraints
• Splittable regions enable fine-grained memory control