Reproduce “Evaluate the Stack Management in Effect Handlers using the libseff C Library” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
ZeHao Yu
Researchers at the University of Edinburgh evaluated four stack management strategies for effect handlers in the libseff C library, including a novel user-level overcommitting mechanism. They found that while fixed-size and segmented stacks offered the lowest basic context switch latency (approximately 1 μs), kernel-based overcommitting provided superior stack expansion efficiency (115 μs), and user-level overcommitting incurred higher overheads (44 μs basic context switch latency) with occasional stability issues under multi-threaded stress.
Key insights
- The novel user-level overcommitting mechanism, while offering custom control over memory, currently exhibits higher context-switching overheads and vulnerability to non-deterministic faults under multi-threaded stress.
- Kernel-based overcommitting emerges as an effective strategy, providing a strong balance of memory flexibility and performance for dynamic loads by efficiently leveraging operating system optimizations for stack expansion.
- Fixed-size and segmented stacks demonstrate the lowest basic context-switching latency but are either prone to overflow/waste or incur overheads associated with segment management and linking.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.