Hey everyone! I've been working on this project for a while now and I'm happy to announce its alpha release.

It's built around Linux's RSEQ features, providing near-zero-overhead per-CPU allocation with a somewhat different philosophy than most allocators.

Current features include:

• Per-CPU RSEQ fast paths with inline assembly critical sections.
• Adaptive EMA-based refill prediction (still not entirely sure if this was a good idea).
• Big allocation support with optional buddy caching.
• Rust GlobalAlloc integration and preload (LD_PRELOAD) support.
• Optional canary and extended-header diagnostics.

RSMalloc is still experimental and very much in its early stages, and it is definitely not production-ready. I'm mostly interested in feedback, design criticism, and hearing about workloads or corner cases I may have overlooked.

GitHub: https://github.com/Metehan120/rsmalloc

crates.io: https://crates.io/crates/rsmalloc

Architecture docs and benchmark notes are included in the repository.