ArXiv paper runs parallel C vs. Rust teams on real IoT industrial firmware; Rust matches C on memory and speed, with Ariel OS beating bare-metal C stack size.
Key Takeaways
Paper (arXiv:2604.25679, authors from STMicro, Inria, Freie Universitat Berlin, Polytechnic Turin) used two concurrent teams building identical MCU firmware over several months.
Hardware measurements found no meaningful difference in memory footprint or execution speed between C and Rust implementations.
Ariel OS (Rust runtime) produced a smaller footprint than the comparable bare-metal C stack, challenging the assumption that C wins on resource-constrained MCUs.
Paper concludes Rust is a viable production choice today for microcontroller firmware, not just a future prospect.
Study scope is a single IoT sensor platform case; generalization to all MCU families or real-time/safety-critical domains requires caution.
Hacker News Comment Review
Practitioners broadly agree Rust is ready for embedded work, citing ergonomics (enums, cargo, no headers) as the practical draw over memory safety arguments alone.
Toolchain coverage remains a real gap: many older or niche MCU platforms lack a Rust-compatible compiler, limiting adoption outside mainstream Cortex-M targets.
The study’s C JSON parser used malloc (via Parson, the code generator’s default) while the Rust version did not; commenters flagged this as a methodology asymmetry that skews the heap comparison, though an author confirmed it was a deliberate non-optimization given available memory.
Notable Comments
@eggy: Chose Ada/SPARK2014 over Rust for Cortex-M mission-critical work; cites lack of stable language spec and immature formal verification tooling in Rust.
@bArray: Flags that a single experiment is thin evidence for a broad conclusion that Rust is “a sound choice today.”
@noosphr: Raises long-term stability concern – Rust’s fast evolution is a risk for firmware needing decade-scale maintenance.
