Multicore Implementation of ML-KEM on Embedded Devices

This paper introduces a cost-effective function-selection procedure, anchored in Amdahl’s Law, to orchestrate selective parallelization of Module-Lattice-Based Key-Encapsulation Mechanism Standard (ML-KEM) scheme on a dual-core ESP32 microcontroller. The procedure quantifies each function’s single-core runtime and the overhead of parallel execution to isolate routines whose concurrent execution yields net time savings, while excluding low-latency tasks. The resulting ESP32 implementation eliminates separate single- and dual-core code paths, simplifies development, and achieves substantial reductions in overall runtime. Experimental evaluation on a dual-core ESP32 microcontroller of ML-KEM confirms the procedure’s suitability for latency-critical applications.