“…In comparison to the software implementation, our hard-ware implementation in this work demonstrates significant performance improvements. Seo et al [6] reported an optimized NEON implementation on an 8-core ARM v8.2 64bit CPU mounted on Jetson AGX Xavier. For Dilithium level 5, their implementation takes 542µs for Keygen, 625µs for Verify and 1001µs for the best-case scenario of Sign.…”