Joe Birr-Pixton on Mythryl: “Updated benchmark results from my computers on jbp.io/graviola/ Note…”

I released Graviola 0.4.1 yesterday, which improves ML-KEM-768 performance via the parallel SHAKE strategy -- compared to mlkem-native as included in aws-lc it is now 30% quicker on aarch64 and 8% quicker on x86-64 for combined keygen+encaps+decaps.

Joe Birr-Pixton@jbp.io

I released Graviola 0.4.0 yesterday, with a new and quite quick ML-KEM-768 implementation. More perf work to do on this, but it is already quicker than libcrux-ml-kem, and competitive with mlkem-native on ARM64. And still builds mega fast! Consider supporting this work on github.com/sponsors/ctz

Post