SIMD implementations of SHA3 algorithms do not seem to work well on the Neoverse processor cores. JDK-8337666 introduced a GPR version for sun.security.provider.SHA3.implCompress0 and the SHA3 version of sun.security.provider.DigestBase.implCompressMultiBlock0, but not for sun.security.provider.SHA3Parallel.doubleKeccak(), that is used by the ML_KEM implementation. Therefore, on Neoverse, the performance of the ML_KEM algorithms is not improved, although it could be, with a similar GPR based implementation as was introduced by the other 2 SHA3 @IntrinsicCandidate methods.