[jdk17u-dev] RFR: 8251216: Implement MD5 intrinsics on AArch64

[Dmitry Chuyko]

As mentioned in the original review MD5 can't be considered a secure hash. But it is still widely used as a lightweight generic hash function, and that's true not only for a legacy software. It is in fact not as lightweight as expected, e.g. hardware accelerated SHA-1 and SHA-256 work much faster on N1. We can sweeten the pill at least for the latest LTS release by 25-30% acceleration with this change.

Original patch applies almost cleanly. There was a minor fuzz, and "code_size2 = 45000" increase had to be re-applied manually.

Micro-benchmark results in openjdk.bench.javax.crypto.full.MessageDigestBench on Graviton 2 show improvements identical to original numbers on Altra.

Testing: tier1, tier2, test/hotspot/jtreg/compiler/intrinsics/sha (aarch64)

-------------

Commit messages:
 - Backport 088b244ec6d9393a1fcd2233fa5b4cf46f9ae0dd

Changes: https://git.openjdk.java.net/jdk17u-dev/pull/110/files
 Webrev: https://webrevs.openjdk.java.net/?repo=jdk17u-dev&pr=110&range=00
  Issue: https://bugs.openjdk.java.net/browse/JDK-8251216
  Stats: 199 lines in 4 files changed: 193 ins; 1 del; 5 mod
  Patch: https://git.openjdk.java.net/jdk17u-dev/pull/110.diff
  Fetch: git fetch https://git.openjdk.java.net/jdk17u-dev pull/110/head:pull/110

PR: https://git.openjdk.java.net/jdk17u-dev/pull/110