RFR: 8281146: Replace StringCoding.hasNegatives with countPositives [v9]

[Claes Redestad]

On Thu, 3 Mar 2022 14:43:52 GMT, Lutz Schmidt <lucy at openjdk.org> wrote:

> I don't think the intrinsic has to be implemented on every target, but AFAICR it does have to be declared as an intrinsic in HotSpot.

Yeah, I got confused. To me it looks like a declaration of intent, and thought the only strict requirement was that HotSpot can't define an intrinsic if there's no annotation. It also seems odd to hardwire it to what HotSpot does - wasn't the purpose of renaming from `@HotSpotIntrinsicCandidate` to decouple things a bit better?

Either way.. I'm re-running tests and benchmarks and seeing if I can improve the aarch64 impl to return a decent approximation rather than always-0 when there are negative bytes.

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

PR: https://git.openjdk.java.net/jdk/pull/7231