RFR: 8266054: VectorAPI rotate operation optimization [v5]

[Jatin Bhateja]

On Sat, 8 May 2021 15:40:53 GMT, Paul Sandoz <psandoz at openjdk.org> wrote:

> Looks good. Someone from the HotSpot side needs to review related changes.
> 
> The way i read the perf numbers is that on non AVX512 systems the numbers are in the noise (no worse, no better), with significant improvement on AVX512.

Hi @PaulSandoz, thanks for your suggestions and review, yes AVX2 speedup is an artifact of the code re-organization, since C2 now directly dismantles the rotates thus this offer better in-lining behavior compared to existing implementation and this is what show up in the performance data.

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

PR: https://git.openjdk.java.net/jdk/pull/3720