[vectorIntrinsics+mask] RFR: 8265109: SVE predicate register allocation support for vectorIntrinsics [v2]

[Ningsheng Jian]

On Thu, 15 Apr 2021 07:08:49 GMT, Jatin Bhateja <jbhateja at openjdk.org> wrote:

>> Yes, aarch64 does not need this for now. I can either revert this change (leaving it to you for future update if required) or update to size 2 for x86. Which do you prefer?
>
> I can see ADLC does parse the preprocessors in source sections, by adding a new #define SVE_ENABLE  1 in aarch64_sve.ad  and checking for the same similar to following handling may allow us to set size = 1 for SVE else 2.
> https://github.com/openjdk/jdk/blob/master/src/hotspot/share/adlc/formssel.cpp#L2279

That sounds good to me. Thanks @jatin-bhateja! Addressed in the new commit. Please let me know whether there are any other concerns.

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

PR: https://git.openjdk.java.net/panama-vector/pull/65