RFR: 8292898: [vectorapi] Unify vector mask cast operation [v4]

Tue Sep 27 05:27:27 UTC 2022

On Tue, 27 Sep 2022 04:07:19 GMT, Quan Anh Mai <qamai at openjdk.org> wrote:

>> I see, thanks! I will change this later.
>
> For AVX3, a mask register is used if the vector species is 512-bit wide or the machine supports AVX512VL, so this code is still relevant in this case.

I verified that new instruction sequence with vector as mask carrier over AVX2 is performant specially for down casting operations, since new sequence is using vector as a mask carrier so will be valid if corresponding 512 bit vector species is supported by the target i.e. it clears the constraints imposed by match_rule_supported vector.

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

PR: https://git.openjdk.org/jdk/pull/10192