[vectorIntrinsics+mask] RFR: 8272869: AArch64: Mark masked rearrange operation as unsupported

[Ningsheng Jian]

On Wed, 1 Sep 2021 01:47:53 GMT, Eric Liu <eliu at openjdk.org> wrote:

> This patch marks masked rearrange operation as unsupported for AArch64,
> the reasons are as follows:
> 
> 1. AArch64 dose not have predicated 'tbl' instruction
> 2. Mid-end has transformed masked VectorRearrange into VectorBlend[1]
>    for those platforms which are missing corresponding instructions.
> 
> Given by this, there's no need to add backend support for masked
> VectorRearrange, and the final generated code as below:
> 
> 
>         tbl     z16.s, {z20.s}, z16.s
>         mov     z17.s, #0
>         sel     z16.s, p2, z16.s, z17.s
> 
> 
> [1] https://github.com/openjdk/panama-vector/blob/vectorIntrinsics%2Bmask/src/hotspot/share/opto/vectorIntrinsics.cpp#L2016

Looks good. Thanks for the patch!

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

Marked as reviewed by njian (Committer).

PR: https://git.openjdk.java.net/panama-vector/pull/118