[vectorIntrinsics] RFR: RFC: Vector API masking support proposal for Arm SVE [v2]

[Xiaohong Gong]

On Mon, 15 Mar 2021 03:58:30 GMT, Xiaohong Gong <xgong at openjdk.org> wrote:

>> src/hotspot/cpu/aarch64/aarch64.ad line 2472:
>> 
>>> 2470: 
>>> 2471: const bool Matcher::match_rule_supported_masked_vector(int opcode, int vlen, BasicType bt) {
>>> 2472:   if (!match_rule_supported(opcode) || !vector_size_supported(bt, vlen)) {
>> 
>> Can we not just use existing routine match_rule_supported_vector here?
>
> This can be improved in future. But I think it's different with `match_rule_supported_vector`. It's used to check whether the optimized predicate version is supported/implemented for an operation.

I guess there might be different masking support status for operations with SVE and AVX-512. So the check that depends on the architecture should be implemented in the backend.

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

PR: https://git.openjdk.java.net/panama-vector/pull/40