[vectorIntrinsics] RFR: 8254826: Add reduction nodes implementation for Arm SVE

[Ningsheng Jian]

On Thu, 15 Oct 2020 16:14:07 GMT, Paul Sandoz <psandoz at openjdk.org> wrote:

>> Contributed-by: <wanghuang3 at huawei.com>, <hexuejin2 at huawei.com>
>> 
>> Add reduction nodes support for Arm SVE. Add the backend codegen for nodes like "AndReductionVNode" "OrReductionVNode"
>> "XorReductionVNode" "MaxReductionVNode" "MinReductionVNode" .
>
> @dongbohe when adding a contributor using email address try enclosing the address in '<' and '>' e.g (view in raw
> textual form:

Hi @PaulSandoz @adinn,

Since Vector API has been integrated into jdk mainline, do you think this kind of enhancements should go to
vectorIntrinsics for now or jdk mainline directly? @XiaohongGong is also implementing some missing part (mask related)
for vector api SVE codegen. So we could either:

1. Get this into vectorIntrinsics for now until we have full/mature SVE codegen support for Vector API. If there's any
conflict with mainline, I can do the merge. So, the reviewers (@adinn ? :-) ) don't need to try and test every such
patch, and we could also test the patches a bit more here in vectorIntrinsics before going to release.

2. Go to jdk mainline directly, as it does not (if no bug introduced) break anything, as we have op_sve_supported check
to make sure no "bad AD file" error, with https://github.com/openjdk/panama-vector/pull/7 merged to jdk mainline.

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

PR: https://git.openjdk.java.net/panama-vector/pull/13