[vectorIntrinsics] RFR: 8263644: Add Extract float nodes implementation for Arm SVE

[Wang Huang]

On Mon, 22 Mar 2021 03:18:57 GMT, Xiaohong Gong <xgong at openjdk.org> wrote:

>> We implement ExtractD & ExtractF Node for Arm SVE backend.
>> * overload encoding functions `sve_lasta` & `sve_lastb`
>> * for future JDK-8263424, we eliminate vector length judgements and use  `rscratch1`
>
> src/hotspot/cpu/aarch64/aarch64_sve.ad line 3063:
> 
>> 3061: %}
>> 3062: 
>> 3063: instruct extractF(vRegF dst, vReg src, immI idx,  pRegGov pTmp, rFlagsReg cr)
> 
> Remove the redundant space between `idx,  pRegGov`?

Sure, thank you for your review.

> src/hotspot/cpu/aarch64/aarch64_sve.ad line 3088:
> 
>> 3086:   format %{ "movzw rscratch1, $idx\n\t"
>> 3087:             "sve_whilele $pTmp, D, zr, rscratch1\n\t"
>> 3088:             "sve_lastb $dst, D, $pTmp, $src\n\t" %}
> 
> Thanks for the implementation! It mostly looks good to me. Just a trivial comment: add a comment at the end like "extract from vector(D)" like what you have done for` extractB/S` ?

OK.I will fix that.

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

PR: https://git.openjdk.java.net/panama-vector/pull/51