RFR: 8285790: AArch64: Merge C2 NEON and SVE matching rules [v5]

[Ningsheng Jian]

On Thu, 4 Aug 2022 08:04:28 GMT, Andrew Haley <aph at openjdk.org> wrote:

>> Hao Sun has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Fix Jtreg failures on NEON with -XX:MaxVectorSize=8
>>   
>>   [Byte|Short|Int|Long|Float|Double]128VectorTests.java failed on NEON
>>   platform with -XX:MaxVectorSize=8.
>>   
>>   The root cause lies in that 128-bit vector MulReduction operations would
>>   pass the check in match_rule_supported_vector(), i.e. returning true if
>>   lengh_in_bytes is 16, which should be invalid since MaxVectorSize=8.
>>   
>>   In this patch, we improve the check logic a bit, i.e. failing fast if
>>   the conditions are not satisfied and falling through to check
>>   vector_size_supported() eventually.
>>   
>>   Similar update is done to the check for Op_MulAddVS2VI.
>
> I just pulled this patch and src/hotspot/cpu/aarch64/aarch64_neon.ad etc. seem still to be there.

Thanks for the review @theRealAph. We will run performance tests on more hardware before integration.

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

PR: https://git.openjdk.org/jdk/pull/9346