RFR: 8281375: Accelerate bitCount operation for AVX2 and AVX512 target.

[Jatin Bhateja]

On Wed, 9 Feb 2022 16:26:44 GMT, Quan Anh Mai <duke at openjdk.java.net> wrote:

> Hi,
> 
> I think you can dump a lane into the constant table and use `vpbroadcasti128` for the look-up table.
> 
> Thanks.

-  POPCOUNT lookup table replicates after 16 bytes, $constantaddress accepts a long value (8bytes) which gets replicated and dumped in the Constant Table.
-  Values loaded from constant table many not be aligned to 64 byte boundary and thus may result into perf degradation due to cache line split when loading them into 64 byte vector.

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

PR: https://git.openjdk.java.net/jdk/pull/7373