RFR: 8296548: Improve MD5 intrinsic for x86_64

[Vladimir Kozlov]

On Wed, 9 Nov 2022 07:57:30 GMT, Yi-Fan Tsai <duke at openjdk.org> wrote:

> The LEA instruction loads the effective address, but MD5 intrinsic uses it for computing values than addresses. This usage potentially uses more cycles than ADDs and reduces the throughput.
> 
> This change replaces
>     LEA:  r1 = r1 + rsi * 1 + t
> with
>     ADDs: r1 += t; r1 += rsi.
> 
> Microbenchmark evaluation shows ~40% performance improvement on Haswell, Broadwell, Skylake, and Cascade Lake. There is ~20% improvement on 2nd gen Epyc.
> 
> No performance change for the same microbenchmark on Ice Lake and 3rd gen Epyc.
> 
> Similar results can be observed with TestMD5Intrinsics and TestMD5MultiBlockIntrinsics. There is ~15% improvement in throughput on Haswell, Broadwell, Skylake, and Cascade Lake.

Yes, please, post performance data.
Note, TestMD5Intrinsics and TestMD5MultiBlockIntrinsics are regression/correctness tests.
Would be nice to have proper JMH benchmarks to show improvement.

@sviswa7 or @jatin-bhateja do you agree with these changes?

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

PR: https://git.openjdk.org/jdk/pull/11054