RFR: 7903367: JMH: Add JMHSample_39_MemoryAccess [v2]

[Vsevolod Tolstopyatov]

On Mon, 17 Oct 2022 19:18:21 GMT, Michael Mirwaldt <duke at openjdk.org> wrote:

>> I propose this sample because overlooking the memory layout of e.g. ArrayLists can be a pitfall for benchmarks.
>> This new sample makes users of JMH aware of this problem. It teaches them how to use JMH right.
>> What do you think?
>
> Michael Mirwaldt has updated the pull request incrementally with one additional commit since the last revision:
> 
>   improved by proposals from first review

jmh-samples/src/main/java/org/openjdk/jmh/samples/JMHSample_39_MemoryAccess.java line 65:

> 63:      * Benchmark                               Mode  Cnt    Score   Error  Units
> 64:      * JMHSample_39_MemoryAccess.sumArray      avgt   25   50.393 ± 0.407  ms/op
> 65:      * JMHSample_39_MemoryAccess.sumArrayList  avgt   25  299.868 ± 1.591  ms/op

It maybe can be even more educational if the benchmarks will expose how *unstable* such measurements are.

Having a third benchmark with the same `ArrayList` with the same `Integer` objects, but **shuffled** yields a completely different result. The educational hint here is that it can be GC who can do the actual shuffle

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

PR: https://git.openjdk.org/jmh/pull/85