On Wed, 30 Aug 2023 15:43:13 GMT, Claes Redestad <[email protected]> wrote:
>> test/micro/org/openjdk/bench/java/net/URLEncodeDecode.java line 86:
>>
>>> 84: tokens[n++] = '*';
>>> 85:
>>> 86: Random r = new Random(mySeed);
>>
>> SplittableRandom allow to have perfectly reproducible sequences
>
> Care to elaborate? `Random` with an explicit seed should be more than enough
> to have perfectly reproducible sequences for our case (`SplittableRandom`
> uses a different algorithm by default with a longer period, but I can't see
> how this'd be relevant to this case.)
>
>
> * If two instances of {@code Random} are created with the same
> * seed, and the same sequence of method calls is made for each, they
> * will generate and return identical sequences of numbers.
Even with the same seed only with large numbers you get statistically a similar
distribution,m (not the same values, but the javadoc said differently which is
quite surprising to me!), but splittable random is actually producing the same
exact sequences. You care about confusing branch predictors here I suppose (or
the JIT, maybe too) but not to add noise or variance with fewer samples nor you
are interested to the security part of it (cause random use the chaos generated
securely by the OS/HW iirc).
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/15448#discussion_r1310489744
