RFR: 8236850: Operations on constant List/Set.of(element) instances does not consistently constant fold

[John Rose]

On Jan 13, 2020, at 11:49 AM, Paul Sandoz <paul.sandoz at oracle.com> wrote:
> 
> For information purposes: an alternative salt might be to use mix32 from Splittable random:
> 
> /**
> * Returns the 32 high bits of Stafford variant 4 mix64 function as int.
> */
> private static int mix32(long z) {
>    z = (z ^ (z >>> 33)) * 0x62a9d9ed799705f5L;
>    return (int)(((z ^ (z >>> 28)) * 0xcb24d0a5c88c35b3L) >>> 32);
> }
> 
> as in:
> 
> SALT = mix32(System.nanoTime());

Thank you for mentioning that!  I get this google hit on it:

https://blogs.oracle.com/dave/a-simple-prng-construction-idiom

There are also similar functions in xxHash.

https://github.com/easyaspi314/xxhash-clean/blob/master/xxhash64-ref.c#L104

Here’s a scary list:

https://en.wikipedia.org/wiki/List_of_hash_functions

I think “doing it right” is overkill for this one place, but I hope we can put
something like that into the JVM or some other low-level place for use
whenever we need a cheap pseudo-random nonce of some sort.

Once there’s an API for it I think we will reach for it a little more often.
Having a common engine means we can afford to tune its predictability
qualities.  In both directions:  resistance to attack, and reproducibility
for debugging.  In the case of Set.of, the former concern doesn’t apply
much, but the latter might.

— John