Re: pgbench - add pseudo-random permutation function

On Mon, Mar 8, 2021 at 11:50:43AM +0100, Fabien COELHO wrote:
>
> > > What are your current thoughts?
> >
> > Thanks for prodding. I still think it's a useful feature. However I
> > don't think I'll have to time to get it done on the current commitfest.
> > I suggest to let it sit in the commitfest to see if somebody else will
> > pick it up -- and if not, we move it to the next one, with apologies to
> > author and reviewers.
> >
> > I may have time to become familiar or at least semi-comfortable with all
> > that weird math in it by then.
>
> Yep.
>
> Generating a parametric good-quality low-cost (but not
> cryptographically-secure) pseudo-random permutations on arbitrary sizes (not
> juste power of two sizes) is not a trivial task, I had to be quite creative
> to achieve it, hence the "weird" maths. I had a lot of bad
> not-really-working ideas before the current status of the patch.
>
> The code could be simplified if we assume that PG_INT128_TYPE will be
> available on all relevant architectures, and accept the feature not to be
> available if not.

Maybe Dean Rasheed can help because of his math background --- CC'ing him.

--
Bruce Momjian <bruce(at)momjian(dot)us> https://momjian.us
EDB https://enterprisedb.com

The usefulness of a cup is in its emptiness, Bruce Lee