Re: Macro customizable hashtable / bitmapscan & aggregation perf

Hi,

Attached is an updated version of the patchset. The main changes are
- address most of Tomas' feedback
- address regression test output changes by adding explicit ORDER BYs,
in a separate commit.
- fix issues around hash tables sized up to PG_UINT32_MAX
- fix a bug in iteration with "concurrent" deletions

> > We didn't really for dynahash (as it basically assumed a fillfactor of 1
> > all the time), not sure if changing it won't also cause issues.
> >
>
> That's a case of glass half-full vs. half-empty, I guess. If we assumed load
> factor 1.0, then I see it as accounting for load factor (while you see it as
> not accounting of it).

Well, that load factor is almost never achieved, because we'd have grown
since... I added a comment about disregarding fill factor and growth
policy to estimate_hashagg_tablesize, which is actually the place where
it'd seemingly make sense to handle it.

Tomas, did you have time to run a benchmark?

I think this is starting to look good.

Regards,

Andres