Re: Poor memory context performance in large hash joins

On 2017-02-24 01:59:01 -0500, Tom Lane wrote:
> Andres Freund <andres(at)anarazel(dot)de> writes:
> > On 2017-02-23 17:28:26 -0500, Tom Lane wrote:
> >> Maybe it's time to convert that to a doubly-linked list.
>
> > Yes, I do think so. Given that we only have that for full blocks, not
> > for small chunks, the cost seems neglegible.
> > That would also, partially, address the performance issue
> > http://archives.postgresql.org/message-id/d15dff83-0b37-28ed-0809-95a5cc7292ad%402ndquadrant.com
> > addresses, in a more realistically backpatchable manner.
>
> Yeah, I was wondering if we could get away with back-patching such a
> change. In principle, nothing outside aset.c should know what's in the
> header of an AllocBlock, but ...

You'd need to go through a fair amount of intentional pain to be
affected by a change AllocBlockData's structure. We could add the
->prev pointer to the end of AllocBlockData's definition to make it less
likely that one would be affected in that unlikely case - but I'm a bit
doubtful it's worth the trouble.

- Andres