On Tue, 31 Dec 2024 at 10:11, James Hunter <james.hunter.pg@gmail.com> wrote:
> Does PostgreSQL currently rescan Hash Joins when they are "no longer
> needed," to free work_mem early? If so, then I would try to reuse this
> existing logic to decide which nodes need work_mem concurrently.
>
> If not, then all nodes that use work_mem actually use it
> "concurrently," because we don't free that work_mem until we call
> ExecutorEnd().
The problem with that statement is that Hash Join isn't the only node
type that uses work_mem. Plenty of other node types do. Have a look
at MultiExecBitmapOr(). You can see logic there that does tbm_free()
after the tbm_union() call. Going by that, it seems there is at least
one place where we might free some work_mem memory before allocating
another lot.
David
