On Mon, Aug 15, 2016 at 6:55 AM, Robert Haas <robertmhaas@gmail.com> wrote:
> A sort of dumb way of handling all this is to assume that once a
> worker joins the hash join, it won't go off and do anything else until
> the hash join is done. Under that assumption, you just need some sort
> of BarrierAttach() operation; workers that have never attached the
> barrier aren't participating in the hash join at all and so they are
> irrelevant - and now you know how many workers you need to await,
> because you can keep a count how many have attached. Perhaps you
> simply turn away any workers that arrive after batch 0 is complete.
Is that really so bad? In general, I don't tend to think of workers as
the cost to worry about. Rather, we should be concerned about the
active use of CPU cores as our major cost.
--
Peter Geoghegan
