Adrian Klaver <adrian.klaver@aklaver.com> writes:
> I redid the test on my 32-bit machine, setting work_mem=16MB, and I got
> comparable results to what I saw on the 64-bit machine. So, what I am
> still am puzzled by is why work_mem seems to make the two paths
> equivalent in time?:
If work_mem is large enough that we never have to go through
tbm_lossify(), then the recheck condition will never be executed,
so its speed doesn't matter.
(So the near-term workaround for Tim is to raise work_mem when
working with tables of this size.)
regards, tom lane