Simon Riggs <simon@2ndquadrant.com> writes:
> So the timing is clearly responsible for the additional time I'm
> personally experiencing and very likely to be that for others also.
Well, that's certainly unsurprising, but the question is why it's such
a large overhead for you when it's not on other apparently-similar kit.
If anyone else is still following this discussion, could you try the
test case Simon gave a bit earlier (select count(*) from 100000-row
table, check \timing with and without explain analyze)? If we could get
a clearer picture of which platforms show the problem and which don't,
maybe we could make some progress.
(BTW, I just tried it on my Apple Powerbook G4, and got results in line
with what I get elsewhere. So I've now tried it on three very different
CPU types and OSes, and not duplicated Simon's problem on any of them.)
regards, tom lane
