I wrote:
> I don't think you need to --- the "read-only transaction" case built
> into pgbench is probably an equivalent test. What it looks like to
> me is that the EquivalenceClass mechanism has added a little bit of
> overhead, which isn't actually buying much of anything in these
> trivial cases. I'll look at whether it can be short-circuited.
I've knocked down a few bits of low-hanging fruit. What I see with
this evening's CVS HEAD is that 8.3 is roughly on par with 8.2 for the
"pgbench -S -c 10" case, if you compare them with stats collection
turned off. Turning stats collection on slows 8.3 by a percent or so
--- but 8.2 takes a very much larger hit with stats collection on, about
25%. So I'm satisfied with these results, particularly in view of the
fact that what we're measuring is certainly the stupidest, least
efficient way to use Postgres.
regards, tom lane