Greg Smith <gsmith@gregsmith.com> writes:
> On Tue, 12 Aug 2008, Tom Lane wrote:
>> This seems like a fairly bad idea, because it introduces a
>> gettimeofday() call per transaction.
> There's already lots of paths through pgbench that introduce gettimeofday
> calls all over the place. I fail to see how this is any different.
You haven't thought about it very hard then. The gettimeofday calls
that are in there are mostly at run startup and shutdown. The ones
that can occur intra-run are associated with
* the seldom-used log-each-transaction option, which pretty obviously is a drag on performance anyway; or
* the seldom-used \sleep command, which also obviously affects pgbench's ability to process transactions fast.
I repeat my concern that transaction rates measured with this patch will
be significantly different from those seen with the old code, and that
confusion will ensue, and that it's not hard to see how to avoid that.
regards, tom lane