On Tue, Feb 16, 2016 at 1:55 AM, Michael Paquier
<michael.paquier@gmail.com> wrote:
> On Tue, Feb 16, 2016 at 9:18 AM, Robert Haas <robertmhaas@gmail.com> wrote:
>> I experimented with trying to do this and ran into a problem: where
>> exactly would you store the evaluated arguments when you don't know
>> how many of them there will be? And even if you did know how many of
>> them there will be, wouldn't that mean that evalFunc or evaluateExpr
>> would have to palloc a buffer of the correct size for each invocation?
>> That's far more heavyweight than the current implementation, and
>> minimizing CPU usage inside pgbench is a concern. It would be
>> interesting to do some pgbench runs with this patch, or the final
>> patch, and see what effect it has on the TPS numbers, if any, and I
>> think we should. But the first concern is to minimize any negative
>> impact, so let's talk about how to do that.
>
> Good point. One simple idea here would be to use a custom pgbench
> script that has no SQL commands and just calculates the values of some
> parameters to measure the impact without depending on the backend,
> with a fixed number of transactions.
Sure, we could do that. But whether it materially changes pgbench -S
results, say, is a lot more important.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company