Re: pgbench - doCustom cleanup

Fabien COELHO <coelho@cri.ensmp.fr>

From: Fabien COELHO <coelho@cri.ensmp.fr>
To: Alvaro Herrera <alvherre@2ndquadrant.com>
Cc: "Jamison, Kirk" <k.jamison@jp.fujitsu.com>, PostgreSQL Developers <pgsql-hackers@lists.postgresql.org>
Date: 2018-11-20T15:21:29Z
Lists: pgsql-hackers
> I didn't quite understand this hunk.  Why does it remove the 
> is_latencies conditional?  (The preceding comment shown here should be 
> updated obviously if this change is correct, but I'm not sure it is.)

Pgbench runs benches a collects performance data about it.

I simplified the code to always collect data, without trying to be clever 
about cases where these data may not be useful so some collection can be 
skipped.

Here the test avoids recording the statement start time, mostly a simple 
assignment and then later another test avoids recording the stats in the 
same case, which are mostly a few adds.

ISTM that this is over optimization and unlikely to be have any measurable 
effects compared to the other tasks performed when executing commands, so 
a simpler code is better.

-- 
Fabien.


Commits

  1. pgbench: doExecuteCommand -> executeMetaCommand

  2. Rework the pgbench state machine code for clarity

  3. instr_time.h: add INSTR_TIME_SET_CURRENT_LAZY

  4. pgbench: introduce a RandomState struct