Re: Instrument checkpoint sync calls

On Mon, Dec 13, 2010 at 3:19 AM, Greg Smith <greg(at)2ndquadrant(dot)com> wrote:
> Robert Haas wrote:
>>
>> I'm wondering why md.c is converting the results from an exact value to a
>> floating
>> point, only to have xlog.c turn around and convert back to an integer.
>>  I think it could just return milliseconds directly, or if you're
>> worried about a checkpoint that takes more than 24 days to complete,
>> seconds and microseconds.
>
> md.c is printing the value as a float, so I converted early to a double and
> then percolated it upward from there.  More an artifact of how the code grew
> from its initial form than an intentional decision. I see your point that
> making elapsed, total_elapsed, ckpt_agg_sync_time, and ckpt_longest_sync all
> the same type of integer that INSTR_TIME_GET_MICROSEC returns would reduce
> the potential for rounding abberations.  I could do another rev of the patch
> tomorrow with that change if you'd prefer it that way.  I don't have a
> strong opinion about that implementation detail; given that xlog.c is
> printing a less fine-grained time anyway (seconds with 3 digits vs. msec
> with 3 digits) it seems unlikely to run into a real problem here.

Yeah, it'd probably be OK anyway, but I think it doesn't really cost
us anything to avoid the unnecessary conversion steps, so count me as
a vote for doing it that way.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company