Re: Performace comparison of indexes over timestamp fields

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: arnaulist(at)andromeiberica(dot)com
Cc: PostgreSQL Performance <pgsql-performance(at)postgresql(dot)org>
Subject: Re: Performace comparison of indexes over timestamp fields
Date: 2007-05-22 14:11:28
Message-ID: 12213.1179843088@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Arnau <arnaulist(at)andromeiberica(dot)com> writes:
> As you can see the time difference are very big
> Timestamp: 318.328 ms
> int8 index: 120.804 ms
> double precision: 57.065 ms

As already suggested elsewhere, you probably weren't sufficiently
careful in taking your measurements.

A look at the code says that int8 comparison ought to be the fastest
of these. If timestamps are implemented as floats (which you didn't
say) their comparison speed ought to be *exactly* the same as floats,
because the comparison functions are line-for-line the same. If
timestamps are implemented as int8 then they should be similar to
int8 comparisons, maybe a tad slower due to an extra level of function
call. But in any case it seems likely that the actual comparison
function calls would be just a small fraction of the runtime.

regards, tom lane

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Chander Ganesan 2007-05-22 16:56:21 Domains versus Check Constraints
Previous Message PFC 2007-05-22 13:55:26 Re: Tips & Tricks for validating hardware/os