Re: poor performance on 7.1 vs 7.0 - Mailing list pgsql-general

From Kyle
Subject Re: poor performance on 7.1 vs 7.0
Date
Msg-id 15096.21846.147769.743842@ip146.usw5.rb1.bel.nwlink.com
Whole thread Raw
In response to Re: poor performance on 7.1 vs 7.0  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: poor performance on 7.1 vs 7.0  (Kyle <kaf@nwlink.com>)
Re: poor performance on 7.1 vs 7.0  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-general
When I profile the postmaster do the forked backends also show up in
the profile?  Or do I need to profile those separately?  (and how?)




Tom Lane wrote:
> Kyle <kaf@nwlink.com> writes:
> > I'm getting poor performance on 7.1 as compared to 7.0.
>
> Hm.  I don't know why 7.1 might be slower for plain SELECTs (the WAL
> changes would certainly not affect that).  Could you compile 7.1 for
> profiling ("make clean; make PROFILE=-pg all" in the src/backend
> subdirectory should do it) and send along a gprof table for your
> test run?
>
> > I've got a btree index on the first 16 of the data elements, and
> > explain verifies that the index is used.
>
> Simple explain doesn't prove that the index is being fully used ---
> the display doesn't distinguish how many index columns are actually
> being looked at.  EXPLAIN VERBOSE output would resolve that doubt.
>
>             regards, tom lane
>


pgsql-general by date:

Previous
From: Fran Fabrizio
Date:
Subject: ALTER COLUMN
Next
From: Tom Lane
Date:
Subject: Re: Pl/PgSQL improvements