From: Ronan Dunklau <ronan.dunklau@aiven.io>
To: Alexander Korotkov <aekorotkov@gmail.com>,
 pgsql-hackers@lists.postgresql.org
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Michael Paquier <michael@paquier.xyz>,
 pgsql-hackers@lists.postgresql.org
Subject: Re: Fix gin index cost estimation
Date: Tue, 06 Dec 2022 11:22:20 +0100
Message-ID: <4787929.31r3eYUQgx@aivenlaptop>
In-Reply-To: <10184681.nUPlyArG6x@aivenlaptop>
References: <3188617.44csPzL39Z@aivenronan>
 <CAPpHfdtFYZjv41djYjNcJUuZhDKUu9eJSthKmmDWE-azAZjn5g@mail.gmail.com>
 <10184681.nUPlyArG6x@aivenlaptop>
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="nextPart4444252.LvFx2qVVIh"
Content-Transfer-Encoding: 7Bit
Archived-At: 
 <https://www.postgresql.org/message-id/4787929.31r3eYUQgx%40aivenlaptop>
Precedence: bulk

This is a multi-part message in MIME format.

--nextPart4444252.LvFx2qVVIh
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="UTF-8"

Le vendredi 2 d=C3=A9cembre 2022, 13:58:27 CET Ronan Dunklau a =C3=A9crit :
> Le vendredi 2 d=C3=A9cembre 2022, 12:33:33 CET Alexander Korotkov a =C3=
=A9crit :
> > Hi, Ronan!
> > Thank you for your patch.  Couple of quick questions.
> > 1) What magic number 50.0 stands for?  I think we at least should make
> > it a macro.
>=20
> This is what is used in other tree-descending  estimation functions, so I
> used that too. Maybe a DEFAULT_PAGE_CPU_COST macro would work for both ? =
If
> so I'll separate this into two patches, one introducing the macro for the
> other estimation functions, and this patch for gin.

The 0001 patch does this.

>=20
> > 2) "We only charge one data page for the startup cost" =E2=80=93 should=
 this
> > be dependent on number of search entries?

In fact there was another problem. The current code estimate two different=
=20
pathes for fetching data pages: in the case of a partial match, it takes in=
to=20
account that all the data pages will have to be fetched. So this is is now=
=20
taken into account for the CPU cost as well.=20

=46or the regular search, we scale the number of data pages by the number o=
f=20
search entries.

Best regards,

=2D-
Ronan Dunklau


--nextPart4444252.LvFx2qVVIh
Content-Disposition: attachment;
 filename="v4-0001-Extract-the-cpu-page-process-cost-multiplier-into.patch"
Content-Transfer-Encoding: 7Bit
Content-Type: text/x-patch; charset="UTF-8";
 name="v4-0001-Extract-the-cpu-page-process-cost-multiplier-into.patch"