Date: Wed, 22 Aug 2018 19:51:12 +0200 (CEST)
From: Andreas Joseph Krogh <andreas@visena.com>
To: Andres Freund <andres@anarazel.de>
Cc: Robert Haas <robertmhaas@gmail.com>,
	Alvaro Herrera <alvherre@2ndquadrant.com>,
	"Jonathan S. Katz" <jkatz@postgresql.org>,
	PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Message-ID: <VisenaEmail.28.47f1027d389da54b.16562b9da65@tc7-visena>
In-Reply-To: <20180822165155.tjbib4pewnbiyy4x@alap3.anarazel.de>
Subject: Sv: Re: Query is over 2x slower with jit=on
MIME-Version: 1.0
Content-Type: multipart/mixed;
	boundary="----=_Part_170_1135327557.1534960272317"
Precedence: bulk

------=_Part_170_1135327557.1534960272317
Content-Type: multipart/related;
	boundary="----=_Part_171_348906196.1534960272317"

------=_Part_171_348906196.1534960272317
Content-Type: multipart/alternative;
	boundary="----=_Part_172_1768968657.1534960272330"

------=_Part_172_1768968657.1534960272330
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

P=C3=A5 onsdag 22. august 2018 kl. 18:51:55, skrev Andres Freund <andres@an=
arazel.de=20
<mailto:andres@anarazel.de>>:
On 2018-08-22 18:39:18 +0200, Andreas Joseph Krogh wrote:
 > Just to be clear; The query=C2=A0really=C2=A0runs slower (wall-clock tim=
e), it's not
 > just the timing.

 I bet it's not actually running slower, it "just" takes longer to start
 up due to the JITing in each worker. I suspect what we should do is to
 multiple the cost limits by the number of workers, to model that.=C2=A0 Bu=
t
 without the fixed instrumentation that's harder to see...
=C2=A0
Well, yes, that might be. By "runs" I meant from me hitting ENTER in psql t=
o=20
the time the query finishes...
=C2=A0
I thought=C2=A0JITing of prepared=C2=A0queries happended once (in "prepare"=
)=C2=A0so it=20
didn't have to do the JITing every time the query is executed. Isn't=20
the=C2=A0previously generated bytecode usable for subsequent queries?
=C2=A0
--
 Andreas Joseph Krogh


------=_Part_172_1768968657.1534960272330
Content-Type: text/html;charset=UTF-8
Content-Transfer-Encoding: quoted-printable

<div>P=C3=A5 onsdag 22. august 2018 kl. 18:51:55, skrev Andres Freund &lt;<=
a href=3D"mailto:andres@anarazel.de">andres@anarazel.de</a>&gt;:</div>

<blockquote style=3D"border-left: 1px solid rgb(204, 204, 204); margin: 0pt=
 0pt 0pt 0.8ex; padding-left: 1ex;">
<div style=3D"display:inline; font-family: monospace; font-size: 12px;">On =
2018-08-22 18:39:18 +0200, Andreas Joseph Krogh wrote:<br>
&gt; Just to be clear; The query=C2=A0really=C2=A0runs slower (wall-clock t=
ime), it's not<br>
&gt; just the timing.<br>
<br>
I bet it's not actually running slower, it &quot;just&quot; takes longer to=
 start<br>
up due to the JITing in each worker. I suspect what we should do is to<br>
multiple the cost limits by the number of workers, to model that.=C2=A0 But=
<br>
without the fixed instrumentation that's harder to see...</div>
</blockquote>

<div>=C2=A0</div>

<div>Well, yes, that might be. By &quot;runs&quot; I meant from me hitting =
ENTER in psql to the time the query finishes...</div>

<div>=C2=A0</div>

<div>I thought=C2=A0JITing of prepared=C2=A0queries happended once (in &quo=
t;prepare&quot;)=C2=A0so it didn't have to do the JITing every time the que=
ry is executed. Isn't the=C2=A0previously generated bytecode usable for sub=
sequent queries?</div>

<div>=C2=A0</div>

<div class=3D"visena-email-signature" data-visena-message-id=3D"&lt;VisenaE=
mail.28.47f1027d389da54b.16562b9da65@tc7-visena>">
<div>--<br>
Andreas Joseph Krogh</div>
</div>
------=_Part_172_1768968657.1534960272330--

------=_Part_171_348906196.1534960272317--

------=_Part_170_1135327557.1534960272317--