MIME-Version: 1.0
References: 
 <CAN55FZ3jXs7XDsP_-v_jUBquRu4uAdheN3xcmW=WhAyKwFLSjg@mail.gmail.com>
 <abGv0ScUWVa6eogw@nathan>
 <CAN55FZ3gdK8dGrEo0M6KFW97OaF8TUbjO_dFoxQKi63davE-jA@mail.gmail.com>
 <abG8R6HkOHyUuyWb@nathan>
 <CAN55FZ0yfETy4UEA5rOJ9S06JSOtiWF8TW_+yi3yjVAcrLqKLA@mail.gmail.com>
 <abHTvkeIK37hj9oS@nathan>
 <CAN55FZ1sn-2tVX_n9C5UNBCfDPjSDOCT4zkMeDsB7MaZ9SUBTw@mail.gmail.com>
 <abL5ylYjmopRRmvy@nathan>
 <CAN55FZ0oYuVHZ=TdJofj0WA3mG77XumVsGJgG8gnav9kexr+wg@mail.gmail.com>
 <CAN55FZ0ocS6cBHEWqHv2s-dK91U6OdVLBqj7VexTehtBtioDbA@mail.gmail.com>
 <abQZmZ0KZesCxfR0@nathan>
In-Reply-To: <abQZmZ0KZesCxfR0@nathan>
From: Nazir Bilal Yavuz <byavuz81@gmail.com>
Date: Fri, 13 Mar 2026 18:58:49 +0300
Message-ID: 
 <CAN55FZ0cTwVUdryaHa+9E81LD53ebxZd19iyCi1=QXisJid0iA@mail.gmail.com>
Subject: Re: Speed up COPY FROM text/CSV parsing using SIMD
To: Nathan Bossart <nathandbossart@gmail.com>
Cc: Manni Wood <manni.wood@enterprisedb.com>, KAZAR Ayoub <ma_kazar@esi.dz>,
	Neil Conway <neil.conway@gmail.com>, Andrew Dunstan <andrew@dunslane.net>,
	Shinya Kato <shinya11.kato@gmail.com>,
	PostgreSQL-development <pgsql-hackers@postgresql.org>
Content-Type: multipart/alternative; boundary="000000000000a2b063064ce9f2ec"
Archived-At: 
 <https://www.postgresql.org/message-id/CAN55FZ0cTwVUdryaHa%2B9E81LD53ebxZd19iyCi1%3DQXisJid0iA%40mail.gmail.com>
Precedence: bulk

--000000000000a2b063064ce9f2ec
Content-Type: text/plain; charset="UTF-8"

Hi,

On Fri, 13 Mar 2026 at 17:05, Nathan Bossart <nathandbossart@gmail.com>
wrote:
>
> On Fri, Mar 13, 2026 at 04:34:49PM +0300, Nazir Bilal Yavuz wrote:
> > On Fri, 13 Mar 2026 at 14:57, Nazir Bilal Yavuz <byavuz81@gmail.com>
wrote:
> >> Unfortunately, v15 causes a regression for a 'csv & wide & 1/3' case
> >> on my end. v14 was taking 8000ms but v15 took ~9100ms. If we add the
> >> tmp_hit_eof variable then the regression disappears. Also, if I use a
> >> struct like below, regression disappears again.
> >
> >> When I removed the tmp_hit_eof variable on v14, I didn't encounter any
> >> regression. I really don't understand why this is happening on my end.
> >> Manni didn't encounter any regression on the benchmark [1].
> >
> > Problem might be related to gcc. I am using Debian Trixie and my
> > current gcc version is 'gcc version 14.2.0 (Debian 14.2.0-19)'. If I
> > compile Postgres with 'Debian clang version 19.1.7 (3+b1)', then there
> > is no regression, which makes more sense IMO.
>
> Let's just re-add the temporary variable for hit_eof.  The struct idea is
> clever, but it's just a little more complicated than I think is necessary
> here.
>
> I've also removed the goto in favor of just duplicating the "out" code,
> like you had before.  I'd like to avoid sporadic #ifndef USE_NO_SIMD uses,
> and goto is out of fashion, anyway.

Thanks! v17 LGTM. I didn't encounter any regressions.

-- 
Regards,
Nazir Bilal Yavuz
Microsoft

--000000000000a2b063064ce9f2ec
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Hi,<br><br>On Fri, 13 Mar 2026 at 17:05, Nathan Bossart &l=
t;<a href=3D"mailto:nathandbossart@gmail.com">nathandbossart@gmail.com</a>&=
gt; wrote:<br>&gt;<br>&gt; On Fri, Mar 13, 2026 at 04:34:49PM +0300, Nazir =
Bilal Yavuz wrote:<br>&gt; &gt; On Fri, 13 Mar 2026 at 14:57, Nazir Bilal Y=
avuz &lt;<a href=3D"mailto:byavuz81@gmail.com">byavuz81@gmail.com</a>&gt; w=
rote:<br>&gt; &gt;&gt; Unfortunately, v15 causes a regression for a &#39;cs=
v &amp; wide &amp; 1/3&#39; case<br>&gt; &gt;&gt; on my end. v14 was taking=
 8000ms but v15 took ~9100ms. If we add the<br>&gt; &gt;&gt; tmp_hit_eof va=
riable then the regression disappears. Also, if I use a<br>&gt; &gt;&gt; st=
ruct like below, regression disappears again.<br>&gt; &gt;<br>&gt; &gt;&gt;=
 When I removed the tmp_hit_eof variable on v14, I didn&#39;t encounter any=
<br>&gt; &gt;&gt; regression. I really don&#39;t understand why this is hap=
pening on my end.<br>&gt; &gt;&gt; Manni didn&#39;t encounter any regressio=
n on the benchmark [1].<br>&gt; &gt;<br>&gt; &gt; Problem might be related =
to gcc. I am using Debian Trixie and my<br>&gt; &gt; current gcc version is=
 &#39;gcc version 14.2.0 (Debian 14.2.0-19)&#39;. If I<br>&gt; &gt; compile=
 Postgres with &#39;Debian clang version 19.1.7 (3+b1)&#39;, then there<br>=
&gt; &gt; is no regression, which makes more sense IMO.<br>&gt;<br>&gt; Let=
&#39;s just re-add the temporary variable for hit_eof.=C2=A0 The struct ide=
a is<br>&gt; clever, but it&#39;s just a little more complicated than I thi=
nk is necessary<br>&gt; here.<br>&gt;<br>&gt; I&#39;ve also removed the got=
o in favor of just duplicating the &quot;out&quot; code,<br>&gt; like you h=
ad before.=C2=A0 I&#39;d like to avoid sporadic #ifndef USE_NO_SIMD uses,<b=
r>&gt; and goto is out of fashion, anyway.<br><br>Thanks! v17 LGTM. I didn&=
#39;t encounter any regressions.<br><br>-- <br>Regards,<br>Nazir Bilal Yavu=
z<br>Microsoft</div>

--000000000000a2b063064ce9f2ec--