Re: Speed up COPY FROM text/CSV parsing using SIMD

From: Nazir Bilal Yavuz <byavuz81(at)gmail(dot)com>
To: KAZAR Ayoub <ma_kazar(at)esi(dot)dz>
Cc: Mark Wong <markwkm(at)gmail(dot)com>, Neil Conway <neil(dot)conway(at)gmail(dot)com>, Manni Wood <manni(dot)wood(at)enterprisedb(dot)com>, Nathan Bossart <nathandbossart(at)gmail(dot)com>, Andrew Dunstan <andrew(at)dunslane(dot)net>, Shinya Kato <shinya11(dot)kato(at)gmail(dot)com>, PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Speed up COPY FROM text/CSV parsing using SIMD
Date: 2026-02-03 10:59:52
Message-ID: CAN55FZ0EiPgUTJmGQXm9gP_VnogdQMkg2yhZSvwoSFJ1uVhFFg@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi,

On Mon, 2 Feb 2026 at 23:45, KAZAR Ayoub <ma_kazar(at)esi(dot)dz> wrote:
>
>>
> Thanks for the benchmark, I'm a bit suspicious about this because I find it illogical or at least highly unexpected for a 1/3 specials workload to perform better than no specials !
> or csv no special regressing ! because it's expected to take the simd path for the whole processing, so it's supposed to perform better than master (at least ...).

I agree with you. I think it is highly unexpected that 1/3 specials
cases perform better than no special cases.

> I wonder what the results look like for COPY TO case on POWER. If you can try, that case is at least even more theoretically predictable.

Yes, that would be helpful.

--
Regards,
Nazir Bilal Yavuz
Microsoft

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Nazir Bilal Yavuz 2026-02-03 11:02:16 Re: Speed up COPY FROM text/CSV parsing using SIMD
Previous Message Tomas Vondra 2026-02-03 10:57:01 Re: Non-deterministic buffer counts reported in execution with EXPLAIN ANALYZE BUFFERS