public inbox for [email protected]  

help / color / mirror / Atom feed
From: lakshmi <[email protected]>
To: Ranier Vilela <[email protected]>
Cc: Bryan Green <[email protected]>
Cc: Pg Hackers <[email protected]>
Subject: Re: Avoid multiple calls to memcpy (src/backend/access/index/genam.c)
Date: Mon, 23 Mar 2026 10:15:59 +0530
Message-ID: <CAEvyyTjT70OVN_NCcHQ+PodVM7bkPccKPzJcAv0u+eW=qpZDMg@mail.gmail.com> (raw)
In-Reply-To: <CAEudQAoF2xfpA7tgW3gusg0dyMTwTDDwAzBqR_bbv08NBkqv-A@mail.gmail.com>
References: <CAEudQApbWon+3Eb9x4WW_D-JkSt2mvfx99dXu9VZ4AeCuTh=fw@mail.gmail.com>
	<CAEudQApEfvhNT1fEPURzVcQH7G0A1ukh_ugoCGaErV6_dbndCQ@mail.gmail.com>
	<CAF+pBj_RS2KErTqQ6ORXjhVzmukG7Ve0wHU1Kq56xjJfFKwVqA@mail.gmail.com>
	<CAEudQAptRymgvmd5hQb2mk-Ft89XcSo_xvC74kv4JBA9v=D4Sg@mail.gmail.com>
	<CAF+pBj-K2bgNQRc9ih01WFmAWUaQtVbS37jLtYdYh5LOwOkF6A@mail.gmail.com>
	<CAF+pBj9M8rCoS-EEBwLiA6hdxm_UNjxMXG9Vc4RU8xifqnQB-g@mail.gmail.com>
	<CAEvyyTjLYVuy0V6gXE9_HgyB4gUftuTZm65gdbacE_OOKk1cqw@mail.gmail.com>
	<CAEudQAoF2xfpA7tgW3gusg0dyMTwTDDwAzBqR_bbv08NBkqv-A@mail.gmail.com>

Hi,

Thanks for the reply.

Here are the details of my test setup.

I ran the tests on Debian Linux (6.1.0-10-amd64, x86_64) on an AMD Ryzen 5
7535U (6 cores / 12 threads).

The PostgreSQL version was 19devel, built from source using gcc 12.2.0 with
enable-debug.

For benchmarking, I used pgbench with the following command:
pgbench -p 55432 -d postgres -c 10 -j 4 -T 60

The database was initialized with scale factor 1 (pgbench -i). I ran
multiple iterations for both the original and patched versions and compared
TPS and latency across runs.

All tests were done on the same system under similar conditions.

Please let me know if you’d like me to try any other scenarios or provide
more details.

Regards,
Lakshmi G

On Fri, Mar 20, 2026 at 4:28 PM Ranier Vilela <[email protected]> wrote:

> Hi.
>
> Em sex., 20 de mar. de 2026 às 06:50, lakshmi <[email protected]>
> escreveu:
>
>> Hi,
>>
>> I tried this change on PostgreSQL (19devel) and ran a few simple tests
>> using pgbench to see how it behaves in practice.
>>
>> I used 10 clients and 4 threads and ran each test for 60 seconds.
>>
>> From my runs, the original version was giving around ~663 TPS with ~15.07
>> ms latency.
>> With the patched version, I observed TPS in the range of ~638–657,
>> averaging around ~648 TPS, with latency slightly higher (~15.2–15.6 ms).
>>
> Thanks for the benchmark.
> Could you share the tests and the environment?
> compiler
> OS
> etc.
>
> best regards,
> Ranier Vilela
>


view thread (17+ messages)  latest in thread

reply
Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: [email protected]
  Cc: [email protected], [email protected], [email protected]
  Subject: Re: Avoid multiple calls to memcpy (src/backend/access/index/genam.c)
  In-Reply-To: <CAEvyyTjT70OVN_NCcHQ+PodVM7bkPccKPzJcAv0u+eW=qpZDMg@mail.gmail.com>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox