Horton

Re: PGPROC alignment (was Re: pgsql: Separate RecoveryConflictReasons from procsignals)

Heikki Linnakangas <hlinnaka@iki.fi>

View thread View on postgr.es
From: Heikki Linnakangas <hlinnaka@iki.fi>
To: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Cc: Andres Freund <andres@anarazel.de>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Date: 2026-02-11T10:03:51Z
Lists: pgsql-hackers

Attachments

On 11/02/2026 06:40, Bertrand Drouvot wrote:
> A few comments:
> 
> 0001:
> 
> + * and (b) to make the multiplication / division to convert between PGPROC *
> + * and ProcNumber be a little cheaper
> 
> Is that correct if PGPROC size is not a power of 2?

You're right, it's not.

> 0002: Good catch!

Committed that.

>> With this, sizeof(PGPROC) == 864 without the explicit alignment to
>> PG_CACHE_LINE_SIZE, and 896 with it.
> 
> I can see 876 -> 896 on my side:
> 
> /*    872      |       4 */    uint32 wait_event_info;
> /* XXX 20-byte padding   */
> 
>                                 /* total size (bytes):  896 */
>                               }

Interesting. I've attached 'pahole bin/postgres' output from my laptop. 
It's Linux on arm64. This is with my v2 patches to rearrange the fields, 
but with the "pg_attribute_aligned(PG_CACHE_LINE_SIZE)" removed.

- Heikki

Commits

  1. Use C11 alignas in typedef definitions

  2. Align PGPROC to cache line boundary

  3. Rearrange fields in PGPROC, for clarity

  4. Split PGPROC 'links' field into two, for clarity

  5. Remove useless store to local variable

  6. Separate RecoveryConflictReasons from procsignals