Horton

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

Bertrand Drouvot <bertranddrouvot.pg@gmail.com>

View thread View on postgr.es
From: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
To: Heikki Linnakangas <hlinnaka@iki.fi>
Cc: Andres Freund <andres@anarazel.de>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Date: 2026-02-21T09:42:13Z
Lists: pgsql-hackers 
Hi,

On Fri, Feb 20, 2026 at 11:03:09PM +0200, Heikki Linnakangas wrote:
> On 11/02/2026 06:40, Bertrand Drouvot wrote:
> > That looks ok to see PGPROC as an "acceptable" one, if not, should we use the
> > union trick?
> 
> It seems acceptable to just not align it if the compiler doesn't support it.
> This is just a performance optimization, after all.

Agreed.

> Attached is new versions the remaining patches. I think these are ready to
> be committed.

Thanks!

One nit, 0001 is adding the typedef:

"
-struct PGPROC
+typedef struct PGPROC
.
.
.
-};
-
-/* NOTE: "typedef struct PGPROC PGPROC" appears in storage/lock.h. */
+       uint32          wait_event_info;        /* proc's wait information */
+} PGPROC;
"

Would that make more sense to add the typedef when we introduce the explicit
alignment in 0002 (like it was done in your previous
v2-0001-Align-PGPROC-to-cache-line-boundary.patch up-thread)?

Regards,

-- 
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com

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