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: Andres Freund <andres@anarazel.de>, Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Cc: "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Date: 2026-02-10T20:53:58Z
Lists: pgsql-hackers

Attachments

On 10/02/2026 21:46, Andres Freund wrote:
> On 2026-02-10 19:15:27 +0000, Bertrand Drouvot wrote:
>> On Tue, Feb 10, 2026 at 01:15:01PM -0500, Andres Freund wrote:
>>> On 2026-02-10 19:14:44 +0200, Heikki Linnakangas wrote:
>>> Yea, I don't think we need to be perfect here. Just a bit less bad. And, as
>>> you say, the current order doesn't make a lot of sense.
>>> Just grouping things like
>>> - pid, pgxactoff, backendType (i.e. barely if ever changing)
>>> - wait_event_info, waitStart (i.e. very frequently changing, but typically
>>>    accessed within one proc)
>>> - sem, lwWaiting, waitLockMode (i.e. stuff that is updated frequently and
>>>    accessed across processes)
>>
>> With an ordering like in the attached (to apply on top of Heikki's patch), we're
>> back to 832 bytes.
> 
> You'd really need to insert padding between the sections to make it work...

Here's my attempt at grouping things more logically. I didn't insert 
padding and also didn't try to avoid alignment padding. I tried to 
optimize for readability rather than size or performance. That said, I 
would assume that grouping things logically like this would also help to 
avoid false sharing. If not, inserting explicit padding seems like a a 
good fix.

I also think we should split 'links' into two fields. For clarity.

With this, sizeof(PGPROC) == 864 without the explicit alignment to 
PG_CACHE_LINE_SIZE, and 896 with it.

- 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