Re: add function for creating/attaching hash table in DSM registry - Mailing list pgsql-hackers

From Nathan Bossart
Subject Re: add function for creating/attaching hash table in DSM registry
Date
Msg-id aFXFCIJtJCy2ENYe@nathan
Whole thread Raw
In response to Re: add function for creating/attaching hash table in DSM registry  (Rahila Syed <rahilasyed90@gmail.com>)
List pgsql-hackers
On Fri, Jun 20, 2025 at 04:53:52PM +0530, Rahila Syed wrote:
> Thank you for implementing these changes.
> The improvements look good and enhance the feature's utility. I have
> already started incorporating
> GetNamedDSA into my code to display memory context statistics.

Great!  Thanks for the reviews.

> A potential future enhancement could be allowing GetNamedDSHASH to accept
> an existing DSA name.
> This would enable the DSHASH to reuse a DSA area instead of creating a new
> one each time.
> I plan to use this registry to store DSA pointers that all belong to the
> same DSA area, and this enhancement
> would be particularly beneficial. If you find this idea useful, I would be
> interested in working on it.

We briefly discussed returning the dshash's DSA upthread [0], and using an
existing DSA seems like a variation on that same idea.  I'm a little wary
of allowing folks to mess with the named dshash internal DSAs.  Not only
does messing with the DSA introduce some risk of breakage, but it could
make it more difficult to change implementation details for the named
dshash code in the future.  I suppose we could document the expected usage
and try to do a few sanity checks, though...

[0] https://postgr.es/m/aEmRcAH1Rmaocz3D%40nathan

-- 
nathan



pgsql-hackers by date:

Previous
From: Anne Struble
Date:
Subject: Memory allocation error DDL invalidation (seen with 15.13 & 16.9)
Next
From: Soumyadeep Chakraborty
Date:
Subject: Re: Improve CRC32C performance on SSE4.2