From: Tom Lane <tgl@sss.pgh.pa.us>
To: Justin Pryzby <pryzby@telsasoft.com>
cc: Peter Geoghegan <pg@bowt.ie>, David Rowley <dgrowleyml@gmail.com>,
        "ldh@laurent-hasson.com" <ldh@laurent-hasson.com>,
        "pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
Subject: Re: Big performance slowdown from 11.2 to 13.3
In-reply-to: <20210722164127.GI19620@telsasoft.com>
References: 
 <MN2PR15MB2560778595BB7AF6E691B26785E39@MN2PR15MB2560.namprd15.prod.outlook.com>
 <MN2PR15MB2560D429866BD65A70574A6C85E49@MN2PR15MB2560.namprd15.prod.outlook.com>
 <CAApHDvqoF7y7SQjGT=940njnAd_CUcUG6oRYChcP_+eJyuNiug@mail.gmail.com>
 <MN2PR15MB25603D3CE06C8F9C312EA58985E49@MN2PR15MB2560.namprd15.prod.outlook.com>
 <MN2PR15MB2560E4574A6E2B5AED2C03CB85E49@MN2PR15MB2560.namprd15.prod.outlook.com>
 <784576.1626968702@sss.pgh.pa.us> <785218.1626969409@sss.pgh.pa.us>
 <CAApHDvo+Gb80BVPKYCwHbLa7a7ZXp9ziq9URrpQaKdvXsd46bg@mail.gmail.com>
 <786600.1626970898@sss.pgh.pa.us>
 <CAH2-WzmftijCna_aY-HyCUfswwfPCC4d2HjKab1cxM0QCFrRUQ@mail.gmail.com>
 <20210722164127.GI19620@telsasoft.com>
Comments: In-reply-to Justin Pryzby <pryzby@telsasoft.com>
	message dated "Thu, 22 Jul 2021 11:41:27 -0500"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <788606.1626973012.1@sss.pgh.pa.us>
Date: Thu, 22 Jul 2021 12:56:53 -0400
Message-ID: <788607.1626973013@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/788607.1626973013%40sss.pgh.pa.us>
Precedence: bulk

Justin Pryzby <pryzby@telsasoft.com> writes:
> Oh.  So the problem seems to be that:

> 1) In v12, HashAgg now obeyes work_mem*hash_mem_multiplier;
> 2) Under windows, work_mem is limited to 2GB.

And more to the point, work_mem*hash_mem_multiplier is *also* limited
to 2GB.  We didn't think that through very carefully.  The point of
the hash_mem_multiplier feature was to allow hash aggregation to still
consume more than the work_mem limit, but we failed to free it from
this 2GB limit.

You're right though that this is Windows-only; on machines with
64-bit "long" there's less of a problem.

			regards, tom lane