Re: Speeding up GIST index creation for tsvectors

From: Amit Khandekar <amitdkhan(dot)pg(at)gmail(dot)com>
To: Pavel Borisov <pashkin(dot)elfe(at)gmail(dot)com>
Cc: PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Subject: Re: Speeding up GIST index creation for tsvectors
Date: 2020-12-13 12:46:54
Message-ID: CAJ3gD9eEoZea-e0XkX=v8ePiT9jRPDX=HgUXePv8HYXT-5XE3g@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Thu, 10 Dec 2020 at 20:43, Pavel Borisov <pashkin(dot)elfe(at)gmail(dot)com> wrote:
>
> Hi, Amit!
> It's really cool to hear about another GiST improvement proposal. I'd like to connect recently committed GiST ordered build discussion here [1] and further improvement proposed [2]
>
> I've tested feature [1] and got 2.5-3 times speed improvement which is much better I believe.

Yeah, I am completely new to the GIST stuff, but I had taken a quick
look at the sortsupport feature for GIST, and found it very
interesting. I believe it's an additional option for making the gist
index builds much faster. But then I thought that my small patch would
still be worthwhile because for tsvector types the non-sort method for
index build would continue to be used by users, and in general we can
extend this small optimization for other gist types also.

> There is an ongoing activity [2] to build support for different data types for GiST. Maybe you will consider it interesting to join.
>
> BTW you may have heard about Gin and Rum [3] indexes which suit text search much, much better (and faster) than GiST. The idea to process data in bigger chunks is good. Still optimize index structure, minimizing disc pages access, etc. seems better in many cases.

Sure. Thanks for the pointers.

--
Thanks,
-Amit Khandekar
Huawei Technologies

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Andrey Borodin 2020-12-13 15:58:18 Re: Speeding up GIST index creation for tsvectors
Previous Message Gilles Darold 2020-12-13 09:17:51 Re: MultiXact\SLRU buffers configuration