Re: postgresql13-llvm jit-13.5-1PGDG.rhel8.x86_64

From: Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>
To: Mladen Gogala <gogala(dot)mladen(at)gmail(dot)com>
Cc: Pgsql Performance <pgsql-performance(at)lists(dot)postgresql(dot)org>
Subject: Re: postgresql13-llvm jit-13.5-1PGDG.rhel8.x86_64
Date: 2021-11-15 05:04:47
Message-ID: CAFj8pRBis=j4fLepacisTxVO=vW2JcquwZk8u98AHKkLRGMhcQ@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Hi

> There is a CentOS8-stream version which solves the problem but I cannot
> use that in the office. I will probably have to wait for another month
> before OL8 has everything that I need in its repositories. Now, the
> question is what kind of an impact will running without llvm-jit have?
> According to the links below, llvm-jit effects are quite spectacular:
>
> https://llvm.org/devmtg/2016-09/slides/Melnik-PostgreSQLLLVM.pdf
>
> https://www.pgcon.org/2017/schedule/events/1092.en.html

When JIT was used on very large query with a lot of CASE expr, then JIT has
a positive effect about 50%. On usual large queries, the effect of JIT was
about 20%. Unfortunately, JIT is sensitive to estimation, and the JIT
sometimes increases seconds to queries, although without JIT this query is
executed in ms. When you use a query that can be well calculated in
parallel, then positive effect of JIT is less.

Regards

Pavel

>
>
>
> --
> Mladen Gogala
> Database Consultant
> Tel: (347) 321-1217
> https://dbwhisperer.wordpress.com
>
>
>
>

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Mladen Gogala 2021-11-15 13:56:54 Re: postgresql13-llvm jit-13.5-1PGDG.rhel8.x86_64
Previous Message Mladen Gogala 2021-11-15 04:21:54 postgresql13-llvm jit-13.5-1PGDG.rhel8.x86_64