JIT in PostgreSQL 12 ?

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

JIT in PostgreSQL 12 ?

Tobias Gierke
Hi,

Browsing the PostgreSQL 12 release notes I noticed that JIT is now
enabled by default. Having not followed PostgreSQL development closely -
does this mean that compilation results are now getting cached and
compilation is no longer executed separately for each worker thread in a
parallel query ?

Cheers,
Tobias



Reply | Threaded
Open this post in threaded view
|

Sv: JIT in PostgreSQL 12 ?

Andreas Joseph Krogh-2
På onsdag 29. mai 2019 kl. 10:02:50, skrev Tobias Gierke <[hidden email]>:
Hi,

Browsing the PostgreSQL 12 release notes I noticed that JIT is now
enabled by default. Having not followed PostgreSQL development closely -
does this mean that compilation results are now getting cached and
compilation is no longer executed separately for each worker thread in a
parallel query ?
 
I don't know, but just want to chime in with my experience with PG-12 and JIT: Execution-time is still way worse then JIT=off for your queries so we'll turn JIT=off until we can mesure performance-gain.
 
--
Andreas Joseph Krogh
CTO / Partner - Visena AS
Mobile: +47 909 56 963
 
Reply | Threaded
Open this post in threaded view
|

Re: Sv: JIT in PostgreSQL 12 ?

Tobias Gierke
Hi,
På onsdag 29. mai 2019 kl. 10:02:50, skrev Tobias Gierke <[hidden email]>:
Hi,

Browsing the PostgreSQL 12 release notes I noticed that JIT is now
enabled by default. Having not followed PostgreSQL development closely -
does this mean that compilation results are now getting cached and
compilation is no longer executed separately for each worker thread in a
parallel query ?
 
I don't know, but just want to chime in with my experience with PG-12 and JIT: Execution-time is still way worse then JIT=off for your queries so we'll turn JIT=off until we can mesure performance-gain.

Hm, that's a bummer. So I guess we'll also have to make sure JIT is turned off when upgrading.

Thanks,
Tobias


Reply | Threaded
Open this post in threaded view
|

Re: Sv: JIT in PostgreSQL 12 ?

Andreas Joseph Krogh-2
På onsdag 29. mai 2019 kl. 11:13:10, skrev Tobias Gierke <[hidden email]>:
Hi,
På onsdag 29. mai 2019 kl. 10:02:50, skrev Tobias Gierke <[hidden email]>:
Hi,

Browsing the PostgreSQL 12 release notes I noticed that JIT is now
enabled by default. Having not followed PostgreSQL development closely -
does this mean that compilation results are now getting cached and
compilation is no longer executed separately for each worker thread in a
parallel query ?
 
I don't know, but just want to chime in with my experience with PG-12 and JIT: Execution-time is still way worse then JIT=off for your queries so we'll turn JIT=off until we can mesure performance-gain.

Hm, that's a bummer. So I guess we'll also have to make sure JIT is turned off when upgrading.

That's what we'll do, unfortunately:-(
 
--
Andreas Joseph Krogh
CTO / Partner - Visena AS
Mobile: +47 909 56 963
 
Reply | Threaded
Open this post in threaded view
|

Re: JIT in PostgreSQL 12 ?

Andreas Karlsson
In reply to this post by Tobias Gierke
On 5/29/19 10:02 AM, Tobias Gierke wrote:
> Browsing the PostgreSQL 12 release notes I noticed that JIT is now
> enabled by default. Having not followed PostgreSQL development closely -
> does this mean that compilation results are now getting cached and
> compilation is no longer executed separately for each worker thread in a
> parallel query ?

No, the compilation still happens once for each worker. PostgreSQL 12
includes some smaller performance improvements for the JIT but nothing
big like that.

Andreas


Reply | Threaded
Open this post in threaded view
|

Re: Sv: JIT in PostgreSQL 12 ?

Justin Pryzby
In reply to this post by Andreas Joseph Krogh-2
On Wed, May 29, 2019 at 10:48:19AM +0200, Andreas Joseph Krogh wrote:
> På onsdag 29. mai 2019 kl. 10:02:50, skrev Tobias Gierke <[hidden email]>
> Hi,
>
>  Browsing the PostgreSQL 12 release notes I noticed that JIT is now
>  enabled by default. Having not followed PostgreSQL development closely -
>  does this mean that compilation results are now getting cached and
>  compilation is no longer executed separately for each worker thread in a
>  parallel query ?

Thanks for starting the conversation.

> I don't know, but just want to chime in with my experience
> with PG-12 and JIT: Execution-time is still way worse then JIT=off for your
> queries so we'll turn JIT=off until we can mesure performance-gain.

That's also been my consistent experience and conclusion.

I gather the presumption has been that JIT will be enabled by default..
..but perhaps this is a "Decision to Recheck Mid-Beta" (?)

Justin