Re: LLVM jit and matview - Mailing list pgsql-hackers

From Dmitry Dolgov
Subject Re: LLVM jit and matview
Date
Msg-id CA+q6zcVFD9601Bfo43dPtLOtWHTYt7A5NqNcsRUMzZX64dBO3A@mail.gmail.com
Whole thread Raw
In response to Re: LLVM jit and matview  (Michael Paquier <michael@paquier.xyz>)
Responses Re: LLVM jit and matview  (Andres Freund <andres@anarazel.de>)
Re: LLVM jit and matview  (Andres Freund <andres@anarazel.de>)
List pgsql-hackers
> On Wed, 25 Jul 2018 at 07:49, Michael Paquier <michael@paquier.xyz> wrote:
>
> On Tue, Jul 24, 2018 at 07:49:56PM -0700, Andres Freund wrote:
> > FWIW, this doesn't crash for me, using a trunk checkout for LLVM. I'll
> > try older ones, as soon as they finish rebuilding. But perhaps you could
> > re-verify that this still is an issue on recent PG checkouts? And which
> > version of LLVM are you guys using?
>
> One of your recent fixes has visibly taken take of the issue (I am too
> lazy to check which one).  I am using llvm 6.0 with Debian SID as far as
> I know, configure links to stuff in /usr/lib/llvm-6.0/.

Looks like I still can reproduce it on the current head 167075be3ab. I was using
LLVM 5.0 package, but will try reproduce on the LLVM master branch as soon
as compilation will be finished (apparently, I need to rebuild it with
LLVM_USE_PERF, otherwise Postgres complains about undefined symbol:
LLVMCreatePerfJITEventListener).


pgsql-hackers by date:

Previous
From: Alexander Korotkov
Date:
Subject: Re: Locking B-tree leafs immediately in exclusive mode
Next
From: Jeremy Finzel
Date:
Subject: Using test_ddl_deparse as an extension