From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Andrew Dunstan <andrew(at)dunslane(dot)net> |
Cc: | PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
Subject: | Re: Fwd: Big Performance drop of Exceptions in UDFs between V11.2 and 13.4 |
Date: | 2021-08-27 17:30:07 |
Message-ID: | 845520.1630085407@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
> First, this apparently only affects build done with NLS. And in fact
> even on release 11 the performance is much better when run on a non-NLS
> build. So there's lots of work to do here.
Wow ... it would not have occurred to me to check that.
Testing that angle using HEAD on Linux (RHEL8), here are times
I see for the OP's slow query:
Non-NLS build, C locale:
Time: 12452.062 ms (00:12.452)
NLS build, en_US.utf8 locale:
Time: 13596.114 ms (00:13.596)
NLS build, after SET lc_messages TO 'es_ES.utf8':
Time: 15190.689 ms (00:15.191)
So there is a cost for translating the error messages on Linux too,
but it's not nearly as awful as on Windows. I wonder if this
boils down to a performance bug in the particular gettext version
you're using?
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Peter Geoghegan | 2021-08-27 17:52:13 | Re: log_autovacuum in Postgres 14 -- ordering issue |
Previous Message | David Christensen | 2021-08-27 17:29:24 | Re: [PATCH] pgbench: add multiconnect option |