IndexBuildHeapScan doesn't use page at a time mode

From: Andres Freund <andres(at)2ndquadrant(dot)com>
To: pgsql-hackers(at)postgresql(dot)org
Subject: IndexBuildHeapScan doesn't use page at a time mode
Date: 2014-02-01 15:44:55
Message-ID: 20140201154455.GD32407@alap3.anarazel.de
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi,

While looking at a profile I noticed that non-concurrent index builds
use the page at a time mode. Since that causes more buffer lwlocks to be
taken, that's a noticeable slowdown according to a completely
unscientific hack.
The reason for not using the page at a time mode is that we only allow
it for mvcc snapshots, but we're using SnapshotAny. Since we're doing
additional visibility checks afterwards and since the table is locked
exlusively it seems safe to me to allow the page at a time mode?

I don't plan to implement this right now, but I do plan to pick it up
somewhere in the next cycle if nobody beats me to it.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services

Browse pgsql-hackers by date

  From Date Subject
Next Message Christian Kruse 2014-02-01 16:03:46 Re: Patch: show xid and xmin in pg_stat_activity and pg_stat_replication
Previous Message Andres Freund 2014-02-01 15:07:45 Re: should we add a XLogRecPtr/LSN SQL type?