I repeated David's original tests not terribly rigorously[*] and got
these numbers:
* Unpatched: 72.396196
* 0001 : 77.279404
* 0001+0002: 20409.415094
* 0002: 816.606613
* control : 22969.140596 (insertion into unpartitioned table)
So while this patch by itself gives a pretty lame increase in tps, it
removes bottlenecks that will appear once we change the locking scheme.
[*] On my laptop, running each test only once for 60s.
--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services