My list of things to do included optimising
GetRunningTransactionLocks(), run once per checkpoint.
I was thinking I needed to try harder to avoid acquiring LWlocks on
all the lock partitions.
ISTM that I don't need to do this - lwlocks on lock partitions are
almost never contended now, so this should go much faster than before.
Any thoughts? Do we think it would benefit from further tweaking?
I'll assume not unless I hear from somebody with a different idea.
--
Simon Riggs http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services