Horton

Re: inefficient loop in StandbyReleaseLockList()

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: "Bossart, Nathan" <bossartn@amazon.com>
Cc: Kyotaro Horiguchi <horikyota.ntt@gmail.com>, "michael@paquier.xyz" <michael@paquier.xyz>, "andres@anarazel.de" <andres@anarazel.de>, "sulamul@gmail.com" <sulamul@gmail.com>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Date: 2021-11-01T17:32:41Z
Lists: pgsql-hackers

Attachments

"Bossart, Nathan" <bossartn@amazon.com> writes:
> Ah, I see it now.  The patch looks good to me, then.

Thanks for looking!  Here's an expanded patch that also takes care
of the other two easy-to-fix cases, nodeAgg.c and llvmjit.c.
AFAICS, llvm_release_context is like StandbyReleaseLockList
in that we don't need to worry about whether the data structure
is valid after an error partway through.  (Maybe we should be
worrying, but I think the callers would need work as well if
that's to be the standard.)

			regards, tom lane

Commits

  1. Doc: add some notes about performance of the List functions.

  2. Avoid O(N^2) behavior in SyncPostCheckpoint().

  3. Avoid some other O(N^2) hazards in list manipulation.

  4. Avoid O(N^2) behavior when the standby process releases many locks.