| From: | Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> |
|---|---|
| To: | Merlin Moncure <mmoncure(at)gmail(dot)com> |
| Cc: | Andres Freund <andres(at)2ndquadrant(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>, Robert Haas <robertmhaas(at)gmail(dot)com> |
| Subject: | Re: Wait free LW_SHARED acquisition - v0.9 |
| Date: | 2014-10-15 03:34:06 |
| Message-ID: | CAA4eK1JiP-yCyNXwLTxA29ErkWC=VNBk4zjF2BZuLkKEaheoew@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Wed, Oct 15, 2014 at 12:06 AM, Merlin Moncure <mmoncure(at)gmail(dot)com> wrote:
>
> A while back, I submitted a minor tweak to the clock sweep so that,
> instead of spinlocking every single buffer header as it swept it just
> did a single TAS as a kind of a trylock and punted to the next buffer
> if the test failed on the principle there's not good reason to hang
> around. You only spin if you passed the first test; that should
> reduce the likelihood of actual spinning to approximately zero. I
> still maintain there's no reason not to do that (I couldn't show a
> benefit but that was because mapping list locking was masking any
> clock sweep contention at that time).
If you feel that can now show the benefit, then I think you can rebase
it for the coming commit fest (which is going to start today).
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Robert Haas | 2014-10-15 04:03:45 | group locking: incomplete patch, just for discussion |
| Previous Message | Alvaro Herrera | 2014-10-15 02:00:09 | Re: [PATCH] Cleanup: unify checks for catalog modification |