[Kst] [PATCH] Improve locking performance and reduce deadlocks
netterfield at astro.utoronto.ca
Wed Aug 16 16:32:32 CEST 2006
Good. We are about to leave on our drive home from Texas, so I won't be able
to test things much until the weekend.
One thing I noticed is that datamode with equations and plugins seemed like a
particularly easy place to get a deadlock.
On Wednesday 16 August 2006 06:44, George Staikos wrote:
> Eli and I spent more time working on locking and we discovered that it's
> not necessary to lock as much as we have been doing. This coincidentally
> fixes many deadlock scenarios (though not all of them), and at the same
> time appears to give a huge performance boost. Patch is attached, but
> needs testing. Please provide some feedback on this patch -without-
> --with-one-lock enabled. I'd like to commit this ASAP. I think this is
> the reason why deadlocks became more visible recently. Meanwhile we're
> working on fixing the rest of the problem (removing recursive locking in
> the tree), and hopefully we'll have a patch for that next week.
More information about the Kst