[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kdevelop-devel
Subject:    Re: Patch: wake up duchainlock writers
From:       David Nolden <zwabel () googlemail ! com>
Date:       2009-12-15 15:49:31
Message-ID: 200912151649.31794.zwabel () googlemail ! com
[Download RAW message or body]

Am Dienstag 15 Dezember 2009 16:05:05 schrieb Hamish Rodda:
> On Wed, 16 Dec 2009 01:50:01 am David Nolden wrote:
> > Am Dienstag 15 Dezember 2009 14:58:51 schrieb Hamish Rodda:
> > > On Tue, 15 Dec 2009 08:47:44 pm David Nolden wrote:
> > I think we can combine the advantages of both approaches, by adding wait-
> > conditions to the spin-lock approach, and letting waiters wait for the
> >  wait- conditions instead of just sleeping.
> 
> Interesting idea... I'm taking a look at reducing the amount of write locks

I'm playing around with that right now, as profiling has shown that the spin-
lock in "referencecounting.h" actually seems to be slower than a mutex, even 
if only 2 threads are involved.

I guess which approach is better ultimately depends on the amount of lock 
contention.

-- 
KDevelop-devel mailing list
KDevelop-devel@kdevelop.org
https://barney.cs.uni-potsdam.de/mailman/listinfo/kdevelop-devel
[prev in list] [next in list] [prev in thread] [next in thread]