Date: Sun, 15 May 2011 19:24:50 +0300 From: Andriy Gapon <avg@FreeBSD.org> To: Max Laier <max@love2party.net> Cc: FreeBSD current <freebsd-current@FreeBSD.org>, Peter Grehan <grehan@FreeBSD.org>, John Baldwin <jhb@FreeBSD.org>, neel@FreeBSD.org Subject: Re: proposed smp_rendezvous change Message-ID: <4DCFFE52.1090002@FreeBSD.org> In-Reply-To: <201105151209.13846.max@love2party.net> References: <4DCD357D.6000109@FreeBSD.org> <4DCFE8FA.6080005@FreeBSD.org> <4DCFEE33.5090808@FreeBSD.org> <201105151209.13846.max@love2party.net>
next in thread | previous in thread | raw e-mail | index | archive | help
on 15/05/2011 19:09 Max Laier said the following: > > I don't think we ever intended to synchronize the local teardown part, and I > believe that is the correct behavior for this API. > > This version is sufficiently close to what I have, so I am resonably sure that > it will work for us. It seems, however, that if we move to check to after > picking up the lock anyway, the generation approach has even less impact and I > am starting to prefer that solution. > > Andriy, is there any reason why you'd prefer your approach over the generation > version? No reason. And I even haven't said that I prefer it :-) I just wanted to show and explain it as apparently there was some misunderstanding about it. I think that generation count approach could even have a little bit better performance while perhaps being a tiny bit less obvious. -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4DCFFE52.1090002>