Date: Mon, 09 Oct 2006 09:41:56 +0200 From: Ivan Voras <ivoras@fer.hr> To: freebsd-current@freebsd.org Subject: Re: [PATCH] MAXCPU alterable in kernel config - needs testers Message-ID: <egcuh6$2p8$1@sea.gmane.org> In-Reply-To: <20061008155350.L29803@demos.bsdclusters.com> References: <2fd864e0610080423q7ba6bdeal656a223e662a5d@mail.gmail.com> <2006 10082011.53649.davidxu@freebsd.org> <egbnv6$crh$1@sea.gmane.org> <20061008135031.G83537@demos.bsdclusters.com> <4529667D.8070108@fer.hr> <20061008155350.L29803@demos.bsdclusters.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Kip Macy wrote: >> Any word on how will they handle migration of threads across sockets (or >> will it be OS's job)? Judging from T1 architecture, I think such event >> would create a very large performance penalty, but I'm not an expert. > > It is the job of the OS to take locality into account in thread > scheduling. Moving between chips You'll just lose the L2 cache locality > just as you would on a normal SMP. I'm reading about the architecture now; it seems 8 threads of a core are divided into two thread groups (each of which has its own ALU) and there's a hierarchy of "closeness" in terms of moving threads between execution units: 1. inside of thread group (instantaneous) 2. between thread groups 3. between cores 4. between CPUs/sockets
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?egcuh6$2p8$1>