Date: Sun, 8 Oct 2006 18:29:45 -0700 (PDT) From: Kip Macy <kmacy@fsmware.com> To: Paul Allen <pallen@ugcs.caltech.edu> Cc: Attilio Rao <attilio@freebsd.org>, freebsd-current@freebsd.org, David Xu <davidxu@freebsd.org>, Ivan Voras <ivoras@fer.hr> Subject: Re: [PATCH] MAXCPU alterable in kernel config - needs testers Message-ID: <20061008182521.U69745@demos.bsdclusters.com> In-Reply-To: <20061009012323.GD3467@riyal.ugcs.caltech.edu> References: <2fd864e0610080423q7ba6bdeal656a223e662a5d@mail.gmail.com> <20061008135031.G83537@demos.bsdclusters.com> <4529667D.8070108@fer.hr> <200610090634.31297.davidxu@freebsd.org> <20061008225150.GK793@funkthat.com> <3bbf2fe10610081555r67265368sf7f12edbf35bff0d@mail.gmail.com> <20061008155817.G29803@demos.bsdclusters.com> <20061009002200.GM793@funkthat.com> <20061009012323.GD3467@riyal.ugcs.caltech.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
> > AFAIK, the linux kernel has generally favored the per-cpu approach. In > that respect, relative underperformance of FreeBSD vs. Linux is an indicator > that per-cpu approaches deserve more weight in the FreeBSD world. As early as 2002 Linux scaled cleanly to 32-way on a powerpc machine. This is, as far as I can tell, a testament to IBM's investment in Linux. FreeBSD is just now moving past subsystem locking. A per-cpu approach will only make sense when finer grained locking is more consistently used throughout the kernel. Ideally in so doing we'll be more consistent about the granularity of locking across subsystems than we have been in the past. -Kip
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20061008182521.U69745>