Date: Thu, 15 Dec 2005 00:31:06 -0500 From: Jonathan Noack <noackjr@alumni.rice.edu> To: Mike Jakubik <mikej@rogers.com> Cc: stable@freebsd.org, Mark Kirkwood <markir@paradise.net.nz> Subject: Re: kernel cpu entries Message-ID: <43A0FF9A.6040505@alumni.rice.edu> In-Reply-To: <43A0FC61.3060704@rogers.com> References: <20051215002618.B4D3B5D07@ptavv.es.net> <43A0E607.2030101@alumni.rice.edu> <43A0E916.7070204@samsco.org> <43A0EC9F.9080800@paradise.net.nz> <43A0FC61.3060704@rogers.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Mike Jakubik wrote: > Mark Kirkwood wrote: >> Is a minor update to the handbook needed in order avoid confusion >> then? e.g. I have been commenting out CPU_I586 on all my PIII systems >> in the (mistaken it would seem) belief that having CPU_I686 only was >> better. > > Agreed, i have always just used I686, assuming it inherited the features > of I586. I think most people will assume this. A quick benchmark (average of five 'dd' runs sending a large file to /dev/null using a blocksize of 32k) similar to that suggested by Matt Dillon (http://www.freebsd.org/projects/ideas/#p-memcpy) shows ~3% increase with I586_CPU on a dual Pentium III. This was on a lightly loaded live system, but I think it shows that using I586_CPU on an i686 CPU does show a performance improvement (at least in certain tests). I have done no testing with blowfish to see how I586_CPU might affect it as I use AES, but it seems it should be recommended for i686 CPUs as well. -Jonathan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?43A0FF9A.6040505>