Date: Sun, 24 May 1998 01:09:16 -0500 (EST) From: "John S. Dyson" <toor@dyson.iquest.net> To: grog@lemis.com (Greg Lehey) Cc: dyson@FreeBSD.ORG, cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG, cvs-sys@FreeBSD.ORG Subject: Re: cvs commit: src/sys/vm vm_page.h Message-ID: <199805240609.BAA02858@dyson.iquest.net> In-Reply-To: <19980524153244.J353@freebie.lemis.com> from Greg Lehey at "May 24, 98 03:32:44 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
> On Sat, 23 May 1998 at 21:25:29 -0700, John Dyson wrote: > > dyson 1998/05/23 21:25:28 PDT > > > > Modified files: > > sys/vm vm_page.h > > Log: > > Support a 16K first level cache for 512K 2nd level. Also, add support > > for 1MB 2nd level cache. > > Does this mean that previously we didn't use the full 1MB second level > cache? > No, it is just a coloring (optimization) issue. Once you color to 256K, coloring more only gives a small incremental improvement over a barely measurable one to begin with. (Actually, the improvement *can* be significant, but the improvement isn't measurable under most conditions.) The support for the 1st level cache isn't too important anyway, since it is often a multi-way associative one, and such caches don't see an advantage when coloring. PPro machines support a 4way 2nd level cache, so fully coloring the cache isn't too valuable. It is most valuable to color a P5 2nd level cache, that is usually direct mapped. John
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199805240609.BAA02858>