Date: Mon, 7 Oct 1996 17:34:26 -0500 (EST) From: "John S. Dyson" <toor@dyson.iquest.net> To: guido@gvr.win.tue.nl (Guido van Rooij) Cc: dyson@freebsd.org, dyson@freefall.freebsd.org, CVS-committers@freefall.freebsd.org, cvs-all@freefall.freebsd.org, cvs-sys@freefall.freebsd.org Subject: Re: cvs commit: src/sys/vm vm_page.h Message-ID: <199610072234.RAA03072@dyson.iquest.net> In-Reply-To: <199610072035.WAA14914@gvr.win.tue.nl> from "Guido van Rooij" at Oct 7, 96 10:35:11 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> John Dyson wrote: > > > > > > Perhaps we could introduce a new word in the kernel config file. > > > Something like > > > cache <n> > > > where <n> is the amount of L2 cache. The L1 cache is processor specific > > > and thus can be obtained via the cpu directive. The rest can then be > > > doen with macros. > > > > > Actually, the L1 cache will be changing on P5 class machines soon. > > hmm. But that would be identifiable by the stepping number, right? > If that is to hard the cache directive above can be etended ;-) > I suspect that we can figure it out... :-). Your comments do show that the current method is inadequate (which I indeed agree with.) The config mod would require the powers that be (config-wizards) to make that decision. I really think that in the shortest term, that DOCUMENTED config variables are probably adequate. I didn't document them because my current naming convention is bogus. In the longer term (probably in release timeframe) an attempt to probe the system interface chips would be useful, upon failure, perhaps default to 256K. (Most reasonably high perf systems since the 386 and certainly the 486 have 256K (maybe some 128K) cache.) It is only the medium 386s that had 64K anyway. Those who have exceptions to the "rule" can optionally recompile the kernel (my opinion.) I don't really want to mess with (but could if strongly requested) changing the number of bins for the coloring dynamically. It is possible to do, since there aren't lots of pointers floating around in the implementation. It would be kind of ugly though. John
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199610072234.RAA03072>