Date: Sun, 29 Sep 1996 15:00:33 -0500 (EST) From: "John S. Dyson" <toor@dyson.iquest.net> To: dyson@freebsd.org Cc: ache@nagual.ru, 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_object.c vm_object.h vm_page.c vm_pageout.c Message-ID: <199609292000.PAA10815@dyson.iquest.net> In-Reply-To: <199609291840.NAA07712@dyson.iquest.net> from "John S. Dyson" at Sep 29, 96 01:40:56 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > > Fix some redundancy in the page coloring code. Also, support > > > contiguous allocation of memory in objects (if the system doesn't > > > have to do anything special), if all else is equal. > > > > BTW, what is "page coloring"? Do you know any papers (URLs) available? > > I just realized that I might have been a bit rude by not answering your initial question as to what "page coloring" is. It is a way of choosing pages more carefully so that a direct mapped cache is used more effectively. It is a way of making a direct mapped cache look more like an associative cache. Unfortunately, our inclusion of page coloring has exacted a small performance hit for certain low level benchmarks (it is because we have optimizations to reuse recently used pages so that they are more likely in the cache.) In the long run, the performance is higher. Specifically, you might try running multiple copies of the heapsort/nsieve benchmark(s) in ftp.nosc.mil:/pub/aburto. John
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199609292000.PAA10815>