Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 26 Jul 1998 21:53:27 +0000 (GMT)
From:      Terry Lambert <tlambert@primenet.com>
To:        green@zone.baldcom.net (Brian Feldman)
Cc:        freebsd-current@FreeBSD.ORG
Subject:   Re: New LINT options: what is VM coloring?
Message-ID:  <199807262153.OAA17869@usr01.primenet.com>
In-Reply-To: <Pine.BSF.4.00.9807251808530.6651-100000@zone.baldcom.net> from "Brian Feldman" at Jul 25, 98 06:11:47 pm

next in thread | previous in thread | raw e-mail | index | archive | help
> Would anyone mind letting me in on exactly what "VM coloring" is, and
> especially whether or not I should define one of these?:
> # Options for the VM subsystem
> #options        PQ_NOOPT                # No coloring
> options         PQ_LARGECACHE           # color for 512k/16k cache
> #options        PQ_HUGECACHE            # color for 1024k/16k cache
> As far as I can tell, it would seem these would be options to set certain
> parameters to optimal as far as the VM subsystem goes, and if so, is it
> safe to use one?

] Page Coloring uses the operating system to control the mapping of
] virtual to physical address by controlling the physical memory page
] which virtual pages are allocated to. Coloring becomes useful when
] the page number is involved in cache indexing, i.e., for large, low
] associativity, first level caches: exactly the trends the industry
] is following. 

[ ... ]

] For these large physically-indexed caches, the use of the low-order
] bits of the physical page number (the color) in cache indexing
] involve the virtual to physical translation in the location of data
] in the cache. While the virtual address space is laid out very
] uniformly, the lack of control over the virtual-to-physical mapping
] by most operating systems causes the physical color (and therefore
] cache placement modulo the page size) to be ``randomly'' selected.
] Thus, depending on the order in which the physical pages are
] assigned, cache miss rate can vary significantly.

See:

	http://umunhum.stanford.edu/res_html/great92/node13.html#SECTION00031000000000000000

Or click in William L. Lynch's name at:

	http://umunhum.stanford.edu/res_html/great92/node12.html

For the full text of the paper.

See also:

	http://www.pdl.cs.cmu.edu/SDI/todd_mowry.html (abstract)
	http://www-flash.stanford.edu/~bugnion/CDPC/  (full paper)



					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199807262153.OAA17869>