Subject: Re: UVM optimalisations / remarks
To: None <eeh@netbsd.org>
From: Charles M. Hannum <abuse@spamalicious.com>
List: tech-kern
Date: 03/26/2002 19:43:37
One idea I tossed around a while back was keeping track of pages that
are zeroed but have not actually been written to.  This actually
requires no more faults -- it's handled entirely by clearing the
"zeroed" bit whenever the page is detected to be modified.

This would allow us to, e.g., allocate stack pages in batches and then
free up ones that aren't actually in use.