[net.arch] Large Main Memories

hutch@sdcsvax.UUCP (Jim Hutchison) (09/06/86)

<>

Zeroing of pages need not be done all at once.

Right now, if you have extra memory lying about (unusual yes, but read on),
it does not get 0'd until somebody grabs a "blank" page.  How about a mode
like "0 on reference"?  There is "copy on write" in some places, why not?
Then you only have to worry about 1 page at a time.  Extra swap space is not
costing you anything, why should extra memory?  Huh?
-- 
    Jim Hutchison   UUCP:	{dcdwest,ucbvax}!sdcsvax!hutch
		    ARPA:	Hutch@sdcsvax.ucsd.edu
	"The fog crept in on little cats feet" -CS

friesen@psivax.UUCP (Stanley Friesen) (09/10/86)

In article <2065@sdcsvax.UUCP> hutch@sdcsvax.UUCP (Jim Hutchison) writes:
 
>Zeroing of pages need not be done all at once.
>Right now, if you have extra memory lying about (unusual yes, but read on),
>it does not get 0'd until somebody grabs a "blank" page.  How about a mode
>like "0 on reference"?  There is "copy on write" in some places, why not?
>Then you only have to worry about 1 page at a time.  Extra swap space is not
>costing you anything, why should extra memory?  Huh?

	As a matter of fact demand-paged UNIX's(Sys Vr2+ & BSD 4.x) do exactly
that! They call them zero-fill pages. It is implemented using page
table entries with the valid bit off and a software flag marking the
entry a zero-fill page. On a page fault from such a page the kernal
allocates a free page frame and zeros it.
---

				Sarima (Stanley Friesen)

UUCP: {ttidca|ihnp4|sdcrdcf|quad1|nrcvax|bellcore|logico}!psivax!friesen
ARPA: ??