*BSD News Article 11503


Return to BSD News archive

Received: by minnie.vk1xwt.ampr.org with NNTP
	id AA1813 ; Tue, 23 Feb 93 14:59:22 EST
Path: sserve!manuel.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!howland.reston.ans.net!gatech!udel!sbcs.sunysb.edu!stark.UUCP!gene
From: gene@stark.uucp (Gene Stark)
Newsgroups: comp.os.386bsd.misc
Subject: Re: A comment on 0.1 + 0.2.1 patchkit's stability
Date: 20 Feb 93 09:08:39
Organization: Gene Stark's home system
Lines: 42
Message-ID: <GENE.93Feb20090839@stark.stark.uucp>
References: <CGD.93Feb17150814@gaia.CS.Berkeley.EDU> <GENE.93Feb18171500@stark.stark.uucp>
	<1m07peINNrn1@hrd769.brooks.af.mil>
NNTP-Posting-Host: stark.uucp
In-reply-to: burgess@hrd769.brooks.af.mil's message of 18 Feb 1993 08:51:26 -0600

burgess@hrd769.brooks.af.mil (Dave Burgess) writes:
)In article <GENE.93Feb18171500@stark.stark.uucp> gene@stark.uucp (Gene Stark) writes:
))
))I was wondering if others were seeing this type of problem.  I have noticed
))that if a process allocates a large amount of memory so that its RSS is
))increased, then those page frames seem to become permanently attached to
))that process, even once they leave the RSS.  For example, try starting
))Emacs and reading in a couple of MB.  Then kill the buffer with the stuff
))in it and wait for things to settle down.  You will see Emacs with a low
))RSS, but active processes do not get the page frames that Emacs once had.
))
)
)  The particular problem with Emacs was recently discussed in *.emacs.*
)somewhere.  As I recall from that interaction, Emacs does not release the
)memory although the buffer has been emptied.  It seems to me that it was
)a 'feature' of Emacs so that it wouldn't have to go to the trouble of
)reallocating memory pages.  
)  On the other hand, it may be that I dreamed the whole thing, too....

It seems to me that the allocation of page frames as part of a process'
working set is something under the control of the OS and not of the process.
Whether or not Emacs releases its memory only impacts its use of *pages*,
not *page frames*.  In the scenario I described, the OS should preempt
the page frames used by Emacs after a short period of inactivity, and
reallocate them to active processes.  Instead, the page frames disappear
from *any* processes working set while Emacs is alive but inactive, and
then magically reappear when Emacs dies.

							- Gene Stark

Oops, I just discovered I am using a fascist news system that requires
additional fodder before it will allow me to post my message.  Oh, well,
I guess I'll have to track down what is rejecting the article and hack
it.  Right now, though, I'll just add this filler.  Hey wait!  Maybe
I can confuse it by using a different character than ">" for the included
text.  Let's try that...




--
							stark@cs.sunysb.edu