From:	CRDGW2::CRDGW2::MRGATE::"SMTP::CRVAX.SRI.COM::RELAY-INFO-VAX" 28-MAY-1991 17:33:04.52
To:	ARISIA::EVERHART
CC:	
Subj:	Re: MEMTRIM

From:	RELAY-INFO-VAX@CRVAX.SRI.COM@SMTP@CRDGW2
To:	Everhart@Arisia@MRGATE

Received:  by crdgw1.ge.com (5.57/GE 1.97)
	 id AA05307; Tue, 28 May 91 16:31:22 EDT
Received: From UCBVAX.BERKELEY.EDU by CRVAX.SRI.COM with TCP; Tue, 28 MAY 91 12:29:03 PDT
Received: by ucbvax.Berkeley.EDU (5.63/1.42)
	id AA09721; Tue, 28 May 91 12:20:16 -0700
Received: from USENET by ucbvax.Berkeley.EDU with netnews
	for info-vax@kl.sri.com (info-vax@kl.sri.com)
	(contact usenet@ucbvax.Berkeley.EDU if you have questions)
Date: 29 May 91 01:13:25 GMT
From: munnari.oz.au!ariel!ucsvc.ucs.unimelb.edu.au!mits.com.au!alistair@uunet.uu.net
Subject: Re: MEMTRIM
Message-Id: <1991May28.201326.3466@mits.com.au>
References: <15MAY91171614@marie.mit.edu>
Sender: info-vax-request@kl.sri.com
To: info-vax@kl.sri.com

In article <15MAY91171614@marie.mit.edu>, chuck@marie.mit.edu 
(CHUCK PARSONS 617-253-4157) writes:
>  I just picked up MEMTRIM from VMSNET.SOURCES and really like it.
>It really helps our batch jobs by freeing up a lot of memory. However

Glad you like it!

>We found that knocking the working set down to 200 is sometimes to heavy
>a hit, processes go into RWMPB wait for as much as a minute or so, and
>if this is DECWINDOWS thats anoying to the user. So I modified MEMTRIM.MAR
>as follows:
>
>;^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>;	ACB with the codde following
>;^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 
>Trim_KAST:
>acb:	.blkb	acb$k_length		;Allocate space for a baseline ACB
>old_ws:	.blkl	1			;saved working set size
>Trim_Code:
>	$adjwsl_s	wsetlm=old_ws	;get current wworking set size
>	blbc	r0,escape		;if error get out
>	pushl	r3			;Save r3
>	subl3	#200,old_ws,r3		;compute the difference between
>					;the current working set size and
>					;the new size of 200 pages 
>	blss	escape			;if he is already under then get out
>; ********** begin new code CCGP may 15 1991
>        ashl    #-1,r3,r3               ;divide by tw0 only do half at a time
>        incl    r3                      ;to make change gentler
>;  *************  end new code
>	mnegl	r3,r3			;negate the difference
>	$adjwsl_s	pagcnt=r3	;Adjust down to 200 pages
>escape:	movl	r5,r0			;set up for deallocation
>	popl	r3			;restore r3
>	jmp	g^exe$deanonpaged	;get out of here
>MemTrimAst_sz	= . - Trim_KAST
>
>	.end
>
>   Expecting that it would knock the workingset down half the distance 
>from current size to 200 pages (+1).  It does work, reducing the size
>gradually, but it seems to not exactly the formula:
>
>      new_size=200+(old_size-200)/2 + 1 

This isn't the formula you have above!  The call to $adwsl specifies
how much to CHANGE the working set by, not the new value.  Your
formula is :-

	new_size = old_size - (old_size-200)/2 + 1

>  that I expect, often the reduction is a good bit less.
>
>  Just wondered if anyone out there could explain it to me. If I've made
>some stupid assembler mistake please be gentle ;-)
>
>thanks CHUCK@MITLNS.MIT.EDU

This seems like a good idea, I may use it a bit myself.

Thanks,

Alistair Grant

Internet:     alistair@mits.com.au    PSI Mail:     psi%0505238730004::alistair
Alternate:    alistair%mits.com.au@ucsvc.ucs.unimelb.edu.au
              grant@decus.com.au
Phone:        +61 3 615 4641	      Fax:          +61 3 615 4408