Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!microsoft!donh
From: donh@microsoft.UUCP (Don HACHERL)
Newsgroups: comp.os.os2.misc
Subject: Re: Virtual Memory Page Size
Message-ID: <72555@microsoft.UUCP>
Date: 25 May 91 00:12:06 GMT
References: <9645@cognos.UUCP> <1991May17.171113.11939@unixg.ubc.ca>
Reply-To: donh@microsoft.UUCP (Don Hacherl)
Organization: Microsoft Corp., Redmond WA
Lines: 29

In article <1991May17.171113.11939@unixg.ubc.ca> ballard@ucs.ubc.ca (Alan Ballard) writes:
>In article <9645@cognos.UUCP> bradc@cognos.UUCP () writes:
>>Is the virtual memory page size (ie. the unit of swapping) adjustable
>>under OS/2 (version 1.3) or is it some fixed size such as 64K?
>
>OS/2 1.x uses SEGMENTED, not paged, virtual memory.  The pieces
>swapped are the code and data segments as defined by the applications.
>Using a larger number of smaller data segments might help. 

Avoiding very large segments can help reduce swap latency on OS/2 1.x,
but you don't want to over do it.  Don't forget that the OS has a
fixed per-segment overhead that's taken out of resident memory, i.e.
more user segments -> less memory available to put them in.  As a rule
of thumb I'd try to tune code segments to be in the range 2-16k.  OS/2
2.0 is paged, though, and its behaviour is different.  Since page
fragments past the end of segments go unused (well, almost, but close
enough) and all swap I/O occurs 4k at a time no matter what the
segment size, you want to have large segments to minimize
fragmentation.  I've looked at a couple of apps that were swap tuned
for 1.x (or Windows) by using tiny segments and they wasted *amazing*
amounts of memory under 2.0.  A good compromise for the two systems
would be in the 16-32k range.

Don
-- 
Don Hacherl		"Sorry, no pithy quotes today"
donh%microsof@uunet.uu.net  or  {uunet|uw-beaver}!microsoft!donh
Disclaimer:  Little of what I say should be taken seriously, and none
	     should be taken as the opinion of my employer.