Path: utzoo!attcan!uunet!lll-winken!lll-lcc!ames!ncar!tank!uwvax!astroatc!nicmad!brown
From: brown@nicmad.UUCP (Vidiot (Vid-e-it))
Newsgroups: comp.sys.ibm.pc
Subject: Re: 640kb limit on DOS?
Message-ID: <3545@nicmad.UUCP>
Date: 15 Dec 88 02:46:04 GMT
References: <8043@ihlpl.ATT.COM> <45900179@uxe.cso.uiuc.edu>
Reply-To: brown@nicmad.UUCP (Vidiot (Vid-e-it))
Organization: Vidiot's Hangout
Lines: 73

In article <45900179@uxe.cso.uiuc.edu> mcdonald@uxe.cso.uiuc.edu writes:
<
<<<With a 386 box, running DOS 3.x, how does one cheat on the 640k limit.
<<<Yes the box has bios shadow ram and exteded memory, but how exactly can
<<<a 386 help you break DOS's 640k limit?
<
<(see below for several answers)
<
<<PS/MS-DOS does not, nor can it ever, access anything except the first 1 MB
<<of memory.  The first 640K being RAM and the rest being various hardware.
<<DOS is designed for the lowest common chip, the 8088, which only knows 1MB.
<
<This is, in a sense, true -- PC-DOS has no way on its own to access more
<than the memory below the screen memory. However, this is not necessarily
<640K. On a PC with only a CGA there is 640+64+32 = 736K available.
<On a PC with NO screen memory ( yes, they exist, a friend of mine has
<a VERY non-clone machine that is that way - he wrote the bios himself) 
<you can have 768K. However, nothing prohibits a program RUNNING UNDER
<PC-DOS from using a 386 in any way it pleases. See below.

PC/MS-DOS will not, on its own, access 736K.  There are drivers out there
to do it for you.  But, that 96K isn't really much, considering the Megs
of memory that can be added to a 386 machine.

MS/PC-DOS is purposely written to stay within the 640K limit.  That is
why they are producing OS/2.  Unfortunatly OS/2 will NOT run existing
DOS applications (expect one in a DOS partition, true?).

I think the original poster would like to be able to load an application,
say like DBASE, that would like to have lots of memory for lots of records,
but can't because DOS won't let him.  This may be a lousy example, but I
hope you understand.  For me, I would like to be able to use QuickBASIC
compiled programs that work with lots of data in the 4MB that I have, but
it won't because DOS won't let me.

<<Even if you run Windows/386, DesqVIEW or VM/386, the DOS partition that is
<<created can only run like an 8088, ie, 640K of user memory for programs.
<
<I believe that this is indeed quite true.
<
<<The only way to get more is to run OS/2 and applications designed for OS/2.
<This is ABSOLUTELY NOT TRUE. NOT TRUE. Read my lips: not true. You can
<run Xenix or some other Unix variant on your 286 or 386. These are not
<so limited. On a 386 some Unixes are real 32bit sustems, some aren't.

Sorry, I should have included the UNIX operating systems.  But the original
poster was talking about DOS, and OS/2 is supposed to be the next generation
DOS.  Excuse me if I don't think so.

<On a 386 you can also run a "386 extender" program in genuine 386 mode.
<This is not quite PURE "running under PC-DOS" - but is indeed IS
<"running under PC-DOS". Its impurity comes from the fact that you have
<something between your 386 code and PC-DOS - either the Phar Lar
<or AI ARchitects extender, or something similar. But to your program,
<it appears to indeed BE PC-DOS: all the usual interrupts work as usual,
<the screen is directly mapped (though it can be anywhere in the 4 gigabyte
<address space), etc. I know this sounds like magic to people used
<to the "classic" 8086. But it is true - the magic is called "paging",
<and it is part of the 386. OF course, there are very few real 386
<(32 bit) programs out there - but there are some commercially
<available. I believe that IBM Interleaf Publisher is such a beast (
<no flames if I am wrong, please - this is by memory.) I have written several
<programs myself.

Doesn't the application have to be written for the AI Architects type code?
The Teradyne CAE package (known as CASE) uses the AI stuff for their 4.0
version of the CAE package.  For a user to get, say QuickBASIC, to work
with 4MB of memory would require the QuickBASIC code to be expressly designed
to work with AI, is this not so?
-- 
	       harvard-\	 att--\
Vidiot            ucbvax!uwvax!astroatc!nicmad!brown
	       rutgers-/      decvax--/
	ARPA/INTERNET: brown%nicmad.UUCP@spool.cs.wisc.edu