Path: utzoo!attcan!uunet!lll-winken!lll-tis!ames!ncar!tank!uxc!uxc.cso.uiuc.edu!uxg.cso.uiuc.edu!uxe.cso.uiuc.edu!mcdonald
From: mcdonald@uxe.cso.uiuc.edu
Newsgroups: comp.sys.ibm.pc
Subject: Re: 640kb limit on DOS?
Message-ID: <45900179@uxe.cso.uiuc.edu>
Date: 13 Dec 88 14:49:00 GMT
References: <8043@ihlpl.ATT.COM>
Lines: 58
Nf-ID: #R:ihlpl.ATT.COM:8043:uxe.cso.uiuc.edu:45900179:000:3076
Nf-From: uxe.cso.uiuc.edu!mcdonald    Dec 13 08:49:00 1988


<<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.

<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.

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.

If you want more space for code or data, but not 32 bit operation,
you can switch a 386 to virtual 8086 mode and use paging to move
any part of memory into the lower 640 K. In this case there is NO
quibble about running "under PC-DOS". What is not clear to me,
despite having a two-inch thick book on the subject in my lap at
this very instant, is whether manipulating the page tables can
be done from virtual 8086 mode or whether you have to switch to 386 mode.
I haven't tried any of this; I like real 386 mode!

If you are a programmer, have a 386, enough memory, and the dough to shell
out, get the Phar Lap extender, and one of the 386 C or Fortran
compilers (from Microway or Metaware).

Doug McDonald (mcdonald@uxe.cso.uiuc.edu)
.