Path: utzoo!attcan!uunet!samsung!brutus.cs.uiuc.edu!ux1.cso.uiuc.edu!aries!mcdonald
From: mcdonald@aries.scs.uiuc.edu (Doug McDonald)
Newsgroups: comp.lang.fortran
Subject: Re: DOS 640K barrier?
Message-ID: <1990Apr7.161228.14105@ux1.cso.uiuc.edu>
Date: 7 Apr 90 16:12:28 GMT
References: <78829@tut.cis.ohio-state.edu> <16870001@hpfinote.HP.COM> <158@tslwat.UUCP>
Sender: usenet@ux1.cso.uiuc.edu (News)
Reply-To: mcdonald@aries.scs.uiuc.edu (Doug McDonald)
Organization: School of Chemical Sciences, Univ. of Illinois at Urbana-Champaign
Lines: 50

In article <158@tslwat.UUCP> louk@tslwat.UUCP (Lou Kates) writes:
>In article <16870001@hpfinote.HP.COM> pnl@hpfinote.HP.COM (Peter Lim) writes:
>>
>> 
>>>We also have OS/2 1.1 on the machine and it has 16 Megs of RAM.  Do I have
>>>to switch to OS/2 to utilize the extra memory for an application.  
>>>I have a program that is written in FORTRAN and it exceeds the 640 K limit.  I
>>>am currently using MS FORTRAN 5.0.
>>>
>>>I have heard of DOS extenders, but aren't they supposed to be quite
>>>expensive?  Would C be better for using the extra memory?  
>> 
>>Only expensive in the fact that they run better on 386(SX) machines because of
>>the edge the 386 architecture has over the 286 architecture.  The basic DOS
>>extenders just put LIM EMS 4.0 memory to full use.  If you have some EMS
>>memory already, then all you have to do is buy the program.  
>>
>>^^^^^^^^^^^^^^^^^^^^^^^^^^
>>I won't believe that totally. The DOS extender itself will easily cost
>>  $400-$500 + debugger $100 + extender compiler $700-$1000 = $1500
>>...
>
>There are a couple misconceptions in the above statements:
>
>1.   DOS extenders  do NOT  put EMS  memory to use. DOS extenders
>     use   ordinary   extended memory,  not  EMS  (also known  as
>     expanded) memory. EMS  is  a paged memory scheme whereas DOS
>     extenders give you a flat address space (not a paged address
>     space) using extended, NOT  EMS,  memory.  They  do this  by
>     switching the processor into its native "protected" mode and
>     switching back to the "real" mode that DOS  runs in whenever
>     you make  a DOS call. You are confusing  DOS  extenders with
>     memory managers like QEMM and 386-to-the-Max  which are used
>     to emulate EMS memory on an extended memory 386.
>

DOS extenders for use on **386's** do indeed do that. But for use
under **286s** you most emphatically do NOT get a **flat** address
space. It is still segmented. It still has all the problems that
the 8086 dies - plus a new, added one: you have to jump through
hoops to access data as code or code as data. The brain-dead
protected mode of the '86s (any of them) makes it impossible
to mark the same segment as both data and code. This is not a problem
with the 386 because one normally simply uses the "Tiny" model
where the code and data segments start at the same place in physical
memory, so a code address and the data address of the same memory
location are the same - and you have up to a gigabyte of code and data. 
But is IS a problem on the 286, eith its teensy segments.

Doug McDonald