🚀 go-pugleaf

In article <DZABEL.92Sep3122216@pc2.cs.tu-berlin.de> dzabel@cs.tu-berlin.de (Dirk Zabel) writes:
>
>But the ONE enhancement which would make the difference between a
>nice toy and a serious tool seems still to be missing: code & data 
>space > 64 K FOR EACH SINGLE MODULE. While this may not be a problem
>if you develop new code (at least not the 64 K maximum code size per
>module), it becomes nearly impossible to port existing software from
>unix (writtten using the Mocka compiler, in my case) which does not
>suffer from these restrictions!
>
Anyone using any system to produce modules with a *code* size >64K
per module should take a very critical look at their program's 
modularity.  Data size is a different matter.

>In the meantime: Is there any other Modula-Compiler for MS-DOS (may
>well require a 386) which does not not have those infamous 64 K -
>limits? I don't have experiences with Topspeed or Logitech, but have
>been told they are not better in this respect.
>
Stony Brook's Optimizing compiler supports the following memory models:
Code:
	Small -	Combined code of all modules <64k
	Large - Each Module <64k; total is limited only by available
		memory.

Data:
	Small - All data including stack & heap total <64k
	Medium- All static data total <64k;
		stack: an additional 64k for dynamic data
		heap: limited only by available memory
	Large -	As in medium but each module can have 64k of
		static data

Their data classes are:
	Static: variables not contained in a procedure, literal
		strings, structured constants.
	Dynamic: variables declared within procedures.
	Heap:	data allocated by NEW or Storage.ALLOCATE.

I believe that these all apply to 8088 and 'better' processors.

David M. Fellows