DOS ain't dead - huge memory model

kerravon

Ligao, Free World North,
22.11.2022, 16:30

huge memory model (Announce)

When I started writing PDOS/86 I didn't know
what "huge" actually was, and I thought it was
for strange people.

I chose "large" for PDOS/86 and when I went to
use my memory management routines, written in
pure C90, originally for MVS, I found that I
was stuck - I could only manage 64k, even
though I was using large memory model.

Rather than change my seemingly valid C90 code,
I instead put a hack into PDOS/86.

About an hour ago the hack was finally killed
off and the problem was solved simply by
recompiling with the "huge" memory model and
a 32-bit size_t. With pure 8086 instructions
from Watcom C, but using PDPCLIB as the C
library.

Code available at http://pdos.org

BFN. Paul.

Guti

25.11.2022, 10:23

@ kerravon

huge memory model

Post reply

> When I started writing PDOS/86 I didn't know
> what "huge" actually was, and I thought it was
> for strange people.
>
> I chose "large" for PDOS/86 and when I went to
> use my memory management routines, written in
> pure C90, originally for MVS, I found that I
> was stuck - I could only manage 64k, even
> though I was using large memory model.
>
> Rather than change my seemingly valid C90 code,
> I instead put a hack into PDOS/86.
>
> About an hour ago the hack was finally killed
> off and the problem was solved simply by
> recompiling with the "huge" memory model and
> a 32-bit size_t. With pure 8086 instructions
> from Watcom C, but using PDPCLIB as the C
> library.
>
> Code available at http://pdos.org
>
> BFN. Paul.

I discovered huge memory model back in the Turbo C++ days. A simple solution to adress all available conventional memory with chunks larger than 64K.

---
Visit my personal blog at https://www.javiergutierrezchamorro.com

kerravon Ligao, Free World North, 25.11.2022, 19:41 @ Guti	huge memory model Post reply
	> I discovered huge memory model back in the Turbo C++ days. A simple > solution to adress all available conventional memory with chunks larger > than 64K. I don't think Turbo C generates proper huge memory model code in its huge option. I think it just allows more than 64k for the DGROUP.

Guti

29.11.2022, 06:55

@ kerravon

huge memory model

Post reply

Indeed no. Turbo C supported huge pointers allowing to allocate blocks larger than 64K.

https://www.vogons.org/viewtopic.php?t=51678

> > I discovered huge memory model back in the Turbo C++ days. A simple
> > solution to adress all available conventional memory with chunks larger
> > than 64K.
>
> I don't think Turbo C generates proper huge
> memory model code in its huge option. I think
> it just allows more than 64k for the DGROUP.

---
Visit my personal blog at https://www.javiergutierrezchamorro.com

kerravon

Ligao, Free World North,
29.11.2022, 08:44

@ Guti

huge memory model

Post reply

> Indeed no. Turbo C supported huge pointers allowing to allocate blocks
> larger than 64K.
>
> https://www.vogons.org/viewtopic.php?t=51678

That seems to require manually defining the
pointers as huge, which is not what I expect
to have to do when I pass "-mh" to the
compiler, and in fact, not what I have to do
when I use Watcom.

Watcom is doing exactly what I want.

Someone else with experience with programming
using huge memory model said that it only adds
a few percent overhead to runtime.

In my case, an OS, it shouldn't even be that,
because the computer should be bottlenecked in
the app, not the OS. And the app can be in any
memory model.

BFN. Paul.