We are talking ‘Big’ Programs here!
by Dave Russ
Way back in the twilight of time itself (about 3 – 4
years ago) in a land not too far from here, there
could be found a new breed of creature. They were
in general a happy band of beings, though they
often complained of frequent headaches. You
would not see them very often, except maybe first
thing in the morning or last thing at night, and if
perchance you stumbled upon ore in its lair it
would seem as if the poor creature was in a trance.
Every now and again they would congregate in
cold halls and back rooms and regale each other
with stories until the small hours. These tales were
not of high adventure or epic endurance but how to
coax an extra 1k of RAM from a small home
computer thus doubling its capacity and the
suchlike. For hour upon hour they would feverishly
tell of how they boldly moved in on video RAM and
BASIC workspace and worked wonders on a tiny
system, that once it had set itself up, left you with
an available workspace that was barely into single
This was the heroic age of the early home
computer world. Now lost forever, I’m glad to say,
though some of the tales still linger on as classics.
However we still have one legacy from those dark
days, and I am sure that it is one of Murphy’s n
laws, and that is whatever system you have, a
program will always expand to fill it. Way back then
all you could do was read routines into available
workspace, wherever it may have been, and
through the magic of having read the manual, link
it all together to give a running program.
Now that we are living in more enlightened times,
and have acquired our 8 bit CP/M computer, we
now have oodles of memory to play with (We
do???), but still we manage to fill it up somehow.
We can now afford to program in high level
languages with little regard for the target object
code size. Fast, tight assembler type code is now
not a prerequisite for an efficient program running
on a small computer.
So we still manage to fill our memory up, but these
days we have better methods of running programs
larger than memory than poking routines into
obscure little corners. This is the purpose of this
awe inspiring (yawn) article.
It may be worth mentioning that I was talking to an
academic type not so long ago who is shortly
publishing a book on Artificial Intelligence. He said
that any machine in design today should be
capable of addressing at least 16 Megabytes of
RAM. This seems rather extravagant, doesn’t it, but
it seems that AI applications are not renowned for
their economy with memory.
Getting back to the point, I’m trying to say in my
own roundabout way that everyone ought to stop
before embarking on a mammoth project and think
ahead a little to the time when the program begins
to get bigger than the available memory. It may
prove difficult to restructure it in six months time,
and so we should consider the use of overlaying or
chaining even at this early stage.
To give an example, I began a test program for
GSX device drivers and at the start I realized that
the main menu would have to contain something
around 40 options. As it stands this is a prime
candidate for overlay techniques, as if I were to add
another 20 options along the way I would never
have fitted it into memory as a single program. So
now I have a single menu/controller program that
reads in and executes options as required. Once
complete the called subprogram is discarded and
the next read into the same memory area.
This approach is by no means revolutionary and is
well established. For example take a look at
Wordstar and diskpen, these make use of overlays,
DR DRAW and DR GRAPH are into double figures
with overlay files in their 8 bit versions.
What to do about it
OK, i hear you shout, so my programs are going to
get bigger, what the hell do I do about it??
Well you have a lot of options but they fall into 2
- Program chaining.
- Program overlays.
Program chaining is simple enough in theory but
could prove difficult in practice to implement.
When a program is complete it runs the next
program in the suite. The first program reads the
second into memory over its own memory area
and then executes it. Usually no trace of the
original program remains.
1. Easy to use if the language used supports
a chain function. Program creation using a
suitable chain function is simpler than overlaying.
2. Control flow between programs is easy to
3. With the use of some clever code a
program management menu could be written
to control a number of applications.