Re: pascal system

["mdj" <mdj.mdj@gmail.com>]

Paul Schlyter wrote:

> Unfortunately, Apple Pascal didn't run on a standard Apple II, sans
> "language card"....

There was a 48k version of the runtime up to 1.1. You couldn't run the
development environment without a language card though.

The operating system switch issues on the Apple II weren't in reality
any worse than switching between DOS and CP/M ...

> Anyway, I remembered one actual advantage of Apple Pascal compared to
> TP (versions 1 to 3): the ability to create modules (called "units" in
> UCSD Pascal jargon - units was a UCSD Pascal extension which was
> absent in the Pascal standard), i.e. separately compiled pieces of
> code.  Early versions of TP required the whole program in one single
> source file available at compilation time.  The limitations due to
> this was in TP overcome by $include statements (which allowed the
> source to be split up into several files, even though it logically
> still was one file) and overlays (if the generated object code became
> too large).  And of course TP was so much faster at compiling than
> Apple Pascal that usuallyTP was still faster even though it had to
> recompile all source code for each new compile.

Yes, and units implemented true shared libraries as well, which saved a
lot of disk space in a big application. Having UNITS link automatically
at runtime was a big advantage, and once you got used to it, the
programs you compiled ended up being fairly small.

There was a version of UCSD Pascal for z-80 systems, but I never used
it, and I don't think there was an Apple CP/M port.

> If you're referring to the 6502 f.p. routines in the Integer Basic
> ROM's -- sorry, they were no help.  They too were 4-byte
> floating-point numbers, the more or less standard 32-bit f.p. format
> used in e.g.  Apple Pascal.  Perhaps you modified them to extend the
> precision?  Or perhaps you instead borrowed the f.p. routines from
> Applesoft?  If the latter: they were written not by Woz but by
> Microsoft - probably by Gates and/or Allen.

Ah, sorry, you're right... the REAL type was essentially the IEEE-754
floating point, I'll dig out the implementation I used, obviously they
were intended for double precision.

This was done as a part of a mathematics program for graphing - which
also used some assembly trickery to calculate trig functions quickly.
In the end, so much code ended up optimised for speed I seriously
contemplated moving the whole thing over to ProDOS, but the DHR code I
was using at the time was written in Pascal, and I couldn't be bothered
porting it.

Matt