Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!wuarchive!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!hplabs!hp-pcd!hpcvra!frankw
From: frankw@hpcvra.CV.HP.COM (Frank Wales)
Newsgroups: comp.sys.handhelds
Subject: Re: What If - someone wrote a HP-48SX C compiler?
Message-ID: <21580037@hpcvra.CV.HP.COM>
Date: 21 Mar 90 00:48:15 GMT
References: <8697@rosevax.Rosemount.COM>
Organization: Hewlett-Packard Co., Corvallis, OR, USA
Lines: 51

[Note: I'm not an HP employee, and don't speak for them, but I do
 have experience pertinent to this topic.]

In some article The Amazing Daniel Winker (daniel@rosevax.Rosemount.COM)
asks about writing an HP-48 C compiler.

Well, I don't want to sound to discouraging, but I think writing a
C compiler for the 48 would be an interesting project, in the sense
of the Chinese curse "may you live in interesting times."

There are certain fundamental difficulties:

	-- the 48 has no file system, so any implementation of
	   open(), close(), read(), write(), etc. would need heavy
	   rework, never mind stdio;
	-- no processes, so no fork(), exec(), etc.;
	-- signal(), setjmp(), and many other commonly used
	   system calls would need work to behave reasonably --
	   some might be impossible to make work;
	-- 99% of the 48's maths functionality has no parallel in C;
	-- fundamental types would be non-standard by most common
	   C implementations -- the 48 CPU has 64-bit registers
	   and 20-bit nibble-based memory addressing, for example, which
	   pushes the usual portability limits;
	-- the 48 operating system is written in a language (RPL) which is
	   like a sort-of polymorphic, sort-of object-oriented, stack-based
	   blend of Forth, LISP and other stuff, and has few things in common
	   with C, certainly at the User level (no pointers, no memory
	   allocation, no reasonable analogues to goto, break or continue).

The last of these is probably the biggest obstacle to a reasonable
implementation of C.  You can either beat the hell out of C to make it
fit the 48, or ignore the OS and take over the hardware.  The former
would involve so many compromises and changes to C that whatever came
out the other end would be seriously unlike any other C you ever saw.
The latter would be nigh on impossible without the internals
documentation, and I would doubt you could get your hands on that.  It
would also defeat much of the obvious object of writing C on the 48,
which is to access its maths capabilities while writing in a popular language.

It would be a genuinely interesting and educational exercise to attempt
to write something which could take C code (ignoring the standard
libraries and providing compatible alternatives which accessed the 48's
resources) and output something which implemented that program on the
48, either by producing RPL code or assembler code.  However, if your
final goal is to bring existing C code to the 48, or to turn the machine
into a C one instead of an RPL one, I feel that you would ultimately be 
frustrated by the obstacles which the design of the HP-48 presents.
--
Frank Wales, Guest of HP Corvallis, [frank@zen.co.uk||frankw@hpcvdq.cv.hp.com]
Zengrange Ltd., Greenfield Road., LEEDS, England, LS9 8DB.   (+1)-503-750-3086