This article first appeared in The Computer Journal #67 (May/June 1994).
At long last, I am ready to present the complete source code for an (I hope) ANSI compliant Forth, CamelForth. As an intellectual exercise -- and to ensure a clear copyright -- I've written this code entirely from scratch. (Do you know how hard it is to not look at excellent code examples?) Of course, my experience with various Forths has no doubt influenced some design decisions.
Due to space limitations, the source code will be presented in four installments (if you can't wait, complete files will be on GEnie):
For CamelForth I'm trying to use exclusively public-domain tools: for the Z80, the Z80MR assembler under CP/M ; for the 8051, the A51 cross-assembler on an IBM PC , and for the 6809, my own metacompiler under F83 for CP/M, IBM PC, or Atari ST.
By "kernel" I mean the set of words that comprises a basic Forth system, including compiler and interpreter. For CamelForth this is the ANS Forth Core word set, plus any non-ANSI words necessary to implement the Core word set. A Forth kernel is usually written partly in machine code (as CODE words), and partly in high-level Forth. The words which are written in machine code are called the "primitives," since, in the final analysis, the entire Forth system is defined in terms of just these words.
Exactly which words should be written in machine code? The selection of the optimal set of primitives is an interesting debate. A smaller set of primitives makes for easier porting, but poorer performance. I've been told that a set of 13 primitives is sufficient to define all of Forth -- a very slow Forth. eForth , designed for easy porting, had a more generous set of 31 primitives. My rules are these:
For Z80 CamelForth I have a set of about 70 primitives. (See Table 1.) Having already decided on the Forth model and CPU usage (see my previous TCJ articles), I followed this development procedure:
With this set of primitives you can begin writing Forth code. Sure, you have to use an assembler instead of a Forth compiler, but -- as Listing 1 suggests -- you can use high-level control flow and nesting to write useful code that would be more difficult to write in assembler.
I've run out of abstractions for today. If you want to learn more about how a Forth kernel works and is written, study Listing 2. It follows the Forth convention for documentation:
WORD-NAME stack in -- stack out description
WORD-NAME is the name by which Forth knows the word. Often these names include peculiar ASCII characters, so an approximation must be used when defining assembler labels (such as ONEPLUS for the Forth word 1+).
stack in are the arguments this word expects to see on the stack, with the topmost stack item always on the right. stack out are the arguments this word will leave on the stack, likewise.
If the word has a return stack effect (other than nesting, that is), an additional return stack comment will be added after "R:"
stack in -- stack out R: stack in -- stack out
ANSI Forth defines a number of useful abbreviations for stack arguments, such as "n" for a signed single-cell number, "u" for an unsigned single-cell number, "c" for a character, and so on. See Table 1.
 Definition of a camel: a horse designed by committee.
 Ting, C. H., eForth Implementation Guide, July 1990, available from Offete Enterprises, 1306 South B Stret, San Mateo, CA 94402 USA.
 Z80MR, a Z80 Macro Assembler by Mike Rubenstein, is public-domain, available on the GEnie CP/M Roundtable as file Z80MR-A.LBR. Warning: do not use the supplied Z1.COM program, use only Z80MR and LOAD. Z1 has a problem with conditional jumps.
 A51, PseudoCorp's freeware Level 1 cross-assembler for the 8051, is available from the Realtime and Control Forth Board, (303) 278-0364, or on the GEnie Forth Roundtable as file A51.ZIP. PseudoCorp's commercial products are advertised here in TCJ.
Source code for Z80 CamelForth is available on this site at http://www.camelforth.com/public_ftp/cam80-12.zip.
Continue with Part 6 | Back to publications page