1 /* A sometimes minimal FORTH compiler and tutorial for Linux / i386 systems. -*- asm -*-
2 By Richard W.M. Jones <rich@annexia.org> http://annexia.org/forth
3 This is PUBLIC DOMAIN (see public domain release statement below).
4 $Id: jonesforth.S,v 1.30 2007-09-25 09:50:54 rich Exp $
6 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
10 INTRODUCTION ----------------------------------------------------------------------
12 FORTH is one of those alien languages which most working programmers regard in the same
13 way as Haskell, LISP, and so on. Something so strange that they'd rather any thoughts
14 of it just go away so they can get on with writing this paying code. But that's wrong
15 and if you care at all about programming then you should at least understand all these
16 languages, even if you will never use them.
18 LISP is the ultimate high-level language, and features from LISP are being added every
19 decade to the more common languages. But FORTH is in some ways the ultimate in low level
20 programming. Out of the box it lacks features like dynamic memory management and even
21 strings. In fact, at its primitive level it lacks even basic concepts like IF-statements
24 Why then would you want to learn FORTH? There are several very good reasons. First
25 and foremost, FORTH is minimal. You really can write a complete FORTH in, say, 2000
26 lines of code. I don't just mean a FORTH program, I mean a complete FORTH operating
27 system, environment and language. You could boot such a FORTH on a bare PC and it would
28 come up with a prompt where you could start doing useful work. The FORTH you have here
29 isn't minimal and uses a Linux process as its 'base PC' (both for the purposes of making
30 it a good tutorial). It's possible to completely understand the system. Who can say they
31 completely understand how Linux works, or gcc?
33 Secondly FORTH has a peculiar bootstrapping property. By that I mean that after writing
34 a little bit of assembly to talk to the hardware and implement a few primitives, all the
35 rest of the language and compiler is written in FORTH itself. Remember I said before
36 that FORTH lacked IF-statements and loops? Well of course it doesn't really because
37 such a lanuage would be useless, but my point was rather that IF-statements and loops are
38 written in FORTH itself.
40 Now of course this is common in other languages as well, and in those languages we call
41 them 'libraries'. For example in C, 'printf' is a library function written in C. But
42 in FORTH this goes way beyond mere libraries. Can you imagine writing C's 'if' in C?
43 And that brings me to my third reason: If you can write 'if' in FORTH, then why restrict
44 yourself to the usual if/while/for/switch constructs? You want a construct that iterates
45 over every other element in a list of numbers? You can add it to the language. What
46 about an operator which pulls in variables directly from a configuration file and makes
47 them available as FORTH variables? Or how about adding Makefile-like dependencies to
48 the language? No problem in FORTH. This concept isn't common in programming languages,
49 but it has a name (in fact two names): "macros" (by which I mean LISP-style macros, not
50 the lame C preprocessor) and "domain specific languages" (DSLs).
52 This tutorial isn't about learning FORTH as the language. I'll point you to some references
53 you should read if you're not familiar with using FORTH. This tutorial is about how to
54 write FORTH. In fact, until you understand how FORTH is written, you'll have only a very
55 superficial understanding of how to use it.
57 So if you're not familiar with FORTH or want to refresh your memory here are some online
60 http://en.wikipedia.org/wiki/Forth_%28programming_language%29
62 http://galileo.phys.virginia.edu/classes/551.jvn.fall01/primer.htm
64 http://wiki.laptop.org/go/Forth_Lessons
66 http://www.albany.net/~hello/simple.htm
68 Here is another "Why FORTH?" essay: http://www.jwdt.com/~paysan/why-forth.html
70 Discussion and criticism of this FORTH here: http://lambda-the-ultimate.org/node/2452
72 ACKNOWLEDGEMENTS ----------------------------------------------------------------------
74 This code draws heavily on the design of LINA FORTH (http://home.hccnet.nl/a.w.m.van.der.horst/lina.html)
75 by Albert van der Horst. Any similarities in the code are probably not accidental.
77 Also I used this document (http://ftp.funet.fi/pub/doc/IOCCC/1992/buzzard.2.design) which really
78 defies easy explanation.
80 PUBLIC DOMAIN ----------------------------------------------------------------------
82 I, the copyright holder of this work, hereby release it into the public domain. This applies worldwide.
84 In case this is not legally possible, I grant any entity the right to use this work for any purpose,
85 without any conditions, unless such conditions are required by law.
87 SETTING UP ----------------------------------------------------------------------
89 Let's get a few housekeeping things out of the way. Firstly because I need to draw lots of
90 ASCII-art diagrams to explain concepts, the best way to look at this is using a window which
91 uses a fixed width font and is at least this wide:
93 <------------------------------------------------------------------------------------------------------------------------>
95 Secondly make sure TABS are set to 8 characters. The following should be a vertical
96 line. If not, sort out your tabs.
102 Thirdly I assume that your screen is at least 50 characters high.
104 ASSEMBLING ----------------------------------------------------------------------
106 If you want to actually run this FORTH, rather than just read it, you will need Linux on an
107 i386. Linux because instead of programming directly to the hardware on a bare PC which I
108 could have done, I went for a simpler tutorial by assuming that the 'hardware' is a Linux
109 process with a few basic system calls (read, write and exit and that's about all). i386
110 is needed because I had to write the assembly for a processor, and i386 is by far the most
111 common. (Of course when I say 'i386', any 32- or 64-bit x86 processor will do. I'm compiling
112 this on a 64 bit AMD Opteron).
114 Again, to assemble this you will need gcc and gas (the GNU assembler). The commands to
115 assemble and run the code (save this file as 'jonesforth.S') are:
117 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
118 cat jonesforth.f - | ./jonesforth
120 If you want to run your own FORTH programs you can do:
122 cat jonesforth.f myprog.f | ./jonesforth
124 If you want to load your own FORTH code and then continue reading user commands, you can do:
126 cat jonesforth.f myfunctions.f - | ./jonesforth
128 ASSEMBLER ----------------------------------------------------------------------
130 (You can just skip to the next section -- you don't need to be able to read assembler to
131 follow this tutorial).
133 However if you do want to read the assembly code here are a few notes about gas (the GNU assembler):
135 (1) Register names are prefixed with '%', so %eax is the 32 bit i386 accumulator. The registers
136 available on i386 are: %eax, %ebx, %ecx, %edx, %esi, %edi, %ebp and %esp, and most of them
137 have special purposes.
139 (2) Add, mov, etc. take arguments in the form SRC,DEST. So mov %eax,%ecx moves %eax -> %ecx
141 (3) Constants are prefixed with '$', and you mustn't forget it! If you forget it then it
142 causes a read from memory instead, so:
143 mov $2,%eax moves number 2 into %eax
144 mov 2,%eax reads the 32 bit word from address 2 into %eax (ie. most likely a mistake)
146 (4) gas has a funky syntax for local labels, where '1f' (etc.) means label '1:' "forwards"
147 and '1b' (etc.) means label '1:' "backwards".
149 (5) 'ja' is "jump if above", 'jb' for "jump if below", 'je' "jump if equal" etc.
151 (6) gas has a reasonably nice .macro syntax, and I use them a lot to make the code shorter and
154 For more help reading the assembler, do "info gas" at the Linux prompt.
156 Now the tutorial starts in earnest.
158 THE DICTIONARY ----------------------------------------------------------------------
160 In FORTH as you will know, functions are called "words", and just as in other languages they
161 have a name and a definition. Here are two FORTH words:
163 : DOUBLE DUP + ; \ name is "DOUBLE", definition is "DUP +"
164 : QUADRUPLE DOUBLE DOUBLE ; \ name is "QUADRUPLE", definition is "DOUBLE DOUBLE"
166 Words, both built-in ones and ones which the programmer defines later, are stored in a dictionary
167 which is just a linked list of dictionary entries.
169 <--- DICTIONARY ENTRY (HEADER) ----------------------->
170 +------------------------+--------+---------- - - - - +----------- - - - -
171 | LINK POINTER | LENGTH/| NAME | DEFINITION
173 +--- (4 bytes) ----------+- byte -+- n bytes - - - - +----------- - - - -
175 I'll come to the definition of the word later. For now just look at the header. The first
176 4 bytes are the link pointer. This points back to the previous word in the dictionary, or, for
177 the first word in the dictionary it is just a NULL pointer. Then comes a length/flags byte.
178 The length of the word can be up to 31 characters (5 bits used) and the top three bits are used
179 for various flags which I'll come to later. This is followed by the name itself, and in this
180 implementation the name is rounded up to a multiple of 4 bytes by padding it with zero bytes.
181 That's just to ensure that the definition starts on a 32 bit boundary.
183 A FORTH variable called LATEST contains a pointer to the most recently defined word, in
184 other words, the head of this linked list.
186 DOUBLE and QUADRUPLE might look like this:
188 pointer to previous word
191 +--|------+---+---+---+---+---+---+---+---+------------- - - - -
192 | LINK | 6 | D | O | U | B | L | E | 0 | (definition ...)
193 +---------+---+---+---+---+---+---+---+---+------------- - - - -
196 +--|------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
197 | LINK | 9 | Q | U | A | D | R | U | P | L | E | 0 | 0 | (definition ...)
198 +---------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
204 You should be able to see from this how you might implement functions to find a word in
205 the dictionary (just walk along the dictionary entries starting at LATEST and matching
206 the names until you either find a match or hit the NULL pointer at the end of the dictionary);
207 and add a word to the dictionary (create a new definition, set its LINK to LATEST, and set
208 LATEST to point to the new word). We'll see precisely these functions implemented in
209 assembly code later on.
211 One interesting consequence of using a linked list is that you can redefine words, and
212 a newer definition of a word overrides an older one. This is an important concept in
213 FORTH because it means that any word (even "built-in" or "standard" words) can be
214 overridden with a new definition, either to enhance it, to make it faster or even to
215 disable it. However because of the way that FORTH words get compiled, which you'll
216 understand below, words defined using the old definition of a word continue to use
217 the old definition. Only words defined after the new definition use the new definition.
219 DIRECT THREADED CODE ----------------------------------------------------------------------
221 Now we'll get to the really crucial bit in understanding FORTH, so go and get a cup of tea
222 or coffee and settle down. It's fair to say that if you don't understand this section, then you
223 won't "get" how FORTH works, and that would be a failure on my part for not explaining it well.
224 So if after reading this section a few times you don't understand it, please email me
227 Let's talk first about what "threaded code" means. Imagine a peculiar version of C where
228 you are only allowed to call functions without arguments. (Don't worry for now that such a
229 language would be completely useless!) So in our peculiar C, code would look like this:
238 and so on. How would a function, say 'f' above, be compiled by a standard C compiler?
239 Probably into assembly code like this. On the right hand side I've written the actual
243 CALL a E8 08 00 00 00
244 CALL b E8 1C 00 00 00
245 CALL c E8 2C 00 00 00
246 ; ignore the return from the function for now
248 "E8" is the x86 machine code to "CALL" a function. In the first 20 years of computing
249 memory was hideously expensive and we might have worried about the wasted space being used
250 by the repeated "E8" bytes. We can save 20% in code size (and therefore, in expensive memory)
251 by compressing this into just:
253 08 00 00 00 Just the function addresses, without
254 1C 00 00 00 the CALL prefix.
257 On a 16-bit machine like the ones which originally ran FORTH the savings are even greater - 33%.
259 [Historical note: If the execution model that FORTH uses looks strange from the following
260 paragraphs, then it was motivated entirely by the need to save memory on early computers.
261 This code compression isn't so important now when our machines have more memory in their L1
262 caches than those early computers had in total, but the execution model still has some
265 Of course this code won't run directly any more. Instead we need to write an interpreter
266 which takes each pair of bytes and calls it.
268 On an i386 machine it turns out that we can write this interpreter rather easily, in just
269 two assembly instructions which turn into just 3 bytes of machine code. Let's store the
270 pointer to the next word to execute in the %esi register:
272 08 00 00 00 <- We're executing this one now. %esi is the _next_ one to execute.
276 The all-important i386 instruction is called LODSL (or in Intel manuals, LODSW). It does
277 two things. Firstly it reads the memory at %esi into the accumulator (%eax). Secondly it
278 increments %esi by 4 bytes. So after LODSL, the situation now looks like this:
280 08 00 00 00 <- We're still executing this one
281 1C 00 00 00 <- %eax now contains this address (0x0000001C)
284 Now we just need to jump to the address in %eax. This is again just a single x86 instruction
285 written JMP *(%eax). And after doing the jump, the situation looks like:
288 1C 00 00 00 <- Now we're executing this subroutine.
291 To make this work, each subroutine is followed by the two instructions 'LODSL; JMP *(%eax)'
292 which literally make the jump to the next subroutine.
294 And that brings us to our first piece of actual code! Well, it's a macro.
303 /* The macro is called NEXT. That's a FORTH-ism. It expands to those two instructions.
305 Every FORTH primitive that we write has to be ended by NEXT. Think of it kind of like
308 The above describes what is known as direct threaded code.
310 To sum up: We compress our function calls down to a list of addresses and use a somewhat
311 magical macro to act as a "jump to next function in the list". We also use one register (%esi)
312 to act as a kind of instruction pointer, pointing to the next function in the list.
314 I'll just give you a hint of what is to come by saying that a FORTH definition such as:
316 : QUADRUPLE DOUBLE DOUBLE ;
318 actually compiles (almost, not precisely but we'll see why in a moment) to a list of
319 function addresses for DOUBLE, DOUBLE and a special function called EXIT to finish off.
321 At this point, REALLY EAGLE-EYED ASSEMBLY EXPERTS are saying "JONES, YOU'VE MADE A MISTAKE!".
323 I lied about JMP *(%eax).
325 INDIRECT THREADED CODE ----------------------------------------------------------------------
327 It turns out that direct threaded code is interesting but only if you want to just execute
328 a list of functions written in assembly language. So QUADRUPLE would work only if DOUBLE
329 was an assembly language function. In the direct threaded code, QUADRUPLE would look like:
332 | addr of DOUBLE --------------------> (assembly code to do the double)
333 +------------------+ NEXT
334 %esi -> | addr of DOUBLE |
337 We can add an extra indirection to allow us to run both words written in assembly language
338 (primitives written for speed) and words written in FORTH themselves as lists of addresses.
340 The extra indirection is the reason for the brackets in JMP *(%eax).
342 Let's have a look at how QUADRUPLE and DOUBLE really look in FORTH:
344 : QUADRUPLE DOUBLE DOUBLE ;
347 | codeword | : DOUBLE DUP + ;
349 | addr of DOUBLE ---------------> +------------------+
350 +------------------+ | codeword |
351 | addr of DOUBLE | +------------------+
352 +------------------+ | addr of DUP --------------> +------------------+
353 | addr of EXIT | +------------------+ | codeword -------+
354 +------------------+ %esi -> | addr of + --------+ +------------------+ |
355 +------------------+ | | assembly to <-----+
356 | addr of EXIT | | | implement DUP |
357 +------------------+ | | .. |
360 | +------------------+
362 +-----> +------------------+
364 +------------------+ |
365 | assembly to <------+
372 This is the part where you may need an extra cup of tea/coffee/favourite caffeinated
373 beverage. What has changed is that I've added an extra pointer to the beginning of
374 the definitions. In FORTH this is sometimes called the "codeword". The codeword is
375 a pointer to the interpreter to run the function. For primitives written in
376 assembly language, the "interpreter" just points to the actual assembly code itself.
377 They don't need interpreting, they just run.
379 In words written in FORTH (like QUADRUPLE and DOUBLE), the codeword points to an interpreter
382 I'll show you the interpreter function shortly, but let's recall our indirect
383 JMP *(%eax) with the "extra" brackets. Take the case where we're executing DOUBLE
384 as shown, and DUP has been called. Note that %esi is pointing to the address of +
386 The assembly code for DUP eventually does a NEXT. That:
388 (1) reads the address of + into %eax %eax points to the codeword of +
389 (2) increments %esi by 4
390 (3) jumps to the indirect %eax jumps to the address in the codeword of +,
391 ie. the assembly code to implement +
396 | addr of DOUBLE ---------------> +------------------+
397 +------------------+ | codeword |
398 | addr of DOUBLE | +------------------+
399 +------------------+ | addr of DUP --------------> +------------------+
400 | addr of EXIT | +------------------+ | codeword -------+
401 +------------------+ | addr of + --------+ +------------------+ |
402 +------------------+ | | assembly to <-----+
403 %esi -> | addr of EXIT | | | implement DUP |
404 +------------------+ | | .. |
407 | +------------------+
409 +-----> +------------------+
411 +------------------+ |
412 now we're | assembly to <-----+
413 executing | implement + |
419 So I hope that I've convinced you that NEXT does roughly what you'd expect. This is
420 indirect threaded code.
422 I've glossed over four things. I wonder if you can guess without reading on what they are?
428 My list of four things are: (1) What does "EXIT" do? (2) which is related to (1) is how do
429 you call into a function, ie. how does %esi start off pointing at part of QUADRUPLE, but
430 then point at part of DOUBLE. (3) What goes in the codeword for the words which are written
431 in FORTH? (4) How do you compile a function which does anything except call other functions
432 ie. a function which contains a number like : DOUBLE 2 * ; ?
434 THE INTERPRETER AND RETURN STACK ------------------------------------------------------------
436 Going at these in no particular order, let's talk about issues (3) and (2), the interpreter
437 and the return stack.
439 Words which are defined in FORTH need a codeword which points to a little bit of code to
440 give them a "helping hand" in life. They don't need much, but they do need what is known
441 as an "interpreter", although it doesn't really "interpret" in the same way that, say,
442 Java bytecode used to be interpreted (ie. slowly). This interpreter just sets up a few
443 machine registers so that the word can then execute at full speed using the indirect
444 threaded model above.
446 One of the things that needs to happen when QUADRUPLE calls DOUBLE is that we save the old
447 %esi ("instruction pointer") and create a new one pointing to the first word in DOUBLE.
448 Because we will need to restore the old %esi at the end of DOUBLE (this is, after all, like
449 a function call), we will need a stack to store these "return addresses" (old values of %esi).
451 As you will have read, when reading the background documentation, FORTH has two stacks,
452 an ordinary stack for parameters, and a return stack which is a bit more mysterious. But
453 our return stack is just the stack I talked about in the previous paragraph, used to save
454 %esi when calling from a FORTH word into another FORTH word.
456 In this FORTH, we are using the normal stack pointer (%esp) for the parameter stack.
457 We will use the i386's "other" stack pointer (%ebp, usually called the "frame pointer")
458 for our return stack.
460 I've got two macros which just wrap up the details of using %ebp for the return stack.
461 You use them as for example "PUSHRSP %eax" (push %eax on the return stack) or "POPRSP %ebx"
462 (pop top of return stack into %ebx).
465 /* Macros to deal with the return stack. */
467 lea -4(%ebp),%ebp // push reg on to return stack
472 mov (%ebp),\reg // pop top of return stack to reg
477 And with that we can now talk about the interpreter.
479 In FORTH the interpreter function is often called DOCOL (I think it means "DO COLON" because
480 all FORTH definitions start with a colon, as in : DOUBLE DUP + ;
482 The "interpreter" (it's not really "interpreting") just needs to push the old %esi on the
483 stack and set %esi to the first word in the definition. Remember that we jumped to the
484 function using JMP *(%eax)? Well a consequence of that is that conveniently %eax contains
485 the address of this codeword, so just by adding 4 to it we get the address of the first
486 data word. Finally after setting up %esi, it just does NEXT which causes that first word
490 /* DOCOL - the interpreter! */
494 PUSHRSP %esi // push %esi on to the return stack
495 addl $4,%eax // %eax points to codeword, so make
496 movl %eax,%esi // %esi point to first data word
500 Just to make this absolutely clear, let's see how DOCOL works when jumping from QUADRUPLE
506 +------------------+ DOUBLE:
507 | addr of DOUBLE ---------------> +------------------+
508 +------------------+ %eax -> | addr of DOCOL |
509 %esi -> | addr of DOUBLE | +------------------+
510 +------------------+ | addr of DUP |
511 | addr of EXIT | +------------------+
512 +------------------+ | etc. |
514 First, the call to DOUBLE calls DOCOL (the codeword of DOUBLE). DOCOL does this: It
515 pushes the old %esi on the return stack. %eax points to the codeword of DOUBLE, so we
516 just add 4 on to it to get our new %esi:
521 +------------------+ DOUBLE:
522 | addr of DOUBLE ---------------> +------------------+
523 top of return +------------------+ %eax -> | addr of DOCOL |
524 stack points -> | addr of DOUBLE | + 4 = +------------------+
525 +------------------+ %esi -> | addr of DUP |
526 | addr of EXIT | +------------------+
527 +------------------+ | etc. |
529 Then we do NEXT, and because of the magic of threaded code that increments %esi again
532 Well, it seems to work.
534 One minor point here. Because DOCOL is the first bit of assembly actually to be defined
535 in this file (the others were just macros), and because I usually compile this code with the
536 text segment starting at address 0, DOCOL has address 0. So if you are disassembling the
537 code and see a word with a codeword of 0, you will immediately know that the word is
538 written in FORTH (it's not an assembler primitive) and so uses DOCOL as the interpreter.
540 STARTING UP ----------------------------------------------------------------------
542 Now let's get down to nuts and bolts. When we start the program we need to set up
543 a few things like the return stack. But as soon as we can, we want to jump into FORTH
544 code (albeit much of the "early" FORTH code will still need to be written as
545 assembly language primitives).
547 This is what the set up code does. Does a tiny bit of house-keeping, sets up the
548 separate return stack (NB: Linux gives us the ordinary parameter stack already), then
549 immediately jumps to a FORTH word called COLD. COLD stands for cold-start. In ISO
550 FORTH (but not in this FORTH), COLD can be called at any time to completely reset
551 the state of FORTH, and there is another word called WARM which does a partial reset.
554 /* ELF entry point. */
559 mov %esp,var_S0 // Store the initial data stack pointer.
560 mov $return_stack,%ebp // Initialise the return stack.
562 mov $cold_start,%esi // Initialise interpreter.
563 NEXT // Run interpreter!
566 cold_start: // High-level code without a codeword.
570 We also allocate some space for the return stack and some space to store user
571 definitions. These are static memory allocations using fixed-size buffers, but it
572 wouldn't be a great deal of work to make them dynamic.
576 /* FORTH return stack. */
577 .set RETURN_STACK_SIZE,8192
579 .space RETURN_STACK_SIZE
580 return_stack: // Initial top of return stack.
582 /* The user definitions area: space for user-defined words and general memory allocations. */
583 .set USER_DEFS_SIZE,65536
586 .space USER_DEFS_SIZE
588 /* This is used as a temporary input buffer when reading from files or the terminal. */
589 .set BUFFER_SIZE,4096
601 BUILT-IN WORDS ----------------------------------------------------------------------
603 Remember our dictionary entries (headers). Let's bring those together with the codeword
604 and data words to see how : DOUBLE DUP + ; really looks in memory.
606 pointer to previous word
609 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
610 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
611 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
614 LINK in next word points to codeword of DUP
616 Initially we can't just write ": DOUBLE DUP + ;" (ie. that literal string) here because we
617 don't yet have anything to read the string, break it up at spaces, parse each word, etc. etc.
618 So instead we will have to define built-in words using the GNU assembler data constructors
619 (like .int, .byte, .string, .ascii and so on -- look them up in the gas info page if you are
622 The long way would be:
623 .int <link to previous word>
625 .ascii "DOUBLE" // string
627 DOUBLE: .int DOCOL // codeword
628 .int DUP // pointer to codeword of DUP
629 .int PLUS // pointer to codeword of +
630 .int EXIT // pointer to codeword of EXIT
632 That's going to get quite tedious rather quickly, so here I define an assembler macro
633 so that I can just write:
635 defword "DOUBLE",6,,DOUBLE
638 and I'll get exactly the same effect.
640 Don't worry too much about the exact implementation details of this macro - it's complicated!
643 /* Flags - these are discussed later. */
646 .set F_LENMASK,0x1f // length mask
648 // Store the chain of links.
651 .macro defword name, namelen, flags=0, label
657 .set link,name_\label
658 .byte \flags+\namelen // flags + length byte
659 .ascii "\name" // the name
663 .int DOCOL // codeword - the interpreter
664 // list of word pointers follow
668 Similarly I want a way to write words written in assembly language. There will quite a few
669 of these to start with because, well, everything has to start in assembly before there's
670 enough "infrastructure" to be able to start writing FORTH words, but also I want to define
671 some common FORTH words in assembly language for speed, even though I could write them in FORTH.
673 This is what DUP looks like in memory:
675 pointer to previous word
678 +--|------+---+---+---+---+------------+
679 | LINK | 3 | D | U | P | code_DUP ---------------------> points to the assembly
680 +---------+---+---+---+---+------------+ code used to write DUP,
681 ^ len codeword which ends with NEXT.
685 Again, for brevity in writing the header I'm going to write an assembler macro called defcode.
688 .macro defcode name, namelen, flags=0, label
694 .set link,name_\label
695 .byte \flags+\namelen // flags + length byte
696 .ascii "\name" // the name
700 .int code_\label // codeword
704 code_\label : // assembler code follows
708 Now some easy FORTH primitives. These are written in assembly for speed. If you understand
709 i386 assembly language then it is worth reading these. However if you don't understand assembly
710 you can skip the details.
714 pop %eax // duplicate top of stack
719 defcode "DROP",4,,DROP
720 pop %eax // drop top of stack
723 defcode "SWAP",4,,SWAP
724 pop %eax // swap top of stack
730 defcode "OVER",4,,OVER
731 mov 4(%esp),%eax // get the second element of stack
732 push %eax // and push it on top
744 defcode "-ROT",4,,NROT
754 incl (%esp) // increment top of stack
758 decl (%esp) // decrement top of stack
761 defcode "4+",2,,INCR4
762 addl $4,(%esp) // add 4 to top of stack
765 defcode "4-",2,,DECR4
766 subl $4,(%esp) // subtract 4 from top of stack
770 pop %eax // get top of stack
771 addl %eax,(%esp) // and add it to next word on stack
775 pop %eax // get top of stack
776 subl %eax,(%esp) // and subtract it from next word on stack
783 push %eax // ignore overflow
791 push %eax // push quotient
799 push %edx // push remainder
802 defcode "=",1,,EQU // top two words are equal?
812 defcode "<>",2,,NEQU // top two words are not equal?
862 defcode "0=",2,,ZEQU // top of stack equals 0?
871 defcode "0<>",3,,ZNEQU // top of stack not 0?
931 defcode "INVERT",6,,INVERT // this is the FORTH bitwise "NOT" function
936 RETURNING FROM FORTH WORDS ----------------------------------------------------------------------
938 Time to talk about what happens when we EXIT a function. In this diagram QUADRUPLE has called
939 DOUBLE, and DOUBLE is about to exit (look at where %esi is pointing):
944 +------------------+ DOUBLE
945 | addr of DOUBLE ---------------> +------------------+
946 +------------------+ | codeword |
947 | addr of DOUBLE | +------------------+
948 +------------------+ | addr of DUP |
949 | addr of EXIT | +------------------+
950 +------------------+ | addr of + |
952 %esi -> | addr of EXIT |
955 What happens when the + function does NEXT? Well, the following code is executed.
958 defcode "EXIT",4,,EXIT
959 POPRSP %esi // pop return stack into %esi
963 EXIT gets the old %esi which we saved from before on the return stack, and puts it in %esi.
964 So after this (but just before NEXT) we get:
969 +------------------+ DOUBLE
970 | addr of DOUBLE ---------------> +------------------+
971 +------------------+ | codeword |
972 %esi -> | addr of DOUBLE | +------------------+
973 +------------------+ | addr of DUP |
974 | addr of EXIT | +------------------+
975 +------------------+ | addr of + |
980 And NEXT just completes the job by, well in this case just by calling DOUBLE again :-)
982 LITERALS ----------------------------------------------------------------------
984 The final point I "glossed over" before was how to deal with functions that do anything
985 apart from calling other functions. For example, suppose that DOUBLE was defined like this:
989 It does the same thing, but how do we compile it since it contains the literal 2? One way
990 would be to have a function called "2" (which you'd have to write in assembler), but you'd need
991 a function for every single literal that you wanted to use.
993 FORTH solves this by compiling the function using a special word called LIT:
995 +---------------------------+-------+-------+-------+-------+-------+
996 | (usual header of DOUBLE) | DOCOL | LIT | 2 | * | EXIT |
997 +---------------------------+-------+-------+-------+-------+-------+
999 LIT is executed in the normal way, but what it does next is definitely not normal. It
1000 looks at %esi (which now points to the literal 2), grabs it, pushes it on the stack, then
1001 manipulates %esi in order to skip the literal as if it had never been there.
1003 What's neat is that the whole grab/manipulate can be done using a single byte single
1004 i386 instruction, our old friend LODSL. Rather than me drawing more ASCII-art diagrams,
1005 see if you can find out how LIT works:
1008 defcode "LIT",3,,LIT
1009 // %esi points to the next command, but in this case it points to the next
1010 // literal 32 bit integer. Get that literal into %eax and increment %esi.
1011 // On x86, it's a convenient single byte instruction! (cf. NEXT macro)
1013 push %eax // push the literal number on to stack
1017 MEMORY ----------------------------------------------------------------------
1019 As important point about FORTH is that it gives you direct access to the lowest levels
1020 of the machine. Manipulating memory directly is done frequently in FORTH, and these are
1021 the primitive words for doing it.
1024 defcode "!",1,,STORE
1025 pop %ebx // address to store at
1026 pop %eax // data to store there
1027 mov %eax,(%ebx) // store it
1030 defcode "@",1,,FETCH
1031 pop %ebx // address to fetch
1032 mov (%ebx),%eax // fetch it
1033 push %eax // push value onto stack
1036 defcode "+!",2,,ADDSTORE
1038 pop %eax // the amount to add
1039 addl %eax,(%ebx) // add it
1042 defcode "-!",2,,SUBSTORE
1044 pop %eax // the amount to subtract
1045 subl %eax,(%ebx) // add it
1048 /* ! and @ (STORE and FETCH) store 32-bit words. It's also useful to be able to read and write bytes.
1049 * I don't know whether FORTH has these words, so I invented my own, called !b and @b.
1050 * Byte-oriented operations only work on architectures which permit them (i386 is one of those).
1051 * UPDATE: writing a byte to the dictionary pointer is called C, in FORTH.
1053 defcode "!b",2,,STOREBYTE
1054 pop %ebx // address to store at
1055 pop %eax // data to store there
1056 movb %al,(%ebx) // store it
1059 defcode "@b",2,,FETCHBYTE
1060 pop %ebx // address to fetch
1062 movb (%ebx),%al // fetch it
1063 push %eax // push value onto stack
1067 BUILT-IN VARIABLES ----------------------------------------------------------------------
1069 These are some built-in variables and related standard FORTH words. Of these, the only one that we
1070 have discussed so far was LATEST, which points to the last (most recently defined) word in the
1071 FORTH dictionary. LATEST is also a FORTH word which pushes the address of LATEST (the variable)
1072 on to the stack, so you can read or write it using @ and ! operators. For example, to print
1073 the current value of LATEST (and this can apply to any FORTH variable) you would do:
1077 To make defining variables shorter, I'm using a macro called defvar, similar to defword and
1078 defcode above. (In fact the defvar macro uses defcode to do the dictionary header).
1081 .macro defvar name, namelen, flags=0, label, initial=0
1082 defcode \name,\namelen,\flags,\label
1092 The built-in variables are:
1094 STATE Is the interpreter executing code (0) or compiling a word (non-zero)?
1095 LATEST Points to the latest (most recently defined) word in the dictionary.
1096 HERE Points to the next free byte of memory. When compiling, compiled words go here.
1097 _X These are three scratch variables, used by some standard dictionary words.
1100 S0 Stores the address of the top of the parameter stack.
1101 BASE The current base for printing and reading numbers.
1104 defvar "STATE",5,,STATE
1105 defvar "HERE",4,,HERE,user_defs_start
1106 defvar "LATEST",6,,LATEST,name_SYSEXIT // SYSEXIT must be last in built-in dictionary
1111 defvar "BASE",4,,BASE,10
1114 BUILT-IN CONSTANTS ----------------------------------------------------------------------
1116 It's also useful to expose a few constants to FORTH. When the word is executed it pushes a
1117 constant value on the stack.
1119 The built-in constants are:
1121 VERSION Is the current version of this FORTH.
1122 R0 The address of the top of the return stack.
1123 DOCOL Pointer to DOCOL.
1124 F_IMMED The IMMEDIATE flag's actual value.
1125 F_HIDDEN The HIDDEN flag's actual value.
1126 F_LENMASK The length mask.
1129 .macro defconst name, namelen, flags=0, label, value
1130 defcode \name,\namelen,\flags,\label
1135 defconst "VERSION",7,,VERSION,JONES_VERSION
1136 defconst "R0",2,,RZ,return_stack
1137 defconst "DOCOL",5,,__DOCOL,DOCOL
1138 defconst "F_IMMED",7,,__F_IMMED,F_IMMED
1139 defconst "F_HIDDEN",8,,__F_HIDDEN,F_HIDDEN
1140 defconst "F_LENMASK",9,,__F_LENMASK,F_LENMASK
1143 RETURN STACK ----------------------------------------------------------------------
1145 These words allow you to access the return stack. Recall that the register %ebp always points to
1146 the top of the return stack.
1150 pop %eax // pop parameter stack into %eax
1151 PUSHRSP %eax // push it on to the return stack
1154 defcode "R>",2,,FROMR
1155 POPRSP %eax // pop return stack on to %eax
1156 push %eax // and push on to parameter stack
1159 defcode "RSP@",4,,RSPFETCH
1163 defcode "RSP!",4,,RSPSTORE
1167 defcode "RDROP",5,,RDROP
1168 lea 4(%ebp),%ebp // pop return stack and throw away
1172 PARAMETER (DATA) STACK ----------------------------------------------------------------------
1174 These functions allow you to manipulate the parameter stack. Recall that Linux sets up the parameter
1175 stack for us, and it is accessed through %esp.
1178 defcode "DSP@",4,,DSPFETCH
1183 defcode "DSP!",4,,DSPSTORE
1188 INPUT AND OUTPUT ----------------------------------------------------------------------
1190 These are our first really meaty/complicated FORTH primitives. I have chosen to write them in
1191 assembler, but surprisingly in "real" FORTH implementations these are often written in terms
1192 of more fundamental FORTH primitives. I chose to avoid that because I think that just obscures
1193 the implementation. After all, you may not understand assembler but you can just think of it
1194 as an opaque block of code that does what it says.
1196 Let's discuss input first.
1198 The FORTH word KEY reads the next byte from stdin (and pushes it on the parameter stack).
1199 So if KEY is called and someone hits the space key, then the number 32 (ASCII code of space)
1200 is pushed on the stack.
1202 In FORTH there is no distinction between reading code and reading input. We might be reading
1203 and compiling code, we might be reading words to execute, we might be asking for the user
1204 to type their name -- ultimately it all comes in through KEY.
1206 The implementation of KEY uses an input buffer of a certain size (defined at the end of the
1207 program). It calls the Linux read(2) system call to fill this buffer and tracks its position
1208 in the buffer using a couple of variables, and if it runs out of input buffer then it refills
1209 it automatically. The other thing that KEY does is if it detects that stdin has closed, it
1210 exits the program, which is why when you hit ^D the FORTH system cleanly exits.
1213 #include <asm-i386/unistd.h>
1215 defcode "KEY",3,,KEY
1217 push %eax // push return value on stack
1229 1: // out of input; use read(2) to fetch more input from stdin
1230 xor %ebx,%ebx // 1st param: stdin
1231 mov $buffer,%ecx // 2nd param: buffer
1233 mov $buffend-buffer,%edx // 3rd param: max length
1234 mov $__NR_read,%eax // syscall: read
1236 test %eax,%eax // If %eax <= 0, then exit.
1238 addl %eax,%ecx // buffer+%eax = bufftop
1242 2: // error or out of input: exit
1244 mov $__NR_exit,%eax // syscall: exit
1248 By contrast, output is much simpler. The FORTH word EMIT writes out a single byte to stdout.
1249 This implementation just uses the write system call. No attempt is made to buffer output, but
1250 it would be a good exercise to add it.
1253 defcode "EMIT",4,,EMIT
1258 mov $1,%ebx // 1st param: stdout
1260 // write needs the address of the byte to write
1262 mov $2f,%ecx // 2nd param: address
1264 mov $1,%edx // 3rd param: nbytes = 1
1266 mov $__NR_write,%eax // write syscall
1271 2: .space 1 // scratch used by EMIT
1274 Back to input, WORD is a FORTH word which reads the next full word of input.
1276 What it does in detail is that it first skips any blanks (spaces, tabs, newlines and so on).
1277 Then it calls KEY to read characters into an internal buffer until it hits a blank. Then it
1278 calculates the length of the word it read and returns the address and the length as
1279 two words on the stack (with address at the top).
1281 Notice that WORD has a single internal buffer which it overwrites each time (rather like
1282 a static C string). Also notice that WORD's internal buffer is just 32 bytes long and
1283 there is NO checking for overflow. 31 bytes happens to be the maximum length of a
1284 FORTH word that we support, and that is what WORD is used for: to read FORTH words when
1285 we are compiling and executing code. The returned strings are not NUL-terminated, so
1286 in some crazy-world you could define FORTH words containing ASCII NULs, although why
1287 you'd want to is a bit beyond me.
1289 WORD is not suitable for just reading strings (eg. user input) because of all the above
1290 peculiarities and limitations.
1292 Note that when executing, you'll see:
1294 which puts "FOO" and length 3 on the stack, but when compiling:
1296 is an error (or at least it doesn't do what you might expect). Later we'll talk about compiling
1297 and immediate mode, and you'll understand why.
1300 defcode "WORD",4,,WORD
1302 push %ecx // push length
1303 push %edi // push base address
1307 /* Search for first non-blank character. Also skip \ comments. */
1309 call _KEY // get next key, returned in %eax
1310 cmpb $'\\',%al // start of a comment?
1311 je 3f // if so, skip the comment
1313 jbe 1b // if so, keep looking
1315 /* Search for the end of the word, storing chars as we go. */
1316 mov $5f,%edi // pointer to return buffer
1318 stosb // add character to return buffer
1319 call _KEY // get next key, returned in %al
1320 cmpb $' ',%al // is blank?
1321 ja 2b // if not, keep looping
1323 /* Return the word (well, the static buffer) and length. */
1325 mov %edi,%ecx // return length of the word
1326 mov $5f,%edi // return address of the word
1329 /* Code to skip \ comments to end of the current line. */
1332 cmpb $'\n',%al // end of line yet?
1337 // A static buffer where WORD returns. Subsequent calls
1338 // overwrite this buffer. Maximum word length is 32 chars.
1342 . (also called DOT) prints the top of the stack as an integer in the current BASE.
1346 pop %eax // Get the number to print into %eax
1347 call _DOT // Easier to do this recursively ...
1350 mov var_BASE,%ecx // Get current BASE
1352 cmp %ecx,%eax // %eax < BASE? If so jump to print immediately.
1354 xor %edx,%edx // %edx:%eax / %ecx -> quotient %eax, remainder %edx
1356 pushl %edx // Print quotient (top half) first ...
1358 popl %eax // ... then loop to print remainder
1360 2: // %eax < BASE so print immediately.
1363 movb (%edx),%al // Note top bits are already zero.
1367 digits: .ascii "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"
1370 Almost the opposite of DOT (but not quite), SNUMBER parses a numeric string such as one returned
1371 by WORD and pushes the number on the parameter stack.
1373 This function does absolutely no error checking, and in particular the length of the string
1374 must be >= 1 bytes, and should contain only digits 0-9. If it doesn't you'll get random results.
1376 This function is only used when reading literal numbers in code, and shouldn't really be used
1377 in user code at all.
1379 defcode "SNUMBER",7,,SNUMBER
1389 imull $10,%eax // %eax *= 10
1392 subb $'0',%bl // ASCII -> digit
1399 DICTIONARY LOOK UPS ----------------------------------------------------------------------
1401 We're building up to our prelude on how FORTH code is compiled, but first we need yet more infrastructure.
1403 The FORTH word FIND takes a string (a word as parsed by WORD -- see above) and looks it up in the
1404 dictionary. What it actually returns is the address of the dictionary header, if it finds it,
1407 So if DOUBLE is defined in the dictionary, then WORD DOUBLE FIND returns the following pointer:
1413 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1414 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1415 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1417 See also >CFA and >DFA.
1419 FIND doesn't find dictionary entries which are flagged as HIDDEN. See below for why.
1422 defcode "FIND",4,,FIND
1423 pop %edi // %edi = address
1424 pop %ecx // %ecx = length
1430 push %esi // Save %esi so we can use it in string comparison.
1432 // Now we start searching backwards through the dictionary for this word.
1433 mov var_LATEST,%edx // LATEST points to name header of the latest word in the dictionary
1435 test %edx,%edx // NULL pointer? (end of the linked list)
1438 // Compare the length expected and the length of the word.
1439 // Note that if the F_HIDDEN flag is set on the word, then by a bit of trickery
1440 // this won't pick the word (the length will appear to be wrong).
1442 movb 4(%edx),%al // %al = flags+length field
1443 andb $(F_HIDDEN|F_LENMASK),%al // %al = name length
1444 cmpb %cl,%al // Length is the same?
1447 // Compare the strings in detail.
1448 push %ecx // Save the length
1449 push %edi // Save the address (repe cmpsb will move this pointer)
1450 lea 5(%edx),%esi // Dictionary string we are checking against.
1451 repe cmpsb // Compare the strings.
1454 jne 2f // Not the same.
1456 // The strings are the same - return the header pointer in %eax
1462 mov (%edx),%edx // Move back through the link field to the previous word
1463 jmp 1b // .. and loop.
1467 xor %eax,%eax // Return zero to indicate not found.
1471 FIND returns the dictionary pointer, but when compiling we need the codeword pointer (recall
1472 that FORTH definitions are compiled into lists of codeword pointers). The standard FORTH
1473 word >CFA turns a dictionary pointer into a codeword pointer.
1475 The example below shows the result of:
1477 WORD DOUBLE FIND >CFA
1479 FIND returns a pointer to this
1480 | >CFA converts it to a pointer to this
1483 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1484 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1485 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1489 Because names vary in length, this isn't just a simple increment.
1491 In this FORTH you cannot easily turn a codeword pointer back into a dictionary entry pointer, but
1492 that is not true in most FORTH implementations where they store a back pointer in the definition
1493 (with an obvious memory/complexity cost). The reason they do this is that it is useful to be
1494 able to go backwards (codeword -> dictionary entry) in order to decompile FORTH definitions.
1496 What does CFA stand for? My best guess is "Code Field Address".
1499 defcode ">CFA",4,,TCFA
1506 add $4,%edi // Skip link pointer.
1507 movb (%edi),%al // Load flags+len into %al.
1508 inc %edi // Skip flags+len byte.
1509 andb $F_LENMASK,%al // Just the length, not the flags.
1510 add %eax,%edi // Skip the name.
1511 addl $3,%edi // The codeword is 4-byte aligned.
1516 Related to >CFA is >DFA which takes a dictionary entry address as returned by FIND and
1517 returns a pointer to the first data field.
1519 FIND returns a pointer to this
1520 | >CFA converts it to a pointer to this
1522 | | >DFA converts it to a pointer to this
1525 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1526 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1527 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1529 (Note to those following the source of FIG-FORTH / ciforth: My >DFA definition is
1530 different from theirs, because they have an extra indirection).
1532 You can see that >DFA is easily defined in FORTH just by adding 4 to the result of >CFA.
1535 defword ">DFA",4,,TDFA
1536 .int TCFA // >CFA (get code field address)
1537 .int INCR4 // 4+ (add 4 to it to get to next word)
1538 .int EXIT // EXIT (return from FORTH word)
1541 COMPILING ----------------------------------------------------------------------
1543 Now we'll talk about how FORTH compiles words. Recall that a word definition looks like this:
1547 and we have to turn this into:
1549 pointer to previous word
1552 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1553 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1554 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
1555 ^ len pad codeword |
1557 LATEST points here points to codeword of DUP
1559 There are several problems to solve. Where to put the new word? How do we read words? How
1560 do we define the words : (COLON) and ; (SEMICOLON)?
1562 FORTH solves this rather elegantly and as you might expect in a very low-level way which
1563 allows you to change how the compiler works on your own code.
1565 FORTH has an INTERPRETER function (a true interpreter this time, not DOCOL) which runs in a
1566 loop, reading words (using WORD), looking them up (using FIND), turning them into codeword
1567 pointers (using >CFA) and deciding what to do with them.
1569 What it does depends on the mode of the interpreter (in variable STATE).
1571 When STATE is zero, the interpreter just runs each word as it looks them up. This is known as
1574 The interesting stuff happens when STATE is non-zero -- compiling mode. In this mode the
1575 interpreter appends the codeword pointer to user memory (the HERE variable points to the next
1576 free byte of user memory).
1578 So you may be able to see how we could define : (COLON). The general plan is:
1580 (1) Use WORD to read the name of the function being defined.
1582 (2) Construct the dictionary entry -- just the header part -- in user memory:
1584 pointer to previous word (from LATEST) +-- Afterwards, HERE points here, where
1585 ^ | the interpreter will start appending
1587 +--|------+---+---+---+---+---+---+---+---+------------+
1588 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL |
1589 +---------+---+---+---+---+---+---+---+---+------------+
1592 (3) Set LATEST to point to the newly defined word, ...
1594 (4) .. and most importantly leave HERE pointing just after the new codeword. This is where
1595 the interpreter will append codewords.
1597 (5) Set STATE to 1. This goes into compile mode so the interpreter starts appending codewords to
1598 our partially-formed header.
1600 After : has run, our input is here:
1605 Next byte returned by KEY will be the 'D' character of DUP
1607 so the interpreter (now it's in compile mode, so I guess it's really the compiler) reads "DUP",
1608 looks it up in the dictionary, gets its codeword pointer, and appends it:
1610 +-- HERE updated to point here.
1613 +---------+---+---+---+---+---+---+---+---+------------+------------+
1614 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP |
1615 +---------+---+---+---+---+---+---+---+---+------------+------------+
1618 Next we read +, get the codeword pointer, and append it:
1620 +-- HERE updated to point here.
1623 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1624 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + |
1625 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1628 The issue is what happens next. Obviously what we _don't_ want to happen is that we
1629 read ";" and compile it and go on compiling everything afterwards.
1631 At this point, FORTH uses a trick. Remember the length byte in the dictionary definition
1632 isn't just a plain length byte, but can also contain flags. One flag is called the
1633 IMMEDIATE flag (F_IMMED in this code). If a word in the dictionary is flagged as
1634 IMMEDIATE then the interpreter runs it immediately _even if it's in compile mode_.
1636 This is how the word ; (SEMICOLON) works -- as a word flagged in the dictionary as IMMEDIATE.
1637 And all it does is append the codeword for EXIT on to the current definition and switch
1638 back to immediate mode (set STATE back to 0). Shortly we'll see the actual definition
1639 of ; and we'll see that it's really a very simple definition, declared IMMEDIATE.
1641 After the interpreter reads ; and executes it 'immediately', we get this:
1643 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1644 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1645 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1652 And that's it, job done, our new definition is compiled, and we're back in immediate mode
1653 just reading and executing words, perhaps including a call to test our new word DOUBLE.
1655 The only last wrinkle in this is that while our word was being compiled, it was in a
1656 half-finished state. We certainly wouldn't want DOUBLE to be called somehow during
1657 this time. There are several ways to stop this from happening, but in FORTH what we
1658 do is flag the word with the HIDDEN flag (F_HIDDEN in this code) just while it is
1659 being compiled. This prevents FIND from finding it, and thus in theory stops any
1660 chance of it being called.
1662 The above explains how compiling, : (COLON) and ; (SEMICOLON) works and in a moment I'm
1663 going to define them. The : (COLON) function can be made a little bit more general by writing
1664 it in two parts. The first part, called CREATE, makes just the header:
1666 +-- Afterwards, HERE points here.
1669 +---------+---+---+---+---+---+---+---+---+
1670 | LINK | 6 | D | O | U | B | L | E | 0 |
1671 +---------+---+---+---+---+---+---+---+---+
1674 and the second part, the actual definition of : (COLON), calls CREATE and appends the
1675 DOCOL codeword, so leaving:
1677 +-- Afterwards, HERE points here.
1680 +---------+---+---+---+---+---+---+---+---+------------+
1681 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL |
1682 +---------+---+---+---+---+---+---+---+---+------------+
1685 CREATE is a standard FORTH word and the advantage of this split is that we can reuse it to
1686 create other types of words (not just ones which contain code, but words which contain variables,
1687 constants and other data).
1690 defcode "CREATE",6,,CREATE
1693 call _WORD // Returns %ecx = length, %edi = pointer to word.
1694 mov %edi,%ebx // %ebx = address of the word
1697 movl var_HERE,%edi // %edi is the address of the header
1698 movl var_LATEST,%eax // Get link pointer
1699 stosl // and store it in the header.
1701 // Length byte and the word itself.
1702 mov %cl,%al // Get the length.
1703 stosb // Store the length/flags byte.
1705 mov %ebx,%esi // %esi = word
1706 rep movsb // Copy the word
1708 addl $3,%edi // Align to next 4 byte boundary.
1711 // Update LATEST and HERE.
1713 movl %eax,var_LATEST
1718 Because I want to define : (COLON) in FORTH, not assembler, we need a few more FORTH words
1721 The first is , (COMMA) which is a standard FORTH word which appends a 32 bit integer to the user
1722 data area pointed to by HERE, and adds 4 to HERE. So the action of , (COMMA) is:
1724 previous value of HERE
1727 +---------+---+---+---+---+---+---+---+---+-- - - - - --+------------+
1728 | LINK | 6 | D | O | U | B | L | E | 0 | | <data> |
1729 +---------+---+---+---+---+---+---+---+---+-- - - - - --+------------+
1734 and <data> is whatever 32 bit integer was at the top of the stack.
1736 , (COMMA) is quite a fundamental operation when compiling. It is used to append codewords
1737 to the current word that is being compiled.
1740 defcode ",",1,,COMMA
1741 pop %eax // Code pointer to store.
1745 movl var_HERE,%edi // HERE
1747 movl %edi,var_HERE // Update HERE (incremented)
1751 Our definitions of : (COLON) and ; (SEMICOLON) will need to switch to and from compile mode.
1753 Immediate mode vs. compile mode is stored in the global variable STATE, and by updating this
1754 variable we can switch between the two modes.
1756 For various reasons which may become apparent later, FORTH defines two standard words called
1757 [ and ] (LBRAC and RBRAC) which switch between modes:
1759 Word Assembler Action Effect
1760 [ LBRAC STATE := 0 Switch to immediate mode.
1761 ] RBRAC STATE := 1 Switch to compile mode.
1763 [ (LBRAC) is an IMMEDIATE word. The reason is as follows: If we are in compile mode and the
1764 interpreter saw [ then it would compile it rather than running it. We would never be able to
1765 switch back to immediate mode! So we flag the word as IMMEDIATE so that even in compile mode
1766 the word runs immediately, switching us back to immediate mode.
1769 defcode "[",1,F_IMMED,LBRAC
1771 movl %eax,var_STATE // Set STATE to 0.
1774 defcode "]",1,,RBRAC
1775 movl $1,var_STATE // Set STATE to 1.
1779 Now we can define : (COLON) using CREATE. It just calls CREATE, appends DOCOL (the codeword), sets
1780 the word HIDDEN and goes into compile mode.
1783 defword ":",1,,COLON
1784 .int CREATE // CREATE the dictionary entry / header
1785 .int LIT, DOCOL, COMMA // Append DOCOL (the codeword).
1786 .int LATEST, FETCH, HIDDEN // Make the word hidden (see below for definition).
1787 .int RBRAC // Go into compile mode.
1788 .int EXIT // Return from the function.
1791 ; (SEMICOLON) is also elegantly simple. Notice the F_IMMED flag.
1794 defword ";",1,F_IMMED,SEMICOLON
1795 .int LIT, EXIT, COMMA // Append EXIT (so the word will return).
1796 .int LATEST, FETCH, HIDDEN // Toggle hidden flag -- unhide the word (see below for definition).
1797 .int LBRAC // Go back to IMMEDIATE mode.
1798 .int EXIT // Return from the function.
1801 EXTENDING THE COMPILER ----------------------------------------------------------------------
1803 Words flagged with IMMEDIATE (F_IMMED) aren't just for the FORTH compiler to use. You can define
1804 your own IMMEDIATE words too, and this is a crucial aspect when extending basic FORTH, because
1805 it allows you in effect to extend the compiler itself. Does gcc let you do that?
1807 Standard FORTH words like IF, WHILE, ." and so on are all written as extensions to the basic
1808 compiler, and are all IMMEDIATE words.
1810 The IMMEDIATE word toggles the F_IMMED (IMMEDIATE flag) on the most recently defined word,
1811 or on the current word if you call it in the middle of a definition.
1815 : MYIMMEDWORD IMMEDIATE
1819 but some FORTH programmers write this instead:
1825 The two usages are equivalent, to a first approximation.
1828 defcode "IMMEDIATE",9,F_IMMED,IMMEDIATE
1829 movl var_LATEST,%edi // LATEST word.
1830 addl $4,%edi // Point to name/flags byte.
1831 xorb $F_IMMED,(%edi) // Toggle the IMMED bit.
1835 'addr HIDDEN' toggles the hidden flag (F_HIDDEN) of the word defined at addr. To hide the
1836 most recently defined word (used above in : and ; definitions) you would do:
1840 Setting this flag stops the word from being found by FIND, and so can be used to make 'private'
1841 words. For example, to break up a large word into smaller parts you might do:
1843 : SUB1 ... subword ... ;
1844 : SUB2 ... subword ... ;
1845 : SUB3 ... subword ... ;
1846 : MAIN ... defined in terms of SUB1, SUB2, SUB3 ... ;
1847 WORD SUB1 FIND HIDDEN \ Hide SUB1
1848 WORD SUB2 FIND HIDDEN \ Hide SUB2
1849 WORD SUB3 FIND HIDDEN \ Hide SUB3
1851 After this, only MAIN is 'exported' or seen by the rest of the program.
1854 defcode "HIDDEN",6,,HIDDEN
1855 pop %edi // Dictionary entry.
1856 addl $4,%edi // Point to name/flags byte.
1857 xorb $F_HIDDEN,(%edi) // Toggle the HIDDEN bit.
1861 ' (TICK) is a standard FORTH word which returns the codeword pointer of the next word.
1863 The common usage is:
1867 which appends the codeword of FOO to the current word we are defining (this only works in compiled code).
1869 You tend to use ' in IMMEDIATE words. For example an alternate (and rather useless) way to define
1870 a literal 2 might be:
1873 ' LIT , \ Appends LIT to the currently-being-defined word
1874 2 , \ Appends the number 2 to the currently-being-defined word
1881 (If you don't understand how LIT2 works, then you should review the material about compiling words
1882 and immediate mode).
1884 This definition of ' uses a cheat which I copied from buzzard92. As a result it only works in
1885 compiled code. It is possible to write a version of ' based on WORD, FIND, >CFA which works in
1889 lodsl // Get the address of the next word and skip it.
1890 pushl %eax // Push it on the stack.
1894 BRANCHING ----------------------------------------------------------------------
1896 It turns out that all you need in order to define looping constructs, IF-statements, etc.
1899 BRANCH is an unconditional branch. 0BRANCH is a conditional branch (it only branches if the
1900 top of stack is zero).
1902 The diagram below shows how BRANCH works in some imaginary compiled word. When BRANCH executes,
1903 %esi starts by pointing to the offset field (compare to LIT above):
1905 +---------------------+-------+---- - - ---+------------+------------+---- - - - ----+------------+
1906 | (Dictionary header) | DOCOL | | BRANCH | offset | (skipped) | word |
1907 +---------------------+-------+---- - - ---+------------+-----|------+---- - - - ----+------------+
1910 | +-----------------------+
1911 %esi added to offset
1913 The offset is added to %esi to make the new %esi, and the result is that when NEXT runs, execution
1914 continues at the branch target. Negative offsets work as expected.
1916 0BRANCH is the same except the branch happens conditionally.
1918 Now standard FORTH words such as IF, THEN, ELSE, WHILE, REPEAT, etc. can be implemented entirely
1919 in FORTH. They are IMMEDIATE words which append various combinations of BRANCH or 0BRANCH
1920 into the word currently being compiled.
1922 As an example, code written like this:
1924 condition-code IF true-part THEN rest-code
1928 condition-code 0BRANCH OFFSET true-part rest-code
1934 defcode "BRANCH",6,,BRANCH
1935 add (%esi),%esi // add the offset to the instruction pointer
1938 defcode "0BRANCH",7,,ZBRANCH
1940 test %eax,%eax // top of stack is zero?
1941 jz code_BRANCH // if so, jump back to the branch function above
1942 lodsl // otherwise we need to skip the offset
1946 PRINTING STRINGS ----------------------------------------------------------------------
1948 LITSTRING and EMITSTRING are primitives used to implement the ." and S" operators
1949 (which are written in FORTH). See the definition of those operators below.
1952 defcode "LITSTRING",9,,LITSTRING
1953 lodsl // get the length of the string
1954 push %eax // push it on the stack
1955 push %esi // push the address of the start of the string
1956 addl %eax,%esi // skip past the string
1957 addl $3,%esi // but round up to next 4 byte boundary
1961 defcode "EMITSTRING",10,,EMITSTRING
1962 mov $1,%ebx // 1st param: stdout
1963 pop %ecx // 2nd param: address of string
1964 pop %edx // 3rd param: length of string
1965 mov $__NR_write,%eax // write syscall
1970 COLD START AND INTERPRETER ----------------------------------------------------------------------
1972 COLD is the first FORTH function called, almost immediately after the FORTH system "boots".
1974 INTERPRETER is the FORTH interpreter ("toploop", "toplevel" or "REPL" might be a more accurate
1975 description -- see: http://en.wikipedia.org/wiki/REPL).
1978 // COLD must not return (ie. must not call EXIT).
1979 defword "COLD",4,,COLD
1980 .int INTERPRETER // call the interpreter loop (never returns)
1981 .int LIT,1,SYSEXIT // hmmm, but in case it does, exit(1).
1983 /* This interpreter is pretty simple, but remember that in FORTH you can always override
1984 * it later with a more powerful one!
1986 defword "INTERPRETER",11,,INTERPRETER
1987 .int INTERPRET,RDROP,INTERPRETER
1989 defcode "INTERPRET",9,,INTERPRET
1990 call _WORD // Returns %ecx = length, %edi = pointer to word.
1992 // Is it in the dictionary?
1994 movl %eax,interpret_is_lit // Not a literal number (not yet anyway ...)
1995 call _FIND // Returns %eax = pointer to header or 0 if not found.
1996 test %eax,%eax // Found?
1999 // In the dictionary. Is it an IMMEDIATE codeword?
2000 mov %eax,%edi // %edi = dictionary entry
2001 movb 4(%edi),%al // Get name+flags.
2002 push %ax // Just save it for now.
2003 call _TCFA // Convert dictionary entry (in %edi) to codeword pointer.
2005 andb $F_IMMED,%al // Is IMMED flag set?
2007 jnz 4f // If IMMED, jump straight to executing.
2011 1: // Not in the dictionary (not a word) so assume it's a literal number.
2012 incl interpret_is_lit
2013 call _SNUMBER // Returns the parsed number in %eax
2015 mov $LIT,%eax // The word is LIT
2017 2: // Are we compiling or executing?
2020 jz 4f // Jump if executing.
2022 // Compiling - just append the word to the current dictionary definition.
2024 mov interpret_is_lit,%ecx // Was it a literal?
2027 mov %ebx,%eax // Yes, so LIT is followed by a number.
2031 4: // Executing - run it!
2032 mov interpret_is_lit,%ecx // Literal?
2033 test %ecx,%ecx // Literal?
2036 // Not a literal, execute it now. This never returns, but the codeword will
2037 // eventually call NEXT which will reenter the loop in INTERPRETER.
2040 5: // Executing a literal, which means push it on the stack.
2047 .int 0 // Flag used to record if reading a literal
2050 ODDS AND ENDS ----------------------------------------------------------------------
2052 CHAR puts the ASCII code of the first character of the following word on the stack. For example
2053 CHAR A puts 65 on the stack.
2055 SYSEXIT exits the process using Linux exit syscall.
2057 In this FORTH, SYSEXIT must be the last word in the built-in (assembler) dictionary because we
2058 initialise the LATEST variable to point to it. This means that if you want to extend the assembler
2059 part, you must put new words before SYSEXIT, or else change how LATEST is initialised.
2062 defcode "CHAR",4,,CHAR
2063 call _WORD // Returns %ecx = length, %edi = pointer to word.
2065 movb (%edi),%al // Get the first character of the word.
2066 push %eax // Push it onto the stack.
2069 // NB: SYSEXIT must be the last entry in the built-in dictionary.
2070 defcode SYSEXIT,7,,SYSEXIT
2076 START OF FORTH CODE ----------------------------------------------------------------------
2078 We've now reached the stage where the FORTH system is running and self-hosting. All further
2079 words can be written as FORTH itself, including words like IF, THEN, .", etc which in most
2080 languages would be considered rather fundamental.
2082 I used to append this here in the assembly file, but I got sick of fighting against gas's
2083 stupid (lack of) multiline string syntax. So now that is in a separate file called jonesforth.f
2085 If you don't already have that file, download it from http://annexia.org/forth in order
2086 to continue the tutorial.
2089 /* END OF jonesforth.S */