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 1.20 2007-09-15 11:21:09 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 Here is another "Why FORTH?" essay: http://www.jwdt.com/~paysan/why-forth.html
68 Discussion and criticism of this FORTH here: http://lambda-the-ultimate.org/node/2452
70 ACKNOWLEDGEMENTS ----------------------------------------------------------------------
72 This code draws heavily on the design of LINA FORTH (http://home.hccnet.nl/a.w.m.van.der.horst/lina.html)
73 by Albert van der Horst. Any similarities in the code are probably not accidental.
75 Also I used this document (http://ftp.funet.fi/pub/doc/IOCCC/1992/buzzard.2.design) which really
76 defies easy explanation.
78 PUBLIC DOMAIN ----------------------------------------------------------------------
80 I, the copyright holder of this work, hereby release it into the public domain. This applies worldwide.
82 In case this is not legally possible, I grant any entity the right to use this work for any purpose,
83 without any conditions, unless such conditions are required by law.
85 SETTING UP ----------------------------------------------------------------------
87 Let's get a few housekeeping things out of the way. Firstly because I need to draw lots of
88 ASCII-art diagrams to explain concepts, the best way to look at this is using a window which
89 uses a fixed width font and is at least this wide:
91 <------------------------------------------------------------------------------------------------------------------------>
93 Secondly make sure TABS are set to 8 characters. The following should be a vertical
94 line. If not, sort out your tabs.
100 Thirdly I assume that your screen is at least 50 characters high.
102 ASSEMBLING ----------------------------------------------------------------------
104 If you want to actually run this FORTH, rather than just read it, you will need Linux on an
105 i386. Linux because instead of programming directly to the hardware on a bare PC which I
106 could have done, I went for a simpler tutorial by assuming that the 'hardware' is a Linux
107 process with a few basic system calls (read, write and exit and that's about all). i386
108 is needed because I had to write the assembly for a processor, and i386 is by far the most
109 common. (Of course when I say 'i386', any 32- or 64-bit x86 processor will do. I'm compiling
110 this on a 64 bit AMD Opteron).
112 Again, to assemble this you will need gcc and gas (the GNU assembler). The commands to
113 assemble and run the code (save this file as 'jonesforth.S') are:
115 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
118 You will see lots of 'Warning: unterminated string; newline inserted' messages from the
119 assembler. That's just because the GNU assembler doesn't have a good syntax for multi-line
120 strings (or rather it used to, but the developers removed it!) so I've abused the syntax
121 slightly to make things readable. Ignore these warnings.
123 If you want to run your own FORTH programs you can do:
125 ./jonesforth < myprog.f
127 If you want to load your own FORTH code and then continue reading user commands, you can do:
129 cat myfunctions.f - | ./jonesforth
131 ASSEMBLER ----------------------------------------------------------------------
133 (You can just skip to the next section -- you don't need to be able to read assembler to
134 follow this tutorial).
136 However if you do want to read the assembly code here are a few notes about gas (the GNU assembler):
138 (1) Register names are prefixed with '%', so %eax is the 32 bit i386 accumulator. The registers
139 available on i386 are: %eax, %ebx, %ecx, %edx, %esi, %edi, %ebp and %esp, and most of them
140 have special purposes.
142 (2) Add, mov, etc. take arguments in the form SRC,DEST. So mov %eax,%ecx moves %eax -> %ecx
144 (3) Constants are prefixed with '$', and you mustn't forget it! If you forget it then it
145 causes a read from memory instead, so:
146 mov $2,%eax moves number 2 into %eax
147 mov 2,%eax reads the 32 bit word from address 2 into %eax (ie. most likely a mistake)
149 (4) gas has a funky syntax for local labels, where '1f' (etc.) means label '1:' "forwards"
150 and '1b' (etc.) means label '1:' "backwards".
152 (5) 'ja' is "jump if above", 'jb' for "jump if below", 'je' "jump if equal" etc.
154 (6) gas has a reasonably nice .macro syntax, and I use them a lot to make the code shorter and
157 For more help reading the assembler, do "info gas" at the Linux prompt.
159 Now the tutorial starts in earnest.
161 THE DICTIONARY ----------------------------------------------------------------------
163 In FORTH as you will know, functions are called "words", and just as in other languages they
164 have a name and a definition. Here are two FORTH words:
166 : DOUBLE DUP + ; \ name is "DOUBLE", definition is "DUP +"
167 : QUADRUPLE DOUBLE DOUBLE ; \ name is "QUADRUPLE", definition is "DOUBLE DOUBLE"
169 Words, both built-in ones and ones which the programmer defines later, are stored in a dictionary
170 which is just a linked list of dictionary entries.
172 <--- DICTIONARY ENTRY (HEADER) ----------------------->
173 +------------------------+--------+---------- - - - - +----------- - - - -
174 | LINK POINTER | LENGTH/| NAME | DEFINITION
176 +--- (4 bytes) ----------+- byte -+- n bytes - - - - +----------- - - - -
178 I'll come to the definition of the word later. For now just look at the header. The first
179 4 bytes are the link pointer. This points back to the previous word in the dictionary, or, for
180 the first word in the dictionary it is just a NULL pointer. Then comes a length/flags byte.
181 The length of the word can be up to 31 characters (5 bits used) and the top three bits are used
182 for various flags which I'll come to later. This is followed by the name itself, and in this
183 implementation the name is rounded up to a multiple of 4 bytes by padding it with zero bytes.
184 That's just to ensure that the definition starts on a 32 bit boundary.
186 A FORTH variable called LATEST contains a pointer to the most recently defined word, in
187 other words, the head of this linked list.
189 DOUBLE and QUADRUPLE might look like this:
191 pointer to previous word
194 +--|------+---+---+---+---+---+---+---+---+------------- - - - -
195 | LINK | 6 | D | O | U | B | L | E | 0 | (definition ...)
196 +---------+---+---+---+---+---+---+---+---+------------- - - - -
199 +--|------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
200 | LINK | 9 | Q | U | A | D | R | U | P | L | E | 0 | 0 | (definition ...)
201 +---------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
207 You should be able to see from this how you might implement functions to find a word in
208 the dictionary (just walk along the dictionary entries starting at LATEST and matching
209 the names until you either find a match or hit the NULL pointer at the end of the dictionary);
210 and add a word to the dictionary (create a new definition, set its LINK to LATEST, and set
211 LATEST to point to the new word). We'll see precisely these functions implemented in
212 assembly code later on.
214 One interesting consequence of using a linked list is that you can redefine words, and
215 a newer definition of a word overrides an older one. This is an important concept in
216 FORTH because it means that any word (even "built-in" or "standard" words) can be
217 overridden with a new definition, either to enhance it, to make it faster or even to
218 disable it. However because of the way that FORTH words get compiled, which you'll
219 understand below, words defined using the old definition of a word continue to use
220 the old definition. Only words defined after the new definition use the new definition.
222 DIRECT THREADED CODE ----------------------------------------------------------------------
224 Now we'll get to the really crucial bit in understanding FORTH, so go and get a cup of tea
225 or coffee and settle down. It's fair to say that if you don't understand this section, then you
226 won't "get" how FORTH works, and that would be a failure on my part for not explaining it well.
227 So if after reading this section a few times you don't understand it, please email me
230 Let's talk first about what "threaded code" means. Imagine a peculiar version of C where
231 you are only allowed to call functions without arguments. (Don't worry for now that such a
232 language would be completely useless!) So in our peculiar C, code would look like this:
241 and so on. How would a function, say 'f' above, be compiled by a standard C compiler?
242 Probably into assembly code like this. On the right hand side I've written the actual
246 CALL a E8 08 00 00 00
247 CALL b E8 1C 00 00 00
248 CALL c E8 2C 00 00 00
249 ; ignore the return from the function for now
251 "E8" is the x86 machine code to "CALL" a function. In the first 20 years of computing
252 memory was hideously expensive and we might have worried about the wasted space being used
253 by the repeated "E8" bytes. We can save 20% in code size (and therefore, in expensive memory)
254 by compressing this into just:
256 08 00 00 00 Just the function addresses, without
257 1C 00 00 00 the CALL prefix.
260 [Historical note: If the execution model that FORTH uses looks strange from the following
261 paragraphs, then it was motivated entirely by the need to save memory on early computers.
262 This code compression isn't so important now when our machines have more memory in their L1
263 caches than those early computers had in total, but the execution model still has some
266 Of course this code won't run directly any more. Instead we need to write an interpreter
267 which takes each pair of bytes and calls it.
269 On an i386 machine it turns out that we can write this interpreter rather easily, in just
270 two assembly instructions which turn into just 3 bytes of machine code. Let's store the
271 pointer to the next word to execute in the %esi register:
273 08 00 00 00 <- We're executing this one now. %esi is the _next_ one to execute.
277 The all-important i386 instruction is called LODSL (or in Intel manuals, LODSW). It does
278 two things. Firstly it reads the memory at %esi into the accumulator (%eax). Secondly it
279 increments %esi by 4 bytes. So after LODSL, the situation now looks like this:
281 08 00 00 00 <- We're still executing this one
282 1C 00 00 00 <- %eax now contains this address (0x0000001C)
285 Now we just need to jump to the address in %eax. This is again just a single x86 instruction
286 written JMP *(%eax). And after doing the jump, the situation looks like:
289 1C 00 00 00 <- Now we're executing this subroutine.
292 To make this work, each subroutine is followed by the two instructions 'LODSL; JMP *(%eax)'
293 which literally make the jump to the next subroutine.
295 And that brings us to our first piece of actual code! Well, it's a macro.
304 /* The macro is called NEXT. That's a FORTH-ism. It expands to those two instructions.
306 Every FORTH primitive that we write has to be ended by NEXT. Think of it kind of like
309 The above describes what is known as direct threaded code.
311 To sum up: We compress our function calls down to a list of addresses and use a somewhat
312 magical macro to act as a "jump to next function in the list". We also use one register (%esi)
313 to act as a kind of instruction pointer, pointing to the next function in the list.
315 I'll just give you a hint of what is to come by saying that a FORTH definition such as:
317 : QUADRUPLE DOUBLE DOUBLE ;
319 actually compiles (almost, not precisely but we'll see why in a moment) to a list of
320 function addresses for DOUBLE, DOUBLE and a special function called EXIT to finish off.
322 At this point, REALLY EAGLE-EYED ASSEMBLY EXPERTS are saying "JONES, YOU'VE MADE A MISTAKE!".
324 I lied about JMP *(%eax).
326 INDIRECT THREADED CODE ----------------------------------------------------------------------
328 It turns out that direct threaded code is interesting but only if you want to just execute
329 a list of functions written in assembly language. So QUADRUPLE would work only if DOUBLE
330 was an assembly language function. In the direct threaded code, QUADRUPLE would look like:
333 | addr of DOUBLE --------------------> (assembly code to do the double)
334 +------------------+ NEXT
335 %esi -> | addr of DOUBLE |
338 We can add an extra indirection to allow us to run both words written in assembly language
339 (primitives written for speed) and words written in FORTH themselves as lists of addresses.
341 The extra indirection is the reason for the brackets in JMP *(%eax).
343 Let's have a look at how QUADRUPLE and DOUBLE really look in FORTH:
345 : QUADRUPLE DOUBLE DOUBLE ;
348 | codeword | : DOUBLE DUP + ;
350 | addr of DOUBLE ---------------> +------------------+
351 +------------------+ | codeword |
352 | addr of DOUBLE | +------------------+
353 +------------------+ | addr of DUP --------------> +------------------+
354 | addr of EXIT | +------------------+ | codeword -------+
355 +------------------+ %esi -> | addr of + --------+ +------------------+ |
356 +------------------+ | | assembly to <-----+
357 | addr of EXIT | | | implement DUP |
358 +------------------+ | | .. |
361 | +------------------+
363 +-----> +------------------+
365 +------------------+ |
366 | assembly to <------+
373 This is the part where you may need an extra cup of tea/coffee/favourite caffeinated
374 beverage. What has changed is that I've added an extra pointer to the beginning of
375 the definitions. In FORTH this is sometimes called the "codeword". The codeword is
376 a pointer to the interpreter to run the function. For primitives written in
377 assembly language, the "interpreter" just points to the actual assembly code itself.
378 They don't need interpreting, they just run.
380 In words written in FORTH (like QUADRUPLE and DOUBLE), the codeword points to an interpreter
383 I'll show you the interpreter function shortly, but let's recall our indirect
384 JMP *(%eax) with the "extra" brackets. Take the case where we're executing DOUBLE
385 as shown, and DUP has been called. Note that %esi is pointing to the address of +
387 The assembly code for DUP eventually does a NEXT. That:
389 (1) reads the address of + into %eax %eax points to the codeword of +
390 (2) increments %esi by 4
391 (3) jumps to the indirect %eax jumps to the address in the codeword of +,
392 ie. the assembly code to implement +
397 | addr of DOUBLE ---------------> +------------------+
398 +------------------+ | codeword |
399 | addr of DOUBLE | +------------------+
400 +------------------+ | addr of DUP --------------> +------------------+
401 | addr of EXIT | +------------------+ | codeword -------+
402 +------------------+ | addr of + --------+ +------------------+ |
403 +------------------+ | | assembly to <-----+
404 %esi -> | addr of EXIT | | | implement DUP |
405 +------------------+ | | .. |
408 | +------------------+
410 +-----> +------------------+
412 +------------------+ |
413 now we're | assembly to <------+
414 executing | implement + |
420 So I hope that I've convinced you that NEXT does roughly what you'd expect. This is
421 indirect threaded code.
423 I've glossed over four things. I wonder if you can guess without reading on what they are?
429 My list of four things are: (1) What does "EXIT" do? (2) which is related to (1) is how do
430 you call into a function, ie. how does %esi start off pointing at part of QUADRUPLE, but
431 then point at part of DOUBLE. (3) What goes in the codeword for the words which are written
432 in FORTH? (4) How do you compile a function which does anything except call other functions
433 ie. a function which contains a number like : DOUBLE 2 * ; ?
435 THE INTERPRETER AND RETURN STACK ------------------------------------------------------------
437 Going at these in no particular order, let's talk about issues (3) and (2), the interpreter
438 and the return stack.
440 Words which are defined in FORTH need a codeword which points to a little bit of code to
441 give them a "helping hand" in life. They don't need much, but they do need what is known
442 as an "interpreter", although it doesn't really "interpret" in the same way that, say,
443 Java bytecode used to be interpreted (ie. slowly). This interpreter just sets up a few
444 machine registers so that the word can then execute at full speed using the indirect
445 threaded model above.
447 One of the things that needs to happen when QUADRUPLE calls DOUBLE is that we save the old
448 %esi ("instruction pointer") and create a new one pointing to the first word in DOUBLE.
449 Because we will need to restore the old %esi at the end of DOUBLE (this is, after all, like
450 a function call), we will need a stack to store these "return addresses" (old values of %esi).
452 As you will have read, when reading the background documentation, FORTH has two stacks,
453 an ordinary stack for parameters, and a return stack which is a bit more mysterious. But
454 our return stack is just the stack I talked about in the previous paragraph, used to save
455 %esi when calling from a FORTH word into another FORTH word.
457 In this FORTH, we are using the normal stack pointer (%esp) for the parameter stack.
458 We will use the i386's "other" stack pointer (%ebp, usually called the "frame pointer")
459 for our return stack.
461 I've got two macros which just wrap up the details of using %ebp for the return stack.
462 You use them as for example "PUSHRSP %eax" (push %eax on the return stack) or "POPRSP %ebx"
463 (pop top of return stack into %ebx).
466 /* Macros to deal with the return stack. */
468 lea -4(%ebp),%ebp // push reg on to return stack
473 mov (%ebp),\reg // pop top of return stack to reg
478 And with that we can now talk about the interpreter.
480 In FORTH the interpreter function is often called DOCOL (I think it means "DO COLON" because
481 all FORTH definitions start with a colon, as in : DOUBLE DUP + ;
483 The "interpreter" (it's not really "interpreting") just needs to push the old %esi on the
484 stack and set %esi to the first word in the definition. Remember that we jumped to the
485 function using JMP *(%eax)? Well a consequence of that is that conveniently %eax contains
486 the address of this codeword, so just by adding 4 to it we get the address of the first
487 data word. Finally after setting up %esi, it just does NEXT which causes that first word
491 /* DOCOL - the interpreter! */
495 PUSHRSP %esi // push %esi on to the return stack
496 addl $4,%eax // %eax points to codeword, so make
497 movl %eax,%esi // %esi point to first data word
501 Just to make this absolutely clear, let's see how DOCOL works when jumping from QUADRUPLE
507 +------------------+ DOUBLE:
508 | addr of DOUBLE ---------------> +------------------+
509 +------------------+ %eax -> | addr of DOCOL |
510 %esi -> | addr of DOUBLE | +------------------+
511 +------------------+ | addr of DUP |
512 | addr of EXIT | +------------------+
513 +------------------+ | etc. |
515 First, the call to DOUBLE calls DOCOL (the codeword of DOUBLE). DOCOL does this: It
516 pushes the old %esi on the return stack. %eax points to the codeword of DOUBLE, so we
517 just add 4 on to it to get our new %esi:
522 +------------------+ DOUBLE:
523 | addr of DOUBLE ---------------> +------------------+
524 top of return +------------------+ %eax -> | addr of DOCOL |
525 stack points -> | addr of DOUBLE | + 4 = +------------------+
526 +------------------+ %esi -> | addr of DUP |
527 | addr of EXIT | +------------------+
528 +------------------+ | etc. |
530 Then we do NEXT, and because of the magic of threaded code that increments %esi again
533 Well, it seems to work.
535 One minor point here. Because DOCOL is the first bit of assembly actually to be defined
536 in this file (the others were just macros), and because I usually compile this code with the
537 text segment starting at address 0, DOCOL has address 0. So if you are disassembling the
538 code and see a word with a codeword of 0, you will immediately know that the word is
539 written in FORTH (it's not an assembler primitive) and so uses DOCOL as the interpreter.
541 STARTING UP ----------------------------------------------------------------------
543 Now let's get down to nuts and bolts. When we start the program we need to set up
544 a few things like the return stack. But as soon as we can, we want to jump into FORTH
545 code (albeit much of the "early" FORTH code will still need to be written as
546 assembly language primitives).
548 This is what the set up code does. Does a tiny bit of house-keeping, sets up the
549 separate return stack (NB: Linux gives us the ordinary parameter stack already), then
550 immediately jumps to a FORTH word called COLD. COLD stands for cold-start. In ISO
551 FORTH (but not in this FORTH), COLD can be called at any time to completely reset
552 the state of FORTH, and there is another word called WARM which does a partial reset.
555 /* ELF entry point. */
560 mov %esp,var_S0 // Store the initial data stack pointer.
561 mov $return_stack,%ebp // Initialise the return stack.
563 mov $cold_start,%esi // Initialise interpreter.
564 NEXT // Run interpreter!
567 cold_start: // High-level code without a codeword.
571 We also allocate some space for the return stack and some space to store user
572 definitions. These are static memory allocations using fixed-size buffers, but it
573 wouldn't be a great deal of work to make them dynamic.
577 /* FORTH return stack. */
578 #define RETURN_STACK_SIZE 8192
580 .space RETURN_STACK_SIZE
581 return_stack: // Initial top of return stack.
583 /* Space for user-defined words. */
584 #define USER_DEFS_SIZE 16384
587 .space USER_DEFS_SIZE
590 BUILT-IN WORDS ----------------------------------------------------------------------
592 Remember our dictionary entries (headers). Let's bring those together with the codeword
593 and data words to see how : DOUBLE DUP + ; really looks in memory.
595 pointer to previous word
598 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
599 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
600 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
603 LINK in next word points to codeword of DUP
605 Initially we can't just write ": DOUBLE DUP + ;" (ie. that literal string) here because we
606 don't yet have anything to read the string, break it up at spaces, parse each word, etc. etc.
607 So instead we will have to define built-in words using the GNU assembler data constructors
608 (like .int, .byte, .string, .ascii and so on -- look them up in the gas info page if you are
611 The long way would be:
612 .int <link to previous word>
614 .ascii "DOUBLE" // string
616 DOUBLE: .int DOCOL // codeword
617 .int DUP // pointer to codeword of DUP
618 .int PLUS // pointer to codeword of +
619 .int EXIT // pointer to codeword of EXIT
621 That's going to get quite tedious rather quickly, so here I define an assembler macro
622 so that I can just write:
624 defword "DOUBLE",6,,DOUBLE
627 and I'll get exactly the same effect.
629 Don't worry too much about the exact implementation details of this macro - it's complicated!
632 /* Flags - these are discussed later. */
634 #define F_HIDDEN 0x20
636 // Store the chain of links.
639 .macro defword name, namelen, flags=0, label
645 .set link,name_\label
646 .byte \flags+\namelen // flags + length byte
647 .ascii "\name" // the name
651 .int DOCOL // codeword - the interpreter
652 // list of word pointers follow
656 Similarly I want a way to write words written in assembly language. There will quite a few
657 of these to start with because, well, everything has to start in assembly before there's
658 enough "infrastructure" to be able to start writing FORTH words, but also I want to define
659 some common FORTH words in assembly language for speed, even though I could write them in FORTH.
661 This is what DUP looks like in memory:
663 pointer to previous word
666 +--|------+---+---+---+---+------------+
667 | LINK | 3 | D | U | P | code_DUP ---------------------> points to the assembly
668 +---------+---+---+---+---+------------+ code used to write DUP,
669 ^ len codeword which ends with NEXT.
673 Again, for brevity in writing the header I'm going to write an assembler macro called defcode.
676 .macro defcode name, namelen, flags=0, label
682 .set link,name_\label
683 .byte \flags+\namelen // flags + length byte
684 .ascii "\name" // the name
688 .int code_\label // codeword
692 code_\label : // assembler code follows
696 Now some easy FORTH primitives. These are written in assembly for speed. If you understand
697 i386 assembly language then it is worth reading these. However if you don't understand assembly
698 you can skip the details.
702 pop %eax // duplicate top of stack
707 defcode "DROP",4,,DROP
708 pop %eax // drop top of stack
711 defcode "SWAP",4,,SWAP
712 pop %eax // swap top of stack
718 defcode "OVER",4,,OVER
719 mov 4(%esp),%eax // get the second element of stack
720 push %eax // and push it on top
732 defcode "-ROT",4,,NROT
742 incl (%esp) // increment top of stack
746 decl (%esp) // decrement top of stack
749 defcode "4+",2,,INCR4
750 addl $4,(%esp) // add 4 to top of stack
753 defcode "4-",2,,DECR4
754 subl $4,(%esp) // subtract 4 from top of stack
758 pop %eax // get top of stack
759 addl %eax,(%esp) // and add it to next word on stack
763 pop %eax // get top of stack
764 subl %eax,(%esp) // and subtract it from next word on stack
771 push %eax // ignore overflow
779 push %eax // push quotient
787 push %edx // push remainder
790 defcode "=",1,,EQU // top two words are equal?
800 defcode "<>",2,,NEQU // top two words are not equal?
810 defcode "0=",2,,ZEQU // top of stack equals 0?
829 defcode "INVERT",6,,INVERT // this is the FORTH "NOT" function
834 RETURNING FROM FORTH WORDS ----------------------------------------------------------------------
836 Time to talk about what happens when we EXIT a function. In this diagram QUADRUPLE has called
837 DOUBLE, and DOUBLE is about to exit (look at where %esi is pointing):
842 +------------------+ DOUBLE
843 | addr of DOUBLE ---------------> +------------------+
844 +------------------+ | codeword |
845 | addr of DOUBLE | +------------------+
846 +------------------+ | addr of DUP |
847 | addr of EXIT | +------------------+
848 +------------------+ | addr of + |
850 %esi -> | addr of EXIT |
853 What happens when the + function does NEXT? Well, the following code is executed.
856 defcode "EXIT",4,,EXIT
857 POPRSP %esi // pop return stack into %esi
861 EXIT gets the old %esi which we saved from before on the return stack, and puts it in %esi.
862 So after this (but just before NEXT) we get:
867 +------------------+ DOUBLE
868 | addr of DOUBLE ---------------> +------------------+
869 +------------------+ | codeword |
870 %esi -> | addr of DOUBLE | +------------------+
871 +------------------+ | addr of DUP |
872 | addr of EXIT | +------------------+
873 +------------------+ | addr of + |
878 And NEXT just completes the job by, well in this case just by calling DOUBLE again :-)
880 LITERALS ----------------------------------------------------------------------
882 The final point I "glossed over" before was how to deal with functions that do anything
883 apart from calling other functions. For example, suppose that DOUBLE was defined like this:
887 It does the same thing, but how do we compile it since it contains the literal 2? One way
888 would be to have a function called "2" (which you'd have to write in assembler), but you'd need
889 a function for every single literal that you wanted to use.
891 FORTH solves this by compiling the function using a special word called LIT:
893 +---------------------------+-------+-------+-------+-------+-------+
894 | (usual header of DOUBLE) | DOCOL | LIT | 2 | * | EXIT |
895 +---------------------------+-------+-------+-------+-------+-------+
897 LIT is executed in the normal way, but what it does next is definitely not normal. It
898 looks at %esi (which now points to the literal 2), grabs it, pushes it on the stack, then
899 manipulates %esi in order to skip the literal as if it had never been there.
901 What's neat is that the whole grab/manipulate can be done using a single byte single
902 i386 instruction, our old friend LODSL. Rather than me drawing more ASCII-art diagrams,
903 see if you can find out how LIT works:
907 // %esi points to the next command, but in this case it points to the next
908 // literal 32 bit integer. Get that literal into %eax and increment %esi.
909 // On x86, it's a convenient single byte instruction! (cf. NEXT macro)
911 push %eax // push the literal number on to stack
915 MEMORY ----------------------------------------------------------------------
917 As important point about FORTH is that it gives you direct access to the lowest levels
918 of the machine. Manipulating memory directly is done frequently in FORTH, and these are
919 the primitive words for doing it.
923 pop %ebx // address to store at
924 pop %eax // data to store there
925 mov %eax,(%ebx) // store it
929 pop %ebx // address to fetch
930 mov (%ebx),%eax // fetch it
931 push %eax // push value onto stack
934 defcode "+!",2,,ADDSTORE
936 pop %eax // the amount to add
937 addl %eax,(%ebx) // add it
940 defcode "-!",2,,SUBSTORE
942 pop %eax // the amount to subtract
943 subl %eax,(%ebx) // add it
946 /* ! and @ (STORE and FETCH) store 32-bit words. It's also useful to be able to read and write bytes.
947 * I don't know whether FORTH has these words, so I invented my own, called !b and @b.
948 * Byte-oriented operations only work on architectures which permit them (i386 is one of those).
949 * UPDATE: writing a byte to the dictionary pointer is called C, in FORTH.
951 defcode "!b",2,,STOREBYTE
952 pop %ebx // address to store at
953 pop %eax // data to store there
954 movb %al,(%ebx) // store it
957 defcode "@b",2,,FETCHBYTE
958 pop %ebx // address to fetch
960 movb (%ebx),%al // fetch it
961 push %eax // push value onto stack
965 BUILT-IN VARIABLES ----------------------------------------------------------------------
967 These are some built-in variables and related standard FORTH words. Of these, the only one that we
968 have discussed so far was LATEST, which points to the last (most recently defined) word in the
969 FORTH dictionary. LATEST is also a FORTH word which pushes the address of LATEST (the variable)
970 on to the stack, so you can read or write it using @ and ! operators. For example, to print
971 the current value of LATEST (and this can apply to any FORTH variable) you would do:
975 To make defining variables shorter, I'm using a macro called defvar, similar to defword and
976 defcode above. (In fact the defvar macro uses defcode to do the dictionary header).
979 .macro defvar name, namelen, flags=0, label, initial=0
980 defcode \name,\namelen,\flags,\label
990 The built-in variables are:
992 STATE Is the interpreter executing code (0) or compiling a word (non-zero)?
993 LATEST Points to the latest (most recently defined) word in the dictionary.
994 HERE Points to the next free byte of memory. When compiling, compiled words go here.
995 _X These are three scratch variables, used by some standard dictionary words.
998 S0 Stores the address of the top of the parameter stack.
999 R0 Stores the address of the top of the return stack.
1000 VERSION Is the current version of this FORTH.
1003 defvar "STATE",5,,STATE
1004 defvar "HERE",4,,HERE,user_defs_start
1005 defvar "LATEST",6,,LATEST,name_SYSEXIT // SYSEXIT must be last in built-in dictionary
1010 defvar "R0",2,,RZ,return_stack
1011 defvar "VERSION",7,,VERSION,JONES_VERSION
1014 RETURN STACK ----------------------------------------------------------------------
1016 These words allow you to access the return stack. Recall that the register %ebp always points to
1017 the top of the return stack.
1021 pop %eax // pop parameter stack into %eax
1022 PUSHRSP %eax // push it on to the return stack
1025 defcode "R>",2,,FROMR
1026 POPRSP %eax // pop return stack on to %eax
1027 push %eax // and push on to parameter stack
1030 defcode "RSP@",4,,RSPFETCH
1034 defcode "RSP!",4,,RSPSTORE
1038 defcode "RDROP",5,,RDROP
1039 lea 4(%ebp),%ebp // pop return stack and throw away
1043 PARAMETER (DATA) STACK ----------------------------------------------------------------------
1045 These functions allow you to manipulate the parameter stack. Recall that Linux sets up the parameter
1046 stack for us, and it is accessed through %esp.
1049 defcode "DSP@",4,,DSPFETCH
1054 defcode "DSP!",4,,DSPSTORE
1059 INPUT AND OUTPUT ----------------------------------------------------------------------
1061 These are our first really meaty/complicated FORTH primitives. I have chosen to write them in
1062 assembler, but surprisingly in "real" FORTH implementations these are often written in terms
1063 of more fundamental FORTH primitives. I chose to avoid that because I think that just obscures
1064 the implementation. After all, you may not understand assembler but you can just think of it
1065 as an opaque block of code that does what it says.
1067 Let's discuss input first.
1069 The FORTH word KEY reads the next byte from stdin (and pushes it on the parameter stack).
1070 So if KEY is called and someone hits the space key, then the number 32 (ASCII code of space)
1071 is pushed on the stack.
1073 In FORTH there is no distinction between reading code and reading input. We might be reading
1074 and compiling code, we might be reading words to execute, we might be asking for the user
1075 to type their name -- ultimately it all comes in through KEY.
1077 The implementation of KEY uses an input buffer of a certain size (defined at the end of the
1078 program). It calls the Linux read(2) system call to fill this buffer and tracks its position
1079 in the buffer using a couple of variables, and if it runs out of input buffer then it refills
1080 it automatically. The other thing that KEY does is if it detects that stdin has closed, it
1081 exits the program, which is why when you hit ^D the FORTH system cleanly exits.
1084 #include <asm-i386/unistd.h>
1086 defcode "KEY",3,,KEY
1088 push %eax // push return value on stack
1100 1: // out of input; use read(2) to fetch more input from stdin
1101 xor %ebx,%ebx // 1st param: stdin
1102 mov $buffer,%ecx // 2nd param: buffer
1104 mov $buffend-buffer,%edx // 3rd param: max length
1105 mov $__NR_read,%eax // syscall: read
1107 test %eax,%eax // If %eax <= 0, then exit.
1109 addl %eax,%ecx // buffer+%eax = bufftop
1113 2: // error or out of input: exit
1115 mov $__NR_exit,%eax // syscall: exit
1119 By contrast, output is much simpler. The FORTH word EMIT writes out a single byte to stdout.
1120 This implementation just uses the write system call. No attempt is made to buffer output, but
1121 it would be a good exercise to add it.
1124 defcode "EMIT",4,,EMIT
1129 mov $1,%ebx // 1st param: stdout
1131 // write needs the address of the byte to write
1133 mov $2f,%ecx // 2nd param: address
1135 mov $1,%edx // 3rd param: nbytes = 1
1137 mov $__NR_write,%eax // write syscall
1142 2: .space 1 // scratch used by EMIT
1145 Back to input, WORD is a FORTH word which reads the next full word of input.
1147 What it does in detail is that it first skips any blanks (spaces, tabs, newlines and so on).
1148 Then it calls KEY to read characters into an internal buffer until it hits a blank. Then it
1149 calculates the length of the word it read and returns the address and the length as
1150 two words on the stack (with address at the top).
1152 Notice that WORD has a single internal buffer which it overwrites each time (rather like
1153 a static C string). Also notice that WORD's internal buffer is just 32 bytes long and
1154 there is NO checking for overflow. 31 bytes happens to be the maximum length of a
1155 FORTH word that we support, and that is what WORD is used for: to read FORTH words when
1156 we are compiling and executing code. The returned strings are not NUL-terminated, so
1157 in some crazy-world you could define FORTH words containing ASCII NULs, although why
1158 you'd want to is a bit beyond me.
1160 WORD is not suitable for just reading strings (eg. user input) because of all the above
1161 peculiarities and limitations.
1163 Note that when executing, you'll see:
1165 which puts "FOO" and length 3 on the stack, but when compiling:
1167 is an error (or at least it doesn't do what you might expect). Later we'll talk about compiling
1168 and immediate mode, and you'll understand why.
1171 defcode "WORD",4,,WORD
1173 push %ecx // push length
1174 push %edi // push base address
1178 /* Search for first non-blank character. Also skip \ comments. */
1180 call _KEY // get next key, returned in %eax
1181 cmpb $'\\',%al // start of a comment?
1182 je 3f // if so, skip the comment
1184 jbe 1b // if so, keep looking
1186 /* Search for the end of the word, storing chars as we go. */
1187 mov $5f,%edi // pointer to return buffer
1189 stosb // add character to return buffer
1190 call _KEY // get next key, returned in %al
1191 cmpb $' ',%al // is blank?
1192 ja 2b // if not, keep looping
1194 /* Return the word (well, the static buffer) and length. */
1196 mov %edi,%ecx // return length of the word
1197 mov $5f,%edi // return address of the word
1200 /* Code to skip \ comments to end of the current line. */
1203 cmpb $'\n',%al // end of line yet?
1208 // A static buffer where WORD returns. Subsequent calls
1209 // overwrite this buffer. Maximum word length is 32 chars.
1213 . (also called DOT) prints the top of the stack as an integer. In real FORTH implementations
1214 it should print it in the current base, but this assembler version is simpler and can only
1217 Remember that you can override even built-in FORTH words easily, so if you want to write a
1218 more advanced DOT then you can do so easily at a later point, and probably in FORTH.
1222 pop %eax // Get the number to print into %eax
1223 call _DOT // Easier to do this recursively ...
1226 mov $10,%ecx // Base 10
1230 xor %edx,%edx // %edx:%eax / %ecx -> quotient %eax, remainder %edx
1245 Almost the opposite of DOT (but not quite), SNUMBER parses a numeric string such as one returned
1246 by WORD and pushes the number on the parameter stack.
1248 This function does absolutely no error checking, and in particular the length of the string
1249 must be >= 1 bytes, and should contain only digits 0-9. If it doesn't you'll get random results.
1251 This function is only used when reading literal numbers in code, and shouldn't really be used
1252 in user code at all.
1254 defcode "SNUMBER",7,,SNUMBER
1264 imull $10,%eax // %eax *= 10
1267 subb $'0',%bl // ASCII -> digit
1274 DICTIONARY LOOK UPS ----------------------------------------------------------------------
1276 We're building up to our prelude on how FORTH code is compiled, but first we need yet more infrastructure.
1278 The FORTH word FIND takes a string (a word as parsed by WORD -- see above) and looks it up in the
1279 dictionary. What it actually returns is the address of the dictionary header, if it finds it,
1282 So if DOUBLE is defined in the dictionary, then WORD DOUBLE FIND returns the following pointer:
1288 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1289 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1290 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1292 See also >CFA which takes a dictionary entry pointer and returns a pointer to the codeword.
1294 FIND doesn't find dictionary entries which are flagged as HIDDEN. See below for why.
1297 defcode "FIND",4,,FIND
1298 pop %edi // %edi = address
1299 pop %ecx // %ecx = length
1305 push %esi // Save %esi so we can use it in string comparison.
1307 // Now we start searching backwards through the dictionary for this word.
1308 mov var_LATEST,%edx // LATEST points to name header of the latest word in the dictionary
1310 test %edx,%edx // NULL pointer? (end of the linked list)
1313 // Compare the length expected and the length of the word.
1314 // Note that if the F_HIDDEN flag is set on the word, then by a bit of trickery
1315 // this won't pick the word (the length will appear to be wrong).
1317 movb 4(%edx),%al // %al = flags+length field
1318 andb $(F_HIDDEN|0x1f),%al // %al = name length
1319 cmpb %cl,%al // Length is the same?
1322 // Compare the strings in detail.
1323 push %ecx // Save the length
1324 push %edi // Save the address (repe cmpsb will move this pointer)
1325 lea 5(%edx),%esi // Dictionary string we are checking against.
1326 repe cmpsb // Compare the strings.
1329 jne 2f // Not the same.
1331 // The strings are the same - return the header pointer in %eax
1337 mov (%edx),%edx // Move back through the link field to the previous word
1338 jmp 1b // .. and loop.
1342 xor %eax,%eax // Return zero to indicate not found.
1346 FIND returns the dictionary pointer, but when compiling we need the codeword pointer (recall
1347 that FORTH definitions are compiled into lists of codeword pointers). The standard FORTH
1348 word >CFA turns a dictionary pointer into a codeword pointer.
1350 The example below shows the result of:
1352 WORD DOUBLE FIND >CFA
1354 FIND returns a pointer to this
1355 | >CFA converts it to a pointer to this
1358 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1359 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1360 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1364 Because names vary in length, this isn't just a simple increment.
1366 In this FORTH you cannot easily turn a codeword pointer back into a dictionary entry pointer, but
1367 that is not true in most FORTH implementations where they store a back pointer in the definition
1368 (with an obvious memory/complexity cost). The reason they do this is that it is useful to be
1369 able to go backwards (codeword -> dictionary entry) in order to decompile FORTH definitions.
1371 What does CFA stand for? My best guess is "Code Field Address".
1374 defcode ">CFA",4,,TCFA
1381 add $4,%edi // Skip link pointer.
1382 movb (%edi),%al // Load flags+len into %al.
1383 inc %edi // Skip flags+len byte.
1384 andb $0x1f,%al // Just the length, not the flags.
1385 add %eax,%edi // Skip the name.
1386 addl $3,%edi // The codeword is 4-byte aligned.
1391 COMPILING ----------------------------------------------------------------------
1393 Now we'll talk about how FORTH compiles words. Recall that a word definition looks like this:
1397 and we have to turn this into:
1399 pointer to previous word
1402 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1403 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1404 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
1405 ^ len pad codeword |
1407 LATEST points here points to codeword of DUP
1409 There are several problems to solve. Where to put the new word? How do we read words? How
1410 do we define the words : (COLON) and ; (SEMICOLON)?
1412 FORTH solves this rather elegantly and as you might expect in a very low-level way which
1413 allows you to change how the compiler works on your own code.
1415 FORTH has an INTERPRETER function (a true interpreter this time, not DOCOL) which runs in a
1416 loop, reading words (using WORD), looking them up (using FIND), turning them into codeword
1417 pointers (using >CFA) and deciding what to do with them.
1419 What it does depends on the mode of the interpreter (in variable STATE).
1421 When STATE is zero, the interpreter just runs each word as it looks them up. This is known as
1424 The interesting stuff happens when STATE is non-zero -- compiling mode. In this mode the
1425 interpreter appends the codeword pointer to user memory (the HERE variable points to the next
1426 free byte of user memory).
1428 So you may be able to see how we could define : (COLON). The general plan is:
1430 (1) Use WORD to read the name of the function being defined.
1432 (2) Construct the dictionary entry -- just the header part -- in user memory:
1434 pointer to previous word (from LATEST) +-- Afterwards, HERE points here, where
1435 ^ | the interpreter will start appending
1437 +--|------+---+---+---+---+---+---+---+---+------------+
1438 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL |
1439 +---------+---+---+---+---+---+---+---+---+------------+
1442 (3) Set LATEST to point to the newly defined word, ...
1444 (4) .. and most importantly leave HERE pointing just after the new codeword. This is where
1445 the interpreter will append codewords.
1447 (5) Set STATE to 1. This goes into compile mode so the interpreter starts appending codewords to
1448 our partially-formed header.
1450 After : has run, our input is here:
1455 Next byte returned by KEY
1457 so the interpreter (now it's in compile mode, so I guess it's really the compiler) reads DUP,
1458 gets its codeword pointer, and appends it:
1460 +-- HERE updated to point here.
1463 +---------+---+---+---+---+---+---+---+---+------------+------------+
1464 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP |
1465 +---------+---+---+---+---+---+---+---+---+------------+------------+
1468 Next we read +, get the codeword pointer, and append it:
1470 +-- HERE updated to point here.
1473 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1474 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + |
1475 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1478 The issue is what happens next. Obviously what we _don't_ want to happen is that we
1479 read ";" and compile it and go on compiling everything afterwards.
1481 At this point, FORTH uses a trick. Remember the length byte in the dictionary definition
1482 isn't just a plain length byte, but can also contain flags. One flag is called the
1483 IMMEDIATE flag (F_IMMED in this code). If a word in the dictionary is flagged as
1484 IMMEDIATE then the interpreter runs it immediately _even if it's in compile mode_.
1486 I hope I don't need to explain that ; (SEMICOLON) is just such a word, flagged as IMMEDIATE.
1487 And all it does is append the codeword for EXIT on to the current definition and switch
1488 back to immediate mode (set STATE back to 0). Shortly we'll see the actual definition
1489 of ; and we'll see that it's really a very simple definition, declared IMMEDIATE.
1491 After the interpreter reads ; and executes it 'immediately', we get this:
1493 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1494 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1495 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1502 And that's it, job done, our new definition is compiled, and we're back in immediate mode
1503 just reading and executing words, perhaps including a call to test our new word DOUBLE.
1505 The only last wrinkle in this is that while our word was being compiled, it was in a
1506 half-finished state. We certainly wouldn't want DOUBLE to be called somehow during
1507 this time. There are several ways to stop this from happening, but in FORTH what we
1508 do is flag the word with the HIDDEN flag (F_HIDDEN in this code) just while it is
1509 being compiled. This prevents FIND from finding it, and thus in theory stops any
1510 chance of it being called.
1512 Compared to the description above, the actual definition of : (COLON) is comparatively simple:
1515 defcode ":",1,,COLON
1517 // Get the word and create a dictionary entry header for it.
1518 call _WORD // Returns %ecx = length, %edi = pointer to word.
1519 mov %edi,%ebx // %ebx = address of the word
1521 movl var_HERE,%edi // %edi is the address of the header
1522 movl var_LATEST,%eax // Get link pointer
1523 stosl // and store it in the header.
1525 mov %cl,%al // Get the length.
1526 orb $F_HIDDEN,%al // Set the HIDDEN flag on this entry.
1527 stosb // Store the length/flags byte.
1529 mov %ebx,%esi // %esi = word
1530 rep movsb // Copy the word
1532 addl $3,%edi // Align to next 4 byte boundary.
1535 movl $DOCOL,%eax // The codeword for user-created words is always DOCOL (the interpreter)
1538 // Header built, so now update LATEST and HERE.
1539 // We'll be compiling words and putting them HERE.
1541 movl %eax,var_LATEST
1544 // And go into compile mode by setting STATE to 1.
1549 , (COMMA) is a standard FORTH word which appends a 32 bit integer (normally a codeword
1550 pointer) to the user data area pointed to by HERE, and adds 4 to HERE.
1553 defcode ",",1,,COMMA
1554 pop %eax // Code pointer to store.
1558 movl var_HERE,%edi // HERE
1560 movl %edi,var_HERE // Update HERE (incremented)
1564 ; (SEMICOLON) is also elegantly simple. Notice the F_IMMED flag.
1567 defcode ";",1,F_IMMED,SEMICOLON
1568 movl $EXIT,%eax // EXIT is the final codeword in compiled words.
1569 call _COMMA // Store it.
1570 call _HIDDEN // Toggle the HIDDEN flag (unhides the new word).
1571 xor %eax,%eax // Set STATE to 0 (back to execute mode).
1576 EXTENDING THE COMPILER ----------------------------------------------------------------------
1578 Words flagged with IMMEDIATE (F_IMMED) aren't just for the FORTH compiler to use. You can define
1579 your own IMMEDIATE words too, and this is a crucial aspect when extending basic FORTH, because
1580 it allows you in effect to extend the compiler itself. Does gcc let you do that?
1582 Standard FORTH words like IF, WHILE, .", [ and so on are all written as extensions to the basic
1583 compiler, and are all IMMEDIATE words.
1585 The IMMEDIATE word toggles the F_IMMED (IMMEDIATE flag) on the most recently defined word,
1586 or on the current word if you call it in the middle of a definition.
1590 : MYIMMEDWORD IMMEDIATE
1594 but some FORTH programmers write this instead:
1600 The two usages are equivalent, to a first approximation.
1603 defcode "IMMEDIATE",9,F_IMMED,IMMEDIATE
1607 movl var_LATEST,%edi // LATEST word.
1608 addl $4,%edi // Point to name/flags byte.
1609 xorb $F_IMMED,(%edi) // Toggle the IMMED bit.
1613 HIDDEN toggles the other flag, F_HIDDEN, of the latest word. Note that words flagged
1614 as hidden are defined but cannot be called, so this is rarely used.
1617 defcode "HIDDEN",6,,HIDDEN
1621 movl var_LATEST,%edi // LATEST word.
1622 addl $4,%edi // Point to name/flags byte.
1623 xorb $F_HIDDEN,(%edi) // Toggle the HIDDEN bit.
1627 ' (TICK) is a standard FORTH word which returns the codeword pointer of the next word.
1629 The common usage is:
1633 which appends the codeword of FOO to the current word we are defining (this only works in compiled code).
1635 You tend to use ' in IMMEDIATE words. For example an alternate (and rather useless) way to define
1636 a literal 2 might be:
1639 ' LIT , \ Appends LIT to the currently-being-defined word
1640 2 , \ Appends the number 2 to the currently-being-defined word
1647 (If you don't understand how LIT2 works, then you should review the material about compiling words
1648 and immediate mode).
1650 This definition of ' uses a cheat which I copied from buzzard92. As a result it only works in
1651 compiled code. It is possible to write a version of ' based on WORD, FIND, >CFA which works in
1655 lodsl // Get the address of the next word and skip it.
1656 pushl %eax // Push it on the stack.
1660 BRANCHING ----------------------------------------------------------------------
1662 It turns out that all you need in order to define looping constructs, IF-statements, etc.
1665 BRANCH is an unconditional branch. 0BRANCH is a conditional branch (it only branches if the
1666 top of stack is zero).
1668 The diagram below shows how BRANCH works in some imaginary compiled word. When BRANCH executes,
1669 %esi starts by pointing to the offset field (compare to LIT above):
1671 +---------------------+-------+---- - - ---+------------+------------+---- - - - ----+------------+
1672 | (Dictionary header) | DOCOL | | BRANCH | offset | (skipped) | word |
1673 +---------------------+-------+---- - - ---+------------+-----|------+---- - - - ----+------------+
1676 | +-----------------------+
1677 %esi added to offset
1679 The offset is added to %esi to make the new %esi, and the result is that when NEXT runs, execution
1680 continues at the branch target. Negative offsets work as expected.
1682 0BRANCH is the same except the branch happens conditionally.
1684 Now standard FORTH words such as IF, THEN, ELSE, WHILE, REPEAT, etc. can be implemented entirely
1685 in FORTH. They are IMMEDIATE words which append various combinations of BRANCH or 0BRANCH
1686 into the word currently being compiled.
1688 As an example, code written like this:
1690 condition-code IF true-part THEN rest-code
1694 condition-code 0BRANCH OFFSET true-part rest-code
1700 defcode "BRANCH",6,,BRANCH
1701 add (%esi),%esi // add the offset to the instruction pointer
1704 defcode "0BRANCH",7,,ZBRANCH
1706 test %eax,%eax // top of stack is zero?
1707 jz code_BRANCH // if so, jump back to the branch function above
1708 lodsl // otherwise we need to skip the offset
1712 PRINTING STRINGS ----------------------------------------------------------------------
1714 LITSTRING and EMITSTRING are primitives used to implement the ." operator (which is
1715 written in FORTH). See the definition of that operator below.
1718 defcode "LITSTRING",9,,LITSTRING
1719 lodsl // get the length of the string
1720 push %eax // push it on the stack
1721 push %esi // push the address of the start of the string
1722 addl %eax,%esi // skip past the string
1723 addl $3,%esi // but round up to next 4 byte boundary
1727 defcode "EMITSTRING",10,,EMITSTRING
1728 mov $1,%ebx // 1st param: stdout
1729 pop %ecx // 2nd param: address of string
1730 pop %edx // 3rd param: length of string
1731 mov $__NR_write,%eax // write syscall
1736 COLD START AND INTERPRETER ----------------------------------------------------------------------
1738 COLD is the first FORTH function called, almost immediately after the FORTH system "boots".
1740 INTERPRETER is the FORTH interpreter ("toploop", "toplevel" or "REPL" might be a more accurate
1741 description -- see: http://en.wikipedia.org/wiki/REPL).
1745 // COLD must not return (ie. must not call EXIT).
1746 defword "COLD",4,,COLD
1747 .int INTERPRETER // call the interpreter loop (never returns)
1748 .int LIT,1,SYSEXIT // hmmm, but in case it does, exit(1).
1750 /* This interpreter is pretty simple, but remember that in FORTH you can always override
1751 * it later with a more powerful one!
1753 defword "INTERPRETER",11,,INTERPRETER
1754 .int INTERPRET,RDROP,INTERPRETER
1756 defcode "INTERPRET",9,,INTERPRET
1757 call _WORD // Returns %ecx = length, %edi = pointer to word.
1759 // Is it in the dictionary?
1761 movl %eax,interpret_is_lit // Not a literal number (not yet anyway ...)
1762 call _FIND // Returns %eax = pointer to header or 0 if not found.
1763 test %eax,%eax // Found?
1766 // In the dictionary. Is it an IMMEDIATE codeword?
1767 mov %eax,%edi // %edi = dictionary entry
1768 movb 4(%edi),%al // Get name+flags.
1769 push %ax // Just save it for now.
1770 call _TCFA // Convert dictionary entry (in %edi) to codeword pointer.
1772 andb $F_IMMED,%al // Is IMMED flag set?
1774 jnz 4f // If IMMED, jump straight to executing.
1778 1: // Not in the dictionary (not a word) so assume it's a literal number.
1779 incl interpret_is_lit
1780 call _SNUMBER // Returns the parsed number in %eax
1782 mov $LIT,%eax // The word is LIT
1784 2: // Are we compiling or executing?
1787 jz 4f // Jump if executing.
1789 // Compiling - just append the word to the current dictionary definition.
1791 mov interpret_is_lit,%ecx // Was it a literal?
1794 mov %ebx,%eax // Yes, so LIT is followed by a number.
1798 4: // Executing - run it!
1799 mov interpret_is_lit,%ecx // Literal?
1800 test %ecx,%ecx // Literal?
1803 // Not a literal, execute it now. This never returns, but the codeword will
1804 // eventually call NEXT which will reenter the loop in INTERPRETER.
1807 5: // Executing a literal, which means push it on the stack.
1814 .int 0 // Flag used to record if reading a literal
1817 ODDS AND ENDS ----------------------------------------------------------------------
1819 CHAR puts the ASCII code of the first character of the following word on the stack. For example
1820 CHAR A puts 65 on the stack.
1822 SYSEXIT exits the process using Linux exit syscall.
1825 defcode "CHAR",4,,CHAR
1826 call _WORD // Returns %ecx = length, %edi = pointer to word.
1828 movb (%edi),%al // Get the first character of the word.
1829 push %eax // Push it onto the stack.
1832 // NB: SYSEXIT must be the last entry in the built-in dictionary.
1833 defcode SYSEXIT,7,,SYSEXIT
1839 START OF FORTH CODE ----------------------------------------------------------------------
1841 We've now reached the stage where the FORTH system is running and self-hosting. All further
1842 words can be written as FORTH itself, including words like IF, THEN, .", etc which in most
1843 languages would be considered rather fundamental.
1845 As a kind of trick, I prefill the input buffer with the initial FORTH code. Once this code
1846 has run (when we get to the "OK" prompt), this input buffer is reused for reading any further
1849 Some notes about the code:
1851 \ (backslash) is the FORTH way to start a comment which goes up to the next newline. However
1852 because this is a C-style string, I have to escape the backslash, which is why they appear as
1855 Similarly, any backslashes in the code are doubled, and " becomes \" (eg. the definition of ."
1856 is written as : .\" ... ;)
1858 I use indenting to show structure. The amount of whitespace has no meaning to FORTH however
1859 except that you must use at least one whitespace character between words, and words themselves
1860 cannot contain whitespace.
1862 FORTH is case-sensitive. Use capslock!
1870 // Multi-line constant gives 'Warning: unterminated string; newline inserted' messages which you can ignore.
1872 \\ Define some character constants
1878 \\ CR prints a carriage return
1881 \\ SPACE prints a space
1882 : SPACE 'SPACE' EMIT ;
1884 \\ Primitive . (DOT) function doesn't follow with a blank, so redefine it to behave like FORTH.
1885 \\ Notice how we can trivially redefine existing functions.
1888 \\ DUP, DROP are defined in assembly for speed, but this is how you might define them
1889 \\ in FORTH. Notice use of the scratch variables _X and _Y.
1890 \\ : DUP _X ! _X @ _X @ ;
1893 \\ The 2... versions of the standard operators work on pairs of stack entries. They're not used
1894 \\ very commonly so not really worth writing in assembler. Here is how they are defined in FORTH.
1898 \\ More standard FORTH words.
1902 \\ [ and ] allow you to break into immediate mode while compiling a word.
1903 : [ IMMEDIATE \\ define [ as an immediate word
1904 0 STATE ! \\ go into immediate mode
1908 1 STATE ! \\ go back to compile mode
1911 \\ LITERAL takes whatever is on the stack and compiles LIT <foo>
1913 ' LIT , \\ compile LIT
1914 , \\ compile the literal itself (from the stack)
1917 \\ condition IF true-part THEN rest
1919 \\ condition 0BRANCH OFFSET true-part rest
1920 \\ where OFFSET is the offset of 'rest'
1921 \\ condition IF true-part ELSE false-part THEN
1923 \\ condition 0BRANCH OFFSET true-part BRANCH OFFSET2 false-part rest
1924 \\ where OFFSET if the offset of false-part and OFFSET2 is the offset of rest
1926 \\ IF is an IMMEDIATE word which compiles 0BRANCH followed by a dummy offset, and places
1927 \\ the address of the 0BRANCH on the stack. Later when we see THEN, we pop that address
1928 \\ off the stack, calculate the offset, and back-fill the offset.
1930 ' 0BRANCH , \\ compile 0BRANCH
1931 HERE @ \\ save location of the offset on the stack
1932 0 , \\ compile a dummy offset
1937 HERE @ SWAP - \\ calculate the offset from the address saved on the stack
1938 SWAP ! \\ store the offset in the back-filled location
1942 ' BRANCH , \\ definite branch to just over the false-part
1943 HERE @ \\ save location of the offset on the stack
1944 0 , \\ compile a dummy offset
1945 SWAP \\ now back-fill the original (IF) offset
1946 DUP \\ same as for THEN word above
1951 \\ BEGIN loop-part condition UNTIL
1953 \\ loop-part condition 0BRANCH OFFSET
1954 \\ where OFFSET points back to the loop-part
1955 \\ This is like do { loop-part } while (condition) in the C language
1957 HERE @ \\ save location on the stack
1961 ' 0BRANCH , \\ compile 0BRANCH
1962 HERE @ - \\ calculate the offset from the address saved on the stack
1963 , \\ compile the offset here
1966 \\ BEGIN loop-part AGAIN
1968 \\ loop-part BRANCH OFFSET
1969 \\ where OFFSET points back to the loop-part
1970 \\ In other words, an infinite loop which can only be returned from with EXIT
1972 ' BRANCH , \\ compile BRANCH
1973 HERE @ - \\ calculate the offset back
1974 , \\ compile the offset here
1977 \\ BEGIN condition WHILE loop-part REPEAT
1979 \\ condition 0BRANCH OFFSET2 loop-part BRANCH OFFSET
1980 \\ where OFFSET points back to condition (the beginning) and OFFSET2 points to after the whole piece of code
1981 \\ So this is like a while (condition) { loop-part } loop in the C language
1983 ' 0BRANCH , \\ compile 0BRANCH
1984 HERE @ \\ save location of the offset2 on the stack
1985 0 , \\ compile a dummy offset2
1989 ' BRANCH , \\ compile BRANCH
1990 SWAP \\ get the original offset (from BEGIN)
1991 HERE @ - , \\ and compile it after BRANCH
1993 HERE @ SWAP - \\ calculate the offset2
1994 SWAP ! \\ and back-fill it in the original location
1997 \\ With the looping constructs, we can now write SPACES, which writes n spaces to stdout.
2000 SPACE \\ print a space
2001 1- \\ until we count down to 0
2006 \\ .S prints the contents of the stack. Very useful for debugging.
2008 DSP@ \\ get current stack pointer
2010 DUP @ . \\ print the stack element
2012 DUP S0 @ 4- = \\ stop when we get to the top
2017 \\ DEPTH returns the depth of the stack.
2018 : DEPTH S0 @ DSP@ - ;
2020 \\ .\" is the print string operator in FORTH. Example: .\" Something to print\"
2021 \\ The space after the operator is the ordinary space required between words.
2022 \\ This is tricky to define because it has to do different things depending on whether
2023 \\ we are compiling or in immediate mode. (Thus the word is marked IMMEDIATE so it can
2024 \\ detect this and do different things).
2025 \\ In immediate mode we just keep reading characters and printing them until we get to
2026 \\ the next double quote.
2027 \\ In compile mode we have the problem of where we're going to store the string (remember
2028 \\ that the input buffer where the string comes from may be overwritten by the time we
2029 \\ come round to running the function). We store the string in the compiled function
2031 \\ ..., LITSTRING, string length, string rounded up to 4 bytes, EMITSTRING, ...
2033 STATE @ \\ compiling?
2035 ' LITSTRING , \\ compile LITSTRING
2036 HERE @ \\ save the address of the length word on the stack
2037 0 , \\ dummy length - we don't know what it is yet
2039 KEY \\ get next character of the string
2042 HERE @ !b \\ store the character in the compiled image
2043 1 HERE +! \\ increment HERE pointer by 1 byte
2045 DROP \\ drop the double quote character at the end
2046 DUP \\ get the saved address of the length word
2047 HERE @ SWAP - \\ calculate the length
2048 4- \\ subtract 4 (because we measured from the start of the length word)
2049 SWAP ! \\ and back-fill the length location
2050 HERE @ \\ round up to next multiple of 4 bytes for the remaining code
2054 ' EMITSTRING , \\ compile the final EMITSTRING
2056 \\ In immediate mode, just read characters and print them until we get
2057 \\ to the ending double quote. Much simpler than the above code!
2060 DUP '\"' = IF EXIT THEN
2066 \\ While compiling, [COMPILE] WORD compiles WORD if it would otherwise be IMMEDIATE.
2067 : [COMPILE] IMMEDIATE
2068 WORD \\ get the next word
2069 FIND \\ find it in the dictionary
2070 >CFA \\ get its codeword
2071 , \\ and compile that
2074 \\ RECURSE makes a recursive call to the current word that is being compiled.
2075 \\ Normally while a word is being compiled, it is marked HIDDEN so that references to the
2076 \\ same word within are calls to the previous definition of the word.
2078 LATEST @ >CFA \\ LATEST points to the word being compiled at the moment
2082 \\ ALLOT is used to allocate (static) memory when compiling. It increases HERE by
2083 \\ the amount given on the stack.
2087 \\ Finally print the welcome prompt.
2088 .\" JONESFORTH VERSION \" VERSION @ . CR
2101 /* END OF jonesforth.S */