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 1.17 2007-09-08 22:10:43 rich Exp $
6 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
8 INTRODUCTION ----------------------------------------------------------------------
10 FORTH is one of those alien languages which most working programmers regard in the same
11 way as Haskell, LISP, and so on. Something so strange that they'd rather any thoughts
12 of it just go away so they can get on with writing this paying code. But that's wrong
13 and if you care at all about programming then you should at least understand all these
14 languages, even if you will never use them.
16 LISP is the ultimate high-level language, and features from LISP are being added every
17 decade to the more common languages. But FORTH is in some ways the ultimate in low level
18 programming. Out of the box it lacks features like dynamic memory management and even
19 strings. In fact, at its primitive level it lacks even basic concepts like IF-statements
22 Why then would you want to learn FORTH? There are several very good reasons. First
23 and foremost, FORTH is minimal. You really can write a complete FORTH in, say, 2000
24 lines of code. I don't just mean a FORTH program, I mean a complete FORTH operating
25 system, environment and language. You could boot such a FORTH on a bare PC and it would
26 come up with a prompt where you could start doing useful work. The FORTH you have here
27 isn't minimal and uses a Linux process as its 'base PC' (both for the purposes of making
28 it a good tutorial). It's possible to completely understand the system. Who can say they
29 completely understand how Linux works, or gcc?
31 Secondly FORTH has a peculiar bootstrapping property. By that I mean that after writing
32 a little bit of assembly to talk to the hardware and implement a few primitives, all the
33 rest of the language and compiler is written in FORTH itself. Remember I said before
34 that FORTH lacked IF-statements and loops? Well of course it doesn't really because
35 such a lanuage would be useless, but my point was rather that IF-statements and loops are
36 written in FORTH itself.
38 Now of course this is common in other languages as well, and in those languages we call
39 them 'libraries'. For example in C, 'printf' is a library function written in C. But
40 in FORTH this goes way beyond mere libraries. Can you imagine writing C's 'if' in C?
41 And that brings me to my third reason: If you can write 'if' in FORTH, then why restrict
42 yourself to the usual if/while/for/switch constructs? You want a construct that iterates
43 over every other element in a list of numbers? You can add it to the language. What
44 about an operator which pulls in variables directly from a configuration file and makes
45 them available as FORTH variables? Or how about adding Makefile-like dependencies to
46 the language? No problem in FORTH. This concept isn't common in programming languages,
47 but it has a name (in fact two names): "macros" (by which I mean LISP-style macros, not
48 the lame C preprocessor) and "domain specific languages" (DSLs).
50 This tutorial isn't about learning FORTH as the language. I'll point you to some references
51 you should read if you're not familiar with using FORTH. This tutorial is about how to
52 write FORTH. In fact, until you understand how FORTH is written, you'll have only a very
53 superficial understanding of how to use it.
55 So if you're not familiar with FORTH or want to refresh your memory here are some online
58 http://en.wikipedia.org/wiki/Forth_%28programming_language%29
60 http://galileo.phys.virginia.edu/classes/551.jvn.fall01/primer.htm
62 http://wiki.laptop.org/go/Forth_Lessons
64 Here is another "Why FORTH?" essay: http://www.jwdt.com/~paysan/why-forth.html
66 ACKNOWLEDGEMENTS ----------------------------------------------------------------------
68 This code draws heavily on the design of LINA FORTH (http://home.hccnet.nl/a.w.m.van.der.horst/lina.html)
69 by Albert van der Horst. Any similarities in the code are probably not accidental.
71 Also I used this document (http://ftp.funet.fi/pub/doc/IOCCC/1992/buzzard.2.design) which really
72 defies easy explanation.
74 PUBLIC DOMAIN ----------------------------------------------------------------------
76 I, the copyright holder of this work, hereby release it into the public domain. This applies worldwide.
78 In case this is not legally possible, I grant any entity the right to use this work for any purpose,
79 without any conditions, unless such conditions are required by law.
81 SETTING UP ----------------------------------------------------------------------
83 Let's get a few housekeeping things out of the way. Firstly because I need to draw lots of
84 ASCII-art diagrams to explain concepts, the best way to look at this is using a window which
85 uses a fixed width font and is at least this wide:
87 <------------------------------------------------------------------------------------------------------------------------>
89 Secondly make sure TABS are set to 8 characters. The following should be a vertical
90 line. If not, sort out your tabs.
96 Thirdly I assume that your screen is at least 50 characters high.
98 ASSEMBLING ----------------------------------------------------------------------
100 If you want to actually run this FORTH, rather than just read it, you will need Linux on an
101 i386. Linux because instead of programming directly to the hardware on a bare PC which I
102 could have done, I went for a simpler tutorial by assuming that the 'hardware' is a Linux
103 process with a few basic system calls (read, write and exit and that's about all). i386
104 is needed because I had to write the assembly for a processor, and i386 is by far the most
105 common. (Of course when I say 'i386', any 32- or 64-bit x86 processor will do. I'm compiling
106 this on a 64 bit AMD Opteron).
108 Again, to assemble this you will need gcc and gas (the GNU assembler). The commands to
109 assemble and run the code (save this file as 'jonesforth.S') are:
111 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
114 You will see lots of 'Warning: unterminated string; newline inserted' messages from the
115 assembler. That's just because the GNU assembler doesn't have a good syntax for multi-line
116 strings (or rather it used to, but the developers removed it!) so I've abused the syntax
117 slightly to make things readable. Ignore these warnings.
119 If you want to run your own FORTH programs you can do:
121 ./jonesforth < myprog.f
123 If you want to load your own FORTH code and then continue reading user commands, you can do:
125 cat myfunctions.f - | ./jonesforth
127 ASSEMBLER ----------------------------------------------------------------------
129 (You can just skip to the next section -- you don't need to be able to read assembler to
130 follow this tutorial).
132 However if you do want to read the assembly code here are a few notes about gas (the GNU assembler):
134 (1) Register names are prefixed with '%', so %eax is the 32 bit i386 accumulator. The registers
135 available on i386 are: %eax, %ebx, %ecx, %edx, %esi, %edi, %ebp and %esp, and most of them
136 have special purposes.
138 (2) Add, mov, etc. take arguments in the form SRC,DEST. So mov %eax,%ecx moves %eax -> %ecx
140 (3) Constants are prefixed with '$', and you mustn't forget it! If you forget it then it
141 causes a read from memory instead, so:
142 mov $2,%eax moves number 2 into %eax
143 mov 2,%eax reads the 32 bit word from address 2 into %eax (ie. most likely a mistake)
145 (4) gas has a funky syntax for local labels, where '1f' (etc.) means label '1:' "forwards"
146 and '1b' (etc.) means label '1:' "backwards".
148 (5) 'ja' is "jump if above", 'jb' for "jump if below", 'je' "jump if equal" etc.
150 (6) gas has a reasonably nice .macro syntax, and I use them a lot to make the code shorter and
153 For more help reading the assembler, do "info gas" at the Linux prompt.
155 Now the tutorial starts in earnest.
157 THE DICTIONARY ----------------------------------------------------------------------
159 In FORTH as you will know, functions are called "words", and just as in other languages they
160 have a name and a definition. Here are two FORTH words:
162 : DOUBLE DUP + ; \ name is "DOUBLE", definition is "DUP +"
163 : QUADRUPLE DOUBLE DOUBLE ; \ name is "QUADRUPLE", definition is "DOUBLE DOUBLE"
165 Words, both built-in ones and ones which the programmer defines later, are stored in a dictionary
166 which is just a linked list of dictionary entries.
168 <--- DICTIONARY ENTRY (HEADER) ----------------------->
169 +------------------------+--------+---------- - - - - +----------- - - - -
170 | LINK POINTER | LENGTH/| NAME | DEFINITION
172 +--- (4 bytes) ----------+- byte -+- n bytes - - - - +----------- - - - -
174 I'll come to the definition of the word later. For now just look at the header. The first
175 4 bytes are the link pointer. This points back to the previous word in the dictionary, or, for
176 the first word in the dictionary it is just a NULL pointer. Then comes a length/flags byte.
177 The length of the word can be up to 31 characters (5 bits used) and the top three bits are used
178 for various flags which I'll come to later. This is followed by the name itself, and in this
179 implementation the name is rounded up to a multiple of 4 bytes by padding it with zero bytes.
180 That's just to ensure that the definition starts on a 32 bit boundary.
182 A FORTH variable called LATEST contains a pointer to the most recently defined word, in
183 other words, the head of this linked list.
185 DOUBLE and QUADRUPLE might look like this:
187 pointer to previous word
190 +--|------+---+---+---+---+---+---+---+---+------------- - - - -
191 | LINK | 6 | D | O | U | B | L | E | 0 | (definition ...)
192 +---------+---+---+---+---+---+---+---+---+------------- - - - -
195 +--|------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
196 | LINK | 9 | Q | U | A | D | R | U | P | L | E | 0 | 0 | (definition ...)
197 +---------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
203 You shoud be able to see from this how you might implement functions to find a word in
204 the dictionary (just walk along the dictionary entries starting at LATEST and matching
205 the names until you either find a match or hit the NULL pointer at the end of the dictionary);
206 and add a word to the dictionary (create a new definition, set its LINK to LATEST, and set
207 LATEST to point to the new word). We'll see precisely these functions implemented in
208 assembly code later on.
210 One interesting consequence of using a linked list is that you can redefine words, and
211 a newer definition of a word overrides an older one. This is an important concept in
212 FORTH because it means that any word (even "built-in" or "standard" words) can be
213 overridden with a new definition, either to enhance it, to make it faster or even to
214 disable it. However because of the way that FORTH words get compiled, which you'll
215 understand below, words defined using the old definition of a word continue to use
216 the old definition. Only words defined after the new definition use the new definition.
218 DIRECT THREADED CODE ----------------------------------------------------------------------
220 Now we'll get to the really crucial bit in understanding FORTH, so go and get a cup of tea
221 or coffee and settle down. It's fair to say that if you don't understand this section, then you
222 won't "get" how FORTH works, and that would be a failure on my part for not explaining it well.
223 So if after reading this section a few times you don't understand it, please email me
226 Let's talk first about what "threaded code" means. Imagine a peculiar version of C where
227 you are only allowed to call functions without arguments. (Don't worry for now that such a
228 language would be completely useless!) So in our peculiar C, code would look like this:
237 and so on. How would a function, say 'f' above, be compiled by a standard C compiler?
238 Probably into assembly code like this. On the right hand side I've written the actual
242 CALL a E8 08 00 00 00
243 CALL b E8 1C 00 00 00
244 CALL c E8 2C 00 00 00
245 ; ignore the return from the function for now
247 "E8" is the x86 machine code to "CALL" a function. In the first 20 years of computing
248 memory was hideously expensive and we might have worried about the wasted space being used
249 by the repeated "E8" bytes. We can save 20% in code size (and therefore, in expensive memory)
250 by compressing this into just:
252 08 00 00 00 Just the function addresses, without
253 1C 00 00 00 the CALL prefix.
256 [Historical note: If the execution model that FORTH uses looks strange from the following
257 paragraphs, then it was motivated entirely by the need to save memory on early computers.
258 This code compression isn't so important now when our machines have more memory in their L1
259 caches than those early computers had in total, but the execution model still has some
262 Of course this code won't run directly any more. Instead we need to write an interpreter
263 which takes each pair of bytes and calls it.
265 On an i386 machine it turns out that we can write this interpreter rather easily, in just
266 two assembly instructions which turn into just 3 bytes of machine code. Let's store the
267 pointer to the next word to execute in the %esi register:
269 08 00 00 00 <- We're executing this one now. %esi is the _next_ one to execute.
273 The all-important i386 instruction is called LODSL (or in Intel manuals, LODSW). It does
274 two things. Firstly it reads the memory at %esi into the accumulator (%eax). Secondly it
275 increments %esi by 4 bytes. So after LODSL, the situation now looks like this:
277 08 00 00 00 <- We're still executing this one
278 1C 00 00 00 <- %eax now contains this address (0x0000001C)
281 Now we just need to jump to the address in %eax. This is again just a single x86 instruction
282 written JMP *(%eax). And after doing the jump, the situation looks like:
285 1C 00 00 00 <- Now we're executing this subroutine.
288 To make this work, each subroutine is followed by the two instructions 'LODSL; JMP *(%eax)'
289 which literally make the jump to the next subroutine.
291 And that brings us to our first piece of actual code! Well, it's a macro.
300 /* The macro is called NEXT. That's a FORTH-ism. It expands to those two instructions.
302 Every FORTH primitive that we write has to be ended by NEXT. Think of it kind of like
305 The above describes what is known as direct threaded code.
307 To sum up: We compress our function calls down to a list of addresses and use a somewhat
308 magical macro to act as a "jump to next function in the list". We also use one register (%esi)
309 to act as a kind of instruction pointer, pointing to the next function in the list.
311 I'll just give you a hint of what is to come by saying that a FORTH definition such as:
313 : QUADRUPLE DOUBLE DOUBLE ;
315 actually compiles (almost, not precisely but we'll see why in a moment) to a list of
316 function addresses for DOUBLE, DOUBLE and a special function called EXIT to finish off.
318 At this point, REALLY EAGLE-EYED ASSEMBLY EXPERTS are saying "JONES, YOU'VE MADE A MISTAKE!".
320 I lied about JMP *(%eax).
322 INDIRECT THREADED CODE ----------------------------------------------------------------------
324 It turns out that direct threaded code is interesting but only if you want to just execute
325 a list of functions written in assembly language. So QUADRUPLE would work only if DOUBLE
326 was an assembly language function. In the direct threaded code, QUADRUPLE would look like:
329 | addr of DOUBLE --------------------> (assembly code to do the double)
330 +------------------+ NEXT
331 %esi -> | addr of DOUBLE |
334 We can add an extra indirection to allow us to run both words written in assembly language
335 (primitives written for speed) and words written in FORTH themselves as lists of addresses.
337 The extra indirection is the reason for the brackets in JMP *(%eax).
339 Let's have a look at how QUADRUPLE and DOUBLE really look in FORTH:
341 : QUADRUPLE DOUBLE DOUBLE ;
344 | codeword | : DOUBLE DUP + ;
346 | addr of DOUBLE ---------------> +------------------+
347 +------------------+ | codeword |
348 | addr of DOUBLE | +------------------+
349 +------------------+ | addr of DUP --------------> +------------------+
350 | addr of EXIT | +------------------+ | codeword -------+
351 +------------------+ %esi -> | addr of + --------+ +------------------+ |
352 +------------------+ | | assembly to <-----+
353 | addr of EXIT | | | implement DUP |
354 +------------------+ | | .. |
357 | +------------------+
359 +-----> +------------------+
361 +------------------+ |
362 | assembly to <------+
369 This is the part where you may need an extra cup of tea/coffee/favourite caffeinated
370 beverage. What has changed is that I've added an extra pointer to the beginning of
371 the definitions. In FORTH this is sometimes called the "codeword". The codeword is
372 a pointer to the interpreter to run the function. For primitives written in
373 assembly language, the "interpreter" just points to the actual assembly code itself.
374 They don't need interpreting, they just run.
376 In words written in FORTH (like QUADRUPLE and DOUBLE), the codeword points to an interpreter
379 I'll show you the interpreter function shortly, but let's recall our indirect
380 JMP *(%eax) with the "extra" brackets. Take the case where we're executing DOUBLE
381 as shown, and DUP has been called. Note that %esi is pointing to the address of +
383 The assembly code for DUP eventually does a NEXT. That:
385 (1) reads the address of + into %eax %eax points to the codeword of +
386 (2) increments %esi by 4
387 (3) jumps to the indirect %eax jumps to the address in the codeword of +,
388 ie. the assembly code to implement +
393 | addr of DOUBLE ---------------> +------------------+
394 +------------------+ | codeword |
395 | addr of DOUBLE | +------------------+
396 +------------------+ | addr of DUP --------------> +------------------+
397 | addr of EXIT | +------------------+ | codeword -------+
398 +------------------+ | addr of + --------+ +------------------+ |
399 +------------------+ | | assembly to <-----+
400 %esi -> | addr of EXIT | | | implement DUP |
401 +------------------+ | | .. |
404 | +------------------+
406 +-----> +------------------+
408 +------------------+ |
409 now we're | assembly to <------+
410 executing | implement + |
416 So I hope that I've convinced you that NEXT does roughly what you'd expect. This is
417 indirect threaded code.
419 I've glossed over four things. I wonder if you can guess without reading on what they are?
425 My list of four things are: (1) What does "EXIT" do? (2) which is related to (1) is how do
426 you call into a function, ie. how does %esi start off pointing at part of QUADRUPLE, but
427 then point at part of DOUBLE. (3) What goes in the codeword for the words which are written
428 in FORTH? (4) How do you compile a function which does anything except call other functions
429 ie. a function which contains a number like : DOUBLE 2 * ; ?
431 THE INTERPRETER AND RETURN STACK ------------------------------------------------------------
433 Going at these in no particular order, let's talk about issues (3) and (2), the interpreter
434 and the return stack.
436 Words which are defined in FORTH need a codeword which points to a little bit of code to
437 give them a "helping hand" in life. They don't need much, but they do need what is known
438 as an "interpreter", although it doesn't really "interpret" in the same way that, say,
439 Java bytecode used to be interpreted (ie. slowly). This interpreter just sets up a few
440 machine registers so that the word can then execute at full speed using the indirect
441 threaded model above.
443 One of the things that needs to happen when QUADRUPLE calls DOUBLE is that we save the old
444 %esi ("instruction pointer") and create a new one pointing to the first word in DOUBLE.
445 Because we will need to restore the old %esi at the end of DOUBLE (this is, after all, like
446 a function call), we will need a stack to store these "return addresses" (old values of %esi).
448 As you will have read, when reading the background documentation, FORTH has two stacks,
449 an ordinary stack for parameters, and a return stack which is a bit more mysterious. But
450 our return stack is just the stack I talked about in the previous paragraph, used to save
451 %esi when calling from a FORTH word into another FORTH word.
453 In this FORTH, we are using the normal stack pointer (%esp) for the parameter stack.
454 We will use the i386's "other" stack pointer (%ebp, usually called the "frame pointer")
455 for our return stack.
457 I've got two macros which just wrap up the details of using %ebp for the return stack.
458 You use them as for example "PUSHRSP %eax" (push %eax on the return stack) or "POPRSP %ebx"
459 (pop top of return stack into %ebx).
462 /* Macros to deal with the return stack. */
464 lea -4(%ebp),%ebp // push reg on to return stack
469 mov (%ebp),\reg // pop top of return stack to reg
474 And with that we can now talk about the interpreter.
476 In FORTH the interpreter function is often called DOCOL (I think it means "DO COLON" because
477 all FORTH definitions start with a colon, as in : DOUBLE DUP + ;
479 The "interpreter" (it's not really "interpreting") just needs to push the old %esi on the
480 stack and set %esi to the first word in the definition. Remember that we jumped to the
481 function using JMP *(%eax)? Well a consequence of that is that conveniently %eax contains
482 the address of this codeword, so just by adding 4 to it we get the address of the first
483 data word. Finally after setting up %esi, it just does NEXT which causes that first word
487 /* DOCOL - the interpreter! */
491 PUSHRSP %esi // push %esi on to the return stack
492 addl $4,%eax // %eax points to codeword, so make
493 movl %eax,%esi // %esi point to first data word
497 Just to make this absolutely clear, let's see how DOCOL works when jumping from QUADRUPLE
503 +------------------+ DOUBLE:
504 | addr of DOUBLE ---------------> +------------------+
505 +------------------+ %eax -> | addr of DOCOL |
506 %esi -> | addr of DOUBLE | +------------------+
507 +------------------+ | addr of DUP |
508 | addr of EXIT | +------------------+
509 +------------------+ | etc. |
511 First, the call to DOUBLE calls DOCOL (the codeword of DOUBLE). DOCOL does this: It
512 pushes the old %esi on the return stack. %eax points to the codeword of DOUBLE, so we
513 just add 4 on to it to get our new %esi:
518 +------------------+ DOUBLE:
519 | addr of DOUBLE ---------------> +------------------+
520 top of return +------------------+ %eax -> | addr of DOCOL |
521 stack points -> | addr of DOUBLE | + 4 = +------------------+
522 +------------------+ %esi -> | addr of DUP |
523 | addr of EXIT | +------------------+
524 +------------------+ | etc. |
526 Then we do NEXT, and because of the magic of threaded code that increments %esi again
529 Well, it seems to work.
531 One minor point here. Because DOCOL is the first bit of assembly actually to be defined
532 in this file (the others were just macros), and because I usually compile this code with the
533 text segment starting at address 0, DOCOL has address 0. So if you are disassembling the
534 code and see a word with a codeword of 0, you will immediately know that the word is
535 written in FORTH (it's not an assembler primitive) and so uses DOCOL as the interpreter.
537 STARTING UP ----------------------------------------------------------------------
539 Now let's get down to nuts and bolts. When we start the program we need to set up
540 a few things like the return stack. But as soon as we can, we want to jump into FORTH
541 code (albeit much of the "early" FORTH code will still need to be written as
542 assembly language primitives).
544 This is what the set up code does. Does a tiny bit of house-keeping, sets up the
545 separate return stack (NB: Linux gives us the ordinary parameter stack already), then
546 immediately jumps to a FORTH word called COLD. COLD stands for cold-start. In ISO
547 FORTH (but not in this FORTH), COLD can be called at any time to completely reset
548 the state of FORTH, and there is another word called WARM which does a partial reset.
551 /* ELF entry point. */
556 mov %esp,var_S0 // Store the initial data stack pointer.
557 mov $return_stack,%ebp // Initialise the return stack.
559 mov $cold_start,%esi // Initialise interpreter.
560 NEXT // Run interpreter!
563 cold_start: // High-level code without a codeword.
567 We also allocate some space for the return stack and some space to store user
568 definitions. These are static memory allocations using fixed-size buffers, but it
569 wouldn't be a great deal of work to make them dynamic.
573 /* FORTH return stack. */
574 #define RETURN_STACK_SIZE 8192
576 .space RETURN_STACK_SIZE
577 return_stack: // Initial top of return stack.
579 /* Space for user-defined words. */
580 #define USER_DEFS_SIZE 16384
583 .space USER_DEFS_SIZE
586 BUILT-IN WORDS ----------------------------------------------------------------------
588 Remember our dictionary entries (headers). Let's bring those together with the codeword
589 and data words to see how : DOUBLE DUP + ; really looks in memory.
591 pointer to previous word
594 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
595 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
596 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
599 LINK in next word points to codeword of DUP
601 Initially we can't just write ": DOUBLE DUP + ;" (ie. that literal string) here because we
602 don't yet have anything to read the string, break it up at spaces, parse each word, etc. etc.
603 So instead we will have to define built-in words using the GNU assembler data constructors
604 (like .int, .byte, .string, .ascii and so on -- look them up in the gas info page if you are
607 The long way would be:
608 .int <link to previous word>
610 .ascii "DOUBLE" // string
612 DOUBLE: .int DOCOL // codeword
613 .int DUP // pointer to codeword of DUP
614 .int PLUS // pointer to codeword of +
615 .int EXIT // pointer to codeword of EXIT
617 That's going to get quite tedious rather quickly, so here I define an assembler macro
618 so that I can just write:
620 defword "DOUBLE",6,,DOUBLE
623 and I'll get exactly the same effect.
625 Don't worry too much about the exact implementation details of this macro - it's complicated!
628 /* Flags - these are discussed later. */
630 #define F_HIDDEN 0x20
632 // Store the chain of links.
635 .macro defword name, namelen, flags=0, label
641 .set link,name_\label
642 .byte \flags+\namelen // flags + length byte
643 .ascii "\name" // the name
647 .int DOCOL // codeword - the interpreter
648 // list of word pointers follow
652 Similarly I want a way to write words written in assembly language. There will quite a few
653 of these to start with because, well, everything has to start in assembly before there's
654 enough "infrastructure" to be able to start writing FORTH words, but also I want to define
655 some common FORTH words in assembly language for speed, even though I could write them in FORTH.
657 This is what DUP looks like in memory:
659 pointer to previous word
662 +--|------+---+---+---+---+------------+
663 | LINK | 3 | D | U | P | code_DUP ---------------------> points to the assembly
664 +---------+---+---+---+---+------------+ code used to write DUP,
665 ^ len codeword which ends with NEXT.
669 Again, for brevity in writing the header I'm going to write an assembler macro called defcode.
672 .macro defcode name, namelen, flags=0, label
678 .set link,name_\label
679 .byte \flags+\namelen // flags + length byte
680 .ascii "\name" // the name
684 .int code_\label // codeword
688 code_\label : // assembler code follows
692 Now some easy FORTH primitives. These are written in assembly for speed. If you understand
693 i386 assembly language then it is worth reading these. However if you don't understand assembly
694 you can skip the details.
698 pop %eax // duplicate top of stack
703 defcode "DROP",4,,DROP
704 pop %eax // drop top of stack
707 defcode "SWAP",4,,SWAP
708 pop %eax // swap top of stack
714 defcode "OVER",4,,OVER
715 mov 4(%esp),%eax // get the second element of stack
716 push %eax // and push it on top
728 defcode "-ROT",4,,NROT
738 incl (%esp) // increment top of stack
742 decl (%esp) // decrement top of stack
745 defcode "4+",2,,INCR4
746 addl $4,(%esp) // add 4 to top of stack
749 defcode "4-",2,,DECR4
750 subl $4,(%esp) // subtract 4 from top of stack
754 pop %eax // get top of stack
755 addl %eax,(%esp) // and add it to next word on stack
759 pop %eax // get top of stack
760 subl %eax,(%esp) // and subtract it from next word on stack
767 push %eax // ignore overflow
775 push %eax // push quotient
783 push %edx // push remainder
786 defcode "=",1,,EQU // top two words are equal?
796 defcode "<>",2,,NEQU // top two words are not equal?
806 defcode "0=",2,,ZEQU // top of stack equals 0?
825 defcode "INVERT",6,,INVERT // this is the FORTH "NOT" function
830 RETURNING FROM FORTH WORDS ----------------------------------------------------------------------
832 Time to talk about what happens when we EXIT a function. In this diagram QUADRUPLE has called
833 DOUBLE, and DOUBLE is about to exit (look at where %esi is pointing):
838 +------------------+ DOUBLE
839 | addr of DOUBLE ---------------> +------------------+
840 +------------------+ | codeword |
841 | addr of DOUBLE | +------------------+
842 +------------------+ | addr of DUP |
843 | addr of EXIT | +------------------+
844 +------------------+ | addr of + |
846 %esi -> | addr of EXIT |
849 What happens when the + function does NEXT? Well, the following code is executed.
852 defcode "EXIT",4,,EXIT
853 POPRSP %esi // pop return stack into %esi
857 EXIT gets the old %esi which we saved from before on the return stack, and puts it in %esi.
858 So after this (but just before NEXT) we get:
863 +------------------+ DOUBLE
864 | addr of DOUBLE ---------------> +------------------+
865 +------------------+ | codeword |
866 %esi -> | addr of DOUBLE | +------------------+
867 +------------------+ | addr of DUP |
868 | addr of EXIT | +------------------+
869 +------------------+ | addr of + |
874 And NEXT just completes the job by, well in this case just by calling DOUBLE again :-)
876 LITERALS ----------------------------------------------------------------------
878 The final point I "glossed over" before was how to deal with functions that do anything
879 apart from calling other functions. For example, suppose that DOUBLE was defined like this:
883 It does the same thing, but how do we compile it since it contains the literal 2? One way
884 would be to have a function called "2" (which you'd have to write in assembler), but you'd need
885 a function for every single literal that you wanted to use.
887 FORTH solves this by compiling the function using a special word called LIT:
889 +---------------------------+-------+-------+-------+-------+-------+
890 | (usual header of DOUBLE) | DOCOL | LIT | 2 | * | EXIT |
891 +---------------------------+-------+-------+-------+-------+-------+
893 LIT is executed in the normal way, but what it does next is definitely not normal. It
894 looks at %esi (which now points to the literal 2), grabs it, pushes it on the stack, then
895 manipulates %esi in order to skip the literal as if it had never been there.
897 What's neat is that the whole grab/manipulate can be done using a single byte single
898 i386 instruction, our old friend LODSL. Rather than me drawing more ASCII-art diagrams,
899 see if you can find out how LIT works:
903 // %esi points to the next command, but in this case it points to the next
904 // literal 32 bit integer. Get that literal into %eax and increment %esi.
905 // On x86, it's a convenient single byte instruction! (cf. NEXT macro)
907 push %eax // push the literal number on to stack
911 MEMORY ----------------------------------------------------------------------
913 As important point about FORTH is that it gives you direct access to the lowest levels
914 of the machine. Manipulating memory directly is done frequently in FORTH, and these are
915 the primitive words for doing it.
919 pop %ebx // address to store at
920 pop %eax // data to store there
921 mov %eax,(%ebx) // store it
925 pop %ebx // address to fetch
926 mov (%ebx),%eax // fetch it
927 push %eax // push value onto stack
930 defcode "+!",2,,ADDSTORE
932 pop %eax // the amount to add
933 addl %eax,(%ebx) // add it
936 defcode "-!",2,,SUBSTORE
938 pop %eax // the amount to subtract
939 subl %eax,(%ebx) // add it
942 /* ! and @ (STORE and FETCH) store 32-bit words. It's also useful to be able to read and write bytes.
943 * I don't know whether FORTH has these words, so I invented my own, called !b and @b.
944 * Byte-oriented operations only work on architectures which permit them (i386 is one of those).
945 * UPDATE: writing a byte to the dictionary pointer is called C, in FORTH.
947 defcode "!b",2,,STOREBYTE
948 pop %ebx // address to store at
949 pop %eax // data to store there
950 movb %al,(%ebx) // store it
953 defcode "@b",2,,FETCHBYTE
954 pop %ebx // address to fetch
956 movb (%ebx),%al // fetch it
957 push %eax // push value onto stack
961 BUILT-IN VARIABLES ----------------------------------------------------------------------
963 These are some built-in variables and related standard FORTH words. Of these, the only one that we
964 have discussed so far was LATEST, which points to the last (most recently defined) word in the
965 FORTH dictionary. LATEST is also a FORTH word which pushes the address of LATEST (the variable)
966 on to the stack, so you can read or write it using @ and ! operators. For example, to print
967 the current value of LATEST (and this can apply to any FORTH variable) you would do:
971 To make defining variables shorter, I'm using a macro called defvar, similar to defword and
972 defcode above. (In fact the defvar macro uses defcode to do the dictionary header).
975 .macro defvar name, namelen, flags=0, label, initial=0
976 defcode \name,\namelen,\flags,\label
986 The built-in variables are:
988 STATE Is the interpreter executing code (0) or compiling a word (non-zero)?
989 LATEST Points to the latest (most recently defined) word in the dictionary.
990 HERE Points to the next free byte of memory. When compiling, compiled words go here.
991 _X These are three scratch variables, used by some standard dictionary words.
994 S0 Stores the address of the top of the parameter stack.
995 R0 Stores the address of the top of the return stack.
998 defvar "STATE",5,,STATE
999 defvar "HERE",4,,HERE,user_defs_start
1000 defvar "LATEST",6,,LATEST,name_SYSEXIT // SYSEXIT must be last in built-in dictionary
1005 defvar "R0",2,,RZ,return_stack
1008 RETURN STACK ----------------------------------------------------------------------
1010 These words allow you to access the return stack. Recall that the register %ebp always points to
1011 the top of the return stack.
1015 pop %eax // pop parameter stack into %eax
1016 PUSHRSP %eax // push it on to the return stack
1019 defcode "R>",2,,FROMR
1020 POPRSP %eax // pop return stack on to %eax
1021 push %eax // and push on to parameter stack
1024 defcode "RSP@",4,,RSPFETCH
1028 defcode "RSP!",4,,RSPSTORE
1032 defcode "RDROP",5,,RDROP
1033 lea 4(%ebp),%ebp // pop return stack and throw away
1037 PARAMETER (DATA) STACK ----------------------------------------------------------------------
1039 These functions allow you to manipulate the parameter stack. Recall that Linux sets up the parameter
1040 stack for us, and it is accessed through %esp.
1043 defcode "DSP@",4,,DSPFETCH
1048 defcode "DSP!",4,,DSPSTORE
1053 INPUT AND OUTPUT ----------------------------------------------------------------------
1055 These are our first really meaty/complicated FORTH primitives. I have chosen to write them in
1056 assembler, but surprisingly in "real" FORTH implementations these are often written in terms
1057 of more fundamental FORTH primitives. I chose to avoid that because I think that just obscures
1058 the implementation. After all, you may not understand assembler but you can just think of it
1059 as an opaque block of code that does what it says.
1061 Let's discuss input first.
1063 The FORTH word KEY reads the next byte from stdin (and pushes it on the parameter stack).
1064 So if KEY is called and someone hits the space key, then the number 32 (ASCII code of space)
1065 is pushed on the stack.
1067 In FORTH there is no distinction between reading code and reading input. We might be reading
1068 and compiling code, we might be reading words to execute, we might be asking for the user
1069 to type their name -- ultimately it all comes in through KEY.
1071 The implementation of KEY uses an input buffer of a certain size (defined at the end of the
1072 program). It calls the Linux read(2) system call to fill this buffer and tracks its position
1073 in the buffer using a couple of variables, and if it runs out of input buffer then it refills
1074 it automatically. The other thing that KEY does is if it detects that stdin has closed, it
1075 exits the program, which is why when you hit ^D the FORTH system cleanly exits.
1078 #include <asm-i386/unistd.h>
1080 defcode "KEY",3,,KEY
1082 push %eax // push return value on stack
1094 1: // out of input; use read(2) to fetch more input from stdin
1095 xor %ebx,%ebx // 1st param: stdin
1096 mov $buffer,%ecx // 2nd param: buffer
1098 mov $buffend-buffer,%edx // 3rd param: max length
1099 mov $__NR_read,%eax // syscall: read
1101 test %eax,%eax // If %eax <= 0, then exit.
1103 addl %eax,%ecx // buffer+%eax = bufftop
1107 2: // error or out of input: exit
1109 mov $__NR_exit,%eax // syscall: exit
1113 By contrast, output is much simpler. The FORTH word EMIT writes out a single byte to stdout.
1114 This implementation just uses the write system call. No attempt is made to buffer output, but
1115 it would be a good exercise to add it.
1118 defcode "EMIT",4,,EMIT
1123 mov $1,%ebx // 1st param: stdout
1125 // write needs the address of the byte to write
1127 mov $2f,%ecx // 2nd param: address
1129 mov $1,%edx // 3rd param: nbytes = 1
1131 mov $__NR_write,%eax // write syscall
1136 2: .space 1 // scratch used by EMIT
1139 Back to input, WORD is a FORTH word which reads the next full word of input.
1141 What it does in detail is that it first skips any blanks (spaces, tabs, newlines and so on).
1142 Then it calls KEY to read characters into an internal buffer until it hits a blank. Then it
1143 calculates the length of the word it read and returns the address and the length as
1144 two words on the stack (with address at the top).
1146 Notice that WORD has a single internal buffer which it overwrites each time (rather like
1147 a static C string). Also notice that WORD's internal buffer is just 32 bytes long and
1148 there is NO checking for overflow. 31 bytes happens to be the maximum length of a
1149 FORTH word that we support, and that is what WORD is used for: to read FORTH words when
1150 we are compiling and executing code. The returned strings are not NUL-terminated, so
1151 in some crazy-world you could define FORTH words containing ASCII NULs, although why
1152 you'd want to is a bit beyond me.
1154 WORD is not suitable for just reading strings (eg. user input) because of all the above
1155 peculiarities and limitations.
1157 Note that when executing, you'll see:
1159 which puts "FOO" and length 3 on the stack, but when compiling:
1161 is an error (or at least it doesn't do what you might expect). Later we'll talk about compiling
1162 and immediate mode, and you'll understand why.
1165 defcode "WORD",4,,WORD
1167 push %ecx // push length
1168 push %edi // push base address
1172 /* Search for first non-blank character. Also skip \ comments. */
1174 call _KEY // get next key, returned in %eax
1175 cmpb $'\\',%al // start of a comment?
1176 je 3f // if so, skip the comment
1178 jbe 1b // if so, keep looking
1180 /* Search for the end of the word, storing chars as we go. */
1181 mov $5f,%edi // pointer to return buffer
1183 stosb // add character to return buffer
1184 call _KEY // get next key, returned in %al
1185 cmpb $' ',%al // is blank?
1186 ja 2b // if not, keep looping
1188 /* Return the word (well, the static buffer) and length. */
1190 mov %edi,%ecx // return length of the word
1191 mov $5f,%edi // return address of the word
1194 /* Code to skip \ comments to end of the current line. */
1197 cmpb $'\n',%al // end of line yet?
1202 // A static buffer where WORD returns. Subsequent calls
1203 // overwrite this buffer. Maximum word length is 32 chars.
1207 . (also called DOT) prints the top of the stack as an integer. In real FORTH implementations
1208 it should print it in the current base, but this assembler version is simpler and can only
1211 Remember that you can override even built-in FORTH words easily, so if you want to write a
1212 more advanced DOT then you can do so easily at a later point, and probably in FORTH.
1216 pop %eax // Get the number to print into %eax
1217 call _DOT // Easier to do this recursively ...
1220 mov $10,%ecx // Base 10
1224 xor %edx,%edx // %edx:%eax / %ecx -> quotient %eax, remainder %edx
1239 Almost the opposite of DOT (but not quite), SNUMBER parses a numeric string such as one returned
1240 by WORD and pushes the number on the parameter stack.
1242 This function does absolutely no error checking, and in particular the length of the string
1243 must be >= 1 bytes, and should contain only digits 0-9. If it doesn't you'll get random results.
1245 This function is only used when reading literal numbers in code, and shouldn't really be used
1246 in user code at all.
1248 defcode "SNUMBER",7,,SNUMBER
1258 imull $10,%eax // %eax *= 10
1261 subb $'0',%bl // ASCII -> digit
1268 DICTIONARY LOOK UPS ----------------------------------------------------------------------
1270 We're building up to our prelude on how FORTH code is compiled, but first we need yet more infrastructure.
1272 The FORTH word FIND takes a string (a word as parsed by WORD -- see above) and looks it up in the
1273 dictionary. What it actually returns is the address of the dictionary header, if it finds it,
1276 So if DOUBLE is defined in the dictionary, then WORD DOUBLE FIND returns the following pointer:
1282 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1283 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1284 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1286 See also >CFA which takes a dictionary entry pointer and returns a pointer to the codeword.
1288 FIND doesn't find dictionary entries which are flagged as HIDDEN. See below for why.
1291 defcode "FIND",4,,FIND
1292 pop %edi // %edi = address
1293 pop %ecx // %ecx = length
1299 push %esi // Save %esi so we can use it in string comparison.
1301 // Now we start searching backwards through the dictionary for this word.
1302 mov var_LATEST,%edx // LATEST points to name header of the latest word in the dictionary
1304 test %edx,%edx // NULL pointer? (end of the linked list)
1307 // Compare the length expected and the length of the word.
1308 // Note that if the F_HIDDEN flag is set on the word, then by a bit of trickery
1309 // this won't pick the word (the length will appear to be wrong).
1311 movb 4(%edx),%al // %al = flags+length field
1312 andb $(F_HIDDEN|0x1f),%al // %al = name length
1313 cmpb %cl,%al // Length is the same?
1316 // Compare the strings in detail.
1317 push %ecx // Save the length
1318 push %edi // Save the address (repe cmpsb will move this pointer)
1319 lea 5(%edx),%esi // Dictionary string we are checking against.
1320 repe cmpsb // Compare the strings.
1323 jne 2f // Not the same.
1325 // The strings are the same - return the header pointer in %eax
1331 mov (%edx),%edx // Move back through the link field to the previous word
1332 jmp 1b // .. and loop.
1336 xor %eax,%eax // Return zero to indicate not found.
1340 FIND returns the dictionary pointer, but when compiling we need the codeword pointer (recall
1341 that FORTH definitions are compiled into lists of codeword pointers). The standard FORTH
1342 word >CFA turns a dictionary pointer into a codeword pointer.
1344 The example below shows the result of:
1346 WORD DOUBLE FIND >CFA
1348 FIND returns a pointer to this
1349 | >CFA converts it to a pointer to this
1352 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1353 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1354 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1358 Because names vary in length, this isn't just a simple increment.
1360 In this FORTH you cannot easily turn a codeword pointer back into a dictionary entry pointer, but
1361 that is not true in most FORTH implementations where they store a back pointer in the definition
1362 (with an obvious memory/complexity cost). The reason they do this is that it is useful to be
1363 able to go backwards (codeword -> dictionary entry) in order to decompile FORTH definitions.
1365 What does CFA stand for? My best guess is "Code Field Address".
1368 defcode ">CFA",4,,TCFA
1375 add $4,%edi // Skip link pointer.
1376 movb (%edi),%al // Load flags+len into %al.
1377 inc %edi // Skip flags+len byte.
1378 andb $0x1f,%al // Just the length, not the flags.
1379 add %eax,%edi // Skip the name.
1380 addl $3,%edi // The codeword is 4-byte aligned.
1385 COMPILING ----------------------------------------------------------------------
1387 Now we'll talk about how FORTH compiles words. Recall that a word definition looks like this:
1391 and we have to turn this into:
1393 pointer to previous word
1396 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1397 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1398 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
1399 ^ len pad codeword |
1401 LATEST points here points to codeword of DUP
1403 There are several problems to solve. Where to put the new word? How do we read words? How
1404 do we define the words : (COLON) and ; (SEMICOLON)?
1406 FORTH solves this rather elegantly and as you might expect in a very low-level way which
1407 allows you to change how the compiler works on your own code.
1409 FORTH has an INTERPRETER function (a true interpreter this time, not DOCOL) which runs in a
1410 loop, reading words (using WORD), looking them up (using FIND), turning them into codeword
1411 pointers (using >CFA) and deciding what to do with them.
1413 What it does depends on the mode of the interpreter (in variable STATE).
1415 When STATE is zero, the interpreter just runs each word as it looks them up. This is known as
1418 The interesting stuff happens when STATE is non-zero -- compiling mode. In this mode the
1419 interpreter appends the codeword pointer to user memory (the HERE variable points to the next
1420 free byte of user memory).
1422 So you may be able to see how we could define : (COLON). The general plan is:
1424 (1) Use WORD to read the name of the function being defined.
1426 (2) Construct the dictionary entry -- just the header part -- in user memory:
1428 pointer to previous word (from LATEST) +-- Afterwards, HERE points here, where
1429 ^ | the interpreter will start appending
1431 +--|------+---+---+---+---+---+---+---+---+------------+
1432 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL |
1433 +---------+---+---+---+---+---+---+---+---+------------+
1436 (3) Set LATEST to point to the newly defined word, ...
1438 (4) .. and most importantly leave HERE pointing just after the new codeword. This is where
1439 the interpreter will append codewords.
1441 (5) Set STATE to 1. This goes into compile mode so the interpreter starts appending codewords to
1442 our partially-formed header.
1444 After : has run, our input is here:
1449 Next byte returned by KEY
1451 so the interpreter (now it's in compile mode, so I guess it's really the compiler) reads DUP,
1452 gets its codeword pointer, and appends it:
1454 +-- HERE updated to point here.
1457 +---------+---+---+---+---+---+---+---+---+------------+------------+
1458 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP |
1459 +---------+---+---+---+---+---+---+---+---+------------+------------+
1462 Next we read +, get the codeword pointer, and append it:
1464 +-- HERE updated to point here.
1467 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1468 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + |
1469 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1472 The issue is what happens next. Obviously what we _don't_ want to happen is that we
1473 read ";" and compile it and go on compiling everything afterwards.
1475 At this point, FORTH uses a trick. Remember the length byte in the dictionary definition
1476 isn't just a plain length byte, but can also contain flags. One flag is called the
1477 IMMEDIATE flag (F_IMMED in this code). If a word in the dictionary is flagged as
1478 IMMEDIATE then the interpreter runs it immediately _even if it's in compile mode_.
1480 I hope I don't need to explain that ; (SEMICOLON) just such a word, flagged as IMMEDIATE.
1481 And all it does is append the codeword for EXIT on to the current definition and switch
1482 back to immediate mode (set STATE back to 0). Shortly we'll see the actual definition
1483 of ; and we'll see that it's really a very simple definition, declared IMMEDIATE.
1485 After the interpreter reads ; and executes it 'immediately', we get this:
1487 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1488 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1489 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1496 And that's it, job done, our new definition is compiled, and we're back in immediate mode
1497 just reading and executing words, perhaps including a call to test our new word DOUBLE.
1499 The only last wrinkle in this is that while our word was being compiled, it was in a
1500 half-finished state. We certainly wouldn't want DOUBLE to be called somehow during
1501 this time. There are several ways to stop this from happening, but in FORTH what we
1502 do is flag the word with the HIDDEN flag (F_HIDDEN in this code) just while it is
1503 being compiled. This prevents FIND from finding it, and thus in theory stops any
1504 chance of it being called.
1506 Compared to the description above, the actual definition of : (COLON) is comparatively simple:
1509 defcode ":",1,,COLON
1511 // Get the word and create a dictionary entry header for it.
1512 call _WORD // Returns %ecx = length, %edi = pointer to word.
1513 mov %edi,%ebx // %ebx = address of the word
1515 movl var_HERE,%edi // %edi is the address of the header
1516 movl var_LATEST,%eax // Get link pointer
1517 stosl // and store it in the header.
1519 mov %cl,%al // Get the length.
1520 orb $F_HIDDEN,%al // Set the HIDDEN flag on this entry.
1521 stosb // Store the length/flags byte.
1523 mov %ebx,%esi // %esi = word
1524 rep movsb // Copy the word
1526 addl $3,%edi // Align to next 4 byte boundary.
1529 movl $DOCOL,%eax // The codeword for user-created words is always DOCOL (the interpreter)
1532 // Header built, so now update LATEST and HERE.
1533 // We'll be compiling words and putting them HERE.
1535 movl %eax,var_LATEST
1538 // And go into compile mode by setting STATE to 1.
1543 , (COMMA) is a standard FORTH word which appends a 32 bit integer (normally a codeword
1544 pointer) to the user data area pointed to by HERE, and adds 4 to HERE.
1547 defcode ",",1,,COMMA
1548 pop %eax // Code pointer to store.
1552 movl var_HERE,%edi // HERE
1554 movl %edi,var_HERE // Update HERE (incremented)
1558 ; (SEMICOLON) is also elegantly simple. Notice the F_IMMED flag.
1561 defcode ";",1,F_IMMED,SEMICOLON
1562 movl $EXIT,%eax // EXIT is the final codeword in compiled words.
1563 call _COMMA // Store it.
1564 call _HIDDEN // Toggle the HIDDEN flag (unhides the new word).
1565 xor %eax,%eax // Set STATE to 0 (back to execute mode).
1570 EXTENDING THE COMPILER ----------------------------------------------------------------------
1572 Words flagged with IMMEDIATE (F_IMMED) aren't just for the FORTH compiler to use. You can define
1573 your own IMMEDIATE words too, and this is a crucial aspect when extending basic FORTH, because
1574 it allows you in effect to extend the compiler itself. Does gcc let you do that?
1576 Standard FORTH words like IF, WHILE, .", [ and so on are all written as extensions to the basic
1577 compiler, and are all IMMEDIATE words.
1579 The IMMEDIATE word toggles the F_IMMED (IMMEDIATE flag) on the most recently defined word,
1580 or on the current word if you call it in the middle of a definition.
1584 : MYIMMEDWORD IMMEDIATE
1588 but some FORTH programmers write this instead:
1594 The two usages are equivalent, to a first approximation.
1597 defcode "IMMEDIATE",9,F_IMMED,IMMEDIATE
1601 movl var_LATEST,%edi // LATEST word.
1602 addl $4,%edi // Point to name/flags byte.
1603 xorb $F_IMMED,(%edi) // Toggle the IMMED bit.
1607 HIDDEN toggles the other flag, F_HIDDEN, of the latest word. Note that words flagged
1608 as hidden are defined but cannot be called, so this is rarely used.
1611 defcode "HIDDEN",6,,HIDDEN
1615 movl var_LATEST,%edi // LATEST word.
1616 addl $4,%edi // Point to name/flags byte.
1617 xorb $F_HIDDEN,(%edi) // Toggle the HIDDEN bit.
1621 ' (TICK) is a standard FORTH word which returns the codeword pointer of the next word.
1623 The common usage is:
1627 which appends the codeword of FOO to the current word we are defining (this only works in compiled code).
1629 You tend to use ' in IMMEDIATE words. For example an alternate (and rather useless) way to define
1630 a literal 2 might be:
1633 ' LIT , \ Appends LIT to the currently-being-defined word
1634 2 , \ Appends the number 2 to the currently-being-defined word
1641 (If you don't understand how LIT2 works, then you should review the material about compiling words
1642 and immediate mode).
1644 This definition of ' uses a cheat which I copied from buzzard92. As a result it only works in
1645 compiled code. It is possible to write a version of ' based on WORD, FIND, >CFA which works in
1649 lodsl // Get the address of the next word and skip it.
1650 pushl %eax // Push it on the stack.
1654 BRANCHING ----------------------------------------------------------------------
1656 It turns out that all you need in order to define looping constructs, IF-statements, etc.
1659 BRANCH is an unconditional branch. 0BRANCH is a conditional branch (it only branches if the
1660 top of stack is zero).
1662 The diagra below shows how BRANCH works in some imaginary compiled word. When BRANCH executes,
1663 %esi starts by pointing to the offset field (compare to LIT above):
1665 +---------------------+-------+---- - - ---+------------+------------+---- - - - ----+------------+
1666 | (Dictionary header) | DOCOL | | BRANCH | offset | (skipped) | word |
1667 +---------------------+-------+---- - - ---+------------+-----|------+---- - - - ----+------------+
1670 | +-----------------------+
1671 %esi added to offset
1673 The offset is added to %esi to make the new %esi, and the result is that when NEXT runs, execution
1674 continues at the branch target. Negative offsets work as expected.
1676 0BRANCH is the same except the branch happens conditionally.
1678 Now standard FORTH words such as IF, THEN, ELSE, WHILE, REPEAT, etc. can be implemented entirely
1679 in FORTH. They are IMMEDIATE words which append various combinations of BRANCH or 0BRANCH
1680 into the word currently being compiled.
1682 As an example, code written like this:
1684 condition-code IF true-part THEN rest-code
1688 condition-code 0BRANCH OFFSET true-part rest-code
1694 defcode "BRANCH",6,,BRANCH
1695 add (%esi),%esi // add the offset to the instruction pointer
1698 defcode "0BRANCH",7,,ZBRANCH
1700 test %eax,%eax // top of stack is zero?
1701 jz code_BRANCH // if so, jump back to the branch function above
1702 lodsl // otherwise we need to skip the offset
1706 PRINTING STRINGS ----------------------------------------------------------------------
1708 LITSTRING and EMITSTRING are primitives used to implement the ." operator (which is
1709 written in FORTH). See the definition of that operator below.
1712 defcode "LITSTRING",9,,LITSTRING
1713 lodsl // get the length of the string
1714 push %eax // push it on the stack
1715 push %esi // push the address of the start of the string
1716 addl %eax,%esi // skip past the string
1717 addl $3,%esi // but round up to next 4 byte boundary
1721 defcode "EMITSTRING",10,,EMITSTRING
1722 mov $1,%ebx // 1st param: stdout
1723 pop %ecx // 2nd param: address of string
1724 pop %edx // 3rd param: length of string
1725 mov $__NR_write,%eax // write syscall
1730 COLD START AND INTERPRETER ----------------------------------------------------------------------
1732 COLD is the first FORTH function called, almost immediately after the FORTH system "boots".
1734 INTERPRETER is the FORTH interpreter ("toploop", "toplevel" or "REPL" might be a more accurate
1735 description -- see: http://en.wikipedia.org/wiki/REPL).
1739 // COLD must not return (ie. must not call EXIT).
1740 defword "COLD",4,,COLD
1741 .int INTERPRETER // call the interpreter loop (never returns)
1742 .int LIT,1,SYSEXIT // hmmm, but in case it does, exit(1).
1744 /* This interpreter is pretty simple, but remember that in FORTH you can always override
1745 * it later with a more powerful one!
1747 defword "INTERPRETER",11,,INTERPRETER
1748 .int INTERPRET,RDROP,INTERPRETER
1750 defcode "INTERPRET",9,,INTERPRET
1751 call _WORD // Returns %ecx = length, %edi = pointer to word.
1753 // Is it in the dictionary?
1755 movl %eax,interpret_is_lit // Not a literal number (not yet anyway ...)
1756 call _FIND // Returns %eax = pointer to header or 0 if not found.
1757 test %eax,%eax // Found?
1760 // In the dictionary. Is it an IMMEDIATE codeword?
1761 mov %eax,%edi // %edi = dictionary entry
1762 movb 4(%edi),%al // Get name+flags.
1763 push %ax // Just save it for now.
1764 call _TCFA // Convert dictionary entry (in %edi) to codeword pointer.
1766 andb $F_IMMED,%al // Is IMMED flag set?
1768 jnz 4f // If IMMED, jump straight to executing.
1772 1: // Not in the dictionary (not a word) so assume it's a literal number.
1773 incl interpret_is_lit
1774 call _SNUMBER // Returns the parsed number in %eax
1776 mov $LIT,%eax // The word is LIT
1778 2: // Are we compiling or executing?
1781 jz 4f // Jump if executing.
1783 // Compiling - just append the word to the current dictionary definition.
1785 mov interpret_is_lit,%ecx // Was it a literal?
1788 mov %ebx,%eax // Yes, so LIT is followed by a number.
1792 4: // Executing - run it!
1793 mov interpret_is_lit,%ecx // Literal?
1794 test %ecx,%ecx // Literal?
1797 // Not a literal, execute it now. This never returns, but the codeword will
1798 // eventually call NEXT which will reenter the loop in INTERPRETER.
1801 5: // Executing a literal, which means push it on the stack.
1808 .int 0 // Flag used to record if reading a literal
1811 ODDS AND ENDS ----------------------------------------------------------------------
1813 CHAR puts the ASCII code of the first character of the following word on the stack. For example
1814 CHAR A puts 65 on the stack.
1816 SYSEXIT pops the status off the stack and exits the process (using Linux exit syscall).
1819 defcode "CHAR",4,,CHAR
1820 call _WORD // Returns %ecx = length, %edi = pointer to word.
1822 movb (%edi),%al // Get the first character of the word.
1823 push %eax // Push it onto the stack.
1826 // NB: SYSEXIT must be the last entry in the built-in dictionary.
1827 defcode SYSEXIT,7,,SYSEXIT
1833 START OF FORTH CODE ----------------------------------------------------------------------
1835 We've now reached the stage where the FORTH system is running and self-hosting. All further
1836 words can be written as FORTH itself, including words like IF, THEN, .", etc which in most
1837 languages would be considered rather fundamental.
1839 As a kind of trick, I prefill the input buffer with the initial FORTH code. Once this code
1840 has run (when we get to the "OK" prompt), this input buffer is reused for reading any further
1843 Some notes about the code:
1845 \ (backslash) is the FORTH way to start a comment which goes up to the next newline. However
1846 because this is a C-style string, I have to escape the backslash, which is why they appear as
1849 Similarly, any backslashes in the code are doubled, and " becomes \" (eg. the definition of ."
1850 is written as : .\" ... ;)
1852 I use indenting to show structure. The amount of whitespace has no meaning to FORTH however
1853 except that you must use at least one whitespace character between words, and words themselves
1854 cannot contain whitespace.
1856 FORTH is case-sensitive. Use capslock!
1864 // Multi-line constant gives 'Warning: unterminated string; newline inserted' messages which you can ignore
1866 \\ Define some character constants
1872 \\ CR prints a carriage return
1875 \\ SPACE prints a space
1876 : SPACE 'SPACE' EMIT ;
1878 \\ Primitive . (DOT) function doesn't follow with a blank, so redefine it to behave like FORTH.
1879 \\ Notice how we can trivially redefine existing functions.
1882 \\ DUP, DROP are defined in assembly for speed, but this is how you might define them
1883 \\ in FORTH. Notice use of the scratch variables _X and _Y.
1884 \\ : DUP _X ! _X @ _X @ ;
1887 \\ The 2... versions of the standard operators work on pairs of stack entries. They're not used
1888 \\ very commonly so not really worth writing in assembler. Here is how they are defined in FORTH.
1892 \\ More standard FORTH words.
1896 \\ [ and ] allow you to break into immediate mode while compiling a word.
1897 : [ IMMEDIATE \\ define [ as an immediate word
1898 0 STATE ! \\ go into immediate mode
1902 1 STATE ! \\ go back to compile mode
1905 \\ LITERAL takes whatever is on the stack and compiles LIT <foo>
1907 ' LIT , \\ compile LIT
1908 , \\ compile the literal itself (from the stack)
1911 \\ condition IF true-part THEN rest
1913 \\ condition 0BRANCH OFFSET true-part rest
1914 \\ where OFFSET is the offset of 'rest'
1915 \\ condition IF true-part ELSE false-part THEN
1917 \\ condition 0BRANCH OFFSET true-part BRANCH OFFSET2 false-part rest
1918 \\ where OFFSET if the offset of false-part and OFFSET2 is the offset of rest
1920 \\ IF is an IMMEDIATE word which compiles 0BRANCH followed by a dummy offset, and places
1921 \\ the address of the 0BRANCH on the stack. Later when we see THEN, we pop that address
1922 \\ off the stack, calculate the offset, and back-fill the offset.
1924 ' 0BRANCH , \\ compile 0BRANCH
1925 HERE @ \\ save location of the offset on the stack
1926 0 , \\ compile a dummy offset
1931 HERE @ SWAP - \\ calculate the offset from the address saved on the stack
1932 SWAP ! \\ store the offset in the back-filled location
1936 ' BRANCH , \\ definite branch to just over the false-part
1937 HERE @ \\ save location of the offset on the stack
1938 0 , \\ compile a dummy offset
1939 SWAP \\ now back-fill the original (IF) offset
1940 DUP \\ same as for THEN word above
1945 \\ BEGIN loop-part condition UNTIL
1947 \\ loop-part condition 0BRANCH OFFSET
1948 \\ where OFFSET points back to the loop-part
1949 \\ This is like do { loop-part } while (condition) in the C language
1951 HERE @ \\ save location on the stack
1955 ' 0BRANCH , \\ compile 0BRANCH
1956 HERE @ - \\ calculate the offset from the address saved on the stack
1957 , \\ compile the offset here
1960 \\ BEGIN loop-part AGAIN
1962 \\ loop-part BRANCH OFFSET
1963 \\ where OFFSET points back to the loop-part
1964 \\ In other words, an infinite loop which can only be returned from with EXIT
1966 ' BRANCH , \\ compile BRANCH
1967 HERE @ - \\ calculate the offset back
1968 , \\ compile the offset here
1971 \\ BEGIN condition WHILE loop-part REPEAT
1973 \\ condition 0BRANCH OFFSET2 loop-part BRANCH OFFSET
1974 \\ where OFFSET points back to condition (the beginning) and OFFSET2 points to after the whole piece of code
1975 \\ So this is like a while (condition) { loop-part } loop in the C language
1977 ' 0BRANCH , \\ compile 0BRANCH
1978 HERE @ \\ save location of the offset2 on the stack
1979 0 , \\ compile a dummy offset2
1983 ' BRANCH , \\ compile BRANCH
1984 SWAP \\ get the original offset (from BEGIN)
1985 HERE @ - , \\ and compile it after BRANCH
1987 HERE @ SWAP - \\ calculate the offset2
1988 SWAP ! \\ and back-fill it in the original location
1991 \\ With the looping constructs, we can now write SPACES, which writes n spaces to stdout.
1994 SPACE \\ print a space
1995 1- \\ until we count down to 0
2000 \\ .S prints the contents of the stack. Very useful for debugging.
2002 DSP@ \\ get current stack pointer
2004 DUP @ . \\ print the stack element
2006 DUP S0 @ 4- = \\ stop when we get to the top
2011 \\ DEPTH returns the depth of the stack.
2012 : DEPTH S0 @ DSP@ - ;
2014 \\ .\" is the print string operator in FORTH. Example: .\" Something to print\"
2015 \\ The space after the operator is the ordinary space required between words.
2016 \\ This is tricky to define because it has to do different things depending on whether
2017 \\ we are compiling or in immediate mode. (Thus the word is marked IMMEDIATE so it can
2018 \\ detect this and do different things).
2019 \\ In immediate mode we just keep reading characters and printing them until we get to
2020 \\ the next double quote.
2021 \\ In compile mode we have the problem of where we're going to store the string (remember
2022 \\ that the input buffer where the string comes from may be overwritten by the time we
2023 \\ come round to running the function). We store the string in the compiled function
2025 \\ LITSTRING, string length, string rounded up to 4 bytes, EMITSTRING, ...
2027 STATE @ \\ compiling?
2029 ' LITSTRING , \\ compile LITSTRING
2030 HERE @ \\ save the address of the length word on the stack
2031 0 , \\ dummy length - we don't know what it is yet
2033 KEY \\ get next character of the string
2036 HERE @ !b \\ store the character in the compiled image
2037 1 HERE +! \\ increment HERE pointer by 1 byte
2039 DROP \\ drop the double quote character at the end
2040 DUP \\ get the saved address of the length word
2041 HERE @ SWAP - \\ calculate the length
2042 4- \\ subtract 4 (because we measured from the start of the length word)
2043 SWAP ! \\ and back-fill the length location
2044 HERE @ \\ round up to next multiple of 4 bytes for the remaining code
2048 ' EMITSTRING , \\ compile the final EMITSTRING
2050 \\ In immediate mode, just read characters and print them until we get
2051 \\ to the ending double quote. Much simpler than the above code!
2054 DUP '\"' = IF EXIT THEN
2060 \\ While compiling, [COMPILE] WORD compiles WORD if it would otherwise be IMMEDIATE.
2061 : [COMPILE] IMMEDIATE
2062 WORD \\ get the next word
2063 FIND \\ find it in the dictionary
2064 >CFA \\ get its codeword
2065 , \\ and compile that
2068 \\ RECURSE makes a recursive call to the current word that is being compiled.
2069 \\ Normally while a word is being compiled, it is marked HIDDEN so that references to the
2070 \\ same word within are calls to the previous definition of the word.
2072 LATEST @ >CFA \\ LATEST points to the word being compiled at the moment
2076 \\ ALLOT is used to allocate (static) memory when compiling. It increases HERE by
2077 \\ the amount given on the stack.
2081 \\ Finally print the welcome prompt.
2094 /* END OF jonesforth.S */