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
4 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
6 INTRODUCTION ----------------------------------------------------------------------
8 FORTH is one of those alien languages which most working programmers regard in the same
9 way as Haskell, LISP, and so on. Something so strange that they'd rather any thoughts
10 of it just go away so they can get on with writing this paying code. But that's wrong
11 and if you care at all about programming then you should at least understand all these
12 languages, even if you will never use them.
14 LISP is the ultimate high-level language, and features from LISP are being added every
15 decade to the more common languages. But FORTH is in some ways the ultimate in low level
16 programming. Out of the box it lacks features like dynamic memory management and even
17 strings. In fact, at its primitive level it lacks even basic concepts like IF-statements
20 Why then would you want to learn FORTH? There are several very good reasons. First
21 and foremost, FORTH is minimal. You really can write a complete FORTH in, say, 2000
22 lines of code. I don't just mean a FORTH program, I mean a complete FORTH operating
23 system, environment and language. You could boot such a FORTH on a bare PC and it would
24 come up with a prompt where you could start doing useful work. The FORTH you have here
25 isn't minimal and uses a Linux process as its 'base PC' (both for the purposes of making
26 it a good tutorial). It's possible to completely understand the system. Who can say they
27 completely understand how Linux works, or gcc?
29 Secondly FORTH has a peculiar bootstrapping property. By that I mean that after writing
30 a little bit of assembly to talk to the hardware and implement a few primitives, all the
31 rest of the language and compiler is written in FORTH itself. Remember I said before
32 that FORTH lacked IF-statements and loops? Well of course it doesn't really because
33 such a lanuage would be useless, but my point was rather that IF-statements and loops are
34 written in FORTH itself.
36 Now of course this is common in other languages as well, and in those languages we call
37 them 'libraries'. For example in C, 'printf' is a library function written in C. But
38 in FORTH this goes way beyond mere libraries. Can you imagine writing C's 'if' in C?
39 And that brings me to my third reason: If you can write 'if' in FORTH, then why restrict
40 yourself to the usual if/while/for/switch constructs? You want a construct that iterates
41 over every other element in a list of numbers? You can add it to the language. What
42 about an operator which pulls in variables directly from a configuration file and makes
43 them available as FORTH variables? Or how about adding Makefile-like dependencies to
44 the language? No problem in FORTH. This concept isn't common in programming languages,
45 but it has a name (in fact two names): "macros" (by which I mean LISP-style macros, not
46 the lame C preprocessor) and "domain specific languages" (DSLs).
48 This tutorial isn't about learning FORTH as the language. I'll point you to some references
49 you should read if you're not familiar with using FORTH. This tutorial is about how to
50 write FORTH. In fact, until you understand how FORTH is written, you'll have only a very
51 superficial understanding of how to use it.
53 So if you're not familiar with FORTH or want to refresh your memory here are some online
56 http://en.wikipedia.org/wiki/Forth_%28programming_language%29
58 http://galileo.phys.virginia.edu/classes/551.jvn.fall01/primer.htm
60 http://wiki.laptop.org/go/Forth_Lessons
62 Here is another "Why FORTH?" essay: http://www.jwdt.com/~paysan/why-forth.html
64 ACKNOWLEDGEMENTS ----------------------------------------------------------------------
66 This code draws heavily on the design of LINA FORTH (http://home.hccnet.nl/a.w.m.van.der.horst/lina.html)
67 by Albert van der Horst. Any similarities in the code are probably not accidental.
69 Also I used this document (http://ftp.funet.fi/pub/doc/IOCCC/1992/buzzard.2.design) which really
70 defies easy explanation.
72 SETTING UP ----------------------------------------------------------------------
74 Let's get a few housekeeping things out of the way. Firstly because I need to draw lots of
75 ASCII-art diagrams to explain concepts, the best way to look at this is using a window which
76 uses a fixed width font and is at least this wide:
78 <------------------------------------------------------------------------------------------------------------------------>
80 Secondly make sure TABS are set to 8 characters. The following should be a vertical
81 line. If not, sort out your tabs.
87 Thirdly I assume that your screen is at least 50 characters high.
89 ASSEMBLING ----------------------------------------------------------------------
91 If you want to actually run this FORTH, rather than just read it, you will need Linux on an
92 i386. Linux because instead of programming directly to the hardware on a bare PC which I
93 could have done, I went for a simpler tutorial by assuming that the 'hardware' is a Linux
94 process with a few basic system calls (read, write and exit and that's about all). i386
95 is needed because I had to write the assembly for a processor, and i386 is by far the most
96 common. (Of course when I say 'i386', any 32- or 64-bit x86 processor will do. I'm compiling
97 this on a 64 bit AMD Opteron).
99 Again, to assemble this you will need gcc and gas (the GNU assembler). The commands to
100 assemble and run the code (save this file as 'jonesforth.S') are:
102 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
105 You will see lots of 'Warning: unterminated string; newline inserted' messages from the
106 assembler. That's just because the GNU assembler doesn't have a good syntax for multi-line
107 strings (or rather it used to, but the developers removed it!) so I've abused the syntax
108 slightly to make things readable. Ignore these warnings.
110 ASSEMBLER ----------------------------------------------------------------------
112 (You can just skip to the next section -- you don't need to be able to read assembler to
113 follow this tutorial).
115 However if you do want to read the assembly code here are a few notes about gas (the GNU assembler):
117 (1) Register names are prefixed with '%', so %eax is the 32 bit i386 accumulator. The registers
118 available on i386 are: %eax, %ebx, %ecx, %edx, %esi, %edi, %ebp and %esp, and most of them
119 have special purposes.
121 (2) Add, mov, etc. take arguments in the form SRC,DEST. So mov %eax,%ecx moves %eax -> %ecx
123 (3) Constants are prefixed with '$', and you mustn't forget it! If you forget it then it
124 causes a read from memory instead, so:
125 mov $2,%eax moves number 2 into %eax
126 mov 2,%eax reads the 32 bit word from address 2 into %eax (ie. most likely a mistake)
128 (4) gas has a funky syntax for local labels, where '1f' (etc.) means label '1:' "forwards"
129 and '1b' (etc.) means label '1:' "backwards".
131 (5) 'ja' is "jump if above", 'jb' for "jump if below", 'je' "jump if equal" etc.
133 (6) gas has a reasonably nice .macro syntax, and I use them a lot to make the code shorter and
136 For more help reading the assembler, do "info gas" at the Linux prompt.
138 Now the tutorial starts in earnest.
140 THE DICTIONARY ----------------------------------------------------------------------
142 In FORTH as you will know, functions are called "words", as just as in other languages they
143 have a name and a definition. Here are two FORTH words:
145 : DOUBLE DUP + ; \ name is "DOUBLE", definition is "DUP +"
146 : QUADRUPLE DOUBLE DOUBLE ; \ name is "QUADRUPLE", definition is "DOUBLE DOUBLE"
148 Words, both built-in ones and ones which the programmer defines later, are stored in a dictionary
149 which is just a linked list of dictionary entries.
151 <--- DICTIONARY ENTRY (HEADER) ----------------------->
152 +------------------------+--------+---------- - - - - +----------- - - - -
153 | LINK POINTER | LENGTH/| NAME | DEFINITION
155 +--- (4 bytes) ----------+- byte -+- n bytes - - - - +----------- - - - -
157 I'll come to the definition of the word later. For now just look at the header. The first
158 4 bytes are the link pointer. This points back to the previous word in the dictionary, or, for
159 the first word in the dictionary it is just a NULL pointer. Then comes a length/flags byte.
160 The length of the word can be up to 31 characters (5 bits used) and the top three bits are used
161 for various flags which I'll come to later. This is followed by the name itself, and in this
162 implementation the name is rounded up to a multiple of 4 bytes by padding it with zero bytes.
163 That's just to ensure that the definition starts on a 32 bit boundary.
165 A FORTH variable called LATEST contains a pointer to the most recently defined word, in
166 other words, the head of this linked list.
168 DOUBLE and QUADRUPLE might look like this:
170 pointer to previous word
173 +--|------+---+---+---+---+---+---+---+---+------------- - - - -
174 | LINK | 6 | D | O | U | B | L | E | 0 | (definition ...)
175 +---------+---+---+---+---+---+---+---+---+------------- - - - -
178 +--|------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
179 | LINK | 9 | Q | U | A | D | R | U | P | L | E | 0 | 0 | (definition ...)
180 +---------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
186 You shoud be able to see from this how you might implement functions to find a word in
187 the dictionary (just walk along the dictionary entries starting at LATEST and matching
188 the names until you either find a match or hit the NULL pointer at the end of the dictionary),
189 and add a word to the dictionary (create a new definition, set its LINK to LATEST, and set
190 LATEST to point to the new word). We'll see precisely these functions implemented in
191 assembly code later on.
193 One interesting consequence of using a linked list is that you can redefine words, and
194 a newer definition of a word overrides an older one. This is an important concept in
195 FORTH because it means that any word (even "built-in" or "standard" words) can be
196 overridden with a new definition, either to enhance it, to make it faster or even to
197 disable it. However because of the way that FORTH words get compiled, which you'll
198 understand below, words defined using the old definition of a word continue to use
199 the old definition. Only words defined after the new definition use the new definition.
201 DIRECT THREADED CODE ----------------------------------------------------------------------
203 Now we'll get to the really crucial bit in understanding FORTH, so go and get a cup of tea
204 or coffee and settle down. It's fair to say that if you don't understand this section, then you
205 won't "get" how FORTH works, and that would be a failure on my part for not explaining it well.
206 So if after reading this section a few times you don't understand it, please email me
209 Let's talk first about what "threaded code" means. Imagine a peculiar version of C where
210 you are only allowed to call functions without arguments. (Don't worry for now that such a
211 language would be completely useless!) So in our peculiar C, code would look like this:
220 and so on. How would a function, say 'f' above, be compiled by a standard C compiler?
221 Probably into assembly code like this. On the right hand side I've written the actual
225 CALL a E8 08 00 00 00
226 CALL b E8 1C 00 00 00
227 CALL c E8 2C 00 00 00
228 ; ignore the return from the function for now
230 "E8" is the x86 machine code to "CALL" a function. In the first 20 years of computing
231 memory was hideously expensive and we might have worried about the wasted space being used
232 by the repeated "E8" bytes. We can save 20% in code size (and therefore, in expensive memory)
233 by compressing this into just:
235 08 00 00 00 Just the function addresses, without
236 1C 00 00 00 the CALL prefix.
239 [Historical note: If the execution model that FORTH uses looks strange from the following
240 paragraphs, then it was motivated entirely by the need to save memory on early computers.
241 This code compression isn't so important now when our machines have more memory in their L1
242 caches than those early computers had in total, but the execution model still has some
245 Of course this code won't run directly any more. Instead we need to write an interpreter
246 which takes each pair of bytes and calls it.
248 On an i386 machine it turns out that we can write this interpreter rather easily, in just
249 two assembly instructions which turn into just 3 bytes of machine code. Let's store the
250 pointer to the next word to execute in the %esi register:
252 08 00 00 00 <- We're executing this one now. %esi is the _next_ one to execute.
256 The all-important x86 instruction is called LODSL (or in Intel manuals, LODSW). It does
257 two things. Firstly it reads the memory at %esi into the accumulator (%eax). Secondly it
258 increments %esi by 4 bytes. So after LODSL, the situation now looks like this:
260 08 00 00 00 <- We're still executing this one
261 1C 00 00 00 <- %eax now contains this address (0x0000001C)
264 Now we just need to jump to the address in %eax. This is again just a single x86 instruction
265 written JMP *(%eax). And after doing the jump, the situation looks like:
268 1C 00 00 00 <- Now we're executing this subroutine.
271 To make this work, each subroutine is followed by the two instructions 'LODSL; JMP *(%eax)'
272 which literally make the jump to the next subroutine.
274 And that brings us to our first piece of actual code! Well, it's a macro.
283 /* The macro is called NEXT. That's a FORTH-ism. It expands to those two instructions.
285 Every FORTH primitive that we write has to be ended by NEXT. Think of it kind of like
288 The above describes what is known as direct threaded code.
290 To sum up: We compress our function calls down to a list of addresses and use a somewhat
291 magical macro to act as a "jump to next function in the list". We also use one register (%esi)
292 to act as a kind of instruction pointer, pointing to the next function in the list.
294 I'll just give you a hint of what is to come by saying that a FORTH definition such as:
296 : QUADRUPLE DOUBLE DOUBLE ;
298 actually compiles (almost, not precisely but we'll see why in a moment) to a list of
299 function addresses for DOUBLE, DOUBLE and a special function called EXIT to finish off.
301 At this point, REALLY EAGLE-EYED ASSEMBLY EXPERTS are saying "JONES, YOU'VE MADE A MISTAKE!".
303 I lied about JMP *(%eax).
305 INDIRECT THREADED CODE ----------------------------------------------------------------------
307 It turns out that direct threaded code is interesting but only if you want to just execute
308 a list of functions written in assembly language. So QUADRUPLE would work only if DOUBLE
309 was an assembly language function. In the direct threaded code, QUADRUPLE would look like:
312 | addr of DOUBLE --------------------> (assembly code to do the double)
313 +------------------+ NEXT
314 %esi -> | addr of DOUBLE |
317 We can add an extra indirection to allow us to run both words written in assembly language
318 (primitives written for speed) and words written in FORTH themselves as lists of addresses.
320 The extra indirection is the reason for the brackets in JMP *(%eax).
322 Let's have a look at how QUADRUPLE and DOUBLE really look in FORTH:
324 : QUADRUPLE DOUBLE DOUBLE ;
327 | codeword | : DOUBLE DUP + ;
329 | addr of DOUBLE ---------------> +------------------+
330 +------------------+ | codeword |
331 | addr of DOUBLE | +------------------+
332 +------------------+ | addr of DUP --------------> +------------------+
333 | addr of EXIT | +------------------+ | codeword -------+
334 +------------------+ %esi -> | addr of + --------+ +------------------+ |
335 +------------------+ | | assembly to <-----+
336 | addr of EXIT | | | implement DUP |
337 +------------------+ | | .. |
340 | +------------------+
342 +-----> +------------------+
344 +------------------+ |
345 | assembly to <------+
352 This is the part where you may need an extra cup of tea/coffee/favourite caffeinated
353 beverage. What has changed is that I've added an extra pointer to the beginning of
354 the definitions. In FORTH this is sometimes called the "codeword". The codeword is
355 a pointer to the interpreter to run the function. For primitives written in
356 assembly language, the "interpreter" just points to the actual assembly code itself.
357 They don't need interpreting, they just run.
359 In words written in FORTH (like QUADRUPLE and DOUBLE), the codeword points to an interpreter
362 I'll show you the interpreter function shortly, but let's recall our indirect
363 JMP *(%eax) with the "extra" brackets. Take the case where we're executing DOUBLE
364 as shown, and DUP has been called. Note that %esi is pointing to the address of +
366 The assembly code for DUP eventually does a NEXT. That:
368 (1) reads the address of + into %eax %eax points to the codeword of +
369 (2) increments %esi by 4
370 (3) jumps to the indirect %eax jumps to the address in the codeword of +,
371 ie. the assembly code to implement +
376 | addr of DOUBLE ---------------> +------------------+
377 +------------------+ | codeword |
378 | addr of DOUBLE | +------------------+
379 +------------------+ | addr of DUP --------------> +------------------+
380 | addr of EXIT | +------------------+ | codeword -------+
381 +------------------+ | addr of + --------+ +------------------+ |
382 +------------------+ | | assembly to <-----+
383 %esi -> | addr of EXIT | | | implement DUP |
384 +------------------+ | | .. |
387 | +------------------+
389 +-----> +------------------+
391 +------------------+ |
392 now we're | assembly to <------+
393 executing | implement + |
399 So I hope that I've convinced you that NEXT does roughly what you'd expect. This is
400 indirect threaded code.
402 I've glossed over four things. I wonder if you can guess without reading on what they are?
408 My list of four things are: (1) What does "EXIT" do? (2) which is related to (1) is how do
409 you call into a function, ie. how does %esi start off pointing at part of QUADRUPLE, but
410 then point at part of DOUBLE. (3) What goes in the codeword for the words which are written
411 in FORTH? (4) How do you compile a function which does anything except call other functions
412 ie. a function which contains a number like : DOUBLE 2 * ; ?
414 THE INTERPRETER AND RETURN STACK ------------------------------------------------------------
416 Going at these in no particular order, let's talk about issues (3) and (2), the interpreter
417 and the return stack.
419 Words which are defined in FORTH need a codeword which points to a little bit of code to
420 give them a "helping hand" in life. They don't need much, but they do need what is known
421 as an "interpreter", although it doesn't really "interpret" in the same way that, say,
422 Java bytecode used to be interpreted (ie. slowly). This interpreter just sets up a few
423 machine registers so that the word can then execute at full speed using the indirect
424 threaded model above.
426 One of the things that needs to happen when QUADRUPLE calls DOUBLE is that we save the old
427 %esi ("instruction pointer") and create a new one pointing to the first word in DOUBLE.
428 Because we will need to restore the old %esi at the end of DOUBLE (this is, after all, like
429 a function call), we will need a stack to store these "return addresses" (old values of %esi).
431 As you will have read, when reading the background documentation, FORTH has two stacks,
432 an ordinary stack for parameters, and a return stack which is a bit more mysterious. But
433 our return stack is just the stack I talked about in the previous paragraph, used to save
434 %esi when calling from a FORTH word into another FORTH word.
436 In this FORTH, we are using the normal stack pointer (%esp) for the parameter stack.
437 We will use the i386's "other" stack pointer (%ebp, usually called the "frame pointer")
438 for our return stack.
440 I've got two macros which just wrap up the details of using %ebp for the return stack.
441 You use them as for example "PUSHRSP %eax" (push %eax on the return stack) or "POPRSP %ebx"
442 (pop top of return stack into %ebx).
445 /* Macros to deal with the return stack. */
447 lea -4(%ebp),%ebp // push reg on to return stack
452 mov (%ebp),\reg // pop top of return stack to reg
457 And with that we can now talk about the interpreter.
459 In FORTH the interpreter function is often called DOCOL (I think it means "DO COLON" because
460 all FORTH definitions start with a colon, as in : DOUBLE DUP + ;
462 The "interpreter" (it's not really "interpreting") just needs to push the old %esi on the
463 stack and set %esi to the first word in the definition. Remember that we jumped to the
464 function using JMP *(%eax)? Well a consequence of that is that conveniently %eax contains
465 the address of this codeword, so just by adding 4 to it we get the address of the first
466 data word. Finally after setting up %esi, it just does NEXT which causes that first word
470 /* DOCOL - the interpreter! */
474 PUSHRSP %esi // push %esi on to the return stack
475 addl $4,%eax // %eax points to codeword, so make
476 movl %eax,%esi // %esi point to first data word
480 Just to make this absolutely clear, let's see how DOCOL works when jumping from QUADRUPLE
486 +------------------+ DOUBLE:
487 | addr of DOUBLE ---------------> +------------------+
488 +------------------+ %eax -> | addr of DOCOL |
489 %esi -> | addr of DOUBLE | +------------------+
490 +------------------+ | addr of DUP -------------->
491 | addr of EXIT | +------------------+
492 +------------------+ | etc. |
494 First, the call to DOUBLE causes DOCOL (the codeword of DOUBLE). DOCOL does this: It
495 pushes the old %esi on the return stack. %eax points to the codeword of DOUBLE, so we
496 just add 4 on to it to get our new %esi:
501 +------------------+ DOUBLE:
502 | addr of DOUBLE ---------------> +------------------+
503 top of return +------------------+ %eax -> | addr of DOCOL |
504 stack points -> | addr of DOUBLE | + 4 = +------------------+
505 +------------------+ %esi -> | addr of DUP -------------->
506 | addr of EXIT | +------------------+
507 +------------------+ | etc. |
509 Then we do NEXT, and because of the magic of threaded code that increments %esi again
512 Well, it seems to work.
514 One minor point here. Because DOCOL is the first bit of assembly actually to be defined
515 in this file (the others were just macros), and because I usually compile this code with the
516 text segment starting at address 0, DOCOL has address 0. So if you are disassembling the
517 code and see a word with a codeword of 0, you will immediately know that the word is
518 written in FORTH (it's not an assembler primitive) and so uses DOCOL as the interpreter.
520 STARTING UP ----------------------------------------------------------------------
522 Now let's get down to nuts and bolts. When we start the program we need to set up
523 a few things like the return stack. But as soon as we can, we want to jump into FORTH
524 code (albeit much of the "early" FORTH code will still need to be written as
525 assembly language primitives).
527 This is what the set up code does. Does a tiny bit of house-keeping, sets up the
528 separate return stack (NB: Linux gives us the ordinary parameter stack already), then
529 immediately jumps to a FORTH word called COLD. COLD stands for cold-start. In ISO
530 FORTH (but not in this FORTH), COLD can be called at any time to completely reset
531 the state of FORTH, and there is another word called WARM which does a partial reset.
534 /* ELF entry point. */
539 mov %esp,var_S0 // Store the initial data stack pointer.
540 mov $return_stack,%ebp // Initialise the return stack.
542 mov $cold_start,%esi // Initialise interpreter.
543 NEXT // Run interpreter!
546 cold_start: // High-level code without a codeword.
550 We also allocate some space for the return stack and some space to store user
551 definitions. These are static memory allocations using fixed-size buffers, but it
552 wouldn't be a great deal of work to make them dynamic.
556 /* FORTH return stack. */
557 #define RETURN_STACK_SIZE 8192
559 .space RETURN_STACK_SIZE
560 return_stack: // Initial top of return stack.
562 /* Space for user-defined words. */
563 #define USER_DEFS_SIZE 16384
566 .space USER_DEFS_SIZE
569 BUILT-IN WORDS ----------------------------------------------------------------------
571 Remember our dictionary entries (headers). Let's bring those together with the codeword
572 and data words to see how : DOUBLE DUP + ; really looks in memory.
574 pointer to previous word
577 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
578 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
579 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
582 LINK in next word points to codeword of DUP
584 Initially we can't just write ": DOUBLE DUP + ;" (ie. that literal string) here because we
585 don't yet have anything to read the string, break it up at spaces, parse each word, etc. etc.
586 So instead we will have to define built-in words using the GNU assembler data constructors
587 (like .int, .byte, .string, .ascii and so on -- look them up in the gas info page if you are
590 The long way would be:
591 .int <link to previous word>
593 .ascii "DOUBLE" // string
595 DOUBLE: .int DOCOL // codeword
596 .int DUP // pointer to codeword of DUP
597 .int PLUS // pointer to codeword of +
598 .int EXIT // pointer to codeword of EXIT
600 That's going to get quite tedious rather quickly, so here I define an assembler macro
601 so that I can just write:
603 defword "DOUBLE",6,,DOUBLE
606 and I'll get exactly the same effect.
608 Don't worry too much about the exact implementation details of this macro - it's complicated!
611 /* Flags - these are discussed later. */
613 #define F_HIDDEN 0x20
615 // Store the chain of links.
618 .macro defword name, namelen, flags=0, label
624 .set link,name_\label
625 .byte \flags+\namelen // flags + length byte
626 .ascii "\name" // the name
630 .int DOCOL // codeword - the interpreter
631 // list of word pointers follow
635 Similarly I want a way to write words written in assembly language. There will quite a few
636 of these to start with because, well, everything has to start in assembly before there's
637 enough "infrastructure" to be able to start writing FORTH words, but also I want to define
638 some common FORTH words in assembly language for speed, even though I could write them in FORTH.
640 This is what DUP looks like in memory:
642 pointer to previous word
645 +--|------+---+---+---+---+------------+
646 | LINK | 3 | D | U | P | code_DUP ---------------------> points to the assembly
647 +---------+---+---+---+---+------------+ code used to write DUP,
648 ^ len codeword which is ended with NEXT.
652 Again, for brevity in writing the header I'm going to write an assembler macro called defcode.
655 .macro defcode name, namelen, flags=0, label
661 .set link,name_\label
662 .byte \flags+\namelen // flags + length byte
663 .ascii "\name" // the name
667 .int code_\label // codeword
671 code_\label : // assembler code follows
675 Now some easy FORTH primitives. These are written in assembly for speed. If you understand
676 i386 assembly language then it is worth reading these. However if you don't understand assembly
677 you can skip the details.
681 pop %eax // duplicate top of stack
686 defcode "DROP",4,,DROP
687 pop %eax // drop top of stack
690 defcode "SWAP",4,,SWAP
691 pop %eax // swap top of stack
697 defcode "OVER",4,,OVER
698 mov 4(%esp),%eax // get the second element of stack
699 push %eax // and push it on top
711 defcode "-ROT",4,,NROT
721 incl (%esp) // increment top of stack
725 decl (%esp) // decrement top of stack
728 defcode "4+",2,,INCR4
729 addl $4,(%esp) // increment top of stack
732 defcode "4-",2,,DECR4
733 subl $4,(%esp) // decrement top of stack
737 pop %eax // get top of stack
738 addl %eax,(%esp) // and add it to next word on stack
742 pop %eax // get top of stack
743 subl %eax,(%esp) // and subtract if from next word on stack
750 push %eax // ignore overflow
758 push %eax // push quotient
766 push %edx // push remainder
769 defcode "=",1,,EQU // top two words are equal?
779 defcode "<>",2,,NEQU // top two words are not equal?
789 defcode "0=",2,,ZEQU // top of stack equals 0?
808 defcode "INVERT",6,,INVERT // this is the FORTH "NOT" function
813 RETURNING FROM FORTH WORDS ----------------------------------------------------------------------
815 Time to talk about what happens when we EXIT a function. In this diagram QUADRUPLE has called
816 DOUBLE, and DOUBLE is about to exit (look at where %esi is pointing):
821 +------------------+ DOUBLE
822 | addr of DOUBLE ---------------> +------------------+
823 +------------------+ | codeword |
824 | addr of DOUBLE | +------------------+
825 +------------------+ | addr of DUP |
826 | addr of EXIT | +------------------+
827 +------------------+ | addr of + |
829 %esi -> | addr of EXIT |
832 What happens when the + function does NEXT? Well, the following code is executed.
835 defcode "EXIT",4,,EXIT
836 POPRSP %esi // pop return stack into %esi
840 EXIT gets the old %esi which we saved from before on the return stack, and puts it in %esi.
841 So after this (but just before NEXT) we get:
846 +------------------+ DOUBLE
847 | addr of DOUBLE ---------------> +------------------+
848 +------------------+ | codeword |
849 %esi -> | addr of DOUBLE | +------------------+
850 +------------------+ | addr of DUP |
851 | addr of EXIT | +------------------+
852 +------------------+ | addr of + |
857 And NEXT just completes the job by, well in this case just by calling DOUBLE again :-)
859 LITERALS ----------------------------------------------------------------------
861 The final point I "glossed over" before was how to deal with functions that do anything
862 apart from calling other functions. For example, suppose that DOUBLE was defined like this:
866 It does the same thing, but how do we compile it since it contains the literal 2? One way
867 would be to have a function called "2" (which you'd have to write in assembler), but you'd need
868 a function for every single literal that you wanted to use.
870 FORTH solves this by compiling the function using a special word called LIT:
872 +---------------------------+-------+-------+-------+-------+-------+
873 | (usual header of DOUBLE) | DOCOL | LIT | 2 | * | EXIT |
874 +---------------------------+-------+-------+-------+-------+-------+
876 LIT is executed in the normal way, but what it does next is definitely not normal. It
877 looks at %esi (which now points to the literal 2), grabs it, pushes it on the stack, then
878 manipulates %esi in order to skip the literal as if it had never been there.
880 What's neat is that the whole grab/manipulate can be done using a single byte single
881 i386 instruction, our old friend LODSL. Rather than me drawing more ASCII-art diagrams,
882 see if you can find out how LIT works:
886 // %esi points to the next command, but in this case it points to the next
887 // literal 32 bit integer. Get that literal into %eax and increment %esi.
888 // On x86, it's a convenient single byte instruction! (cf. NEXT macro)
890 push %eax // push the literal number on to stack
894 MEMORY ----------------------------------------------------------------------
896 As important point about FORTH is that it gives you direct access to the lowest levels
897 of the machine. Manipulating memory directly is done frequently in FORTH, and these are
898 the primitive words for doing it.
902 pop %ebx // address to store at
903 pop %eax // data to store there
904 mov %eax,(%ebx) // store it
908 pop %ebx // address to fetch
909 mov (%ebx),%eax // fetch it
910 push %eax // push value onto stack
913 defcode "+!",2,,ADDSTORE
915 pop %eax // the amount to add
916 addl %eax,(%ebx) // add it
919 defcode "-!",2,,SUBSTORE
921 pop %eax // the amount to subtract
922 subl %eax,(%ebx) // add it
925 /* ! and @ (STORE and FETCH) store 32-bit words. It's also useful to be able to read and write bytes.
926 * I don't know whether FORTH has these words, so I invented my own, called !b and @b.
927 * Byte-oriented operations only work on architectures which permit them (i386 is one of those).
928 * UPDATE: writing a byte to the dictionary pointer is called C, in FORTH.
930 defcode "!b",2,,STOREBYTE
931 pop %ebx // address to store at
932 pop %eax // data to store there
933 movb %al,(%ebx) // store it
936 defcode "@b",2,,FETCHBYTE
937 pop %ebx // address to fetch
939 movb (%ebx),%al // fetch it
940 push %eax // push value onto stack
944 BUILT-IN VARIABLES ----------------------------------------------------------------------
946 These are some built-in variables and related standard FORTH words. Of these, the only one that we
947 have discussed so far was LATEST, which points to the last (most recently defined) word in the
948 FORTH dictionary. LATEST is also a FORTH word which pushes the address of LATEST (the variable)
949 on to the stack, so you can read or write it using @ and ! operators. For example, to print
950 the current value of LATEST (and this can apply to any FORTH variable) you would do:
954 To make defining variables shorter, I'm using a macro called defvar, similar to defword and
955 defcode above. (In fact the defvar macro uses defcode to do the dictionary header).
958 .macro defvar name, namelen, flags=0, label, initial=0
959 defcode \name,\namelen,\flags,\label
969 The built-in variables are:
971 STATE Is the interpreter executing code (0) or compiling a word (non-zero)?
972 LATEST Points to the latest (most recently defined) word in the dictionary.
973 HERE When compiling, compiled words go here.
974 _X These are three scratch variables, used by some standard dictionary words.
977 S0 Stores the address of the top of the parameter stack.
978 R0 Stores the address of the top of the return stack.
981 defvar "STATE",5,,STATE
982 defvar "HERE",4,,HERE,user_defs_start
983 defvar "LATEST",6,,LATEST,name_SYSEXIT // SYSEXIT must be last in built-in dictionary
988 defvar "R0",2,,RZ,return_stack
991 RETURN STACK ----------------------------------------------------------------------
993 These words allow you to access the return stack. Recall that the register %ebp always points to
994 the top of the return stack.
997 defcode "DSP@",4,,DSPFETCH
1002 defcode "DSP!",4,,DSPSTORE
1007 pop %eax // pop parameter stack into %eax
1008 PUSHRSP %eax // push it on to the return stack
1011 defcode "R>",2,,FROMR
1012 POPRSP %eax // pop return stack on to %eax
1013 push %eax // and push on to parameter stack
1016 defcode "RSP@",4,,RSPFETCH
1020 defcode "RSP!",4,,RSPSTORE
1024 defcode "RDROP",5,,RDROP
1025 lea 4(%ebp),%ebp // pop return stack and throw away
1029 INPUT AND OUTPUT ----------------------------------------------------------------------
1036 #include <asm-i386/unistd.h>
1038 defcode "KEY",3,,KEY
1040 push %eax // push return value on stack
1052 1: // out of input; use read(2) to fetch more input from stdin
1053 xor %ebx,%ebx // 1st param: stdin
1054 mov $buffer,%ecx // 2nd param: buffer
1056 mov $buffend-buffer,%edx // 3rd param: max length
1057 mov $__NR_read,%eax // syscall: read
1059 test %eax,%eax // If %eax <= 0, then exit.
1061 addl %eax,%ecx // buffer+%eax = bufftop
1065 2: // error or out of input: exit
1067 mov $__NR_exit,%eax // syscall: exit
1070 defcode "EMIT",4,,EMIT
1075 mov $1,%ebx // 1st param: stdout
1077 // write needs the address of the byte to write
1079 mov $2f,%ecx // 2nd param: address
1081 mov $1,%edx // 3rd param: nbytes = 1
1083 mov $__NR_write,%eax // write syscall
1088 2: .space 1 // scratch used by EMIT
1090 defcode "WORD",4,,WORD
1092 push %ecx // push length
1093 push %edi // push base address
1097 /* Search for first non-blank character. Also skip \ comments. */
1099 call _KEY // get next key, returned in %eax
1100 cmpb $'\\',%al // start of a comment?
1101 je 3f // if so, skip the comment
1103 jbe 1b // if so, keep looking
1105 /* Search for the end of the word, storing chars as we go. */
1106 mov $5f,%edi // pointer to return buffer
1108 stosb // add character to return buffer
1109 call _KEY // get next key, returned in %al
1110 cmpb $' ',%al // is blank?
1111 ja 2b // if not, keep looping
1113 /* Return the word (well, the static buffer) and length. */
1115 mov %edi,%ecx // return length of the word
1116 mov $5f,%edi // return address of the word
1119 /* Code to skip \ comments to end of the current line. */
1122 cmpb $'\n',%al // end of line yet?
1127 // A static buffer where WORD returns. Subsequent calls
1128 // overwrite this buffer. Maximum word length is 32 chars.
1131 defcode "EMITSTRING",10,,EMITSTRING
1132 mov $1,%ebx // 1st param: stdout
1133 pop %ecx // 2nd param: address of string
1134 pop %edx // 3rd param: length of string
1136 mov $__NR_write,%eax // write syscall
1142 pop %eax // Get the number to print into %eax
1143 call _DOT // Easier to do this recursively ...
1146 mov $10,%ecx // Base 10
1150 xor %edx,%edx // %edx:%eax / %ecx -> quotient %eax, remainder %edx
1164 // Parse a number from a string on the stack -- almost the opposite of . (DOT)
1165 // Note that there is absolutely no error checking. In particular the length of the
1166 // string must be >= 1 bytes.
1167 defcode "SNUMBER",7,,SNUMBER
1177 imull $10,%eax // %eax *= 10
1180 subb $'0',%bl // ASCII -> digit
1186 defcode "FIND",4,,FIND
1187 pop %edi // %edi = address
1188 pop %ecx // %ecx = length
1194 push %esi // Save %esi so we can use it in string comparison.
1196 // Now we start searching backwards through the dictionary for this word.
1197 mov var_LATEST,%edx // LATEST points to name header of the latest word in the dictionary
1199 test %edx,%edx // NULL pointer? (end of the linked list)
1202 // Compare the length expected and the length of the word.
1203 // Note that if the F_HIDDEN flag is set on the word, then by a bit of trickery
1204 // this won't pick the word (the length will appear to be wrong).
1206 movb 4(%edx),%al // %al = flags+length field
1207 andb $(F_HIDDEN|0x1f),%al // %al = name length
1208 cmpb %cl,%al // Length is the same?
1211 // Compare the strings in detail.
1212 push %ecx // Save the length
1213 push %edi // Save the address (repe cmpsb will move this pointer)
1214 lea 5(%edx),%esi // Dictionary string we are checking against.
1215 repe cmpsb // Compare the strings.
1218 jne 2f // Not the same.
1220 // The strings are the same - return the header pointer in %eax
1226 mov (%edx),%edx // Move back through the link field to the previous word
1227 jmp 1b // .. and loop.
1231 xor %eax,%eax // Return zero to indicate not found.
1234 defcode ">CFA",4,,TCFA // DEA -> Codeword address
1241 add $4,%edi // Skip link pointer.
1242 movb (%edi),%al // Load flags+len into %al.
1243 inc %edi // Skip flags+len byte.
1244 andb $0x1f,%al // Just the length, not the flags.
1245 add %eax,%edi // Skip the name.
1246 addl $3,%edi // The codeword is 4-byte aligned.
1250 defcode "CHAR",4,,CHAR
1251 call _WORD // Returns %ecx = length, %edi = pointer to word.
1253 movb (%edi),%al // Get the first character of the word.
1254 push %eax // Push it onto the stack.
1257 defcode ":",1,,COLON
1259 // Get the word and create a dictionary entry header for it.
1260 call _WORD // Returns %ecx = length, %edi = pointer to word.
1261 mov %edi,%ebx // %ebx = address of the word
1263 movl var_HERE,%edi // %edi is the address of the header
1264 movl var_LATEST,%eax // Get link pointer
1265 stosl // and store it in the header.
1267 mov %cl,%al // Get the length.
1268 orb $F_HIDDEN,%al // Set the HIDDEN flag on this entry.
1269 stosb // Store the length/flags byte.
1271 mov %ebx,%esi // %esi = word
1272 rep movsb // Copy the word
1274 addl $3,%edi // Align to next 4 byte boundary.
1277 movl $DOCOL,%eax // The codeword for user-created words is always DOCOL (the interpreter)
1280 // Header built, so now update LATEST and HERE.
1281 // We'll be compiling words and putting them HERE.
1283 movl %eax,var_LATEST
1286 // And go into compile mode by setting STATE to 1.
1290 defcode ",",1,,COMMA
1291 pop %eax // Code pointer to store.
1295 movl var_HERE,%edi // HERE
1297 movl %edi,var_HERE // Update HERE (incremented)
1300 defcode ";",1,F_IMMED,SEMICOLON
1301 movl $EXIT,%eax // EXIT is the final codeword in compiled words.
1302 call _COMMA // Store it.
1303 call _HIDDEN // Toggle the HIDDEN flag (unhides the new word).
1304 xor %eax,%eax // Set STATE to 0 (back to execute mode).
1308 defcode "IMMEDIATE",9,F_IMMED,IMMEDIATE
1312 movl var_LATEST,%edi // LATEST word.
1313 addl $4,%edi // Point to name/flags byte.
1314 xorb $F_IMMED,(%edi) // Toggle the IMMED bit.
1317 defcode "HIDDEN",6,,HIDDEN
1321 movl var_LATEST,%edi // LATEST word.
1322 addl $4,%edi // Point to name/flags byte.
1323 xorb $F_HIDDEN,(%edi) // Toggle the HIDDEN bit.
1326 /* This definiton of ' (TICK) is strictly cheating - it also only works in compiled code. */
1328 lodsl // Get the address of the next word and skip it.
1329 pushl %eax // Push it on the stack.
1332 defcode "BRANCH",6,,BRANCH
1333 add (%esi),%esi // add the offset to the instruction pointer
1336 defcode "0BRANCH",7,,ZBRANCH
1338 test %eax,%eax // top of stack is zero?
1339 jz code_BRANCH // if so, jump back to the branch function above
1340 lodsl // otherwise we need to skip the offset
1343 defcode "LITSTRING",9,,LITSTRING
1344 lodsl // get the length of the string
1345 push %eax // push it on the stack
1346 push %esi // push the address of the start of the string
1347 addl %eax,%esi // skip past the string
1348 addl $3,%esi // but round up to next 4 byte boundary
1352 // COLD must not return (ie. must not call EXIT).
1353 defword "COLD",4,,COLD
1354 .int INTERPRETER // call the interpreter loop (never returns)
1355 .int LIT,1,SYSEXIT // hmmm, but in case it does, exit(1).
1357 /* This interpreter is pretty simple, but remember that in FORTH you can always override
1358 * it later with a more powerful one!
1360 defword "INTERPRETER",11,,INTERPRETER
1361 .int INTERPRET,RDROP,INTERPRETER
1363 defcode "INTERPRET",9,,INTERPRET
1364 call _WORD // Returns %ecx = length, %edi = pointer to word.
1366 // Is it in the dictionary?
1368 movl %eax,interpret_is_lit // Not a literal number (not yet anyway ...)
1369 call _FIND // Returns %eax = pointer to header or 0 if not found.
1370 test %eax,%eax // Found?
1373 // In the dictionary. Is it an IMMEDIATE codeword?
1374 mov %eax,%edi // %edi = dictionary entry
1375 movb 4(%edi),%al // Get name+flags.
1376 push %ax // Just save it for now.
1377 call _TCFA // Convert dictionary entry (in %edi) to codeword pointer.
1379 andb $F_IMMED,%al // Is IMMED flag set?
1381 jnz 4f // If IMMED, jump straight to executing.
1385 1: // Not in the dictionary (not a word) so assume it's a literal number.
1386 incl interpret_is_lit
1387 call _SNUMBER // Returns the parsed number in %eax
1389 mov $LIT,%eax // The word is LIT
1391 2: // Are we compiling or executing?
1394 jz 4f // Jump if executing.
1396 // Compiling - just append the word to the current dictionary definition.
1398 mov interpret_is_lit,%ecx // Was it a literal?
1401 mov %ebx,%eax // Yes, so LIT is followed by a number.
1405 4: // Executing - run it!
1406 mov interpret_is_lit,%ecx // Literal?
1407 test %ecx,%ecx // Literal?
1410 // Not a literal, execute it now. This never returns, but the codeword will
1411 // eventually call NEXT which will reenter the loop in INTERPRETER.
1414 5: // Executing a literal, which means push it on the stack.
1421 .int 0 // Flag used to record if reading a literal
1423 // NB: SYSEXIT must be the last entry in the built-in dictionary.
1424 defcode SYSEXIT,7,,SYSEXIT
1429 /*----------------------------------------------------------------------
1430 * Input buffer & initial input.
1435 // XXX gives 'Warning: unterminated string; newline inserted' messages which you can ignore
1437 \\ Define some character constants
1443 \\ CR prints a carriage return
1446 \\ SPACE prints a space
1447 : SPACE 'SPACE' EMIT ;
1449 \\ Primitive . (DOT) function doesn't follow with a blank, so redefine it to behave like FORTH.
1450 \\ Notice how we can trivially redefine existing functions.
1453 \\ DUP, DROP are defined in assembly for speed, but this is how you might define them
1454 \\ in FORTH. Notice use of the scratch variables _X and _Y.
1455 \\ : DUP _X ! _X @ _X @ ;
1458 \\ The 2... versions of the standard operators work on pairs of stack entries. They're not used
1459 \\ very commonly so not really worth writing in assembler. Here is how they are defined in FORTH.
1463 \\ More standard FORTH words.
1467 \\ [ and ] allow you to break into immediate mode while compiling a word.
1468 : [ IMMEDIATE \\ define [ as an immediate word
1469 0 STATE ! \\ go into immediate mode
1473 1 STATE ! \\ go back to compile mode
1476 \\ LITERAL takes whatever is on the stack and compiles LIT <foo>
1478 ' LIT , \\ compile LIT
1479 , \\ compile the literal itself (from the stack)
1482 \\ condition IF true-part THEN rest
1484 \\ condition 0BRANCH OFFSET true-part rest
1485 \\ where OFFSET is the offset of 'rest'
1486 \\ condition IF true-part ELSE false-part THEN
1488 \\ condition 0BRANCH OFFSET true-part BRANCH OFFSET2 false-part rest
1489 \\ where OFFSET if the offset of false-part and OFFSET2 is the offset of rest
1491 \\ IF is an IMMEDIATE word which compiles 0BRANCH followed by a dummy offset, and places
1492 \\ the address of the 0BRANCH on the stack. Later when we see THEN, we pop that address
1493 \\ off the stack, calculate the offset, and back-fill the offset.
1495 ' 0BRANCH , \\ compile 0BRANCH
1496 HERE @ \\ save location of the offset on the stack
1497 0 , \\ compile a dummy offset
1502 HERE @ SWAP - \\ calculate the offset from the address saved on the stack
1503 SWAP ! \\ store the offset in the back-filled location
1507 ' BRANCH , \\ definite branch to just over the false-part
1508 HERE @ \\ save location of the offset on the stack
1509 0 , \\ compile a dummy offset
1510 SWAP \\ now back-fill the original (IF) offset
1511 DUP \\ same as for THEN word above
1516 \\ BEGIN loop-part condition UNTIL
1518 \\ loop-part condition 0BRANCH OFFSET
1519 \\ where OFFSET points back to the loop-part
1520 \\ This is like do { loop-part } while (condition) in the C language
1522 HERE @ \\ save location on the stack
1526 ' 0BRANCH , \\ compile 0BRANCH
1527 HERE @ - \\ calculate the offset from the address saved on the stack
1528 , \\ compile the offset here
1531 \\ BEGIN loop-part AGAIN
1533 \\ loop-part BRANCH OFFSET
1534 \\ where OFFSET points back to the loop-part
1535 \\ In other words, an infinite loop which can only be returned from with EXIT
1537 ' BRANCH , \\ compile BRANCH
1538 HERE @ - \\ calculate the offset back
1539 , \\ compile the offset here
1542 \\ BEGIN condition WHILE loop-part REPEAT
1544 \\ condition 0BRANCH OFFSET2 loop-part BRANCH OFFSET
1545 \\ where OFFSET points back to condition (the beginning) and OFFSET2 points to after the whole piece of code
1546 \\ So this is like a while (condition) { loop-part } loop in the C language
1548 ' 0BRANCH , \\ compile 0BRANCH
1549 HERE @ \\ save location of the offset2 on the stack
1550 0 , \\ compile a dummy offset2
1554 ' BRANCH , \\ compile BRANCH
1555 SWAP \\ get the original offset (from BEGIN)
1556 HERE @ - , \\ and compile it after BRANCH
1558 HERE @ SWAP - \\ calculate the offset2
1559 SWAP ! \\ and back-fill it in the original location
1562 \\ With the looping constructs, we can now write SPACES, which writes n spaces to stdout.
1565 SPACE \\ print a space
1566 1- \\ until we count down to 0
1571 \\ .S prints the contents of the stack. Very useful for debugging.
1573 DSP@ \\ get current stack pointer
1575 DUP @ . \\ print the stack element
1577 DUP S0 @ 4- = \\ stop when we get to the top
1582 \\ DEPTH returns the depth of the stack.
1583 : DEPTH S0 @ DSP@ - ;
1585 \\ .\" is the print string operator in FORTH. Example: .\" Something to print\"
1586 \\ The space after the operator is the ordinary space required between words.
1587 \\ This is tricky to define because it has to do different things depending on whether
1588 \\ we are compiling or in immediate mode. (Thus the word is marked IMMEDIATE so it can
1589 \\ detect this and do different things).
1590 \\ In immediate mode we just keep reading characters and printing them until we get to
1591 \\ the next double quote.
1592 \\ In compile mode we have the problem of where we're going to store the string (remember
1593 \\ that the input buffer where the string comes from may be overwritten by the time we
1594 \\ come round to running the function). We store the string in the compiled function
1596 \\ LITSTRING, string length, string rounded up to 4 bytes, EMITSTRING, ...
1598 STATE @ \\ compiling?
1600 ' LITSTRING , \\ compile LITSTRING
1601 HERE @ \\ save the address of the length word on the stack
1602 0 , \\ dummy length - we don't know what it is yet
1604 KEY \\ get next character of the string
1607 HERE @ !b \\ store the character in the compiled image
1608 1 HERE +! \\ increment HERE pointer by 1 byte
1610 DROP \\ drop the double quote character at the end
1611 DUP \\ get the saved address of the length word
1612 HERE @ SWAP - \\ calculate the length
1613 4- \\ subtract 4 (because we measured from the start of the length word)
1614 SWAP ! \\ and back-fill the length location
1615 HERE @ \\ round up to next multiple of 4 bytes for the remaining code
1619 ' EMITSTRING , \\ compile the final EMITSTRING
1621 \\ In immediate mode, just read characters and print them until we get
1622 \\ to the ending double quote. Much simpler than the above code!
1625 DUP '\"' = IF EXIT THEN
1631 \\ While compiling, [COMPILE] WORD compiles WORD if it would otherwise be IMMEDIATE.
1632 : [COMPILE] IMMEDIATE
1633 WORD \\ get the next word
1634 FIND \\ find it in the dictionary
1635 >CFA \\ get its codeword
1636 , \\ and compile that
1639 \\ RECURSE makes a recursive call to the current word that is being compiled.
1640 \\ Normally while a word is being compiled, it is marked HIDDEN so that references to the
1641 \\ same word within are calls to the previous definition of the word.
1643 LATEST @ >CFA \\ LATEST points to the word being compiled at the moment
1647 \\ ALLOT is used to allocate (static) memory when compiling. It increases HERE by
1648 \\ the amount given on the stack.
1652 \\ Finally print the welcome prompt.