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).
5 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
7 INTRODUCTION ----------------------------------------------------------------------
9 FORTH is one of those alien languages which most working programmers regard in the same
10 way as Haskell, LISP, and so on. Something so strange that they'd rather any thoughts
11 of it just go away so they can get on with writing this paying code. But that's wrong
12 and if you care at all about programming then you should at least understand all these
13 languages, even if you will never use them.
15 LISP is the ultimate high-level language, and features from LISP are being added every
16 decade to the more common languages. But FORTH is in some ways the ultimate in low level
17 programming. Out of the box it lacks features like dynamic memory management and even
18 strings. In fact, at its primitive level it lacks even basic concepts like IF-statements
21 Why then would you want to learn FORTH? There are several very good reasons. First
22 and foremost, FORTH is minimal. You really can write a complete FORTH in, say, 2000
23 lines of code. I don't just mean a FORTH program, I mean a complete FORTH operating
24 system, environment and language. You could boot such a FORTH on a bare PC and it would
25 come up with a prompt where you could start doing useful work. The FORTH you have here
26 isn't minimal and uses a Linux process as its 'base PC' (both for the purposes of making
27 it a good tutorial). It's possible to completely understand the system. Who can say they
28 completely understand how Linux works, or gcc?
30 Secondly FORTH has a peculiar bootstrapping property. By that I mean that after writing
31 a little bit of assembly to talk to the hardware and implement a few primitives, all the
32 rest of the language and compiler is written in FORTH itself. Remember I said before
33 that FORTH lacked IF-statements and loops? Well of course it doesn't really because
34 such a lanuage would be useless, but my point was rather that IF-statements and loops are
35 written in FORTH itself.
37 Now of course this is common in other languages as well, and in those languages we call
38 them 'libraries'. For example in C, 'printf' is a library function written in C. But
39 in FORTH this goes way beyond mere libraries. Can you imagine writing C's 'if' in C?
40 And that brings me to my third reason: If you can write 'if' in FORTH, then why restrict
41 yourself to the usual if/while/for/switch constructs? You want a construct that iterates
42 over every other element in a list of numbers? You can add it to the language. What
43 about an operator which pulls in variables directly from a configuration file and makes
44 them available as FORTH variables? Or how about adding Makefile-like dependencies to
45 the language? No problem in FORTH. This concept isn't common in programming languages,
46 but it has a name (in fact two names): "macros" (by which I mean LISP-style macros, not
47 the lame C preprocessor) and "domain specific languages" (DSLs).
49 This tutorial isn't about learning FORTH as the language. I'll point you to some references
50 you should read if you're not familiar with using FORTH. This tutorial is about how to
51 write FORTH. In fact, until you understand how FORTH is written, you'll have only a very
52 superficial understanding of how to use it.
54 So if you're not familiar with FORTH or want to refresh your memory here are some online
57 http://en.wikipedia.org/wiki/Forth_%28programming_language%29
59 http://galileo.phys.virginia.edu/classes/551.jvn.fall01/primer.htm
61 http://wiki.laptop.org/go/Forth_Lessons
63 Here is another "Why FORTH?" essay: http://www.jwdt.com/~paysan/why-forth.html
65 ACKNOWLEDGEMENTS ----------------------------------------------------------------------
67 This code draws heavily on the design of LINA FORTH (http://home.hccnet.nl/a.w.m.van.der.horst/lina.html)
68 by Albert van der Horst. Any similarities in the code are probably not accidental.
70 Also I used this document (http://ftp.funet.fi/pub/doc/IOCCC/1992/buzzard.2.design) which really
71 defies easy explanation.
73 PUBLIC DOMAIN ----------------------------------------------------------------------
75 I, the copyright holder of this work, hereby release it into the public domain. This applies worldwide.
77 In case this is not legally possible, I grant any entity the right to use this work for any purpose,
78 without any conditions, unless such conditions are required by law.
80 SETTING UP ----------------------------------------------------------------------
82 Let's get a few housekeeping things out of the way. Firstly because I need to draw lots of
83 ASCII-art diagrams to explain concepts, the best way to look at this is using a window which
84 uses a fixed width font and is at least this wide:
86 <------------------------------------------------------------------------------------------------------------------------>
88 Secondly make sure TABS are set to 8 characters. The following should be a vertical
89 line. If not, sort out your tabs.
95 Thirdly I assume that your screen is at least 50 characters high.
97 ASSEMBLING ----------------------------------------------------------------------
99 If you want to actually run this FORTH, rather than just read it, you will need Linux on an
100 i386. Linux because instead of programming directly to the hardware on a bare PC which I
101 could have done, I went for a simpler tutorial by assuming that the 'hardware' is a Linux
102 process with a few basic system calls (read, write and exit and that's about all). i386
103 is needed because I had to write the assembly for a processor, and i386 is by far the most
104 common. (Of course when I say 'i386', any 32- or 64-bit x86 processor will do. I'm compiling
105 this on a 64 bit AMD Opteron).
107 Again, to assemble this you will need gcc and gas (the GNU assembler). The commands to
108 assemble and run the code (save this file as 'jonesforth.S') are:
110 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -o jonesforth jonesforth.S
113 You will see lots of 'Warning: unterminated string; newline inserted' messages from the
114 assembler. That's just because the GNU assembler doesn't have a good syntax for multi-line
115 strings (or rather it used to, but the developers removed it!) so I've abused the syntax
116 slightly to make things readable. Ignore these warnings.
118 If you want to run your own FORTH programs you can do:
120 ./jonesforth < myprog.f
122 If you want to load your own FORTH code and then continue reading user commands, you can do:
124 cat myfunctions.f - | ./jonesforth
126 ASSEMBLER ----------------------------------------------------------------------
128 (You can just skip to the next section -- you don't need to be able to read assembler to
129 follow this tutorial).
131 However if you do want to read the assembly code here are a few notes about gas (the GNU assembler):
133 (1) Register names are prefixed with '%', so %eax is the 32 bit i386 accumulator. The registers
134 available on i386 are: %eax, %ebx, %ecx, %edx, %esi, %edi, %ebp and %esp, and most of them
135 have special purposes.
137 (2) Add, mov, etc. take arguments in the form SRC,DEST. So mov %eax,%ecx moves %eax -> %ecx
139 (3) Constants are prefixed with '$', and you mustn't forget it! If you forget it then it
140 causes a read from memory instead, so:
141 mov $2,%eax moves number 2 into %eax
142 mov 2,%eax reads the 32 bit word from address 2 into %eax (ie. most likely a mistake)
144 (4) gas has a funky syntax for local labels, where '1f' (etc.) means label '1:' "forwards"
145 and '1b' (etc.) means label '1:' "backwards".
147 (5) 'ja' is "jump if above", 'jb' for "jump if below", 'je' "jump if equal" etc.
149 (6) gas has a reasonably nice .macro syntax, and I use them a lot to make the code shorter and
152 For more help reading the assembler, do "info gas" at the Linux prompt.
154 Now the tutorial starts in earnest.
156 THE DICTIONARY ----------------------------------------------------------------------
158 In FORTH as you will know, functions are called "words", as just as in other languages they
159 have a name and a definition. Here are two FORTH words:
161 : DOUBLE DUP + ; \ name is "DOUBLE", definition is "DUP +"
162 : QUADRUPLE DOUBLE DOUBLE ; \ name is "QUADRUPLE", definition is "DOUBLE DOUBLE"
164 Words, both built-in ones and ones which the programmer defines later, are stored in a dictionary
165 which is just a linked list of dictionary entries.
167 <--- DICTIONARY ENTRY (HEADER) ----------------------->
168 +------------------------+--------+---------- - - - - +----------- - - - -
169 | LINK POINTER | LENGTH/| NAME | DEFINITION
171 +--- (4 bytes) ----------+- byte -+- n bytes - - - - +----------- - - - -
173 I'll come to the definition of the word later. For now just look at the header. The first
174 4 bytes are the link pointer. This points back to the previous word in the dictionary, or, for
175 the first word in the dictionary it is just a NULL pointer. Then comes a length/flags byte.
176 The length of the word can be up to 31 characters (5 bits used) and the top three bits are used
177 for various flags which I'll come to later. This is followed by the name itself, and in this
178 implementation the name is rounded up to a multiple of 4 bytes by padding it with zero bytes.
179 That's just to ensure that the definition starts on a 32 bit boundary.
181 A FORTH variable called LATEST contains a pointer to the most recently defined word, in
182 other words, the head of this linked list.
184 DOUBLE and QUADRUPLE might look like this:
186 pointer to previous word
189 +--|------+---+---+---+---+---+---+---+---+------------- - - - -
190 | LINK | 6 | D | O | U | B | L | E | 0 | (definition ...)
191 +---------+---+---+---+---+---+---+---+---+------------- - - - -
194 +--|------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
195 | LINK | 9 | Q | U | A | D | R | U | P | L | E | 0 | 0 | (definition ...)
196 +---------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
202 You shoud be able to see from this how you might implement functions to find a word in
203 the dictionary (just walk along the dictionary entries starting at LATEST and matching
204 the names until you either find a match or hit the NULL pointer at the end of the dictionary),
205 and add a word to the dictionary (create a new definition, set its LINK to LATEST, and set
206 LATEST to point to the new word). We'll see precisely these functions implemented in
207 assembly code later on.
209 One interesting consequence of using a linked list is that you can redefine words, and
210 a newer definition of a word overrides an older one. This is an important concept in
211 FORTH because it means that any word (even "built-in" or "standard" words) can be
212 overridden with a new definition, either to enhance it, to make it faster or even to
213 disable it. However because of the way that FORTH words get compiled, which you'll
214 understand below, words defined using the old definition of a word continue to use
215 the old definition. Only words defined after the new definition use the new definition.
217 DIRECT THREADED CODE ----------------------------------------------------------------------
219 Now we'll get to the really crucial bit in understanding FORTH, so go and get a cup of tea
220 or coffee and settle down. It's fair to say that if you don't understand this section, then you
221 won't "get" how FORTH works, and that would be a failure on my part for not explaining it well.
222 So if after reading this section a few times you don't understand it, please email me
225 Let's talk first about what "threaded code" means. Imagine a peculiar version of C where
226 you are only allowed to call functions without arguments. (Don't worry for now that such a
227 language would be completely useless!) So in our peculiar C, code would look like this:
236 and so on. How would a function, say 'f' above, be compiled by a standard C compiler?
237 Probably into assembly code like this. On the right hand side I've written the actual
241 CALL a E8 08 00 00 00
242 CALL b E8 1C 00 00 00
243 CALL c E8 2C 00 00 00
244 ; ignore the return from the function for now
246 "E8" is the x86 machine code to "CALL" a function. In the first 20 years of computing
247 memory was hideously expensive and we might have worried about the wasted space being used
248 by the repeated "E8" bytes. We can save 20% in code size (and therefore, in expensive memory)
249 by compressing this into just:
251 08 00 00 00 Just the function addresses, without
252 1C 00 00 00 the CALL prefix.
255 [Historical note: If the execution model that FORTH uses looks strange from the following
256 paragraphs, then it was motivated entirely by the need to save memory on early computers.
257 This code compression isn't so important now when our machines have more memory in their L1
258 caches than those early computers had in total, but the execution model still has some
261 Of course this code won't run directly any more. Instead we need to write an interpreter
262 which takes each pair of bytes and calls it.
264 On an i386 machine it turns out that we can write this interpreter rather easily, in just
265 two assembly instructions which turn into just 3 bytes of machine code. Let's store the
266 pointer to the next word to execute in the %esi register:
268 08 00 00 00 <- We're executing this one now. %esi is the _next_ one to execute.
272 The all-important x86 instruction is called LODSL (or in Intel manuals, LODSW). It does
273 two things. Firstly it reads the memory at %esi into the accumulator (%eax). Secondly it
274 increments %esi by 4 bytes. So after LODSL, the situation now looks like this:
276 08 00 00 00 <- We're still executing this one
277 1C 00 00 00 <- %eax now contains this address (0x0000001C)
280 Now we just need to jump to the address in %eax. This is again just a single x86 instruction
281 written JMP *(%eax). And after doing the jump, the situation looks like:
284 1C 00 00 00 <- Now we're executing this subroutine.
287 To make this work, each subroutine is followed by the two instructions 'LODSL; JMP *(%eax)'
288 which literally make the jump to the next subroutine.
290 And that brings us to our first piece of actual code! Well, it's a macro.
299 /* The macro is called NEXT. That's a FORTH-ism. It expands to those two instructions.
301 Every FORTH primitive that we write has to be ended by NEXT. Think of it kind of like
304 The above describes what is known as direct threaded code.
306 To sum up: We compress our function calls down to a list of addresses and use a somewhat
307 magical macro to act as a "jump to next function in the list". We also use one register (%esi)
308 to act as a kind of instruction pointer, pointing to the next function in the list.
310 I'll just give you a hint of what is to come by saying that a FORTH definition such as:
312 : QUADRUPLE DOUBLE DOUBLE ;
314 actually compiles (almost, not precisely but we'll see why in a moment) to a list of
315 function addresses for DOUBLE, DOUBLE and a special function called EXIT to finish off.
317 At this point, REALLY EAGLE-EYED ASSEMBLY EXPERTS are saying "JONES, YOU'VE MADE A MISTAKE!".
319 I lied about JMP *(%eax).
321 INDIRECT THREADED CODE ----------------------------------------------------------------------
323 It turns out that direct threaded code is interesting but only if you want to just execute
324 a list of functions written in assembly language. So QUADRUPLE would work only if DOUBLE
325 was an assembly language function. In the direct threaded code, QUADRUPLE would look like:
328 | addr of DOUBLE --------------------> (assembly code to do the double)
329 +------------------+ NEXT
330 %esi -> | addr of DOUBLE |
333 We can add an extra indirection to allow us to run both words written in assembly language
334 (primitives written for speed) and words written in FORTH themselves as lists of addresses.
336 The extra indirection is the reason for the brackets in JMP *(%eax).
338 Let's have a look at how QUADRUPLE and DOUBLE really look in FORTH:
340 : QUADRUPLE DOUBLE DOUBLE ;
343 | codeword | : DOUBLE DUP + ;
345 | addr of DOUBLE ---------------> +------------------+
346 +------------------+ | codeword |
347 | addr of DOUBLE | +------------------+
348 +------------------+ | addr of DUP --------------> +------------------+
349 | addr of EXIT | +------------------+ | codeword -------+
350 +------------------+ %esi -> | addr of + --------+ +------------------+ |
351 +------------------+ | | assembly to <-----+
352 | addr of EXIT | | | implement DUP |
353 +------------------+ | | .. |
356 | +------------------+
358 +-----> +------------------+
360 +------------------+ |
361 | assembly to <------+
368 This is the part where you may need an extra cup of tea/coffee/favourite caffeinated
369 beverage. What has changed is that I've added an extra pointer to the beginning of
370 the definitions. In FORTH this is sometimes called the "codeword". The codeword is
371 a pointer to the interpreter to run the function. For primitives written in
372 assembly language, the "interpreter" just points to the actual assembly code itself.
373 They don't need interpreting, they just run.
375 In words written in FORTH (like QUADRUPLE and DOUBLE), the codeword points to an interpreter
378 I'll show you the interpreter function shortly, but let's recall our indirect
379 JMP *(%eax) with the "extra" brackets. Take the case where we're executing DOUBLE
380 as shown, and DUP has been called. Note that %esi is pointing to the address of +
382 The assembly code for DUP eventually does a NEXT. That:
384 (1) reads the address of + into %eax %eax points to the codeword of +
385 (2) increments %esi by 4
386 (3) jumps to the indirect %eax jumps to the address in the codeword of +,
387 ie. the assembly code to implement +
392 | addr of DOUBLE ---------------> +------------------+
393 +------------------+ | codeword |
394 | addr of DOUBLE | +------------------+
395 +------------------+ | addr of DUP --------------> +------------------+
396 | addr of EXIT | +------------------+ | codeword -------+
397 +------------------+ | addr of + --------+ +------------------+ |
398 +------------------+ | | assembly to <-----+
399 %esi -> | addr of EXIT | | | implement DUP |
400 +------------------+ | | .. |
403 | +------------------+
405 +-----> +------------------+
407 +------------------+ |
408 now we're | assembly to <------+
409 executing | implement + |
415 So I hope that I've convinced you that NEXT does roughly what you'd expect. This is
416 indirect threaded code.
418 I've glossed over four things. I wonder if you can guess without reading on what they are?
424 My list of four things are: (1) What does "EXIT" do? (2) which is related to (1) is how do
425 you call into a function, ie. how does %esi start off pointing at part of QUADRUPLE, but
426 then point at part of DOUBLE. (3) What goes in the codeword for the words which are written
427 in FORTH? (4) How do you compile a function which does anything except call other functions
428 ie. a function which contains a number like : DOUBLE 2 * ; ?
430 THE INTERPRETER AND RETURN STACK ------------------------------------------------------------
432 Going at these in no particular order, let's talk about issues (3) and (2), the interpreter
433 and the return stack.
435 Words which are defined in FORTH need a codeword which points to a little bit of code to
436 give them a "helping hand" in life. They don't need much, but they do need what is known
437 as an "interpreter", although it doesn't really "interpret" in the same way that, say,
438 Java bytecode used to be interpreted (ie. slowly). This interpreter just sets up a few
439 machine registers so that the word can then execute at full speed using the indirect
440 threaded model above.
442 One of the things that needs to happen when QUADRUPLE calls DOUBLE is that we save the old
443 %esi ("instruction pointer") and create a new one pointing to the first word in DOUBLE.
444 Because we will need to restore the old %esi at the end of DOUBLE (this is, after all, like
445 a function call), we will need a stack to store these "return addresses" (old values of %esi).
447 As you will have read, when reading the background documentation, FORTH has two stacks,
448 an ordinary stack for parameters, and a return stack which is a bit more mysterious. But
449 our return stack is just the stack I talked about in the previous paragraph, used to save
450 %esi when calling from a FORTH word into another FORTH word.
452 In this FORTH, we are using the normal stack pointer (%esp) for the parameter stack.
453 We will use the i386's "other" stack pointer (%ebp, usually called the "frame pointer")
454 for our return stack.
456 I've got two macros which just wrap up the details of using %ebp for the return stack.
457 You use them as for example "PUSHRSP %eax" (push %eax on the return stack) or "POPRSP %ebx"
458 (pop top of return stack into %ebx).
461 /* Macros to deal with the return stack. */
463 lea -4(%ebp),%ebp // push reg on to return stack
468 mov (%ebp),\reg // pop top of return stack to reg
473 And with that we can now talk about the interpreter.
475 In FORTH the interpreter function is often called DOCOL (I think it means "DO COLON" because
476 all FORTH definitions start with a colon, as in : DOUBLE DUP + ;
478 The "interpreter" (it's not really "interpreting") just needs to push the old %esi on the
479 stack and set %esi to the first word in the definition. Remember that we jumped to the
480 function using JMP *(%eax)? Well a consequence of that is that conveniently %eax contains
481 the address of this codeword, so just by adding 4 to it we get the address of the first
482 data word. Finally after setting up %esi, it just does NEXT which causes that first word
486 /* DOCOL - the interpreter! */
490 PUSHRSP %esi // push %esi on to the return stack
491 addl $4,%eax // %eax points to codeword, so make
492 movl %eax,%esi // %esi point to first data word
496 Just to make this absolutely clear, let's see how DOCOL works when jumping from QUADRUPLE
502 +------------------+ DOUBLE:
503 | addr of DOUBLE ---------------> +------------------+
504 +------------------+ %eax -> | addr of DOCOL |
505 %esi -> | addr of DOUBLE | +------------------+
506 +------------------+ | addr of DUP -------------->
507 | addr of EXIT | +------------------+
508 +------------------+ | etc. |
510 First, the call to DOUBLE causes DOCOL (the codeword of DOUBLE). DOCOL does this: It
511 pushes the old %esi on the return stack. %eax points to the codeword of DOUBLE, so we
512 just add 4 on to it to get our new %esi:
517 +------------------+ DOUBLE:
518 | addr of DOUBLE ---------------> +------------------+
519 top of return +------------------+ %eax -> | addr of DOCOL |
520 stack points -> | addr of DOUBLE | + 4 = +------------------+
521 +------------------+ %esi -> | addr of DUP -------------->
522 | addr of EXIT | +------------------+
523 +------------------+ | etc. |
525 Then we do NEXT, and because of the magic of threaded code that increments %esi again
528 Well, it seems to work.
530 One minor point here. Because DOCOL is the first bit of assembly actually to be defined
531 in this file (the others were just macros), and because I usually compile this code with the
532 text segment starting at address 0, DOCOL has address 0. So if you are disassembling the
533 code and see a word with a codeword of 0, you will immediately know that the word is
534 written in FORTH (it's not an assembler primitive) and so uses DOCOL as the interpreter.
536 STARTING UP ----------------------------------------------------------------------
538 Now let's get down to nuts and bolts. When we start the program we need to set up
539 a few things like the return stack. But as soon as we can, we want to jump into FORTH
540 code (albeit much of the "early" FORTH code will still need to be written as
541 assembly language primitives).
543 This is what the set up code does. Does a tiny bit of house-keeping, sets up the
544 separate return stack (NB: Linux gives us the ordinary parameter stack already), then
545 immediately jumps to a FORTH word called COLD. COLD stands for cold-start. In ISO
546 FORTH (but not in this FORTH), COLD can be called at any time to completely reset
547 the state of FORTH, and there is another word called WARM which does a partial reset.
550 /* ELF entry point. */
555 mov %esp,var_S0 // Store the initial data stack pointer.
556 mov $return_stack,%ebp // Initialise the return stack.
558 mov $cold_start,%esi // Initialise interpreter.
559 NEXT // Run interpreter!
562 cold_start: // High-level code without a codeword.
566 We also allocate some space for the return stack and some space to store user
567 definitions. These are static memory allocations using fixed-size buffers, but it
568 wouldn't be a great deal of work to make them dynamic.
572 /* FORTH return stack. */
573 #define RETURN_STACK_SIZE 8192
575 .space RETURN_STACK_SIZE
576 return_stack: // Initial top of return stack.
578 /* Space for user-defined words. */
579 #define USER_DEFS_SIZE 16384
582 .space USER_DEFS_SIZE
585 BUILT-IN WORDS ----------------------------------------------------------------------
587 Remember our dictionary entries (headers). Let's bring those together with the codeword
588 and data words to see how : DOUBLE DUP + ; really looks in memory.
590 pointer to previous word
593 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
594 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
595 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
598 LINK in next word points to codeword of DUP
600 Initially we can't just write ": DOUBLE DUP + ;" (ie. that literal string) here because we
601 don't yet have anything to read the string, break it up at spaces, parse each word, etc. etc.
602 So instead we will have to define built-in words using the GNU assembler data constructors
603 (like .int, .byte, .string, .ascii and so on -- look them up in the gas info page if you are
606 The long way would be:
607 .int <link to previous word>
609 .ascii "DOUBLE" // string
611 DOUBLE: .int DOCOL // codeword
612 .int DUP // pointer to codeword of DUP
613 .int PLUS // pointer to codeword of +
614 .int EXIT // pointer to codeword of EXIT
616 That's going to get quite tedious rather quickly, so here I define an assembler macro
617 so that I can just write:
619 defword "DOUBLE",6,,DOUBLE
622 and I'll get exactly the same effect.
624 Don't worry too much about the exact implementation details of this macro - it's complicated!
627 /* Flags - these are discussed later. */
629 #define F_HIDDEN 0x20
631 // Store the chain of links.
634 .macro defword name, namelen, flags=0, label
640 .set link,name_\label
641 .byte \flags+\namelen // flags + length byte
642 .ascii "\name" // the name
646 .int DOCOL // codeword - the interpreter
647 // list of word pointers follow
651 Similarly I want a way to write words written in assembly language. There will quite a few
652 of these to start with because, well, everything has to start in assembly before there's
653 enough "infrastructure" to be able to start writing FORTH words, but also I want to define
654 some common FORTH words in assembly language for speed, even though I could write them in FORTH.
656 This is what DUP looks like in memory:
658 pointer to previous word
661 +--|------+---+---+---+---+------------+
662 | LINK | 3 | D | U | P | code_DUP ---------------------> points to the assembly
663 +---------+---+---+---+---+------------+ code used to write DUP,
664 ^ len codeword which is ended with NEXT.
668 Again, for brevity in writing the header I'm going to write an assembler macro called defcode.
671 .macro defcode name, namelen, flags=0, label
677 .set link,name_\label
678 .byte \flags+\namelen // flags + length byte
679 .ascii "\name" // the name
683 .int code_\label // codeword
687 code_\label : // assembler code follows
691 Now some easy FORTH primitives. These are written in assembly for speed. If you understand
692 i386 assembly language then it is worth reading these. However if you don't understand assembly
693 you can skip the details.
697 pop %eax // duplicate top of stack
702 defcode "DROP",4,,DROP
703 pop %eax // drop top of stack
706 defcode "SWAP",4,,SWAP
707 pop %eax // swap top of stack
713 defcode "OVER",4,,OVER
714 mov 4(%esp),%eax // get the second element of stack
715 push %eax // and push it on top
727 defcode "-ROT",4,,NROT
737 incl (%esp) // increment top of stack
741 decl (%esp) // decrement top of stack
744 defcode "4+",2,,INCR4
745 addl $4,(%esp) // increment top of stack
748 defcode "4-",2,,DECR4
749 subl $4,(%esp) // decrement top of stack
753 pop %eax // get top of stack
754 addl %eax,(%esp) // and add it to next word on stack
758 pop %eax // get top of stack
759 subl %eax,(%esp) // and subtract if from next word on stack
766 push %eax // ignore overflow
774 push %eax // push quotient
782 push %edx // push remainder
785 defcode "=",1,,EQU // top two words are equal?
795 defcode "<>",2,,NEQU // top two words are not equal?
805 defcode "0=",2,,ZEQU // top of stack equals 0?
824 defcode "INVERT",6,,INVERT // this is the FORTH "NOT" function
829 RETURNING FROM FORTH WORDS ----------------------------------------------------------------------
831 Time to talk about what happens when we EXIT a function. In this diagram QUADRUPLE has called
832 DOUBLE, and DOUBLE is about to exit (look at where %esi is pointing):
837 +------------------+ DOUBLE
838 | addr of DOUBLE ---------------> +------------------+
839 +------------------+ | codeword |
840 | addr of DOUBLE | +------------------+
841 +------------------+ | addr of DUP |
842 | addr of EXIT | +------------------+
843 +------------------+ | addr of + |
845 %esi -> | addr of EXIT |
848 What happens when the + function does NEXT? Well, the following code is executed.
851 defcode "EXIT",4,,EXIT
852 POPRSP %esi // pop return stack into %esi
856 EXIT gets the old %esi which we saved from before on the return stack, and puts it in %esi.
857 So after this (but just before NEXT) we get:
862 +------------------+ DOUBLE
863 | addr of DOUBLE ---------------> +------------------+
864 +------------------+ | codeword |
865 %esi -> | addr of DOUBLE | +------------------+
866 +------------------+ | addr of DUP |
867 | addr of EXIT | +------------------+
868 +------------------+ | addr of + |
873 And NEXT just completes the job by, well in this case just by calling DOUBLE again :-)
875 LITERALS ----------------------------------------------------------------------
877 The final point I "glossed over" before was how to deal with functions that do anything
878 apart from calling other functions. For example, suppose that DOUBLE was defined like this:
882 It does the same thing, but how do we compile it since it contains the literal 2? One way
883 would be to have a function called "2" (which you'd have to write in assembler), but you'd need
884 a function for every single literal that you wanted to use.
886 FORTH solves this by compiling the function using a special word called LIT:
888 +---------------------------+-------+-------+-------+-------+-------+
889 | (usual header of DOUBLE) | DOCOL | LIT | 2 | * | EXIT |
890 +---------------------------+-------+-------+-------+-------+-------+
892 LIT is executed in the normal way, but what it does next is definitely not normal. It
893 looks at %esi (which now points to the literal 2), grabs it, pushes it on the stack, then
894 manipulates %esi in order to skip the literal as if it had never been there.
896 What's neat is that the whole grab/manipulate can be done using a single byte single
897 i386 instruction, our old friend LODSL. Rather than me drawing more ASCII-art diagrams,
898 see if you can find out how LIT works:
902 // %esi points to the next command, but in this case it points to the next
903 // literal 32 bit integer. Get that literal into %eax and increment %esi.
904 // On x86, it's a convenient single byte instruction! (cf. NEXT macro)
906 push %eax // push the literal number on to stack
910 MEMORY ----------------------------------------------------------------------
912 As important point about FORTH is that it gives you direct access to the lowest levels
913 of the machine. Manipulating memory directly is done frequently in FORTH, and these are
914 the primitive words for doing it.
918 pop %ebx // address to store at
919 pop %eax // data to store there
920 mov %eax,(%ebx) // store it
924 pop %ebx // address to fetch
925 mov (%ebx),%eax // fetch it
926 push %eax // push value onto stack
929 defcode "+!",2,,ADDSTORE
931 pop %eax // the amount to add
932 addl %eax,(%ebx) // add it
935 defcode "-!",2,,SUBSTORE
937 pop %eax // the amount to subtract
938 subl %eax,(%ebx) // add it
941 /* ! and @ (STORE and FETCH) store 32-bit words. It's also useful to be able to read and write bytes.
942 * I don't know whether FORTH has these words, so I invented my own, called !b and @b.
943 * Byte-oriented operations only work on architectures which permit them (i386 is one of those).
944 * UPDATE: writing a byte to the dictionary pointer is called C, in FORTH.
946 defcode "!b",2,,STOREBYTE
947 pop %ebx // address to store at
948 pop %eax // data to store there
949 movb %al,(%ebx) // store it
952 defcode "@b",2,,FETCHBYTE
953 pop %ebx // address to fetch
955 movb (%ebx),%al // fetch it
956 push %eax // push value onto stack
960 BUILT-IN VARIABLES ----------------------------------------------------------------------
962 These are some built-in variables and related standard FORTH words. Of these, the only one that we
963 have discussed so far was LATEST, which points to the last (most recently defined) word in the
964 FORTH dictionary. LATEST is also a FORTH word which pushes the address of LATEST (the variable)
965 on to the stack, so you can read or write it using @ and ! operators. For example, to print
966 the current value of LATEST (and this can apply to any FORTH variable) you would do:
970 To make defining variables shorter, I'm using a macro called defvar, similar to defword and
971 defcode above. (In fact the defvar macro uses defcode to do the dictionary header).
974 .macro defvar name, namelen, flags=0, label, initial=0
975 defcode \name,\namelen,\flags,\label
985 The built-in variables are:
987 STATE Is the interpreter executing code (0) or compiling a word (non-zero)?
988 LATEST Points to the latest (most recently defined) word in the dictionary.
989 HERE Points to the next free byte of memory. When compiling, compiled words go here.
990 _X These are three scratch variables, used by some standard dictionary words.
993 S0 Stores the address of the top of the parameter stack.
994 R0 Stores the address of the top of the return stack.
997 defvar "STATE",5,,STATE
998 defvar "HERE",4,,HERE,user_defs_start
999 defvar "LATEST",6,,LATEST,name_SYSEXIT // SYSEXIT must be last in built-in dictionary
1004 defvar "R0",2,,RZ,return_stack
1007 RETURN STACK ----------------------------------------------------------------------
1009 These words allow you to access the return stack. Recall that the register %ebp always points to
1010 the top of the return stack.
1014 pop %eax // pop parameter stack into %eax
1015 PUSHRSP %eax // push it on to the return stack
1018 defcode "R>",2,,FROMR
1019 POPRSP %eax // pop return stack on to %eax
1020 push %eax // and push on to parameter stack
1023 defcode "RSP@",4,,RSPFETCH
1027 defcode "RSP!",4,,RSPSTORE
1031 defcode "RDROP",5,,RDROP
1032 lea 4(%ebp),%ebp // pop return stack and throw away
1036 PARAMETER (DATA) STACK ----------------------------------------------------------------------
1038 These functions allow you to manipulate the parameter stack. Recall that Linux sets up the parameter
1039 stack for us, and it is accessed through %esp.
1042 defcode "DSP@",4,,DSPFETCH
1047 defcode "DSP!",4,,DSPSTORE
1052 INPUT AND OUTPUT ----------------------------------------------------------------------
1054 These are our first really meaty/complicated FORTH primitives. I have chosen to write them in
1055 assembler, but surprisingly in "real" FORTH implementations these are often written in terms
1056 of more fundamental FORTH primitives. I chose to avoid that because I think that just obscures
1057 the implementation. After all, you may not understand assembler but you can just think of it
1058 as an opaque block of code that does what it says.
1060 Let's discuss input first.
1062 The FORTH word KEY reads the next byte from stdin (and pushes it on the parameter stack).
1063 So if KEY is called and someone hits the space key, then the number 32 (ASCII code of space)
1064 is pushed on the stack.
1066 In FORTH there is no distinction between reading code and reading input. We might be reading
1067 and compiling code, we might be reading words to execute, we might be asking for the user
1068 to type their name -- ultimately it all comes in through KEY.
1070 The implementation of KEY uses an input buffer so a certain size (defined at the end of the
1071 program). It calls the Linux read(2) system call to fill this buffer and tracks its position
1072 in the buffer using a couple of variables, and if it runs out of input buffer then it refills
1073 it automatically. The other thing that KEY does is if it detects that stdin has closed, it
1074 exits the program, which is why when you hit ^D the FORTH system cleanly exits.
1077 #include <asm-i386/unistd.h>
1079 defcode "KEY",3,,KEY
1081 push %eax // push return value on stack
1093 1: // out of input; use read(2) to fetch more input from stdin
1094 xor %ebx,%ebx // 1st param: stdin
1095 mov $buffer,%ecx // 2nd param: buffer
1097 mov $buffend-buffer,%edx // 3rd param: max length
1098 mov $__NR_read,%eax // syscall: read
1100 test %eax,%eax // If %eax <= 0, then exit.
1102 addl %eax,%ecx // buffer+%eax = bufftop
1106 2: // error or out of input: exit
1108 mov $__NR_exit,%eax // syscall: exit
1112 By contrast, output is much simpler. The FORTH word EMIT writes out a single byte to stdout.
1113 This implementation just uses the write system call. No attempt is made to buffer output, but
1114 it would be a good exercise to add it.
1117 defcode "EMIT",4,,EMIT
1122 mov $1,%ebx // 1st param: stdout
1124 // write needs the address of the byte to write
1126 mov $2f,%ecx // 2nd param: address
1128 mov $1,%edx // 3rd param: nbytes = 1
1130 mov $__NR_write,%eax // write syscall
1135 2: .space 1 // scratch used by EMIT
1138 Back to input, WORD is a FORTH word which reads the next full word of input.
1140 What it does in detail is that it first skips any blanks (spaces, tabs, newlines and so on).
1141 Then it calls KEY to read characters into an internal buffer until it hits a blank. Then it
1142 calculates the length of the word it read and returns the address and the length as
1143 two words on the stack (with address at the top).
1145 Notice that WORD has a single internal buffer which it overwrites each time (rather like
1146 a static C string). Also notice that WORD's internal buffer is just 32 bytes long and
1147 there is NO checking for overflow. 31 bytes happens to be the maximum length of a
1148 FORTH word that we support, and that is what WORD is used for: to read FORTH words when
1149 we are compiling and executing code. The returned strings are not NUL-terminated, so
1150 in some crazy-world you could define FORTH words containing ASCII NULs, although why
1151 you'd want to is a bit beyond me.
1153 WORD is not suitable for just reading strings (eg. user input) because of all the above
1154 peculiarities and limitations.
1156 Note that when executing, you'll see:
1158 which puts "FOO" and length 3 on the stack, but when compiling:
1160 is an error (or at least it doesn't do what you might expect). Later we'll talk about compiling
1161 and immediate mode, and you'll understand why.
1164 defcode "WORD",4,,WORD
1166 push %ecx // push length
1167 push %edi // push base address
1171 /* Search for first non-blank character. Also skip \ comments. */
1173 call _KEY // get next key, returned in %eax
1174 cmpb $'\\',%al // start of a comment?
1175 je 3f // if so, skip the comment
1177 jbe 1b // if so, keep looking
1179 /* Search for the end of the word, storing chars as we go. */
1180 mov $5f,%edi // pointer to return buffer
1182 stosb // add character to return buffer
1183 call _KEY // get next key, returned in %al
1184 cmpb $' ',%al // is blank?
1185 ja 2b // if not, keep looping
1187 /* Return the word (well, the static buffer) and length. */
1189 mov %edi,%ecx // return length of the word
1190 mov $5f,%edi // return address of the word
1193 /* Code to skip \ comments to end of the current line. */
1196 cmpb $'\n',%al // end of line yet?
1201 // A static buffer where WORD returns. Subsequent calls
1202 // overwrite this buffer. Maximum word length is 32 chars.
1206 . (also called DOT) prints the top of the stack as an integer. In real FORTH implementations
1207 it should print it in the current base, but this assembler version is simpler and can only
1210 Remember that you can override even built-in FORTH words easily, so if you want to write a
1211 more advanced DOT then you can do so easily at a later point, and probably in FORTH.
1215 pop %eax // Get the number to print into %eax
1216 call _DOT // Easier to do this recursively ...
1219 mov $10,%ecx // Base 10
1223 xor %edx,%edx // %edx:%eax / %ecx -> quotient %eax, remainder %edx
1238 Almost the opposite of DOT (but not quite), SNUMBER parses a numeric string such as one returned
1239 by WORD and pushes the number on the parameter stack.
1241 This function does absolutely no error checking, and in particular the length of the string
1242 must be >= 1 bytes, and should contain only digits 0-9. If it doesn't you'll get random results.
1244 This function is only used when reading literal numbers in code, and shouldn't really be used
1245 in user code at all.
1247 defcode "SNUMBER",7,,SNUMBER
1257 imull $10,%eax // %eax *= 10
1260 subb $'0',%bl // ASCII -> digit
1267 DICTIONARY LOOK UPS ----------------------------------------------------------------------
1269 We're building up to our prelude on how FORTH code is compiled, but first we need yet more infrastructure.
1271 The FORTH word FIND takes a string (a word as parsed by WORD -- see above) and looks it up in the
1272 dictionary. What it actually returns is the address of the dictionary header, if it finds it,
1275 So if DOUBLE is defined in the dictionary, then WORD DOUBLE FIND returns the following pointer:
1281 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1282 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1283 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1285 See also >CFA which takes a dictionary entry pointer and returns a pointer to the codeword.
1287 FIND doesn't find dictionary entries which are flagged as HIDDEN. See below for why.
1290 defcode "FIND",4,,FIND
1291 pop %edi // %edi = address
1292 pop %ecx // %ecx = length
1298 push %esi // Save %esi so we can use it in string comparison.
1300 // Now we start searching backwards through the dictionary for this word.
1301 mov var_LATEST,%edx // LATEST points to name header of the latest word in the dictionary
1303 test %edx,%edx // NULL pointer? (end of the linked list)
1306 // Compare the length expected and the length of the word.
1307 // Note that if the F_HIDDEN flag is set on the word, then by a bit of trickery
1308 // this won't pick the word (the length will appear to be wrong).
1310 movb 4(%edx),%al // %al = flags+length field
1311 andb $(F_HIDDEN|0x1f),%al // %al = name length
1312 cmpb %cl,%al // Length is the same?
1315 // Compare the strings in detail.
1316 push %ecx // Save the length
1317 push %edi // Save the address (repe cmpsb will move this pointer)
1318 lea 5(%edx),%esi // Dictionary string we are checking against.
1319 repe cmpsb // Compare the strings.
1322 jne 2f // Not the same.
1324 // The strings are the same - return the header pointer in %eax
1330 mov (%edx),%edx // Move back through the link field to the previous word
1331 jmp 1b // .. and loop.
1335 xor %eax,%eax // Return zero to indicate not found.
1339 FIND returns the dictionary pointer, but when compiling we need the codeword pointer (recall
1340 that FORTH definitions are compiled into lists of codeword pointers).
1342 In the example below, WORD DOUBLE FIND >CFA
1344 FIND returns a pointer to this
1345 | >CFA converts it to a pointer to this
1348 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1349 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1350 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1354 Because names vary in length, this isn't just a simple increment.
1356 In this FORTH you cannot easily turn a codeword pointer back into a dictionary entry pointer, but
1357 that is not true in most FORTH implementations where they store a back pointer in the definition
1358 (with an obvious memory/complexity cost). The reason they do this is that it is useful to be
1359 able to go backwards (codeword -> dictionary entry) in order to decompile FORTH definitions.
1362 defcode ">CFA",4,,TCFA
1369 add $4,%edi // Skip link pointer.
1370 movb (%edi),%al // Load flags+len into %al.
1371 inc %edi // Skip flags+len byte.
1372 andb $0x1f,%al // Just the length, not the flags.
1373 add %eax,%edi // Skip the name.
1374 addl $3,%edi // The codeword is 4-byte aligned.
1379 COMPILING ----------------------------------------------------------------------
1381 Now we'll talk about how FORTH compiles words. Recall that a word definition looks like this:
1385 and we have to turn this into:
1387 pointer to previous word
1390 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1391 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1392 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
1393 ^ len pad codeword |
1395 LATEST points here points to codeword of DUP
1397 There are several problems to solve. Where to put the new word? How do we read words? How
1398 do we define : (COLON) and ; (SEMICOLON)?
1400 FORTH solves this rather elegantly and as you might expect in a very low-level way which
1401 allows you to change how the compiler works in your own code.
1403 FORTH has an INTERPRETER function (a true interpreter this time, not DOCOL) which runs in a
1404 loop, reading words (using WORD), looking them up (using FIND), turning them into codeword
1405 points (using >CFA) and deciding what to do with them. What it does depends on the mode
1406 of the interpreter (in variable STATE). When STATE is zero, the interpreter just runs
1407 each word as it looks them up. (Known as immediate mode).
1409 The interesting stuff happens when STATE is non-zero -- compiling mode. In this mode the
1410 interpreter just appends the codeword pointers to user memory (the HERE variable points to
1411 the next free byte of user memory).
1413 So you may be able to see how we could define : (COLON). The general plan is:
1415 (1) Use WORD to read the name of the function being defined.
1417 (2) Construct the dictionary entry header in user memory:
1419 pointer to previous word (from LATEST) +-- Afterwards, HERE points here, where
1420 ^ | the interpreter will start appending
1422 +--|------+---+---+---+---+---+---+---+---+------------+
1423 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL |
1424 +---------+---+---+---+---+---+---+---+---+------------+
1427 (3) Set LATEST to point to the newly defined word and most importantly leave HERE pointing
1428 just after the new codeword. This is where the interpreter will append codewords.
1430 (4) Set STATE to 1. Go into compile mode so the interpreter starts appending codewords.
1432 After : has run, our input is here:
1437 Next byte returned by KEY
1439 so the interpreter (now it's in compile mode, so I guess it's really the compiler) reads DUP,
1440 gets its codeword pointer, and appends it:
1442 +-- HERE updated to point here.
1445 +---------+---+---+---+---+---+---+---+---+------------+------------+
1446 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP |
1447 +---------+---+---+---+---+---+---+---+---+------------+------------+
1450 Next we read +, get the codeword pointer, and append it:
1452 +-- HERE updated to point here.
1455 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1456 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + |
1457 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1460 The issue is what happens next. Obviously what we _don't_ want to happen is that we
1461 read ; and compile it and go on compiling everything afterwards.
1463 At this point, FORTH uses a trick. Remember the length byte in the dictionary definition
1464 isn't just a plain length byte, but can also contain flags. One flag is called the
1465 IMMEDIATE flag (F_IMMED in this code). If a word in the dictionary is flagged as
1466 IMMEDIATE then the interpreter runs it immediately _even if it's in compile mode_.
1468 I hope I don't need to explain that ; (SEMICOLON) is an IMMEDIATE flagged word. And
1469 all it does is append the codeword for EXIT on to the current definition and switch
1470 back to immediate mode (set STATE back to 0). After executing ; we get this:
1472 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1473 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1474 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1479 And that's it, job done, our new definition is compiled.
1481 The only last wrinkle in this is that while our word was being compiled, it was in a
1482 half-finished state. We certainly wouldn't want DOUBLE to be called somehow during
1483 this time. There are several ways to stop this from happening, but in FORTH what we
1484 do is flag the word with the HIDDEN flag (F_HIDDEN in this code) just while it is
1485 being compiled. This prevents FIND from finding it, and thus in theory stops any
1486 chance of it being called.
1488 Compared to the description above, the actual definition of : (COLON) is comparatively simple:
1491 defcode ":",1,,COLON
1493 // Get the word and create a dictionary entry header for it.
1494 call _WORD // Returns %ecx = length, %edi = pointer to word.
1495 mov %edi,%ebx // %ebx = address of the word
1497 movl var_HERE,%edi // %edi is the address of the header
1498 movl var_LATEST,%eax // Get link pointer
1499 stosl // and store it in the header.
1501 mov %cl,%al // Get the length.
1502 orb $F_HIDDEN,%al // Set the HIDDEN flag on this entry.
1503 stosb // Store the length/flags byte.
1505 mov %ebx,%esi // %esi = word
1506 rep movsb // Copy the word
1508 addl $3,%edi // Align to next 4 byte boundary.
1511 movl $DOCOL,%eax // The codeword for user-created words is always DOCOL (the interpreter)
1514 // Header built, so now update LATEST and HERE.
1515 // We'll be compiling words and putting them HERE.
1517 movl %eax,var_LATEST
1520 // And go into compile mode by setting STATE to 1.
1525 , (COMMA) is a standard FORTH word which appends a 32 bit integer (normally a codeword
1526 pointer) to the user data area pointed to by HERE, and adds 4 to HERE.
1529 defcode ",",1,,COMMA
1530 pop %eax // Code pointer to store.
1534 movl var_HERE,%edi // HERE
1536 movl %edi,var_HERE // Update HERE (incremented)
1540 ; (SEMICOLON) is also elegantly simple. Notice the F_IMMED flag.
1543 defcode ";",1,F_IMMED,SEMICOLON
1544 movl $EXIT,%eax // EXIT is the final codeword in compiled words.
1545 call _COMMA // Store it.
1546 call _HIDDEN // Toggle the HIDDEN flag (unhides the new word).
1547 xor %eax,%eax // Set STATE to 0 (back to execute mode).
1552 IMMEDIATE mode words aren't just for the FORTH compiler to use. You can define your
1553 own IMMEDIATE words too. The IMMEDIATE word toggles the F_IMMED (IMMEDIATE flag) on the
1554 most recently defined word, or on the current word if you call it in the middle of a
1559 : MYIMMEDWORD IMMEDIATE
1563 but some FORTH programmers write this instead:
1569 The two are basically equivalent.
1572 defcode "IMMEDIATE",9,F_IMMED,IMMEDIATE
1576 movl var_LATEST,%edi // LATEST word.
1577 addl $4,%edi // Point to name/flags byte.
1578 xorb $F_IMMED,(%edi) // Toggle the IMMED bit.
1582 HIDDEN toggles the other flag, F_HIDDEN, of the latest word. Note that words flagged
1583 as hidden are defined but cannot be called, so this is rarely used.
1586 defcode "HIDDEN",6,,HIDDEN
1590 movl var_LATEST,%edi // LATEST word.
1591 addl $4,%edi // Point to name/flags byte.
1592 xorb $F_HIDDEN,(%edi) // Toggle the HIDDEN bit.
1596 ' (TICK) is a standard FORTH word which returns the codeword pointer of the next word.
1598 The common usage is:
1602 which appends the codeword of FOO to the current word we are defining (this only works in compiled code).
1604 You tend to use ' in IMMEDIATE words. For example an alternate (and rather useless) way to define
1605 a literal 2 might be:
1608 ' LIT , \ Appends LIT to the currently-being-defined word
1609 2 , \ Appends the number 2 to the currently-being-defined word
1616 (If you don't understand how LIT2 works, then you should review the material about compiling words
1617 and immediate mode).
1619 This definition of ' uses a cheat which I copied from buzzard92. As a result it only works in
1623 lodsl // Get the address of the next word and skip it.
1624 pushl %eax // Push it on the stack.
1628 BRANCHING ----------------------------------------------------------------------
1630 It turns out that all you need in order to define looping constructs, IF-statements, etc.
1633 BRANCH is an unconditional branch. 0BRANCH is a conditional branch (it only branches if the
1634 top of stack is zero).
1636 This is how BRANCH works. When BRANCH executes, %esi starts by pointing to the offset:
1638 +---------------------+-------+---- - - ---+------------+------------+---- - - - ----+------------+
1639 | (Dictionary header) | DOCOL | | BRANCH | offset | (skipped) | word |
1640 +---------------------+-------+---- - - ---+------------+-----|------+---- - - - ----+------------+
1643 | +-----------------------+
1644 %esi added to offset
1646 The offset is added to %esi to make the new %esi, and the result is that when NEXT runs, execution
1647 continues at the branch target. Negative offsets work as expected.
1649 0BRANCH is the same except the branch happens conditionally.
1651 Now standard FORTH words such as IF, THEN, ELSE, WHILE, REPEAT, etc. are implemented entirely
1652 in FORTH. They are IMMEDIATE words which append various combinations of BRANCH or 0BRANCH
1653 into the word currently being compiled.
1655 As an example, code written like this:
1657 condition-code IF true-part THEN rest-code
1661 condition-code 0BRANCH OFFSET true-part rest-code
1667 defcode "BRANCH",6,,BRANCH
1668 add (%esi),%esi // add the offset to the instruction pointer
1671 defcode "0BRANCH",7,,ZBRANCH
1673 test %eax,%eax // top of stack is zero?
1674 jz code_BRANCH // if so, jump back to the branch function above
1675 lodsl // otherwise we need to skip the offset
1679 PRINTING STRINGS ----------------------------------------------------------------------
1681 LITSTRING and EMITSTRING are primitives used to implement the ." operator (which is
1682 written in FORTH). See the definition of that operator below.
1685 defcode "LITSTRING",9,,LITSTRING
1686 lodsl // get the length of the string
1687 push %eax // push it on the stack
1688 push %esi // push the address of the start of the string
1689 addl %eax,%esi // skip past the string
1690 addl $3,%esi // but round up to next 4 byte boundary
1694 defcode "EMITSTRING",10,,EMITSTRING
1695 mov $1,%ebx // 1st param: stdout
1696 pop %ecx // 2nd param: address of string
1697 pop %edx // 3rd param: length of string
1698 mov $__NR_write,%eax // write syscall
1703 COLD START AND INTERPRETER ----------------------------------------------------------------------
1705 COLD is the first FORTH function called, almost immediately after the FORTH system "boots".
1707 INTERPRETER is the FORTH interpreter ("toploop", "toplevel" or REPL might be a more accurate
1712 // COLD must not return (ie. must not call EXIT).
1713 defword "COLD",4,,COLD
1714 .int INTERPRETER // call the interpreter loop (never returns)
1715 .int LIT,1,SYSEXIT // hmmm, but in case it does, exit(1).
1717 /* This interpreter is pretty simple, but remember that in FORTH you can always override
1718 * it later with a more powerful one!
1720 defword "INTERPRETER",11,,INTERPRETER
1721 .int INTERPRET,RDROP,INTERPRETER
1723 defcode "INTERPRET",9,,INTERPRET
1724 call _WORD // Returns %ecx = length, %edi = pointer to word.
1726 // Is it in the dictionary?
1728 movl %eax,interpret_is_lit // Not a literal number (not yet anyway ...)
1729 call _FIND // Returns %eax = pointer to header or 0 if not found.
1730 test %eax,%eax // Found?
1733 // In the dictionary. Is it an IMMEDIATE codeword?
1734 mov %eax,%edi // %edi = dictionary entry
1735 movb 4(%edi),%al // Get name+flags.
1736 push %ax // Just save it for now.
1737 call _TCFA // Convert dictionary entry (in %edi) to codeword pointer.
1739 andb $F_IMMED,%al // Is IMMED flag set?
1741 jnz 4f // If IMMED, jump straight to executing.
1745 1: // Not in the dictionary (not a word) so assume it's a literal number.
1746 incl interpret_is_lit
1747 call _SNUMBER // Returns the parsed number in %eax
1749 mov $LIT,%eax // The word is LIT
1751 2: // Are we compiling or executing?
1754 jz 4f // Jump if executing.
1756 // Compiling - just append the word to the current dictionary definition.
1758 mov interpret_is_lit,%ecx // Was it a literal?
1761 mov %ebx,%eax // Yes, so LIT is followed by a number.
1765 4: // Executing - run it!
1766 mov interpret_is_lit,%ecx // Literal?
1767 test %ecx,%ecx // Literal?
1770 // Not a literal, execute it now. This never returns, but the codeword will
1771 // eventually call NEXT which will reenter the loop in INTERPRETER.
1774 5: // Executing a literal, which means push it on the stack.
1781 .int 0 // Flag used to record if reading a literal
1784 ODDS AND ENDS ----------------------------------------------------------------------
1786 CHAR puts the ASCII code of the first character of the following word on the stack. For example
1787 CHAR A puts 65 on the stack.
1789 SYSEXIT pops the status off the stack and exits the process (using Linux exit syscall).
1792 defcode "CHAR",4,,CHAR
1793 call _WORD // Returns %ecx = length, %edi = pointer to word.
1795 movb (%edi),%al // Get the first character of the word.
1796 push %eax // Push it onto the stack.
1799 // NB: SYSEXIT must be the last entry in the built-in dictionary.
1800 defcode SYSEXIT,7,,SYSEXIT
1806 START OF FORTH CODE ----------------------------------------------------------------------
1808 We've now reached the stage where the FORTH system is running and self-hosting. All further
1809 words can be written as FORTH itself, including words like IF, THEN, .", etc which in most
1810 languages would be considered rather fundamental.
1812 As a kind of trick, I prefill the input buffer with the initial FORTH code. Once this code
1813 has run (when we get to the "OK" prompt), this input buffer is reused for reading user input.
1815 Some notes about the code:
1817 \ (backslash) is the FORTH way to start a comment which goes up to the next newline. However
1818 because this is a C-style string, I have to escape the backslash, which is why they appear as
1821 Similarly, any backslashes in the code are doubled, and " becomes \" (eg. the definition of ."
1822 is written as : .\" ... ;)
1824 I use indenting to show structure. The amount of whitespace has no meaning to FORTH however
1825 except that you must use at least one whitespace character between words, and words themselves
1826 cannot contain whitespace.
1828 FORTH is case-sensitive. Use capslock!
1836 // Multi-line constant gives 'Warning: unterminated string; newline inserted' messages which you can ignore
1838 \\ Define some character constants
1844 \\ CR prints a carriage return
1847 \\ SPACE prints a space
1848 : SPACE 'SPACE' EMIT ;
1850 \\ Primitive . (DOT) function doesn't follow with a blank, so redefine it to behave like FORTH.
1851 \\ Notice how we can trivially redefine existing functions.
1854 \\ DUP, DROP are defined in assembly for speed, but this is how you might define them
1855 \\ in FORTH. Notice use of the scratch variables _X and _Y.
1856 \\ : DUP _X ! _X @ _X @ ;
1859 \\ The 2... versions of the standard operators work on pairs of stack entries. They're not used
1860 \\ very commonly so not really worth writing in assembler. Here is how they are defined in FORTH.
1864 \\ More standard FORTH words.
1868 \\ [ and ] allow you to break into immediate mode while compiling a word.
1869 : [ IMMEDIATE \\ define [ as an immediate word
1870 0 STATE ! \\ go into immediate mode
1874 1 STATE ! \\ go back to compile mode
1877 \\ LITERAL takes whatever is on the stack and compiles LIT <foo>
1879 ' LIT , \\ compile LIT
1880 , \\ compile the literal itself (from the stack)
1883 \\ condition IF true-part THEN rest
1885 \\ condition 0BRANCH OFFSET true-part rest
1886 \\ where OFFSET is the offset of 'rest'
1887 \\ condition IF true-part ELSE false-part THEN
1889 \\ condition 0BRANCH OFFSET true-part BRANCH OFFSET2 false-part rest
1890 \\ where OFFSET if the offset of false-part and OFFSET2 is the offset of rest
1892 \\ IF is an IMMEDIATE word which compiles 0BRANCH followed by a dummy offset, and places
1893 \\ the address of the 0BRANCH on the stack. Later when we see THEN, we pop that address
1894 \\ off the stack, calculate the offset, and back-fill the offset.
1896 ' 0BRANCH , \\ compile 0BRANCH
1897 HERE @ \\ save location of the offset on the stack
1898 0 , \\ compile a dummy offset
1903 HERE @ SWAP - \\ calculate the offset from the address saved on the stack
1904 SWAP ! \\ store the offset in the back-filled location
1908 ' BRANCH , \\ definite branch to just over the false-part
1909 HERE @ \\ save location of the offset on the stack
1910 0 , \\ compile a dummy offset
1911 SWAP \\ now back-fill the original (IF) offset
1912 DUP \\ same as for THEN word above
1917 \\ BEGIN loop-part condition UNTIL
1919 \\ loop-part condition 0BRANCH OFFSET
1920 \\ where OFFSET points back to the loop-part
1921 \\ This is like do { loop-part } while (condition) in the C language
1923 HERE @ \\ save location on the stack
1927 ' 0BRANCH , \\ compile 0BRANCH
1928 HERE @ - \\ calculate the offset from the address saved on the stack
1929 , \\ compile the offset here
1932 \\ BEGIN loop-part AGAIN
1934 \\ loop-part BRANCH OFFSET
1935 \\ where OFFSET points back to the loop-part
1936 \\ In other words, an infinite loop which can only be returned from with EXIT
1938 ' BRANCH , \\ compile BRANCH
1939 HERE @ - \\ calculate the offset back
1940 , \\ compile the offset here
1943 \\ BEGIN condition WHILE loop-part REPEAT
1945 \\ condition 0BRANCH OFFSET2 loop-part BRANCH OFFSET
1946 \\ where OFFSET points back to condition (the beginning) and OFFSET2 points to after the whole piece of code
1947 \\ So this is like a while (condition) { loop-part } loop in the C language
1949 ' 0BRANCH , \\ compile 0BRANCH
1950 HERE @ \\ save location of the offset2 on the stack
1951 0 , \\ compile a dummy offset2
1955 ' BRANCH , \\ compile BRANCH
1956 SWAP \\ get the original offset (from BEGIN)
1957 HERE @ - , \\ and compile it after BRANCH
1959 HERE @ SWAP - \\ calculate the offset2
1960 SWAP ! \\ and back-fill it in the original location
1963 \\ With the looping constructs, we can now write SPACES, which writes n spaces to stdout.
1966 SPACE \\ print a space
1967 1- \\ until we count down to 0
1972 \\ .S prints the contents of the stack. Very useful for debugging.
1974 DSP@ \\ get current stack pointer
1976 DUP @ . \\ print the stack element
1978 DUP S0 @ 4- = \\ stop when we get to the top
1983 \\ DEPTH returns the depth of the stack.
1984 : DEPTH S0 @ DSP@ - ;
1986 \\ .\" is the print string operator in FORTH. Example: .\" Something to print\"
1987 \\ The space after the operator is the ordinary space required between words.
1988 \\ This is tricky to define because it has to do different things depending on whether
1989 \\ we are compiling or in immediate mode. (Thus the word is marked IMMEDIATE so it can
1990 \\ detect this and do different things).
1991 \\ In immediate mode we just keep reading characters and printing them until we get to
1992 \\ the next double quote.
1993 \\ In compile mode we have the problem of where we're going to store the string (remember
1994 \\ that the input buffer where the string comes from may be overwritten by the time we
1995 \\ come round to running the function). We store the string in the compiled function
1997 \\ LITSTRING, string length, string rounded up to 4 bytes, EMITSTRING, ...
1999 STATE @ \\ compiling?
2001 ' LITSTRING , \\ compile LITSTRING
2002 HERE @ \\ save the address of the length word on the stack
2003 0 , \\ dummy length - we don't know what it is yet
2005 KEY \\ get next character of the string
2008 HERE @ !b \\ store the character in the compiled image
2009 1 HERE +! \\ increment HERE pointer by 1 byte
2011 DROP \\ drop the double quote character at the end
2012 DUP \\ get the saved address of the length word
2013 HERE @ SWAP - \\ calculate the length
2014 4- \\ subtract 4 (because we measured from the start of the length word)
2015 SWAP ! \\ and back-fill the length location
2016 HERE @ \\ round up to next multiple of 4 bytes for the remaining code
2020 ' EMITSTRING , \\ compile the final EMITSTRING
2022 \\ In immediate mode, just read characters and print them until we get
2023 \\ to the ending double quote. Much simpler than the above code!
2026 DUP '\"' = IF EXIT THEN
2032 \\ While compiling, [COMPILE] WORD compiles WORD if it would otherwise be IMMEDIATE.
2033 : [COMPILE] IMMEDIATE
2034 WORD \\ get the next word
2035 FIND \\ find it in the dictionary
2036 >CFA \\ get its codeword
2037 , \\ and compile that
2040 \\ RECURSE makes a recursive call to the current word that is being compiled.
2041 \\ Normally while a word is being compiled, it is marked HIDDEN so that references to the
2042 \\ same word within are calls to the previous definition of the word.
2044 LATEST @ >CFA \\ LATEST points to the word being compiled at the moment
2048 \\ ALLOT is used to allocate (static) memory when compiling. It increases HERE by
2049 \\ the amount given on the stack.
2053 \\ Finally print the welcome prompt.
2066 /* END OF jonesforth.S */
git.annexia.org / jonesforth.git / blob