1 /* A sometimes minimal FORTH compiler and tutorial for Linux / i386 systems. -*- asm -*-
2 By Richard W.M. Jones <rich@annexia.org> http://annexia.org/forth
3 This is PUBLIC DOMAIN (see public domain release statement below).
4 $Id: jonesforth.S,v 1.43 2007-10-10 13:01:05 rich Exp $
6 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -Wl,--build-id=none -o jonesforth jonesforth.S
10 INTRODUCTION ----------------------------------------------------------------------
12 FORTH is one of those alien languages which most working programmers regard in the same
13 way as Haskell, LISP, and so on. Something so strange that they'd rather any thoughts
14 of it just go away so they can get on with writing this paying code. But that's wrong
15 and if you care at all about programming then you should at least understand all these
16 languages, even if you will never use them.
18 LISP is the ultimate high-level language, and features from LISP are being added every
19 decade to the more common languages. But FORTH is in some ways the ultimate in low level
20 programming. Out of the box it lacks features like dynamic memory management and even
21 strings. In fact, at its primitive level it lacks even basic concepts like IF-statements
24 Why then would you want to learn FORTH? There are several very good reasons. First
25 and foremost, FORTH is minimal. You really can write a complete FORTH in, say, 2000
26 lines of code. I don't just mean a FORTH program, I mean a complete FORTH operating
27 system, environment and language. You could boot such a FORTH on a bare PC and it would
28 come up with a prompt where you could start doing useful work. The FORTH you have here
29 isn't minimal and uses a Linux process as its 'base PC' (both for the purposes of making
30 it a good tutorial). It's possible to completely understand the system. Who can say they
31 completely understand how Linux works, or gcc?
33 Secondly FORTH has a peculiar bootstrapping property. By that I mean that after writing
34 a little bit of assembly to talk to the hardware and implement a few primitives, all the
35 rest of the language and compiler is written in FORTH itself. Remember I said before
36 that FORTH lacked IF-statements and loops? Well of course it doesn't really because
37 such a lanuage would be useless, but my point was rather that IF-statements and loops are
38 written in FORTH itself.
40 Now of course this is common in other languages as well, and in those languages we call
41 them 'libraries'. For example in C, 'printf' is a library function written in C. But
42 in FORTH this goes way beyond mere libraries. Can you imagine writing C's 'if' in C?
43 And that brings me to my third reason: If you can write 'if' in FORTH, then why restrict
44 yourself to the usual if/while/for/switch constructs? You want a construct that iterates
45 over every other element in a list of numbers? You can add it to the language. What
46 about an operator which pulls in variables directly from a configuration file and makes
47 them available as FORTH variables? Or how about adding Makefile-like dependencies to
48 the language? No problem in FORTH. How about modifying the FORTH compiler to allow
49 complex inlining strategies -- simple. This concept isn't common in programming languages,
50 but it has a name (in fact two names): "macros" (by which I mean LISP-style macros, not
51 the lame C preprocessor) and "domain specific languages" (DSLs).
53 This tutorial isn't about learning FORTH as the language. I'll point you to some references
54 you should read if you're not familiar with using FORTH. This tutorial is about how to
55 write FORTH. In fact, until you understand how FORTH is written, you'll have only a very
56 superficial understanding of how to use it.
58 So if you're not familiar with FORTH or want to refresh your memory here are some online
61 http://en.wikipedia.org/wiki/Forth_%28programming_language%29
63 http://galileo.phys.virginia.edu/classes/551.jvn.fall01/primer.htm
65 http://wiki.laptop.org/go/Forth_Lessons
67 http://www.albany.net/~hello/simple.htm
69 Here is another "Why FORTH?" essay: http://www.jwdt.com/~paysan/why-forth.html
71 Discussion and criticism of this FORTH here: http://lambda-the-ultimate.org/node/2452
73 ACKNOWLEDGEMENTS ----------------------------------------------------------------------
75 This code draws heavily on the design of LINA FORTH (http://home.hccnet.nl/a.w.m.van.der.horst/lina.html)
76 by Albert van der Horst. Any similarities in the code are probably not accidental.
78 Some parts of this FORTH are also based on this IOCCC entry from 1992:
79 http://ftp.funet.fi/pub/doc/IOCCC/1992/buzzard.2.design.
80 I was very proud when Sean Barrett, the original author of the IOCCC entry, commented in the LtU thread
81 http://lambda-the-ultimate.org/node/2452#comment-36818 about this FORTH.
83 And finally I'd like to acknowledge the (possibly forgotten?) authors of ARTIC FORTH because their
84 original program which I still have on original cassette tape kept nagging away at me all these years.
85 http://en.wikipedia.org/wiki/Artic_Software
87 PUBLIC DOMAIN ----------------------------------------------------------------------
89 I, the copyright holder of this work, hereby release it into the public domain. This applies worldwide.
91 In case this is not legally possible, I grant any entity the right to use this work for any purpose,
92 without any conditions, unless such conditions are required by law.
94 SETTING UP ----------------------------------------------------------------------
96 Let's get a few housekeeping things out of the way. Firstly because I need to draw lots of
97 ASCII-art diagrams to explain concepts, the best way to look at this is using a window which
98 uses a fixed width font and is at least this wide:
100 <------------------------------------------------------------------------------------------------------------------------>
102 Secondly make sure TABS are set to 8 characters. The following should be a vertical
103 line. If not, sort out your tabs.
109 Thirdly I assume that your screen is at least 50 characters high.
111 ASSEMBLING ----------------------------------------------------------------------
113 If you want to actually run this FORTH, rather than just read it, you will need Linux on an
114 i386. Linux because instead of programming directly to the hardware on a bare PC which I
115 could have done, I went for a simpler tutorial by assuming that the 'hardware' is a Linux
116 process with a few basic system calls (read, write and exit and that's about all). i386
117 is needed because I had to write the assembly for a processor, and i386 is by far the most
118 common. (Of course when I say 'i386', any 32- or 64-bit x86 processor will do. I'm compiling
119 this on a 64 bit AMD Opteron).
121 Again, to assemble this you will need gcc and gas (the GNU assembler). The commands to
122 assemble and run the code (save this file as 'jonesforth.S') are:
124 gcc -m32 -nostdlib -static -Wl,-Ttext,0 -Wl,--build-id=none -o jonesforth jonesforth.S
125 cat jonesforth.f - | ./jonesforth
127 If you want to run your own FORTH programs you can do:
129 cat jonesforth.f myprog.f | ./jonesforth
131 If you want to load your own FORTH code and then continue reading user commands, you can do:
133 cat jonesforth.f myfunctions.f - | ./jonesforth
135 ASSEMBLER ----------------------------------------------------------------------
137 (You can just skip to the next section -- you don't need to be able to read assembler to
138 follow this tutorial).
140 However if you do want to read the assembly code here are a few notes about gas (the GNU assembler):
142 (1) Register names are prefixed with '%', so %eax is the 32 bit i386 accumulator. The registers
143 available on i386 are: %eax, %ebx, %ecx, %edx, %esi, %edi, %ebp and %esp, and most of them
144 have special purposes.
146 (2) Add, mov, etc. take arguments in the form SRC,DEST. So mov %eax,%ecx moves %eax -> %ecx
148 (3) Constants are prefixed with '$', and you mustn't forget it! If you forget it then it
149 causes a read from memory instead, so:
150 mov $2,%eax moves number 2 into %eax
151 mov 2,%eax reads the 32 bit word from address 2 into %eax (ie. most likely a mistake)
153 (4) gas has a funky syntax for local labels, where '1f' (etc.) means label '1:' "forwards"
154 and '1b' (etc.) means label '1:' "backwards". Notice that these labels might be mistaken
155 for hex numbers (eg. you might confuse 1b with $0x1b).
157 (5) 'ja' is "jump if above", 'jb' for "jump if below", 'je' "jump if equal" etc.
159 (6) gas has a reasonably nice .macro syntax, and I use them a lot to make the code shorter and
162 For more help reading the assembler, do "info gas" at the Linux prompt.
164 Now the tutorial starts in earnest.
166 THE DICTIONARY ----------------------------------------------------------------------
168 In FORTH as you will know, functions are called "words", and just as in other languages they
169 have a name and a definition. Here are two FORTH words:
171 : DOUBLE DUP + ; \ name is "DOUBLE", definition is "DUP +"
172 : QUADRUPLE DOUBLE DOUBLE ; \ name is "QUADRUPLE", definition is "DOUBLE DOUBLE"
174 Words, both built-in ones and ones which the programmer defines later, are stored in a dictionary
175 which is just a linked list of dictionary entries.
177 <--- DICTIONARY ENTRY (HEADER) ----------------------->
178 +------------------------+--------+---------- - - - - +----------- - - - -
179 | LINK POINTER | LENGTH/| NAME | DEFINITION
181 +--- (4 bytes) ----------+- byte -+- n bytes - - - - +----------- - - - -
183 I'll come to the definition of the word later. For now just look at the header. The first
184 4 bytes are the link pointer. This points back to the previous word in the dictionary, or, for
185 the first word in the dictionary it is just a NULL pointer. Then comes a length/flags byte.
186 The length of the word can be up to 31 characters (5 bits used) and the top three bits are used
187 for various flags which I'll come to later. This is followed by the name itself, and in this
188 implementation the name is rounded up to a multiple of 4 bytes by padding it with zero bytes.
189 That's just to ensure that the definition starts on a 32 bit boundary.
191 A FORTH variable called LATEST contains a pointer to the most recently defined word, in
192 other words, the head of this linked list.
194 DOUBLE and QUADRUPLE might look like this:
196 pointer to previous word
199 +--|------+---+---+---+---+---+---+---+---+------------- - - - -
200 | LINK | 6 | D | O | U | B | L | E | 0 | (definition ...)
201 +---------+---+---+---+---+---+---+---+---+------------- - - - -
204 +--|------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
205 | LINK | 9 | Q | U | A | D | R | U | P | L | E | 0 | 0 | (definition ...)
206 +---------+---+---+---+---+---+---+---+---+---+---+---+---+------------- - - - -
212 You should be able to see from this how you might implement functions to find a word in
213 the dictionary (just walk along the dictionary entries starting at LATEST and matching
214 the names until you either find a match or hit the NULL pointer at the end of the dictionary);
215 and add a word to the dictionary (create a new definition, set its LINK to LATEST, and set
216 LATEST to point to the new word). We'll see precisely these functions implemented in
217 assembly code later on.
219 One interesting consequence of using a linked list is that you can redefine words, and
220 a newer definition of a word overrides an older one. This is an important concept in
221 FORTH because it means that any word (even "built-in" or "standard" words) can be
222 overridden with a new definition, either to enhance it, to make it faster or even to
223 disable it. However because of the way that FORTH words get compiled, which you'll
224 understand below, words defined using the old definition of a word continue to use
225 the old definition. Only words defined after the new definition use the new definition.
227 DIRECT THREADED CODE ----------------------------------------------------------------------
229 Now we'll get to the really crucial bit in understanding FORTH, so go and get a cup of tea
230 or coffee and settle down. It's fair to say that if you don't understand this section, then you
231 won't "get" how FORTH works, and that would be a failure on my part for not explaining it well.
232 So if after reading this section a few times you don't understand it, please email me
235 Let's talk first about what "threaded code" means. Imagine a peculiar version of C where
236 you are only allowed to call functions without arguments. (Don't worry for now that such a
237 language would be completely useless!) So in our peculiar C, code would look like this:
246 and so on. How would a function, say 'f' above, be compiled by a standard C compiler?
247 Probably into assembly code like this. On the right hand side I've written the actual
251 CALL a E8 08 00 00 00
252 CALL b E8 1C 00 00 00
253 CALL c E8 2C 00 00 00
254 ; ignore the return from the function for now
256 "E8" is the x86 machine code to "CALL" a function. In the first 20 years of computing
257 memory was hideously expensive and we might have worried about the wasted space being used
258 by the repeated "E8" bytes. We can save 20% in code size (and therefore, in expensive memory)
259 by compressing this into just:
261 08 00 00 00 Just the function addresses, without
262 1C 00 00 00 the CALL prefix.
265 On a 16-bit machine like the ones which originally ran FORTH the savings are even greater - 33%.
267 [Historical note: If the execution model that FORTH uses looks strange from the following
268 paragraphs, then it was motivated entirely by the need to save memory on early computers.
269 This code compression isn't so important now when our machines have more memory in their L1
270 caches than those early computers had in total, but the execution model still has some
273 Of course this code won't run directly on the CPU any more. Instead we need to write an
274 interpreter which takes each set of bytes and calls it.
276 On an i386 machine it turns out that we can write this interpreter rather easily, in just
277 two assembly instructions which turn into just 3 bytes of machine code. Let's store the
278 pointer to the next word to execute in the %esi register:
280 08 00 00 00 <- We're executing this one now. %esi is the _next_ one to execute.
284 The all-important i386 instruction is called LODSL (or in Intel manuals, LODSW). It does
285 two things. Firstly it reads the memory at %esi into the accumulator (%eax). Secondly it
286 increments %esi by 4 bytes. So after LODSL, the situation now looks like this:
288 08 00 00 00 <- We're still executing this one
289 1C 00 00 00 <- %eax now contains this address (0x0000001C)
292 Now we just need to jump to the address in %eax. This is again just a single x86 instruction
293 written JMP *(%eax). And after doing the jump, the situation looks like:
296 1C 00 00 00 <- Now we're executing this subroutine.
299 To make this work, each subroutine is followed by the two instructions 'LODSL; JMP *(%eax)'
300 which literally make the jump to the next subroutine.
302 And that brings us to our first piece of actual code! Well, it's a macro.
311 /* The macro is called NEXT. That's a FORTH-ism. It expands to those two instructions.
313 Every FORTH primitive that we write has to be ended by NEXT. Think of it kind of like
316 The above describes what is known as direct threaded code.
318 To sum up: We compress our function calls down to a list of addresses and use a somewhat
319 magical macro to act as a "jump to next function in the list". We also use one register (%esi)
320 to act as a kind of instruction pointer, pointing to the next function in the list.
322 I'll just give you a hint of what is to come by saying that a FORTH definition such as:
324 : QUADRUPLE DOUBLE DOUBLE ;
326 actually compiles (almost, not precisely but we'll see why in a moment) to a list of
327 function addresses for DOUBLE, DOUBLE and a special function called EXIT to finish off.
329 At this point, REALLY EAGLE-EYED ASSEMBLY EXPERTS are saying "JONES, YOU'VE MADE A MISTAKE!".
331 I lied about JMP *(%eax).
333 INDIRECT THREADED CODE ----------------------------------------------------------------------
335 It turns out that direct threaded code is interesting but only if you want to just execute
336 a list of functions written in assembly language. So QUADRUPLE would work only if DOUBLE
337 was an assembly language function. In the direct threaded code, QUADRUPLE would look like:
340 | addr of DOUBLE --------------------> (assembly code to do the double)
341 +------------------+ NEXT
342 %esi -> | addr of DOUBLE |
345 We can add an extra indirection to allow us to run both words written in assembly language
346 (primitives written for speed) and words written in FORTH themselves as lists of addresses.
348 The extra indirection is the reason for the brackets in JMP *(%eax).
350 Let's have a look at how QUADRUPLE and DOUBLE really look in FORTH:
352 : QUADRUPLE DOUBLE DOUBLE ;
355 | codeword | : DOUBLE DUP + ;
357 | addr of DOUBLE ---------------> +------------------+
358 +------------------+ | codeword |
359 | addr of DOUBLE | +------------------+
360 +------------------+ | addr of DUP --------------> +------------------+
361 | addr of EXIT | +------------------+ | codeword -------+
362 +------------------+ %esi -> | addr of + --------+ +------------------+ |
363 +------------------+ | | assembly to <-----+
364 | addr of EXIT | | | implement DUP |
365 +------------------+ | | .. |
368 | +------------------+
370 +-----> +------------------+
372 +------------------+ |
373 | assembly to <------+
380 This is the part where you may need an extra cup of tea/coffee/favourite caffeinated
381 beverage. What has changed is that I've added an extra pointer to the beginning of
382 the definitions. In FORTH this is sometimes called the "codeword". The codeword is
383 a pointer to the interpreter to run the function. For primitives written in
384 assembly language, the "interpreter" just points to the actual assembly code itself.
385 They don't need interpreting, they just run.
387 In words written in FORTH (like QUADRUPLE and DOUBLE), the codeword points to an interpreter
390 I'll show you the interpreter function shortly, but let's recall our indirect
391 JMP *(%eax) with the "extra" brackets. Take the case where we're executing DOUBLE
392 as shown, and DUP has been called. Note that %esi is pointing to the address of +
394 The assembly code for DUP eventually does a NEXT. That:
396 (1) reads the address of + into %eax %eax points to the codeword of +
397 (2) increments %esi by 4
398 (3) jumps to the indirect %eax jumps to the address in the codeword of +,
399 ie. the assembly code to implement +
404 | addr of DOUBLE ---------------> +------------------+
405 +------------------+ | codeword |
406 | addr of DOUBLE | +------------------+
407 +------------------+ | addr of DUP --------------> +------------------+
408 | addr of EXIT | +------------------+ | codeword -------+
409 +------------------+ | addr of + --------+ +------------------+ |
410 +------------------+ | | assembly to <-----+
411 %esi -> | addr of EXIT | | | implement DUP |
412 +------------------+ | | .. |
415 | +------------------+
417 +-----> +------------------+
419 +------------------+ |
420 now we're | assembly to <-----+
421 executing | implement + |
427 So I hope that I've convinced you that NEXT does roughly what you'd expect. This is
428 indirect threaded code.
430 I've glossed over four things. I wonder if you can guess without reading on what they are?
436 My list of four things are: (1) What does "EXIT" do? (2) which is related to (1) is how do
437 you call into a function, ie. how does %esi start off pointing at part of QUADRUPLE, but
438 then point at part of DOUBLE. (3) What goes in the codeword for the words which are written
439 in FORTH? (4) How do you compile a function which does anything except call other functions
440 ie. a function which contains a number like : DOUBLE 2 * ; ?
442 THE INTERPRETER AND RETURN STACK ------------------------------------------------------------
444 Going at these in no particular order, let's talk about issues (3) and (2), the interpreter
445 and the return stack.
447 Words which are defined in FORTH need a codeword which points to a little bit of code to
448 give them a "helping hand" in life. They don't need much, but they do need what is known
449 as an "interpreter", although it doesn't really "interpret" in the same way that, say,
450 Java bytecode used to be interpreted (ie. slowly). This interpreter just sets up a few
451 machine registers so that the word can then execute at full speed using the indirect
452 threaded model above.
454 One of the things that needs to happen when QUADRUPLE calls DOUBLE is that we save the old
455 %esi ("instruction pointer") and create a new one pointing to the first word in DOUBLE.
456 Because we will need to restore the old %esi at the end of DOUBLE (this is, after all, like
457 a function call), we will need a stack to store these "return addresses" (old values of %esi).
459 As you will have seen in the background documentation, FORTH has two stacks, an ordinary
460 stack for parameters, and a return stack which is a bit more mysterious. But our return
461 stack is just the stack I talked about in the previous paragraph, used to save %esi when
462 calling from a FORTH word into another FORTH word.
464 In this FORTH, we are using the normal stack pointer (%esp) for the parameter stack.
465 We will use the i386's "other" stack pointer (%ebp, usually called the "frame pointer")
466 for our return stack.
468 I've got two macros which just wrap up the details of using %ebp for the return stack.
469 You use them as for example "PUSHRSP %eax" (push %eax on the return stack) or "POPRSP %ebx"
470 (pop top of return stack into %ebx).
473 /* Macros to deal with the return stack. */
475 lea -4(%ebp),%ebp // push reg on to return stack
480 mov (%ebp),\reg // pop top of return stack to reg
485 And with that we can now talk about the interpreter.
487 In FORTH the interpreter function is often called DOCOL (I think it means "DO COLON" because
488 all FORTH definitions start with a colon, as in : DOUBLE DUP + ;
490 The "interpreter" (it's not really "interpreting") just needs to push the old %esi on the
491 stack and set %esi to the first word in the definition. Remember that we jumped to the
492 function using JMP *(%eax)? Well a consequence of that is that conveniently %eax contains
493 the address of this codeword, so just by adding 4 to it we get the address of the first
494 data word. Finally after setting up %esi, it just does NEXT which causes that first word
498 /* DOCOL - the interpreter! */
502 PUSHRSP %esi // push %esi on to the return stack
503 addl $4,%eax // %eax points to codeword, so make
504 movl %eax,%esi // %esi point to first data word
508 Just to make this absolutely clear, let's see how DOCOL works when jumping from QUADRUPLE
514 +------------------+ DOUBLE:
515 | addr of DOUBLE ---------------> +------------------+
516 +------------------+ %eax -> | addr of DOCOL |
517 %esi -> | addr of DOUBLE | +------------------+
518 +------------------+ | addr of DUP |
519 | addr of EXIT | +------------------+
520 +------------------+ | etc. |
522 First, the call to DOUBLE calls DOCOL (the codeword of DOUBLE). DOCOL does this: It
523 pushes the old %esi on the return stack. %eax points to the codeword of DOUBLE, so we
524 just add 4 on to it to get our new %esi:
529 +------------------+ DOUBLE:
530 | addr of DOUBLE ---------------> +------------------+
531 top of return +------------------+ %eax -> | addr of DOCOL |
532 stack points -> | addr of DOUBLE | + 4 = +------------------+
533 +------------------+ %esi -> | addr of DUP |
534 | addr of EXIT | +------------------+
535 +------------------+ | etc. |
537 Then we do NEXT, and because of the magic of threaded code that increments %esi again
540 Well, it seems to work.
542 One minor point here. Because DOCOL is the first bit of assembly actually to be defined
543 in this file (the others were just macros), and because I usually compile this code with the
544 text segment starting at address 0, DOCOL has address 0. So if you are disassembling the
545 code and see a word with a codeword of 0, you will immediately know that the word is
546 written in FORTH (it's not an assembler primitive) and so uses DOCOL as the interpreter.
548 STARTING UP ----------------------------------------------------------------------
550 Now let's get down to nuts and bolts. When we start the program we need to set up
551 a few things like the return stack. But as soon as we can, we want to jump into FORTH
552 code (albeit much of the "early" FORTH code will still need to be written as
553 assembly language primitives).
555 This is what the set up code does. Does a tiny bit of house-keeping, sets up the
556 separate return stack (NB: Linux gives us the ordinary parameter stack already), then
557 immediately jumps to a FORTH word called QUIT. Despite its name, QUIT doesn't quit
558 anything. It resets some internal state and starts reading and interpreting commands.
559 (The reason it is called QUIT is because you can call QUIT from your own FORTH code
560 to "quit" your program and go back to interpreting).
563 /* Assembler entry point. */
568 mov %esp,var_S0 // Save the initial data stack pointer in FORTH variable S0.
569 mov $return_stack_top,%ebp // Initialise the return stack.
570 call set_up_data_segment
572 mov $cold_start,%esi // Initialise interpreter.
573 NEXT // Run interpreter!
576 cold_start: // High-level code without a codeword.
580 BUILT-IN WORDS ----------------------------------------------------------------------
582 Remember our dictionary entries (headers)? Let's bring those together with the codeword
583 and data words to see how : DOUBLE DUP + ; really looks in memory.
585 pointer to previous word
588 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
589 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
590 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
593 LINK in next word points to codeword of DUP
595 Initially we can't just write ": DOUBLE DUP + ;" (ie. that literal string) here because we
596 don't yet have anything to read the string, break it up at spaces, parse each word, etc. etc.
597 So instead we will have to define built-in words using the GNU assembler data constructors
598 (like .int, .byte, .string, .ascii and so on -- look them up in the gas info page if you are
601 The long way would be:
603 .int <link to previous word>
605 .ascii "DOUBLE" // string
607 DOUBLE: .int DOCOL // codeword
608 .int DUP // pointer to codeword of DUP
609 .int PLUS // pointer to codeword of +
610 .int EXIT // pointer to codeword of EXIT
612 That's going to get quite tedious rather quickly, so here I define an assembler macro
613 so that I can just write:
615 defword "DOUBLE",6,,DOUBLE
618 and I'll get exactly the same effect.
620 Don't worry too much about the exact implementation details of this macro - it's complicated!
623 /* Flags - these are discussed later. */
626 .set F_LENMASK,0x1f // length mask
628 // Store the chain of links.
631 .macro defword name, namelen, flags=0, label
637 .set link,name_\label
638 .byte \flags+\namelen // flags + length byte
639 .ascii "\name" // the name
640 .align 4 // padding to next 4 byte boundary
643 .int DOCOL // codeword - the interpreter
644 // list of word pointers follow
648 Similarly I want a way to write words written in assembly language. There will quite a few
649 of these to start with because, well, everything has to start in assembly before there's
650 enough "infrastructure" to be able to start writing FORTH words, but also I want to define
651 some common FORTH words in assembly language for speed, even though I could write them in FORTH.
653 This is what DUP looks like in memory:
655 pointer to previous word
658 +--|------+---+---+---+---+------------+
659 | LINK | 3 | D | U | P | code_DUP ---------------------> points to the assembly
660 +---------+---+---+---+---+------------+ code used to write DUP,
661 ^ len codeword which ends with NEXT.
665 Again, for brevity in writing the header I'm going to write an assembler macro called defcode.
666 As with defword above, don't worry about the complicated details of the macro.
669 .macro defcode name, namelen, flags=0, label
675 .set link,name_\label
676 .byte \flags+\namelen // flags + length byte
677 .ascii "\name" // the name
678 .align 4 // padding to next 4 byte boundary
681 .int code_\label // codeword
685 code_\label : // assembler code follows
689 Now some easy FORTH primitives. These are written in assembly for speed. If you understand
690 i386 assembly language then it is worth reading these. However if you don't understand assembly
691 you can skip the details.
694 defcode "DROP",4,,DROP
695 pop %eax // drop top of stack
698 defcode "SWAP",4,,SWAP
699 pop %eax // swap top two elements on stack
706 mov (%esp),%eax // duplicate top of stack
710 defcode "OVER",4,,OVER
711 mov 4(%esp),%eax // get the second element of stack
712 push %eax // and push it on top
724 defcode "-ROT",4,,NROT
733 defcode "?DUP",4,,QDUP // duplicate top of stack if non-zero
741 incl (%esp) // increment top of stack
745 decl (%esp) // decrement top of stack
748 defcode "4+",2,,INCR4
749 addl $4,(%esp) // add 4 to top of stack
752 defcode "4-",2,,DECR4
753 subl $4,(%esp) // subtract 4 from top of stack
757 pop %eax // get top of stack
758 addl %eax,(%esp) // and add it to next word on stack
762 pop %eax // get top of stack
763 subl %eax,(%esp) // and subtract it from next word on stack
770 push %eax // ignore overflow
774 In this FORTH, only /MOD is primitive. Later we will define the / and MOD words in
775 terms of the primitive /MOD. The design of the i386 assembly instruction idiv which
776 leaves both quotient and remainder makes this the obvious choice.
779 defcode "/MOD",4,,DIVMOD
784 push %edx // push remainder
785 push %eax // push quotient
789 Lots of comparison operations like =, <, >, etc..
791 ANS FORTH says that the comparison words should return all (binary) 1's for
792 TRUE and all 0's for FALSE. However this is a bit of a strange convention
793 so this FORTH breaks it and returns the more normal (for C programmers ...)
794 1 meaning TRUE and 0 meaning FALSE.
797 defcode "=",1,,EQU // top two words are equal?
806 defcode "<>",2,,NEQU // top two words are not equal?
851 defcode "0=",2,,ZEQU // top of stack equals 0?
859 defcode "0<>",3,,ZNEQU // top of stack not 0?
867 defcode "0<",2,,ZLT // comparisons with 0
899 defcode "AND",3,,AND // bitwise AND
904 defcode "OR",2,,OR // bitwise OR
909 defcode "XOR",3,,XOR // bitwise XOR
914 defcode "INVERT",6,,INVERT // this is the FORTH bitwise "NOT" function (cf. NEGATE and NOT)
919 RETURNING FROM FORTH WORDS ----------------------------------------------------------------------
921 Time to talk about what happens when we EXIT a function. In this diagram QUADRUPLE has called
922 DOUBLE, and DOUBLE is about to exit (look at where %esi is pointing):
927 +------------------+ DOUBLE
928 | addr of DOUBLE ---------------> +------------------+
929 +------------------+ | codeword |
930 | addr of DOUBLE | +------------------+
931 +------------------+ | addr of DUP |
932 | addr of EXIT | +------------------+
933 +------------------+ | addr of + |
935 %esi -> | addr of EXIT |
938 What happens when the + function does NEXT? Well, the following code is executed.
941 defcode "EXIT",4,,EXIT
942 POPRSP %esi // pop return stack into %esi
946 EXIT gets the old %esi which we saved from before on the return stack, and puts it in %esi.
947 So after this (but just before NEXT) we get:
952 +------------------+ DOUBLE
953 | addr of DOUBLE ---------------> +------------------+
954 +------------------+ | codeword |
955 %esi -> | addr of DOUBLE | +------------------+
956 +------------------+ | addr of DUP |
957 | addr of EXIT | +------------------+
958 +------------------+ | addr of + |
963 And NEXT just completes the job by, well, in this case just by calling DOUBLE again :-)
965 LITERALS ----------------------------------------------------------------------
967 The final point I "glossed over" before was how to deal with functions that do anything
968 apart from calling other functions. For example, suppose that DOUBLE was defined like this:
972 It does the same thing, but how do we compile it since it contains the literal 2? One way
973 would be to have a function called "2" (which you'd have to write in assembler), but you'd need
974 a function for every single literal that you wanted to use.
976 FORTH solves this by compiling the function using a special word called LIT:
978 +---------------------------+-------+-------+-------+-------+-------+
979 | (usual header of DOUBLE) | DOCOL | LIT | 2 | * | EXIT |
980 +---------------------------+-------+-------+-------+-------+-------+
982 LIT is executed in the normal way, but what it does next is definitely not normal. It
983 looks at %esi (which now points to the number 2), grabs it, pushes it on the stack, then
984 manipulates %esi in order to skip the number as if it had never been there.
986 What's neat is that the whole grab/manipulate can be done using a single byte single
987 i386 instruction, our old friend LODSL. Rather than me drawing more ASCII-art diagrams,
988 see if you can find out how LIT works:
992 // %esi points to the next command, but in this case it points to the next
993 // literal 32 bit integer. Get that literal into %eax and increment %esi.
994 // On x86, it's a convenient single byte instruction! (cf. NEXT macro)
996 push %eax // push the literal number on to stack
1000 MEMORY ----------------------------------------------------------------------
1002 As important point about FORTH is that it gives you direct access to the lowest levels
1003 of the machine. Manipulating memory directly is done frequently in FORTH, and these are
1004 the primitive words for doing it.
1007 defcode "!",1,,STORE
1008 pop %ebx // address to store at
1009 pop %eax // data to store there
1010 mov %eax,(%ebx) // store it
1013 defcode "@",1,,FETCH
1014 pop %ebx // address to fetch
1015 mov (%ebx),%eax // fetch it
1016 push %eax // push value onto stack
1019 defcode "+!",2,,ADDSTORE
1021 pop %eax // the amount to add
1022 addl %eax,(%ebx) // add it
1025 defcode "-!",2,,SUBSTORE
1027 pop %eax // the amount to subtract
1028 subl %eax,(%ebx) // add it
1032 ! and @ (STORE and FETCH) store 32-bit words. It's also useful to be able to read and write bytes
1033 so we also define standard words C@ and C!.
1035 Byte-oriented operations only work on architectures which permit them (i386 is one of those).
1038 defcode "C!",2,,STOREBYTE
1039 pop %ebx // address to store at
1040 pop %eax // data to store there
1041 movb %al,(%ebx) // store it
1044 defcode "C@",2,,FETCHBYTE
1045 pop %ebx // address to fetch
1047 movb (%ebx),%al // fetch it
1048 push %eax // push value onto stack
1051 /* C@C! is a useful byte copy primitive. */
1052 defcode "C@C!",4,,CCOPY
1053 movl 4(%esp),%ebx // source address
1054 movb (%ebx),%al // get source character
1055 pop %edi // destination address
1056 stosb // copy to destination
1057 push %edi // increment destination address
1058 incl 4(%esp) // increment source address
1061 /* and CMOVE is a block copy operation. */
1062 defcode "CMOVE",5,,CMOVE
1063 mov %esi,%edx // preserve %esi
1065 pop %edi // destination address
1066 pop %esi // source address
1067 rep movsb // copy source to destination
1068 mov %edx,%esi // restore %esi
1072 BUILT-IN VARIABLES ----------------------------------------------------------------------
1074 These are some built-in variables and related standard FORTH words. Of these, the only one that we
1075 have discussed so far was LATEST, which points to the last (most recently defined) word in the
1076 FORTH dictionary. LATEST is also a FORTH word which pushes the address of LATEST (the variable)
1077 on to the stack, so you can read or write it using @ and ! operators. For example, to print
1078 the current value of LATEST (and this can apply to any FORTH variable) you would do:
1082 To make defining variables shorter, I'm using a macro called defvar, similar to defword and
1083 defcode above. (In fact the defvar macro uses defcode to do the dictionary header).
1086 .macro defvar name, namelen, flags=0, label, initial=0
1087 defcode \name,\namelen,\flags,\label
1097 The built-in variables are:
1099 STATE Is the interpreter executing code (0) or compiling a word (non-zero)?
1100 LATEST Points to the latest (most recently defined) word in the dictionary.
1101 HERE Points to the next free byte of memory. When compiling, compiled words go here.
1102 S0 Stores the address of the top of the parameter stack.
1103 BASE The current base for printing and reading numbers.
1106 defvar "STATE",5,,STATE
1107 defvar "HERE",4,,HERE
1108 defvar "LATEST",6,,LATEST,name_SYSCALL0 // SYSCALL0 must be last in built-in dictionary
1110 defvar "BASE",4,,BASE,10
1113 BUILT-IN CONSTANTS ----------------------------------------------------------------------
1115 It's also useful to expose a few constants to FORTH. When the word is executed it pushes a
1116 constant value on the stack.
1118 The built-in constants are:
1120 VERSION Is the current version of this FORTH.
1121 R0 The address of the top of the return stack.
1122 DOCOL Pointer to DOCOL.
1123 F_IMMED The IMMEDIATE flag's actual value.
1124 F_HIDDEN The HIDDEN flag's actual value.
1125 F_LENMASK The length mask in the flags/len byte.
1127 SYS_* and the numeric codes of various Linux syscalls (from <asm/unistd.h>)
1130 //#include <asm-i386/unistd.h> // you might need this instead
1131 #include <asm/unistd.h>
1133 .macro defconst name, namelen, flags=0, label, value
1134 defcode \name,\namelen,\flags,\label
1139 defconst "VERSION",7,,VERSION,JONES_VERSION
1140 defconst "R0",2,,RZ,return_stack_top
1141 defconst "DOCOL",5,,__DOCOL,DOCOL
1142 defconst "F_IMMED",7,,__F_IMMED,F_IMMED
1143 defconst "F_HIDDEN",8,,__F_HIDDEN,F_HIDDEN
1144 defconst "F_LENMASK",9,,__F_LENMASK,F_LENMASK
1146 defconst "SYS_EXIT",8,,SYS_EXIT,__NR_exit
1147 defconst "SYS_OPEN",8,,SYS_OPEN,__NR_open
1148 defconst "SYS_CLOSE",9,,SYS_CLOSE,__NR_close
1149 defconst "SYS_READ",8,,SYS_READ,__NR_read
1150 defconst "SYS_WRITE",9,,SYS_WRITE,__NR_write
1151 defconst "SYS_CREAT",9,,SYS_CREAT,__NR_creat
1152 defconst "SYS_BRK",7,,SYS_BRK,__NR_brk
1154 defconst "O_RDONLY",8,,__O_RDONLY,0
1155 defconst "O_WRONLY",8,,__O_WRONLY,1
1156 defconst "O_RDWR",6,,__O_RDWR,2
1157 defconst "O_CREAT",7,,__O_CREAT,0100
1158 defconst "O_EXCL",6,,__O_EXCL,0200
1159 defconst "O_TRUNC",7,,__O_TRUNC,01000
1160 defconst "O_APPEND",8,,__O_APPEND,02000
1161 defconst "O_NONBLOCK",10,,__O_NONBLOCK,04000
1164 RETURN STACK ----------------------------------------------------------------------
1166 These words allow you to access the return stack. Recall that the register %ebp always points to
1167 the top of the return stack.
1171 pop %eax // pop parameter stack into %eax
1172 PUSHRSP %eax // push it on to the return stack
1175 defcode "R>",2,,FROMR
1176 POPRSP %eax // pop return stack on to %eax
1177 push %eax // and push on to parameter stack
1180 defcode "RSP@",4,,RSPFETCH
1184 defcode "RSP!",4,,RSPSTORE
1188 defcode "RDROP",5,,RDROP
1189 addl $4,%ebp // pop return stack and throw away
1193 PARAMETER (DATA) STACK ----------------------------------------------------------------------
1195 These functions allow you to manipulate the parameter stack. Recall that Linux sets up the parameter
1196 stack for us, and it is accessed through %esp.
1199 defcode "DSP@",4,,DSPFETCH
1204 defcode "DSP!",4,,DSPSTORE
1209 INPUT AND OUTPUT ----------------------------------------------------------------------
1211 These are our first really meaty/complicated FORTH primitives. I have chosen to write them in
1212 assembler, but surprisingly in "real" FORTH implementations these are often written in terms
1213 of more fundamental FORTH primitives. I chose to avoid that because I think that just obscures
1214 the implementation. After all, you may not understand assembler but you can just think of it
1215 as an opaque block of code that does what it says.
1217 Let's discuss input first.
1219 The FORTH word KEY reads the next byte from stdin (and pushes it on the parameter stack).
1220 So if KEY is called and someone hits the space key, then the number 32 (ASCII code of space)
1221 is pushed on the stack.
1223 In FORTH there is no distinction between reading code and reading input. We might be reading
1224 and compiling code, we might be reading words to execute, we might be asking for the user
1225 to type their name -- ultimately it all comes in through KEY.
1227 The implementation of KEY uses an input buffer of a certain size (defined at the end of this
1228 file). It calls the Linux read(2) system call to fill this buffer and tracks its position
1229 in the buffer using a couple of variables, and if it runs out of input buffer then it refills
1230 it automatically. The other thing that KEY does is if it detects that stdin has closed, it
1231 exits the program, which is why when you hit ^D the FORTH system cleanly exits.
1236 +-------------------------------+--------------------------------------+
1237 | INPUT READ FROM STDIN ....... | unused part of the buffer |
1238 +-------------------------------+--------------------------------------+
1241 currkey (next character to read)
1243 <---------------------- BUFFER_SIZE (4096 bytes) ---------------------->
1246 defcode "KEY",3,,KEY
1248 push %eax // push return value on stack
1253 jge 1f // exhausted the input buffer?
1255 mov (%ebx),%al // get next key from input buffer
1257 mov %ebx,(currkey) // increment currkey
1260 1: // Out of input; use read(2) to fetch more input from stdin.
1261 xor %ebx,%ebx // 1st param: stdin
1262 mov $buffer,%ecx // 2nd param: buffer
1264 mov $BUFFER_SIZE,%edx // 3rd param: max length
1265 mov $__NR_read,%eax // syscall: read
1267 test %eax,%eax // If %eax <= 0, then exit.
1269 addl %eax,%ecx // buffer+%eax = bufftop
1273 2: // Error or end of input: exit the program.
1275 mov $__NR_exit,%eax // syscall: exit
1281 .int buffer // Current place in input buffer (next character to read).
1283 .int buffer // Last valid data in input buffer + 1.
1286 By contrast, output is much simpler. The FORTH word EMIT writes out a single byte to stdout.
1287 This implementation just uses the write system call. No attempt is made to buffer output, but
1288 it would be a good exercise to add it.
1291 defcode "EMIT",4,,EMIT
1296 mov $1,%ebx // 1st param: stdout
1298 // write needs the address of the byte to write
1299 mov %al,emit_scratch
1300 mov $emit_scratch,%ecx // 2nd param: address
1302 mov $1,%edx // 3rd param: nbytes = 1
1304 mov $__NR_write,%eax // write syscall
1308 .data // NB: easier to fit in the .data section
1310 .space 1 // scratch used by EMIT
1313 Back to input, WORD is a FORTH word which reads the next full word of input.
1315 What it does in detail is that it first skips any blanks (spaces, tabs, newlines and so on).
1316 Then it calls KEY to read characters into an internal buffer until it hits a blank. Then it
1317 calculates the length of the word it read and returns the address and the length as
1318 two words on the stack (with the length at the top of stack).
1320 Notice that WORD has a single internal buffer which it overwrites each time (rather like
1321 a static C string). Also notice that WORD's internal buffer is just 32 bytes long and
1322 there is NO checking for overflow. 31 bytes happens to be the maximum length of a
1323 FORTH word that we support, and that is what WORD is used for: to read FORTH words when
1324 we are compiling and executing code. The returned strings are not NUL-terminated.
1326 Start address+length is the normal way to represent strings in FORTH (not ending in an
1327 ASCII NUL character as in C), and so FORTH strings can contain any character including NULs
1328 and can be any length.
1330 WORD is not suitable for just reading strings (eg. user input) because of all the above
1331 peculiarities and limitations.
1333 Note that when executing, you'll see:
1335 which puts "FOO" and length 3 on the stack, but when compiling:
1337 is an error (or at least it doesn't do what you might expect). Later we'll talk about compiling
1338 and immediate mode, and you'll understand why.
1341 defcode "WORD",4,,WORD
1343 push %edi // push base address
1344 push %ecx // push length
1348 /* Search for first non-blank character. Also skip \ comments. */
1350 call _KEY // get next key, returned in %eax
1351 cmpb $'\\',%al // start of a comment?
1352 je 3f // if so, skip the comment
1354 jbe 1b // if so, keep looking
1356 /* Search for the end of the word, storing chars as we go. */
1357 mov $word_buffer,%edi // pointer to return buffer
1359 stosb // add character to return buffer
1360 call _KEY // get next key, returned in %al
1361 cmpb $' ',%al // is blank?
1362 ja 2b // if not, keep looping
1364 /* Return the word (well, the static buffer) and length. */
1365 sub $word_buffer,%edi
1366 mov %edi,%ecx // return length of the word
1367 mov $word_buffer,%edi // return address of the word
1370 /* Code to skip \ comments to end of the current line. */
1373 cmpb $'\n',%al // end of line yet?
1377 .data // NB: easier to fit in the .data section
1378 // A static buffer where WORD returns. Subsequent calls
1379 // overwrite this buffer. Maximum word length is 32 chars.
1384 As well as reading in words we'll need to read in numbers and for that we are using a function
1385 called NUMBER. This parses a numeric string such as one returned by WORD and pushes the
1386 number on the parameter stack.
1388 The function uses the variable BASE as the base (radix) for conversion, so for example if
1389 BASE is 2 then we expect a binary number. Normally BASE is 10.
1391 If the word starts with a '-' character then the returned value is negative.
1393 If the string can't be parsed as a number (or contains characters outside the current BASE)
1394 then we need to return an error indication. So NUMBER actually returns two items on the stack.
1395 At the top of stack we return the number of unconverted characters (ie. if 0 then all characters
1396 were converted, so there is no error). Second from top of stack is the parsed number or a
1397 partial value if there was an error.
1399 defcode "NUMBER",6,,NUMBER
1400 pop %ecx // length of string
1401 pop %edi // start address of string
1403 push %eax // parsed number
1404 push %ecx // number of unparsed characters (0 = no error)
1411 test %ecx,%ecx // trying to parse a zero-length string is an error, but will return 0.
1414 movl var_BASE,%edx // get BASE (in %dl)
1416 // Check if first character is '-'.
1417 movb (%edi),%bl // %bl = first character in string
1419 push %eax // push 0 on stack
1420 cmpb $'-',%bl // negative number?
1423 push %ebx // push <> 0 on stack, indicating negative
1426 pop %ebx // error: string is only '-'.
1430 // Loop reading digits.
1431 1: imull %edx,%eax // %eax *= BASE
1432 movb (%edi),%bl // %bl = next character in string
1435 // Convert 0-9, A-Z to a number 0-35.
1436 2: subb $'0',%bl // < '0'?
1438 cmp $10,%bl // <= '9'?
1440 subb $17,%bl // < 'A'? (17 is 'A'-'0')
1444 3: cmp %dl,%bl // >= BASE?
1447 // OK, so add it to %eax and loop.
1452 // Negate the result if first character was '-' (saved on the stack).
1461 DICTIONARY LOOK UPS ----------------------------------------------------------------------
1463 We're building up to our prelude on how FORTH code is compiled, but first we need yet more infrastructure.
1465 The FORTH word FIND takes a string (a word as parsed by WORD -- see above) and looks it up in the
1466 dictionary. What it actually returns is the address of the dictionary header, if it finds it,
1469 So if DOUBLE is defined in the dictionary, then WORD DOUBLE FIND returns the following pointer:
1475 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1476 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1477 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1479 See also >CFA and >DFA.
1481 FIND doesn't find dictionary entries which are flagged as HIDDEN. See below for why.
1484 defcode "FIND",4,,FIND
1485 pop %ecx // %ecx = length
1486 pop %edi // %edi = address
1488 push %eax // %eax = address of dictionary entry (or NULL)
1492 push %esi // Save %esi so we can use it in string comparison.
1494 // Now we start searching backwards through the dictionary for this word.
1495 mov var_LATEST,%edx // LATEST points to name header of the latest word in the dictionary
1496 1: test %edx,%edx // NULL pointer? (end of the linked list)
1499 // Compare the length expected and the length of the word.
1500 // Note that if the F_HIDDEN flag is set on the word, then by a bit of trickery
1501 // this won't pick the word (the length will appear to be wrong).
1503 movb 4(%edx),%al // %al = flags+length field
1504 andb $(F_HIDDEN|F_LENMASK),%al // %al = name length
1505 cmpb %cl,%al // Length is the same?
1508 // Compare the strings in detail.
1509 push %ecx // Save the length
1510 push %edi // Save the address (repe cmpsb will move this pointer)
1511 lea 5(%edx),%esi // Dictionary string we are checking against.
1512 repe cmpsb // Compare the strings.
1515 jne 2f // Not the same.
1517 // The strings are the same - return the header pointer in %eax
1522 2: mov (%edx),%edx // Move back through the link field to the previous word
1523 jmp 1b // .. and loop.
1527 xor %eax,%eax // Return zero to indicate not found.
1531 FIND returns the dictionary pointer, but when compiling we need the codeword pointer (recall
1532 that FORTH definitions are compiled into lists of codeword pointers). The standard FORTH
1533 word >CFA turns a dictionary pointer into a codeword pointer.
1535 The example below shows the result of:
1537 WORD DOUBLE FIND >CFA
1539 FIND returns a pointer to this
1540 | >CFA converts it to a pointer to this
1543 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1544 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1545 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1550 Because names vary in length, this isn't just a simple increment.
1552 In this FORTH you cannot easily turn a codeword pointer back into a dictionary entry pointer, but
1553 that is not true in most FORTH implementations where they store a back pointer in the definition
1554 (with an obvious memory/complexity cost). The reason they do this is that it is useful to be
1555 able to go backwards (codeword -> dictionary entry) in order to decompile FORTH definitions
1558 What does CFA stand for? My best guess is "Code Field Address".
1561 defcode ">CFA",4,,TCFA
1568 add $4,%edi // Skip link pointer.
1569 movb (%edi),%al // Load flags+len into %al.
1570 inc %edi // Skip flags+len byte.
1571 andb $F_LENMASK,%al // Just the length, not the flags.
1572 add %eax,%edi // Skip the name.
1573 addl $3,%edi // The codeword is 4-byte aligned.
1578 Related to >CFA is >DFA which takes a dictionary entry address as returned by FIND and
1579 returns a pointer to the first data field.
1581 FIND returns a pointer to this
1582 | >CFA converts it to a pointer to this
1584 | | >DFA converts it to a pointer to this
1587 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1588 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1589 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1592 (Note to those following the source of FIG-FORTH / ciforth: My >DFA definition is
1593 different from theirs, because they have an extra indirection).
1595 You can see that >DFA is easily defined in FORTH just by adding 4 to the result of >CFA.
1598 defword ">DFA",4,,TDFA
1599 .int TCFA // >CFA (get code field address)
1600 .int INCR4 // 4+ (add 4 to it to get to next word)
1601 .int EXIT // EXIT (return from FORTH word)
1604 COMPILING ----------------------------------------------------------------------
1606 Now we'll talk about how FORTH compiles words. Recall that a word definition looks like this:
1610 and we have to turn this into:
1612 pointer to previous word
1615 +--|------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1616 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1617 +---------+---+---+---+---+---+---+---+---+------------+--|---------+------------+------------+
1618 ^ len pad codeword |
1620 LATEST points here points to codeword of DUP
1622 There are several problems to solve. Where to put the new word? How do we read words? How
1623 do we define the words : (COLON) and ; (SEMICOLON)?
1625 FORTH solves this rather elegantly and as you might expect in a very low-level way which
1626 allows you to change how the compiler works on your own code.
1628 FORTH has an INTERPRET function (a true interpreter this time, not DOCOL) which runs in a
1629 loop, reading words (using WORD), looking them up (using FIND), turning them into codeword
1630 pointers (using >CFA) and deciding what to do with them.
1632 What it does depends on the mode of the interpreter (in variable STATE).
1634 When STATE is zero, the interpreter just runs each word as it looks them up. This is known as
1637 The interesting stuff happens when STATE is non-zero -- compiling mode. In this mode the
1638 interpreter appends the codeword pointer to user memory (the HERE variable points to the next
1639 free byte of user memory -- see DATA SEGMENT section below).
1641 So you may be able to see how we could define : (COLON). The general plan is:
1643 (1) Use WORD to read the name of the function being defined.
1645 (2) Construct the dictionary entry -- just the header part -- in user memory:
1647 pointer to previous word (from LATEST) +-- Afterwards, HERE points here, where
1648 ^ | the interpreter will start appending
1650 +--|------+---+---+---+---+---+---+---+---+------------+
1651 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL |
1652 +---------+---+---+---+---+---+---+---+---+------------+
1655 (3) Set LATEST to point to the newly defined word, ...
1657 (4) .. and most importantly leave HERE pointing just after the new codeword. This is where
1658 the interpreter will append codewords.
1660 (5) Set STATE to 1. This goes into compile mode so the interpreter starts appending codewords to
1661 our partially-formed header.
1663 After : has run, our input is here:
1668 Next byte returned by KEY will be the 'D' character of DUP
1670 so the interpreter (now it's in compile mode, so I guess it's really the compiler) reads "DUP",
1671 looks it up in the dictionary, gets its codeword pointer, and appends it:
1673 +-- HERE updated to point here.
1676 +---------+---+---+---+---+---+---+---+---+------------+------------+
1677 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP |
1678 +---------+---+---+---+---+---+---+---+---+------------+------------+
1681 Next we read +, get the codeword pointer, and append it:
1683 +-- HERE updated to point here.
1686 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1687 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + |
1688 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+
1691 The issue is what happens next. Obviously what we _don't_ want to happen is that we
1692 read ";" and compile it and go on compiling everything afterwards.
1694 At this point, FORTH uses a trick. Remember the length byte in the dictionary definition
1695 isn't just a plain length byte, but can also contain flags. One flag is called the
1696 IMMEDIATE flag (F_IMMED in this code). If a word in the dictionary is flagged as
1697 IMMEDIATE then the interpreter runs it immediately _even if it's in compile mode_.
1699 This is how the word ; (SEMICOLON) works -- as a word flagged in the dictionary as IMMEDIATE.
1701 And all it does is append the codeword for EXIT on to the current definition and switch
1702 back to immediate mode (set STATE back to 0). Shortly we'll see the actual definition
1703 of ; and we'll see that it's really a very simple definition, declared IMMEDIATE.
1705 After the interpreter reads ; and executes it 'immediately', we get this:
1707 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1708 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL | DUP | + | EXIT |
1709 +---------+---+---+---+---+---+---+---+---+------------+------------+------------+------------+
1715 And that's it, job done, our new definition is compiled, and we're back in immediate mode
1716 just reading and executing words, perhaps including a call to test our new word DOUBLE.
1718 The only last wrinkle in this is that while our word was being compiled, it was in a
1719 half-finished state. We certainly wouldn't want DOUBLE to be called somehow during
1720 this time. There are several ways to stop this from happening, but in FORTH what we
1721 do is flag the word with the HIDDEN flag (F_HIDDEN in this code) just while it is
1722 being compiled. This prevents FIND from finding it, and thus in theory stops any
1723 chance of it being called.
1725 The above explains how compiling, : (COLON) and ; (SEMICOLON) works and in a moment I'm
1726 going to define them. The : (COLON) function can be made a little bit more general by writing
1727 it in two parts. The first part, called CREATE, makes just the header:
1729 +-- Afterwards, HERE points here.
1732 +---------+---+---+---+---+---+---+---+---+
1733 | LINK | 6 | D | O | U | B | L | E | 0 |
1734 +---------+---+---+---+---+---+---+---+---+
1737 and the second part, the actual definition of : (COLON), calls CREATE and appends the
1738 DOCOL codeword, so leaving:
1740 +-- Afterwards, HERE points here.
1743 +---------+---+---+---+---+---+---+---+---+------------+
1744 | LINK | 6 | D | O | U | B | L | E | 0 | DOCOL |
1745 +---------+---+---+---+---+---+---+---+---+------------+
1748 CREATE is a standard FORTH word and the advantage of this split is that we can reuse it to
1749 create other types of words (not just ones which contain code, but words which contain variables,
1750 constants and other data).
1753 defcode "CREATE",6,,CREATE
1755 // Get the name length and address.
1756 pop %ecx // %ecx = length
1757 pop %ebx // %ebx = address of name
1760 movl var_HERE,%edi // %edi is the address of the header
1761 movl var_LATEST,%eax // Get link pointer
1762 stosl // and store it in the header.
1764 // Length byte and the word itself.
1765 mov %cl,%al // Get the length.
1766 stosb // Store the length/flags byte.
1768 mov %ebx,%esi // %esi = word
1769 rep movsb // Copy the word
1771 addl $3,%edi // Align to next 4 byte boundary.
1774 // Update LATEST and HERE.
1776 movl %eax,var_LATEST
1781 Because I want to define : (COLON) in FORTH, not assembler, we need a few more FORTH words
1784 The first is , (COMMA) which is a standard FORTH word which appends a 32 bit integer to the user
1785 memory pointed to by HERE, and adds 4 to HERE. So the action of , (COMMA) is:
1787 previous value of HERE
1790 +---------+---+---+---+---+---+---+---+---+-- - - - - --+------------+
1791 | LINK | 6 | D | O | U | B | L | E | 0 | | <data> |
1792 +---------+---+---+---+---+---+---+---+---+-- - - - - --+------------+
1797 and <data> is whatever 32 bit integer was at the top of the stack.
1799 , (COMMA) is quite a fundamental operation when compiling. It is used to append codewords
1800 to the current word that is being compiled.
1803 defcode ",",1,,COMMA
1804 pop %eax // Code pointer to store.
1808 movl var_HERE,%edi // HERE
1810 movl %edi,var_HERE // Update HERE (incremented)
1814 Our definitions of : (COLON) and ; (SEMICOLON) will need to switch to and from compile mode.
1816 Immediate mode vs. compile mode is stored in the global variable STATE, and by updating this
1817 variable we can switch between the two modes.
1819 For various reasons which may become apparent later, FORTH defines two standard words called
1820 [ and ] (LBRAC and RBRAC) which switch between modes:
1822 Word Assembler Action Effect
1823 [ LBRAC STATE := 0 Switch to immediate mode.
1824 ] RBRAC STATE := 1 Switch to compile mode.
1826 [ (LBRAC) is an IMMEDIATE word. The reason is as follows: If we are in compile mode and the
1827 interpreter saw [ then it would compile it rather than running it. We would never be able to
1828 switch back to immediate mode! So we flag the word as IMMEDIATE so that even in compile mode
1829 the word runs immediately, switching us back to immediate mode.
1832 defcode "[",1,F_IMMED,LBRAC
1834 movl %eax,var_STATE // Set STATE to 0.
1837 defcode "]",1,,RBRAC
1838 movl $1,var_STATE // Set STATE to 1.
1842 Now we can define : (COLON) using CREATE. It just calls CREATE, appends DOCOL (the codeword), sets
1843 the word HIDDEN and goes into compile mode.
1846 defword ":",1,,COLON
1847 .int WORD // Get the name of the new word
1848 .int CREATE // CREATE the dictionary entry / header
1849 .int LIT, DOCOL, COMMA // Append DOCOL (the codeword).
1850 .int LATEST, FETCH, HIDDEN // Make the word hidden (see below for definition).
1851 .int RBRAC // Go into compile mode.
1852 .int EXIT // Return from the function.
1855 ; (SEMICOLON) is also elegantly simple. Notice the F_IMMED flag.
1858 defword ";",1,F_IMMED,SEMICOLON
1859 .int LIT, EXIT, COMMA // Append EXIT (so the word will return).
1860 .int LATEST, FETCH, HIDDEN // Toggle hidden flag -- unhide the word (see below for definition).
1861 .int LBRAC // Go back to IMMEDIATE mode.
1862 .int EXIT // Return from the function.
1865 EXTENDING THE COMPILER ----------------------------------------------------------------------
1867 Words flagged with IMMEDIATE (F_IMMED) aren't just for the FORTH compiler to use. You can define
1868 your own IMMEDIATE words too, and this is a crucial aspect when extending basic FORTH, because
1869 it allows you in effect to extend the compiler itself. Does gcc let you do that?
1871 Standard FORTH words like IF, WHILE, ." and so on are all written as extensions to the basic
1872 compiler, and are all IMMEDIATE words.
1874 The IMMEDIATE word toggles the F_IMMED (IMMEDIATE flag) on the most recently defined word,
1875 or on the current word if you call it in the middle of a definition.
1879 : MYIMMEDWORD IMMEDIATE
1883 but some FORTH programmers write this instead:
1889 The two usages are equivalent, to a first approximation.
1892 defcode "IMMEDIATE",9,F_IMMED,IMMEDIATE
1893 movl var_LATEST,%edi // LATEST word.
1894 addl $4,%edi // Point to name/flags byte.
1895 xorb $F_IMMED,(%edi) // Toggle the IMMED bit.
1899 'addr HIDDEN' toggles the hidden flag (F_HIDDEN) of the word defined at addr. To hide the
1900 most recently defined word (used above in : and ; definitions) you would do:
1904 'HIDE word' toggles the flag on a named 'word'.
1906 Setting this flag stops the word from being found by FIND, and so can be used to make 'private'
1907 words. For example, to break up a large word into smaller parts you might do:
1909 : SUB1 ... subword ... ;
1910 : SUB2 ... subword ... ;
1911 : SUB3 ... subword ... ;
1912 : MAIN ... defined in terms of SUB1, SUB2, SUB3 ... ;
1917 After this, only MAIN is 'exported' or seen by the rest of the program.
1920 defcode "HIDDEN",6,,HIDDEN
1921 pop %edi // Dictionary entry.
1922 addl $4,%edi // Point to name/flags byte.
1923 xorb $F_HIDDEN,(%edi) // Toggle the HIDDEN bit.
1926 defword "HIDE",4,,HIDE
1927 .int WORD // Get the word (after HIDE).
1928 .int FIND // Look up in the dictionary.
1929 .int HIDDEN // Set F_HIDDEN flag.
1930 .int EXIT // Return.
1933 ' (TICK) is a standard FORTH word which returns the codeword pointer of the next word.
1935 The common usage is:
1939 which appends the codeword of FOO to the current word we are defining (this only works in compiled code).
1941 You tend to use ' in IMMEDIATE words. For example an alternate (and rather useless) way to define
1942 a literal 2 might be:
1945 ' LIT , \ Appends LIT to the currently-being-defined word
1946 2 , \ Appends the number 2 to the currently-being-defined word
1953 (If you don't understand how LIT2 works, then you should review the material about compiling words
1954 and immediate mode).
1956 This definition of ' uses a cheat which I copied from buzzard92. As a result it only works in
1957 compiled code. It is possible to write a version of ' based on WORD, FIND, >CFA which works in
1961 lodsl // Get the address of the next word and skip it.
1962 pushl %eax // Push it on the stack.
1966 BRANCHING ----------------------------------------------------------------------
1968 It turns out that all you need in order to define looping constructs, IF-statements, etc.
1971 BRANCH is an unconditional branch. 0BRANCH is a conditional branch (it only branches if the
1972 top of stack is zero).
1974 The diagram below shows how BRANCH works in some imaginary compiled word. When BRANCH executes,
1975 %esi starts by pointing to the offset field (compare to LIT above):
1977 +---------------------+-------+---- - - ---+------------+------------+---- - - - ----+------------+
1978 | (Dictionary header) | DOCOL | | BRANCH | offset | (skipped) | word |
1979 +---------------------+-------+---- - - ---+------------+-----|------+---- - - - ----+------------+
1982 | +-----------------------+
1983 %esi added to offset
1985 The offset is added to %esi to make the new %esi, and the result is that when NEXT runs, execution
1986 continues at the branch target. Negative offsets work as expected.
1988 0BRANCH is the same except the branch happens conditionally.
1990 Now standard FORTH words such as IF, THEN, ELSE, WHILE, REPEAT, etc. can be implemented entirely
1991 in FORTH. They are IMMEDIATE words which append various combinations of BRANCH or 0BRANCH
1992 into the word currently being compiled.
1994 As an example, code written like this:
1996 condition-code IF true-part THEN rest-code
2000 condition-code 0BRANCH OFFSET true-part rest-code
2006 defcode "BRANCH",6,,BRANCH
2007 add (%esi),%esi // add the offset to the instruction pointer
2010 defcode "0BRANCH",7,,ZBRANCH
2012 test %eax,%eax // top of stack is zero?
2013 jz code_BRANCH // if so, jump back to the branch function above
2014 lodsl // otherwise we need to skip the offset
2018 LITERAL STRINGS ----------------------------------------------------------------------
2020 LITSTRING is a primitive used to implement the ." and S" operators (which are written in
2021 FORTH). See the definition of those operators later.
2023 TELL just prints a string. It's more efficient to define this in assembly because we
2024 can make it a single Linux syscall.
2027 defcode "LITSTRING",9,,LITSTRING
2028 lodsl // get the length of the string
2029 push %esi // push the address of the start of the string
2030 push %eax // push it on the stack
2031 addl %eax,%esi // skip past the string
2032 addl $3,%esi // but round up to next 4 byte boundary
2036 defcode "TELL",4,,TELL
2037 mov $1,%ebx // 1st param: stdout
2038 pop %edx // 3rd param: length of string
2039 pop %ecx // 2nd param: address of string
2040 mov $__NR_write,%eax // write syscall
2045 QUIT AND INTERPRET ----------------------------------------------------------------------
2047 QUIT is the first FORTH function called, almost immediately after the FORTH system "boots".
2048 As explained before, QUIT doesn't "quit" anything. It does some initialisation (in particular
2049 it clears the return stack) and it calls INTERPRET in a loop to interpret commands. The
2050 reason it is called QUIT is because you can call it from your own FORTH words in order to
2051 "quit" your program and start again at the user prompt.
2053 INTERPRET is the FORTH interpreter ("toploop", "toplevel" or "REPL" might be a more accurate
2054 description -- see: http://en.wikipedia.org/wiki/REPL).
2057 // QUIT must not return (ie. must not call EXIT).
2058 defword "QUIT",4,,QUIT
2059 .int RZ,RSPSTORE // R0 RSP!, clear the return stack
2060 .int INTERPRET // interpret the next word
2061 .int BRANCH,-8 // and loop (indefinitely)
2064 This interpreter is pretty simple, but remember that in FORTH you can always override
2065 it later with a more powerful one!
2067 defcode "INTERPRET",9,,INTERPRET
2068 call _WORD // Returns %ecx = length, %edi = pointer to word.
2070 // Is it in the dictionary?
2072 movl %eax,interpret_is_lit // Not a literal number (not yet anyway ...)
2073 call _FIND // Returns %eax = pointer to header or 0 if not found.
2074 test %eax,%eax // Found?
2077 // In the dictionary. Is it an IMMEDIATE codeword?
2078 mov %eax,%edi // %edi = dictionary entry
2079 movb 4(%edi),%al // Get name+flags.
2080 push %ax // Just save it for now.
2081 call _TCFA // Convert dictionary entry (in %edi) to codeword pointer.
2083 andb $F_IMMED,%al // Is IMMED flag set?
2085 jnz 4f // If IMMED, jump straight to executing.
2089 1: // Not in the dictionary (not a word) so assume it's a literal number.
2090 incl interpret_is_lit
2091 call _NUMBER // Returns the parsed number in %eax, %ecx > 0 if error
2095 mov $LIT,%eax // The word is LIT
2097 2: // Are we compiling or executing?
2100 jz 4f // Jump if executing.
2102 // Compiling - just append the word to the current dictionary definition.
2104 mov interpret_is_lit,%ecx // Was it a literal?
2107 mov %ebx,%eax // Yes, so LIT is followed by a number.
2111 4: // Executing - run it!
2112 mov interpret_is_lit,%ecx // Literal?
2113 test %ecx,%ecx // Literal?
2116 // Not a literal, execute it now. This never returns, but the codeword will
2117 // eventually call NEXT which will reenter the loop in QUIT.
2120 5: // Executing a literal, which means push it on the stack.
2124 6: // Parse error (not a known word or a number in the current BASE).
2125 // Print an error message followed by up to 40 characters of context.
2126 mov $2,%ebx // 1st param: stderr
2127 mov $errmsg,%ecx // 2nd param: error message
2128 mov $errmsgend-errmsg,%edx // 3rd param: length of string
2129 mov $__NR_write,%eax // write syscall
2132 mov (currkey),%ecx // the error occurred just before currkey position
2134 sub $buffer,%edx // %edx = currkey - buffer (length in buffer before currkey)
2135 cmp $40,%edx // if > 40, then print only 40 characters
2138 7: sub %edx,%ecx // %ecx = start of area to print, %edx = length
2139 mov $__NR_write,%eax // write syscall
2142 mov $errmsgnl,%ecx // newline
2144 mov $__NR_write,%eax // write syscall
2150 errmsg: .ascii "PARSE ERROR: "
2152 errmsgnl: .ascii "\n"
2154 .data // NB: easier to fit in the .data section
2157 .int 0 // Flag used to record if reading a literal
2160 ODDS AND ENDS ----------------------------------------------------------------------
2162 CHAR puts the ASCII code of the first character of the following word on the stack. For example
2163 CHAR A puts 65 on the stack.
2165 EXECUTE is used to run execution tokens. See the discussion of execution tokens in the
2166 FORTH code for more details.
2168 SYSCALL0, SYSCALL1, SYSCALL2, SYSCALL3 make a standard Linux system call. (See <asm/unistd.h>
2169 for a list of system call numbers). As their name suggests these forms take between 0 and 3
2170 syscall parameters, plus the system call number.
2172 In this FORTH, SYSCALL0 must be the last word in the built-in (assembler) dictionary because we
2173 initialise the LATEST variable to point to it. This means that if you want to extend the assembler
2174 part, you must put new words before SYSCALL0, or else change how LATEST is initialised.
2177 defcode "CHAR",4,,CHAR
2178 call _WORD // Returns %ecx = length, %edi = pointer to word.
2180 movb (%edi),%al // Get the first character of the word.
2181 push %eax // Push it onto the stack.
2184 defcode "EXECUTE",7,,EXECUTE
2185 pop %eax // Get xt into %eax
2186 jmp *(%eax) // and jump to it.
2187 // After xt runs its NEXT will continue executing the current word.
2189 defcode "SYSCALL3",8,,SYSCALL3
2190 pop %eax // System call number (see <asm/unistd.h>)
2191 pop %ebx // First parameter.
2192 pop %ecx // Second parameter
2193 pop %edx // Third parameter
2195 push %eax // Result (negative for -errno)
2198 defcode "SYSCALL2",8,,SYSCALL2
2199 pop %eax // System call number (see <asm/unistd.h>)
2200 pop %ebx // First parameter.
2201 pop %ecx // Second parameter
2203 push %eax // Result (negative for -errno)
2206 defcode "SYSCALL1",8,,SYSCALL1
2207 pop %eax // System call number (see <asm/unistd.h>)
2208 pop %ebx // First parameter.
2210 push %eax // Result (negative for -errno)
2213 defcode "SYSCALL0",8,,SYSCALL0
2214 pop %eax // System call number (see <asm/unistd.h>)
2216 push %eax // Result (negative for -errno)
2220 DATA SEGMENT ----------------------------------------------------------------------
2222 Here we set up the Linux data segment, used for user definitions and variously known as just
2223 the 'data segment', 'user memory' or 'user definitions area'. It is an area of memory which
2224 grows upwards and stores both newly-defined FORTH words and global variables of various
2227 It is completely analogous to the C heap, except there is no generalised 'malloc' and 'free'
2228 (but as with everything in FORTH, writing such functions would just be a Simple Matter
2229 Of Programming). Instead in normal use the data segment just grows upwards as new FORTH
2230 words are defined/appended to it.
2232 There are various "features" of the GNU toolchain which make setting up the data segment
2233 more complicated than it really needs to be. One is the GNU linker which inserts a random
2234 "build ID" segment. Another is Address Space Randomization which means we can't tell
2235 where the kernel will choose to place the data segment (or the stack for that matter).
2237 Therefore writing this set_up_data_segment assembler routine is a little more complicated
2238 than it really needs to be. We ask the Linux kernel where it thinks the data segment starts
2239 using the brk(2) system call, then ask it to reserve some initial space (also using brk(2)).
2241 You don't need to worry about this code.
2244 .set INITIAL_DATA_SEGMENT_SIZE,65536
2245 set_up_data_segment:
2246 xor %ebx,%ebx // Call brk(0)
2249 movl %eax,var_HERE // Initialise HERE to point at beginning of data segment.
2250 addl $INITIAL_DATA_SEGMENT_SIZE,%eax // Reserve nn bytes of memory for initial data segment.
2251 movl %eax,%ebx // Call brk(HERE+INITIAL_DATA_SEGMENT_SIZE)
2257 We allocate static buffers for the return static and input buffer (used when
2258 reading in files and text that the user types in).
2260 .set RETURN_STACK_SIZE,8192
2261 .set BUFFER_SIZE,4096
2264 /* FORTH return stack. */
2267 .space RETURN_STACK_SIZE
2268 return_stack_top: // Initial top of return stack.
2270 /* This is used as a temporary input buffer when reading from files or the terminal. */
2276 START OF FORTH CODE ----------------------------------------------------------------------
2278 We've now reached the stage where the FORTH system is running and self-hosting. All further
2279 words can be written as FORTH itself, including words like IF, THEN, .", etc which in most
2280 languages would be considered rather fundamental.
2282 I used to append this here in the assembly file, but I got sick of fighting against gas's
2283 crack-smoking (lack of) multiline string syntax. So now that is in a separate file called
2286 If you don't already have that file, download it from http://annexia.org/forth in order
2287 to continue the tutorial.
2290 /* END OF jonesforth.S */