File:  [gforth] / gforth / Attic / gforth.ds
Revision 1.6: download - view: text, annotated - select for diffs
Wed Jan 18 18:41:37 1995 UTC (24 years, 10 months ago) by anton
Branches: MAIN
CVS tags: HEAD
worked a bit on m68k.h and power.h
moved hyperbolic functions and falog to primitives

    1: \input texinfo   @c -*-texinfo-*-
    2: @comment The source is gforth.ds, from which gforth.texi is generated
    3: @comment %**start of header (This is for running Texinfo on a region.)
    4: @setfilename gforth.info
    5: @settitle GNU Forth Manual
    6: @comment @setchapternewpage odd
    7: @comment %**end of header (This is for running Texinfo on a region.)
    8: 
    9: @ifinfo
   10: This file documents GNU Forth 0.0
   11: 
   12: Copyright @copyright{} 1994 GNU Forth Development Group
   13: 
   14:      Permission is granted to make and distribute verbatim copies of
   15:      this manual provided the copyright notice and this permission notice
   16:      are preserved on all copies.
   17:      
   18: @ignore
   19:      Permission is granted to process this file through TeX and print the
   20:      results, provided the printed document carries a copying permission
   21:      notice identical to this one except for the removal of this paragraph
   22:      (this paragraph not being relevant to the printed manual).
   23:      
   24: @end ignore
   25:      Permission is granted to copy and distribute modified versions of this
   26:      manual under the conditions for verbatim copying, provided also that the
   27:      sections entitled "Distribution" and "General Public License" are
   28:      included exactly as in the original, and provided that the entire
   29:      resulting derived work is distributed under the terms of a permission
   30:      notice identical to this one.
   31:      
   32:      Permission is granted to copy and distribute translations of this manual
   33:      into another language, under the above conditions for modified versions,
   34:      except that the sections entitled "Distribution" and "General Public
   35:      License" may be included in a translation approved by the author instead
   36:      of in the original English.
   37: @end ifinfo
   38: 
   39: @titlepage
   40: @sp 10
   41: @center @titlefont{GNU Forth Manual}
   42: @sp 2
   43: @center for version 0.0
   44: @sp 2
   45: @center Anton Ertl
   46: 
   47: @comment  The following two commands start the copyright page.
   48: @page
   49: @vskip 0pt plus 1filll
   50: Copyright @copyright{} 1994 GNU Forth Development Group
   51: 
   52: @comment !! Published by ... or You can get a copy of this manual ...
   53: 
   54:      Permission is granted to make and distribute verbatim copies of
   55:      this manual provided the copyright notice and this permission notice
   56:      are preserved on all copies.
   57:      
   58:      Permission is granted to copy and distribute modified versions of this
   59:      manual under the conditions for verbatim copying, provided also that the
   60:      sections entitled "Distribution" and "General Public License" are
   61:      included exactly as in the original, and provided that the entire
   62:      resulting derived work is distributed under the terms of a permission
   63:      notice identical to this one.
   64:      
   65:      Permission is granted to copy and distribute translations of this manual
   66:      into another language, under the above conditions for modified versions,
   67:      except that the sections entitled "Distribution" and "General Public
   68:      License" may be included in a translation approved by the author instead
   69:      of in the original English.
   70: @end titlepage
   71: 
   72: 
   73: @node Top, License, (dir), (dir)
   74: @ifinfo
   75: GNU Forth is a free implementation of ANS Forth available on many
   76: personal machines. This manual corresponds to version 0.0.
   77: @end ifinfo
   78: 
   79: @menu
   80: * License::                     
   81: * Goals::                       About the GNU Forth Project
   82: * Other Books::                 Things you might want to read
   83: * Invocation::                  Starting GNU Forth
   84: * Words::                       Forth words available in GNU Forth
   85: * ANS conformance::             Implementation-defined options etc.
   86: * Model::                       The abstract machine of GNU Forth
   87: * Emacs and GForth::            The GForth Mode
   88: * Internals::                   Implementation details
   89: * Bugs::                        How to report them
   90: * Pedigree::                    Ancestors of GNU Forth
   91: * Word Index::                  An item for each Forth word
   92: * Node Index::                  An item for each node
   93: @end menu
   94: 
   95: @node License, Goals, Top, Top
   96: @unnumbered License
   97: !! Insert GPL here
   98: 
   99: @iftex
  100: @unnumbered Preface
  101: This manual documents GNU Forth. The reader is expected to know
  102: Forth. This manual is primarily a reference manual. @xref{Other Books}
  103: for introductory material.
  104: @end iftex
  105: 
  106: @node    Goals, Other Books, License, Top
  107: @comment node-name,     next,           previous, up
  108: @chapter Goals of GNU Forth
  109: @cindex Goals
  110: The goal of the GNU Forth Project is to develop a standard model for
  111: ANSI Forth. This can be split into several subgoals:
  112: 
  113: @itemize @bullet
  114: @item
  115: GNU Forth should conform to the ANSI Forth standard.
  116: @item
  117: It should be a model, i.e. it should define all the
  118: implementation-dependent things.
  119: @item
  120: It should become standard, i.e. widely accepted and used. This goal
  121: is the most difficult one.
  122: @end itemize
  123: 
  124: To achieve these goals GNU Forth should be
  125: @itemize @bullet
  126: @item
  127: Similar to previous models (fig-Forth, F83)
  128: @item
  129: Powerful. It should provide for all the things that are considered
  130: necessary today and even some that are not yet considered necessary.
  131: @item
  132: Efficient. It should not get the reputation of being exceptionally
  133: slow.
  134: @item
  135: Free.
  136: @item
  137: Available on many machines/easy to port.
  138: @end itemize
  139: 
  140: Have we achieved these goals? GNU Forth conforms to the ANS Forth
  141: standard; it may be considered a model, but we have not yet documented
  142: which parts of the model are stable and which parts we are likely to
  143: change; it certainly has not yet become a de facto standard. It has some
  144: similarities and some differences to previous models; It has some
  145: powerful features, but not yet everything that we envisioned; on RISCs
  146: it is as fast as interpreters programmed in assembly, on
  147: register-starved machines it is not so fast, but still faster than any
  148: other C-based interpretive implementation; it is free and available on
  149: many machines.
  150: 
  151: @node Other Books, Invocation, Goals, Top
  152: @chapter Other books on ANS Forth
  153: 
  154: As the standard is relatively new, there are not many books out yet. It
  155: is not recommended to learn Forth by using GNU Forth and a book that is
  156: not written for ANS Forth, as you will not know your mistakes from the
  157: deviations of the book.
  158: 
  159: There is, of course, the standard, the definite reference if you want to
  160: write ANS Forth programs. It will be available in printed form from
  161: Global Engineering Documents !! somtime in spring or summer 1994. If you
  162: are lucky, you can still get dpANS6 (the draft that was approved as
  163: standard) by aftp from ftp.uu.net:/vendor/minerva/x3j14.
  164: 
  165: @cite{Forth: The new model} by Jack Woehr (!! Publisher) is an
  166: introductory book based on a draft version of the standard. It does not
  167: cover the whole standard. It also contains interesting background
  168: information (Jack Woehr was in the ANS Forth Technical Committe). It is
  169: not appropriate for complete newbies, but programmers experienced in
  170: other languages should find it ok.
  171: 
  172: @node Invocation, Words, Other Books, Top
  173: @chapter Invocation
  174: 
  175: You will usually just say @code{gforth}. In many other cases the default
  176: GNU Forth image will be invoked like this:
  177: 
  178: @example
  179: gforth [files] [-e forth-code]
  180: @end example
  181: 
  182: executing the contents of the files and the Forth code in the order they
  183: are given.
  184: 
  185: In general, the command line looks like this:
  186: 
  187: @example
  188: gforth [initialization options] [image-specific options]
  189: @end example
  190: 
  191: The initialization options must come before the rest of the command
  192: line. They are:
  193: 
  194: @table @code
  195: @item --image-file @var{file}
  196: Loads the Forth image @var{file} instead of the default
  197: @file{gforth.fi}.
  198: 
  199: @item --path @var{path}
  200: Uses @var{path} for searching the image file and Forth source code
  201: files instead of the default in the environment variable
  202: @code{GFORTHPATH} or the path specified at installation time (typically
  203: @file{/usr/local/lib/gforth:.}). A path is given as a @code{:}-separated
  204: list.
  205: 
  206: @item --dictionary-size @var{size}
  207: @item -m @var{size}
  208: Allocate @var{size} space for the Forth dictionary space instead of
  209: using the default specified in the image (typically 256K). The
  210: @var{size} specification consists of an integer and a unit (e.g.,
  211: @code{4M}). The unit can be one of @code{b} (bytes), @code{e} (element
  212: size, in this case Cells), @code{k} (kilobytes), and @code{M}
  213: (Megabytes). If no unit is specified, @code{e} is used.
  214: 
  215: @item --data-stack-size @var{size}
  216: @item -d @var{size}
  217: Allocate @var{size} space for the data stack instead of using the
  218: default specified in the image (typically 16K).
  219: 
  220: @item --return-stack-size @var{size}
  221: @item -r @var{size}
  222: Allocate @var{size} space for the return stack instead of using the
  223: default specified in the image (typically 16K).
  224: 
  225: @item --fp-stack-size @var{size}
  226: @item -f @var{size}
  227: Allocate @var{size} space for the floating point stack instead of
  228: using the default specified in the image (typically 16K). In this case
  229: the unit specifier @code{e} refers to floating point numbers.
  230: 
  231: @item --locals-stack-size @var{size}
  232: @item -l @var{size}
  233: Allocate @var{size} space for the locals stack instead of using the
  234: default specified in the image (typically 16K).
  235: 
  236: @end table
  237: 
  238: As explained above, the image-specific command-line arguments for the
  239: default image @file{gforth.fi} consist of a sequence of filenames and
  240: @code{-e @var{forth-code}} options that are interpreted in the seqence
  241: in which they are given. The @code{-e @var{forth-code}} or
  242: @code{--evaluate @var{forth-code}} option evaluates the forth
  243: code. This option takes only one argument; if you want to evaluate more
  244: Forth words, you have to quote them or use several @code{-e}s. To exit
  245: after processing the command line (instead of entering interactive mode)
  246: append @code{-e bye} to the command line.
  247: 
  248: Not yet implemented:
  249: On startup the system first executes the system initialization file
  250: (unless the option @code{--no-init-file} is given; note that the system
  251: resulting from using this option may not be ANS Forth conformant). Then
  252: the user initialization file @file{.gforth.fs} is executed, unless the
  253: option @code{--no-rc} is given; this file is first searched in @file{.},
  254: then in @file{~}, then in the normal path (see above).
  255: 
  256: @node Words, ANS conformance, Invocation, Top
  257: @chapter Forth Words
  258: 
  259: @menu
  260: * Notation::                    
  261: * Arithmetic::                  
  262: * Stack Manipulation::          
  263: * Memory access::               
  264: * Control Structures::          
  265: * Locals::                      
  266: * Defining Words::              
  267: * Wordlists::                   
  268: * Files::                       
  269: * Blocks::                      
  270: * Other I/O::                   
  271: * Programming Tools::           
  272: * Threading Words::             
  273: @end menu
  274: 
  275: @node Notation, Arithmetic, Words, Words
  276: @section Notation
  277: 
  278: The Forth words are described in this section in the glossary notation
  279: that has become a de-facto standard for Forth texts, i.e.
  280: 
  281: @format
  282: @var{word}     @var{Stack effect}   @var{wordset}   @var{pronunciation}
  283: @end format
  284: @var{Description}
  285: 
  286: @table @var
  287: @item word
  288: The name of the word. BTW, GNU Forth is case insensitive, so you can
  289: type the words in in lower case.
  290: 
  291: @item Stack effect
  292: The stack effect is written in the notation @code{@var{before} --
  293: @var{after}}, where @var{before} and @var{after} describe the top of
  294: stack entries before and after the execution of the word. The rest of
  295: the stack is not touched by the word. The top of stack is rightmost,
  296: i.e., a stack sequence is written as it is typed in. Note that GNU Forth
  297: uses a separate floating point stack, but a unified stack
  298: notation. Also, return stack effects are not shown in @var{stack
  299: effect}, but in @var{Description}. The name of a stack item describes
  300: the type and/or the function of the item. See below for a discussion of
  301: the types.
  302: 
  303: @item pronunciation
  304: How the word is pronounced
  305: 
  306: @item wordset
  307: The ANS Forth standard is divided into several wordsets. A standard
  308: system need not support all of them. So, the fewer wordsets your program
  309: uses the more portable it will be in theory. However, we suspect that
  310: most ANS Forth systems on personal machines will feature all
  311: wordsets. Words that are not defined in the ANS standard have
  312: @code{gforth} as wordset.
  313: 
  314: @item Description
  315: A description of the behaviour of the word.
  316: @end table
  317: 
  318: The type of a stack item is specified by the character(s) the name
  319: starts with:
  320: 
  321: @table @code
  322: @item f
  323: Bool, i.e. @code{false} or @code{true}.
  324: @item c
  325: Char
  326: @item w
  327: Cell, can contain an integer or an address
  328: @item n
  329: signed integer
  330: @item u
  331: unsigned integer
  332: @item d
  333: double sized signed integer
  334: @item ud
  335: double sized unsigned integer
  336: @item r
  337: Float
  338: @item a_
  339: Cell-aligned address
  340: @item c_
  341: Char-aligned address (note that a Char is two bytes in Windows NT)
  342: @item f_
  343: Float-aligned address
  344: @item df_
  345: Address aligned for IEEE double precision float
  346: @item sf_
  347: Address aligned for IEEE single precision float
  348: @item xt
  349: Execution token, same size as Cell
  350: @item wid
  351: Wordlist ID, same size as Cell
  352: @item f83name
  353: Pointer to a name structure
  354: @end table
  355: 
  356: @node Arithmetic, Stack Manipulation, Notation, Words
  357: @section Arithmetic
  358: Forth arithmetic is not checked, i.e., you will not hear about integer
  359: overflow on addition or multiplication, you may hear about division by
  360: zero if you are lucky. The operator is written after the operands, but
  361: the operands are still in the original order. I.e., the infix @code{2-1}
  362: corresponds to @code{2 1 -}. Forth offers a variety of division
  363: operators. If you perform division with potentially negative operands,
  364: you do not want to use @code{/} or @code{/mod} with its undefined
  365: behaviour, but rather @code{fm/mod} or @code{sm/mod} (probably the
  366: former, @pxref{Mixed precision}).
  367: 
  368: @menu
  369: * Single precision::            
  370: * Bitwise operations::          
  371: * Mixed precision::             operations with single and double-cell integers
  372: * Double precision::            Double-cell integer arithmetic
  373: * Floating Point::              
  374: @end menu
  375: 
  376: @node Single precision, Bitwise operations, Arithmetic, Arithmetic
  377: @subsection Single precision
  378: doc-+
  379: doc--
  380: doc-*
  381: doc-/
  382: doc-mod
  383: doc-/mod
  384: doc-negate
  385: doc-abs
  386: doc-min
  387: doc-max
  388: 
  389: @node Bitwise operations, Mixed precision, Single precision, Arithmetic
  390: @subsection Bitwise operations
  391: doc-and
  392: doc-or
  393: doc-xor
  394: doc-invert
  395: doc-2*
  396: doc-2/
  397: 
  398: @node Mixed precision, Double precision, Bitwise operations, Arithmetic
  399: @subsection Mixed precision
  400: doc-m+
  401: doc-*/
  402: doc-*/mod
  403: doc-m*
  404: doc-um*
  405: doc-m*/
  406: doc-um/mod
  407: doc-fm/mod
  408: doc-sm/rem
  409: 
  410: @node Double precision, Floating Point, Mixed precision, Arithmetic
  411: @subsection Double precision
  412: doc-d+
  413: doc-d-
  414: doc-dnegate
  415: doc-dabs
  416: doc-dmin
  417: doc-dmax
  418: 
  419: @node Floating Point,  , Double precision, Arithmetic
  420: @subsection Floating Point
  421: 
  422: Angles in floating point operations are given in radians (a full circle
  423: has 2 pi radians). Note, that gforth has a separate floating point
  424: stack, but we use the unified notation.
  425: 
  426: Floating point numbers have a number of unpleasant surprises for the
  427: unwary (e.g., floating point addition is not associative) and even a few
  428: for the wary. You should not use them unless you know what you are doing
  429: or you don't care that the results you get are totally bogus. If you
  430: want to learn about the problems of floating point numbers (and how to
  431: avoid them), you might start with @cite{David (?) Goldberg, What Every
  432: Computer Scientist Should Know About Floating-Point Arithmetic, ACM
  433: Computing Surveys 23(1):5@minus{}48, March 1991}.
  434: 
  435: doc-f+
  436: doc-f-
  437: doc-f*
  438: doc-f/
  439: doc-fnegate
  440: doc-fabs
  441: doc-fmax
  442: doc-fmin
  443: doc-floor
  444: doc-fround
  445: doc-f**
  446: doc-fsqrt
  447: doc-fexp
  448: doc-fexpm1
  449: doc-fln
  450: doc-flnp1
  451: doc-flog
  452: doc-falog
  453: doc-fsin
  454: doc-fcos
  455: doc-fsincos
  456: doc-ftan
  457: doc-fasin
  458: doc-facos
  459: doc-fatan
  460: doc-fatan2
  461: doc-fsinh
  462: doc-fcosh
  463: doc-ftanh
  464: doc-fasinh
  465: doc-facosh
  466: doc-fatanh
  467: 
  468: @node Stack Manipulation, Memory access, Arithmetic, Words
  469: @section Stack Manipulation
  470: 
  471: gforth has a data stack (aka parameter stack) for characters, cells,
  472: addresses, and double cells, a floating point stack for floating point
  473: numbers, a return stack for storing the return addresses of colon
  474: definitions and other data, and a locals stack for storing local
  475: variables. Note that while every sane Forth has a separate floating
  476: point stack, this is not strictly required; an ANS Forth system could
  477: theoretically keep floating point numbers on the data stack. As an
  478: additional difficulty, you don't know how many cells a floating point
  479: number takes. It is reportedly possible to write words in a way that
  480: they work also for a unified stack model, but we do not recommend trying
  481: it. Instead, just say that your program has an environmental dependency
  482: on a separate FP stack.
  483: 
  484: Also, a Forth system is allowed to keep the local variables on the
  485: return stack. This is reasonable, as local variables usually eliminate
  486: the need to use the return stack explicitly. So, if you want to produce
  487: a standard complying program and if you are using local variables in a
  488: word, forget about return stack manipulations in that word (see the
  489: standard document for the exact rules).
  490: 
  491: @menu
  492: * Data stack::                  
  493: * Floating point stack::        
  494: * Return stack::                
  495: * Locals stack::                
  496: * Stack pointer manipulation::  
  497: @end menu
  498: 
  499: @node Data stack, Floating point stack, Stack Manipulation, Stack Manipulation
  500: @subsection Data stack
  501: doc-drop
  502: doc-nip
  503: doc-dup
  504: doc-over
  505: doc-tuck
  506: doc-swap
  507: doc-rot
  508: doc--rot
  509: doc-?dup
  510: doc-pick
  511: doc-roll
  512: doc-2drop
  513: doc-2nip
  514: doc-2dup
  515: doc-2over
  516: doc-2tuck
  517: doc-2swap
  518: doc-2rot
  519: 
  520: @node Floating point stack, Return stack, Data stack, Stack Manipulation
  521: @subsection Floating point stack
  522: doc-fdrop
  523: doc-fnip
  524: doc-fdup
  525: doc-fover
  526: doc-ftuck
  527: doc-fswap
  528: doc-frot
  529: 
  530: @node Return stack, Locals stack, Floating point stack, Stack Manipulation
  531: @subsection Return stack
  532: doc->r
  533: doc-r>
  534: doc-r@
  535: doc-rdrop
  536: doc-2>r
  537: doc-2r>
  538: doc-2r@
  539: doc-2rdrop
  540: 
  541: @node Locals stack, Stack pointer manipulation, Return stack, Stack Manipulation
  542: @subsection Locals stack
  543: 
  544: @node Stack pointer manipulation,  , Locals stack, Stack Manipulation
  545: @subsection Stack pointer manipulation
  546: doc-sp@
  547: doc-sp!
  548: doc-fp@
  549: doc-fp!
  550: doc-rp@
  551: doc-rp!
  552: doc-lp@
  553: doc-lp!
  554: 
  555: @node Memory access, Control Structures, Stack Manipulation, Words
  556: @section Memory access
  557: 
  558: @menu
  559: * Stack-Memory transfers::      
  560: * Address arithmetic::          
  561: * Memory block access::         
  562: @end menu
  563: 
  564: @node Stack-Memory transfers, Address arithmetic, Memory access, Memory access
  565: @subsection Stack-Memory transfers
  566: 
  567: doc-@
  568: doc-!
  569: doc-+!
  570: doc-c@
  571: doc-c!
  572: doc-2@
  573: doc-2!
  574: doc-f@
  575: doc-f!
  576: doc-sf@
  577: doc-sf!
  578: doc-df@
  579: doc-df!
  580: 
  581: @node Address arithmetic, Memory block access, Stack-Memory transfers, Memory access
  582: @subsection Address arithmetic
  583: 
  584: ANS Forth does not specify the sizes of the data types. Instead, it
  585: offers a number of words for computing sizes and doing address
  586: arithmetic. Basically, address arithmetic is performed in terms of
  587: address units (aus); on most systems the address unit is one byte. Note
  588: that a character may have more than one au, so @code{chars} is no noop
  589: (on systems where it is a noop, it compiles to nothing).
  590: 
  591: ANS Forth also defines words for aligning addresses for specific
  592: addresses. Many computers require that accesses to specific data types
  593: must only occur at specific addresses; e.g., that cells may only be
  594: accessed at addresses divisible by 4. Even if a machine allows unaligned
  595: accesses, it can usually perform aligned accesses faster. 
  596: 
  597: For the performance-concious: alignment operations are usually only
  598: necessary during the definition of a data structure, not during the
  599: (more frequent) accesses to it.
  600: 
  601: ANS Forth defines no words for character-aligning addresses. This is not
  602: an oversight, but reflects the fact that addresses that are not
  603: char-aligned have no use in the standard and therefore will not be
  604: created.
  605: 
  606: The standard guarantees that addresses returned by @code{CREATE}d words
  607: are cell-aligned; in addition, gforth guarantees that these addresses
  608: are aligned for all purposes.
  609: 
  610: doc-chars
  611: doc-char+
  612: doc-cells
  613: doc-cell+
  614: doc-align
  615: doc-aligned
  616: doc-floats
  617: doc-float+
  618: doc-falign
  619: doc-faligned
  620: doc-sfloats
  621: doc-sfloat+
  622: doc-sfalign
  623: doc-sfaligned
  624: doc-dfloats
  625: doc-dfloat+
  626: doc-dfalign
  627: doc-dfaligned
  628: doc-address-unit-bits
  629: 
  630: @node Memory block access,  , Address arithmetic, Memory access
  631: @subsection Memory block access
  632: 
  633: doc-move
  634: doc-erase
  635: 
  636: While the previous words work on address units, the rest works on
  637: characters.
  638: 
  639: doc-cmove
  640: doc-cmove>
  641: doc-fill
  642: doc-blank
  643: 
  644: @node Control Structures, Locals, Memory access, Words
  645: @section Control Structures
  646: 
  647: Control structures in Forth cannot be used in interpret state, only in
  648: compile state, i.e., in a colon definition. We do not like this
  649: limitation, but have not seen a satisfying way around it yet, although
  650: many schemes have been proposed.
  651: 
  652: @menu
  653: * Selection::                   
  654: * Simple Loops::                
  655: * Counted Loops::               
  656: * Arbitrary control structures::  
  657: * Calls and returns::           
  658: * Exception Handling::          
  659: @end menu
  660: 
  661: @node Selection, Simple Loops, Control Structures, Control Structures
  662: @subsection Selection
  663: 
  664: @example
  665: @var{flag}
  666: IF
  667:   @var{code}
  668: ENDIF
  669: @end example
  670: or
  671: @example
  672: @var{flag}
  673: IF
  674:   @var{code1}
  675: ELSE
  676:   @var{code2}
  677: ENDIF
  678: @end example
  679: 
  680: You can use @code{THEN} instead of @code{ENDIF}. Indeed, @code{THEN} is
  681: standard, and @code{ENDIF} is not, although it is quite popular. We
  682: recommend using @code{ENDIF}, because it is less confusing for people
  683: who also know other languages (and is not prone to reinforcing negative
  684: prejudices against Forth in these people). Adding @code{ENDIF} to a
  685: system that only supplies @code{THEN} is simple:
  686: @example
  687: : endif   POSTPONE then ; immediate
  688: @end example
  689: 
  690: [According to @cite{Webster's New Encyclopedic Dictionary}, @dfn{then
  691: (adv.)}  has the following meanings:
  692: @quotation
  693: ... 2b: following next after in order ... 3d: as a necessary consequence
  694: (if you were there, then you saw them).
  695: @end quotation
  696: Forth's @code{THEN} has the meaning 2b, whereas @code{THEN} in Pascal
  697: and many other programming languages has the meaning 3d.]
  698: 
  699: We also provide the words @code{?dup-if} and @code{?dup-0=-if}, so you
  700: can avoid using @code{?dup}.
  701: 
  702: @example
  703: @var{n}
  704: CASE
  705:   @var{n1} OF @var{code1} ENDOF
  706:   @var{n2} OF @var{code2} ENDOF
  707:   @dots{}
  708: ENDCASE
  709: @end example
  710: 
  711: Executes the first @var{codei}, where the @var{ni} is equal to
  712: @var{n}. A default case can be added by simply writing the code after
  713: the last @code{ENDOF}. It may use @var{n}, which is on top of the stack,
  714: but must not consume it.
  715: 
  716: @node Simple Loops, Counted Loops, Selection, Control Structures
  717: @subsection Simple Loops
  718: 
  719: @example
  720: BEGIN
  721:   @var{code1}
  722:   @var{flag}
  723: WHILE
  724:   @var{code2}
  725: REPEAT
  726: @end example
  727: 
  728: @var{code1} is executed and @var{flag} is computed. If it is true,
  729: @var{code2} is executed and the loop is restarted; If @var{flag} is false, execution continues after the @code{REPEAT}.
  730: 
  731: @example
  732: BEGIN
  733:   @var{code}
  734:   @var{flag}
  735: UNTIL
  736: @end example
  737: 
  738: @var{code} is executed. The loop is restarted if @code{flag} is false.
  739: 
  740: @example
  741: BEGIN
  742:   @var{code}
  743: AGAIN
  744: @end example
  745: 
  746: This is an endless loop.
  747: 
  748: @node Counted Loops, Arbitrary control structures, Simple Loops, Control Structures
  749: @subsection Counted Loops
  750: 
  751: The basic counted loop is:
  752: @example
  753: @var{limit} @var{start}
  754: ?DO
  755:   @var{body}
  756: LOOP
  757: @end example
  758: 
  759: This performs one iteration for every integer, starting from @var{start}
  760: and up to, but excluding @var{limit}. The counter, aka index, can be
  761: accessed with @code{i}. E.g., the loop
  762: @example
  763: 10 0 ?DO
  764:   i .
  765: LOOP
  766: @end example
  767: prints
  768: @example
  769: 0 1 2 3 4 5 6 7 8 9
  770: @end example
  771: The index of the innermost loop can be accessed with @code{i}, the index
  772: of the next loop with @code{j}, and the index of the third loop with
  773: @code{k}.
  774: 
  775: The loop control data are kept on the return stack, so there are some
  776: restrictions on mixing return stack accesses and counted loop
  777: words. E.g., if you put values on the return stack outside the loop, you
  778: cannot read them inside the loop. If you put values on the return stack
  779: within a loop, you have to remove them before the end of the loop and
  780: before accessing the index of the loop.
  781: 
  782: There are several variations on the counted loop:
  783: 
  784: @code{LEAVE} leaves the innermost counted loop immediately.
  785: 
  786: @code{LOOP} can be replaced with @code{@var{n} +LOOP}; this updates the
  787: index by @var{n} instead of by 1. The loop is terminated when the border
  788: between @var{limit-1} and @var{limit} is crossed. E.g.:
  789: 
  790: @code{4 0 ?DO  i .  2 +LOOP}   prints @code{0 2}
  791: 
  792: @code{4 1 ?DO  i .  2 +LOOP}   prints @code{1 3}
  793: 
  794: The behaviour of @code{@var{n} +LOOP} is peculiar when @var{n} is negative:
  795: 
  796: @code{-1 0 ?DO  i .  -1 +LOOP}  prints @code{0 -1}
  797: 
  798: @code{ 0 0 ?DO  i .  -1 +LOOP}  prints nothing
  799: 
  800: Therefore we recommend avoiding using @code{@var{n} +LOOP} with negative
  801: @var{n}. One alternative is @code{@var{n} S+LOOP}, where the negative
  802: case behaves symmetrical to the positive case:
  803: 
  804: @code{-2 0 ?DO  i .  -1 +LOOP}  prints @code{0 -1}
  805: 
  806: @code{-1 0 ?DO  i .  -1 +LOOP}  prints @code{0}
  807: 
  808: @code{ 0 0 ?DO  i .  -1 +LOOP}  prints nothing
  809: 
  810: The loop is terminated when the border between @var{limit@minus{}sgn(n)} and
  811: @var{limit} is crossed. However, @code{S+LOOP} is not part of the ANS
  812: Forth standard.
  813: 
  814: @code{?DO} can be replaced by @code{DO}. @code{DO} enters the loop even
  815: when the start and the limit value are equal. We do not recommend using
  816: @code{DO}. It will just give you maintenance troubles.
  817: 
  818: @code{UNLOOP} is used to prepare for an abnormal loop exit, e.g., via
  819: @code{EXIT}. @code{UNLOOP} removes the loop control parameters from the
  820: return stack so @code{EXIT} can get to its return address.
  821: 
  822: Another counted loop is
  823: @example
  824: @var{n}
  825: FOR
  826:   @var{body}
  827: NEXT
  828: @end example
  829: This is the preferred loop of native code compiler writers who are too
  830: lazy to optimize @code{?DO} loops properly. In GNU Forth, this loop
  831: iterates @var{n+1} times; @code{i} produces values starting with @var{n}
  832: and ending with 0. Other Forth systems may behave differently, even if
  833: they support @code{FOR} loops.
  834: 
  835: @node Arbitrary control structures, Calls and returns, Counted Loops, Control Structures
  836: @subsection Arbitrary control structures
  837: 
  838: ANS Forth permits and supports using control structures in a non-nested
  839: way. Information about incomplete control structures is stored on the
  840: control-flow stack. This stack may be implemented on the Forth data
  841: stack, and this is what we have done in gforth.
  842: 
  843: An @i{orig} entry represents an unresolved forward branch, a @i{dest}
  844: entry represents a backward branch target. A few words are the basis for
  845: building any control structure possible (except control structures that
  846: need storage, like calls, coroutines, and backtracking).
  847: 
  848: doc-if
  849: doc-ahead
  850: doc-then
  851: doc-begin
  852: doc-until
  853: doc-again
  854: doc-cs-pick
  855: doc-cs-roll
  856: 
  857: On many systems control-flow stack items take one word, in gforth they
  858: currently take three (this may change in the future). Therefore it is a
  859: really good idea to manipulate the control flow stack with
  860: @code{cs-pick} and @code{cs-roll}, not with data stack manipulation
  861: words.
  862: 
  863: Some standard control structure words are built from these words:
  864: 
  865: doc-else
  866: doc-while
  867: doc-repeat
  868: 
  869: Counted loop words constitute a separate group of words:
  870: 
  871: doc-?do
  872: doc-do
  873: doc-for
  874: doc-loop
  875: doc-s+loop
  876: doc-+loop
  877: doc-next
  878: doc-leave
  879: doc-?leave
  880: doc-unloop
  881: doc-undo
  882: 
  883: The standard does not allow using @code{cs-pick} and @code{cs-roll} on
  884: @i{do-sys}. Our system allows it, but it's your job to ensure that for
  885: every @code{?DO} etc. there is exactly one @code{UNLOOP} on any path
  886: through the definition (@code{LOOP} etc. compile an @code{UNLOOP} on the
  887: fall-through path). Also, you have to ensure that all @code{LEAVE}s are
  888: resolved (by using one of the loop-ending words or @code{UNDO}).
  889: 
  890: Another group of control structure words are
  891: 
  892: doc-case
  893: doc-endcase
  894: doc-of
  895: doc-endof
  896: 
  897: @i{case-sys} and @i{of-sys} cannot be processed using @code{cs-pick} and
  898: @code{cs-roll}.
  899: 
  900: @subsubsection Programming Style
  901: 
  902: In order to ensure readability we recommend that you do not create
  903: arbitrary control structures directly, but define new control structure
  904: words for the control structure you want and use these words in your
  905: program.
  906: 
  907: E.g., instead of writing
  908: 
  909: @example
  910: begin
  911:   ...
  912: if [ 1 cs-roll ]
  913:   ...
  914: again then
  915: @end example
  916: 
  917: we recommend defining control structure words, e.g.,
  918: 
  919: @example
  920: : while ( dest -- orig dest )
  921:  POSTPONE if
  922:  1 cs-roll ; immediate
  923: 
  924: : repeat ( orig dest -- )
  925:  POSTPONE again
  926:  POSTPONE then ; immediate
  927: @end example
  928: 
  929: and then using these to create the control structure:
  930: 
  931: @example
  932: begin
  933:   ...
  934: while
  935:   ...
  936: repeat
  937: @end example
  938: 
  939: That's much easier to read, isn't it? Of course, @code{BEGIN} and
  940: @code{WHILE} are predefined, so in this example it would not be
  941: necessary to define them.
  942: 
  943: @node Calls and returns, Exception Handling, Arbitrary control structures, Control Structures
  944: @subsection Calls and returns
  945: 
  946: A definition can be called simply be writing the name of the
  947: definition. When the end of the definition is reached, it returns. An earlier return can be forced using
  948: 
  949: doc-exit
  950: 
  951: Don't forget to clean up the return stack and @code{UNLOOP} any
  952: outstanding @code{?DO}...@code{LOOP}s before @code{EXIT}ing. The
  953: primitive compiled by @code{EXIT} is
  954: 
  955: doc-;s
  956: 
  957: @node Exception Handling,  , Calls and returns, Control Structures
  958: @subsection Exception Handling
  959: 
  960: doc-catch
  961: doc-throw
  962: 
  963: @node Locals, Defining Words, Control Structures, Words
  964: @section Locals
  965: 
  966: Local variables can make Forth programming more enjoyable and Forth
  967: programs easier to read. Unfortunately, the locals of ANS Forth are
  968: laden with restrictions. Therefore, we provide not only the ANS Forth
  969: locals wordset, but also our own, more powerful locals wordset (we
  970: implemented the ANS Forth locals wordset through our locals wordset).
  971: 
  972: @menu
  973: * gforth locals::               
  974: * ANS Forth locals::            
  975: @end menu
  976: 
  977: @node gforth locals, ANS Forth locals, Locals, Locals
  978: @subsection gforth locals
  979: 
  980: Locals can be defined with
  981: 
  982: @example
  983: @{ local1 local2 ... -- comment @}
  984: @end example
  985: or
  986: @example
  987: @{ local1 local2 ... @}
  988: @end example
  989: 
  990: E.g.,
  991: @example
  992: : max @{ n1 n2 -- n3 @}
  993:  n1 n2 > if
  994:    n1
  995:  else
  996:    n2
  997:  endif ;
  998: @end example
  999: 
 1000: The similarity of locals definitions with stack comments is intended. A
 1001: locals definition often replaces the stack comment of a word. The order
 1002: of the locals corresponds to the order in a stack comment and everything
 1003: after the @code{--} is really a comment.
 1004: 
 1005: This similarity has one disadvantage: It is too easy to confuse locals
 1006: declarations with stack comments, causing bugs and making them hard to
 1007: find. However, this problem can be avoided by appropriate coding
 1008: conventions: Do not use both notations in the same program. If you do,
 1009: they should be distinguished using additional means, e.g. by position.
 1010: 
 1011: The name of the local may be preceded by a type specifier, e.g.,
 1012: @code{F:} for a floating point value:
 1013: 
 1014: @example
 1015: : CX* @{ F: Ar F: Ai F: Br F: Bi -- Cr Ci @}
 1016: \ complex multiplication
 1017:  Ar Br f* Ai Bi f* f-
 1018:  Ar Bi f* Ai Br f* f+ ;
 1019: @end example
 1020: 
 1021: GNU Forth currently supports cells (@code{W:}, @code{W^}), doubles
 1022: (@code{D:}, @code{D^}), floats (@code{F:}, @code{F^}) and characters
 1023: (@code{C:}, @code{C^}) in two flavours: a value-flavoured local (defined
 1024: with @code{W:}, @code{D:} etc.) produces its value and can be changed
 1025: with @code{TO}. A variable-flavoured local (defined with @code{W^} etc.)
 1026: produces its address (which becomes invalid when the variable's scope is
 1027: left). E.g., the standard word @code{emit} can be defined in therms of
 1028: @code{type} like this:
 1029: 
 1030: @example
 1031: : emit @{ C^ char* -- @}
 1032:     char* 1 type ;
 1033: @end example
 1034: 
 1035: A local without type specifier is a @code{W:} local. Both flavours of
 1036: locals are initialized with values from the data or FP stack.
 1037: 
 1038: Currently there is no way to define locals with user-defined data
 1039: structures, but we are working on it.
 1040: 
 1041: GNU Forth allows defining locals everywhere in a colon definition. This poses the following questions:
 1042: 
 1043: @menu
 1044: * Where are locals visible by name?::  
 1045: * How long do locals live? ::   
 1046: * Programming Style::           
 1047: * Implementation::              
 1048: @end menu
 1049: 
 1050: @node Where are locals visible by name?, How long do locals live?, gforth locals, gforth locals
 1051: @subsubsection Where are locals visible by name?
 1052: 
 1053: Basically, the answer is that locals are visible where you would expect
 1054: it in block-structured languages, and sometimes a little longer. If you
 1055: want to restrict the scope of a local, enclose its definition in
 1056: @code{SCOPE}...@code{ENDSCOPE}.
 1057: 
 1058: doc-scope
 1059: doc-endscope
 1060: 
 1061: These words behave like control structure words, so you can use them
 1062: with @code{CS-PICK} and @code{CS-ROLL} to restrict the scope in
 1063: arbitrary ways.
 1064: 
 1065: If you want a more exact answer to the visibility question, here's the
 1066: basic principle: A local is visible in all places that can only be
 1067: reached through the definition of the local@footnote{In compiler
 1068: construction terminology, all places dominated by the definition of the
 1069: local.}. In other words, it is not visible in places that can be reached
 1070: without going through the definition of the local. E.g., locals defined
 1071: in @code{IF}...@code{ENDIF} are visible until the @code{ENDIF}, locals
 1072: defined in @code{BEGIN}...@code{UNTIL} are visible after the
 1073: @code{UNTIL} (until, e.g., a subsequent @code{ENDSCOPE}).
 1074: 
 1075: The reasoning behind this solution is: We want to have the locals
 1076: visible as long as it is meaningful. The user can always make the
 1077: visibility shorter by using explicit scoping. In a place that can
 1078: only be reached through the definition of a local, the meaning of a
 1079: local name is clear. In other places it is not: How is the local
 1080: initialized at the control flow path that does not contain the
 1081: definition? Which local is meant, if the same name is defined twice in
 1082: two independent control flow paths?
 1083: 
 1084: This should be enough detail for nearly all users, so you can skip the
 1085: rest of this section. If you relly must know all the gory details and
 1086: options, read on.
 1087: 
 1088: In order to implement this rule, the compiler has to know which places
 1089: are unreachable. It knows this automatically after @code{AHEAD},
 1090: @code{AGAIN}, @code{EXIT} and @code{LEAVE}; in other cases (e.g., after
 1091: most @code{THROW}s), you can use the word @code{UNREACHABLE} to tell the
 1092: compiler that the control flow never reaches that place. If
 1093: @code{UNREACHABLE} is not used where it could, the only consequence is
 1094: that the visibility of some locals is more limited than the rule above
 1095: says. If @code{UNREACHABLE} is used where it should not (i.e., if you
 1096: lie to the compiler), buggy code will be produced.
 1097: 
 1098: Another problem with this rule is that at @code{BEGIN}, the compiler
 1099: does not know which locals will be visible on the incoming
 1100: back-edge. All problems discussed in the following are due to this
 1101: ignorance of the compiler (we discuss the problems using @code{BEGIN}
 1102: loops as examples; the discussion also applies to @code{?DO} and other
 1103: loops). Perhaps the most insidious example is:
 1104: @example
 1105: AHEAD
 1106: BEGIN
 1107:   x
 1108: [ 1 CS-ROLL ] THEN
 1109:   @{ x @}
 1110:   ...
 1111: UNTIL
 1112: @end example
 1113: 
 1114: This should be legal according to the visibility rule. The use of
 1115: @code{x} can only be reached through the definition; but that appears
 1116: textually below the use.
 1117: 
 1118: From this example it is clear that the visibility rules cannot be fully
 1119: implemented without major headaches. Our implementation treats common
 1120: cases as advertised and the exceptions are treated in a safe way: The
 1121: compiler makes a reasonable guess about the locals visible after a
 1122: @code{BEGIN}; if it is too pessimistic, the
 1123: user will get a spurious error about the local not being defined; if the
 1124: compiler is too optimistic, it will notice this later and issue a
 1125: warning. In the case above the compiler would complain about @code{x}
 1126: being undefined at its use. You can see from the obscure examples in
 1127: this section that it takes quite unusual control structures to get the
 1128: compiler into trouble, and even then it will often do fine.
 1129: 
 1130: If the @code{BEGIN} is reachable from above, the most optimistic guess
 1131: is that all locals visible before the @code{BEGIN} will also be
 1132: visible after the @code{BEGIN}. This guess is valid for all loops that
 1133: are entered only through the @code{BEGIN}, in particular, for normal
 1134: @code{BEGIN}...@code{WHILE}...@code{REPEAT} and
 1135: @code{BEGIN}...@code{UNTIL} loops and it is implemented in our
 1136: compiler. When the branch to the @code{BEGIN} is finally generated by
 1137: @code{AGAIN} or @code{UNTIL}, the compiler checks the guess and
 1138: warns the user if it was too optimisitic:
 1139: @example
 1140: IF
 1141:   @{ x @}
 1142: BEGIN
 1143:   \ x ? 
 1144: [ 1 cs-roll ] THEN
 1145:   ...
 1146: UNTIL
 1147: @end example
 1148: 
 1149: Here, @code{x} lives only until the @code{BEGIN}, but the compiler
 1150: optimistically assumes that it lives until the @code{THEN}. It notices
 1151: this difference when it compiles the @code{UNTIL} and issues a
 1152: warning. The user can avoid the warning, and make sure that @code{x}
 1153: is not used in the wrong area by using explicit scoping:
 1154: @example
 1155: IF
 1156:   SCOPE
 1157:   @{ x @}
 1158:   ENDSCOPE
 1159: BEGIN
 1160: [ 1 cs-roll ] THEN
 1161:   ...
 1162: UNTIL
 1163: @end example
 1164: 
 1165: Since the guess is optimistic, there will be no spurious error messages
 1166: about undefined locals.
 1167: 
 1168: If the @code{BEGIN} is not reachable from above (e.g., after
 1169: @code{AHEAD} or @code{EXIT}), the compiler cannot even make an
 1170: optimistic guess, as the locals visible after the @code{BEGIN} may be
 1171: defined later. Therefore, the compiler assumes that no locals are
 1172: visible after the @code{BEGIN}. However, the useer can use
 1173: @code{ASSUME-LIVE} to make the compiler assume that the same locals are
 1174: visible at the BEGIN as at the point where the item was created.
 1175: 
 1176: doc-assume-live
 1177: 
 1178: E.g.,
 1179: @example
 1180: @{ x @}
 1181: AHEAD
 1182: ASSUME-LIVE
 1183: BEGIN
 1184:   x
 1185: [ 1 CS-ROLL ] THEN
 1186:   ...
 1187: UNTIL
 1188: @end example
 1189: 
 1190: Other cases where the locals are defined before the @code{BEGIN} can be
 1191: handled by inserting an appropriate @code{CS-ROLL} before the
 1192: @code{ASSUME-LIVE} (and changing the control-flow stack manipulation
 1193: behind the @code{ASSUME-LIVE}).
 1194: 
 1195: Cases where locals are defined after the @code{BEGIN} (but should be
 1196: visible immediately after the @code{BEGIN}) can only be handled by
 1197: rearranging the loop. E.g., the ``most insidious'' example above can be
 1198: arranged into:
 1199: @example
 1200: BEGIN
 1201:   @{ x @}
 1202:   ... 0=
 1203: WHILE
 1204:   x
 1205: REPEAT
 1206: @end example
 1207: 
 1208: @node How long do locals live?, Programming Style, Where are locals visible by name?, gforth locals
 1209: @subsubsection How long do locals live?
 1210: 
 1211: The right answer for the lifetime question would be: A local lives at
 1212: least as long as it can be accessed. For a value-flavoured local this
 1213: means: until the end of its visibility. However, a variable-flavoured
 1214: local could be accessed through its address far beyond its visibility
 1215: scope. Ultimately, this would mean that such locals would have to be
 1216: garbage collected. Since this entails un-Forth-like implementation
 1217: complexities, I adopted the same cowardly solution as some other
 1218: languages (e.g., C): The local lives only as long as it is visible;
 1219: afterwards its address is invalid (and programs that access it
 1220: afterwards are erroneous).
 1221: 
 1222: @node Programming Style, Implementation, How long do locals live?, gforth locals
 1223: @subsubsection Programming Style
 1224: 
 1225: The freedom to define locals anywhere has the potential to change
 1226: programming styles dramatically. In particular, the need to use the
 1227: return stack for intermediate storage vanishes. Moreover, all stack
 1228: manipulations (except @code{PICK}s and @code{ROLL}s with run-time
 1229: determined arguments) can be eliminated: If the stack items are in the
 1230: wrong order, just write a locals definition for all of them; then
 1231: write the items in the order you want.
 1232: 
 1233: This seems a little far-fetched and eliminating stack manipulations is
 1234: unlikely to become a conscious programming objective. Still, the number
 1235: of stack manipulations will be reduced dramatically if local variables
 1236: are used liberally (e.g., compare @code{max} in @ref{gforth locals} with
 1237: a traditional implementation of @code{max}).
 1238: 
 1239: This shows one potential benefit of locals: making Forth programs more
 1240: readable. Of course, this benefit will only be realized if the
 1241: programmers continue to honour the principle of factoring instead of
 1242: using the added latitude to make the words longer.
 1243: 
 1244: Using @code{TO} can and should be avoided.  Without @code{TO},
 1245: every value-flavoured local has only a single assignment and many
 1246: advantages of functional languages apply to Forth. I.e., programs are
 1247: easier to analyse, to optimize and to read: It is clear from the
 1248: definition what the local stands for, it does not turn into something
 1249: different later.
 1250: 
 1251: E.g., a definition using @code{TO} might look like this:
 1252: @example
 1253: : strcmp @{ addr1 u1 addr2 u2 -- n @}
 1254:  u1 u2 min 0
 1255:  ?do
 1256:    addr1 c@ addr2 c@ - ?dup
 1257:    if
 1258:      unloop exit
 1259:    then
 1260:    addr1 char+ TO addr1
 1261:    addr2 char+ TO addr2
 1262:  loop
 1263:  u1 u2 - ;
 1264: @end example
 1265: Here, @code{TO} is used to update @code{addr1} and @code{addr2} at
 1266: every loop iteration. @code{strcmp} is a typical example of the
 1267: readability problems of using @code{TO}. When you start reading
 1268: @code{strcmp}, you think that @code{addr1} refers to the start of the
 1269: string. Only near the end of the loop you realize that it is something
 1270: else.
 1271: 
 1272: This can be avoided by defining two locals at the start of the loop that
 1273: are initialized with the right value for the current iteration.
 1274: @example
 1275: : strcmp @{ addr1 u1 addr2 u2 -- n @}
 1276:  addr1 addr2
 1277:  u1 u2 min 0 
 1278:  ?do @{ s1 s2 @}
 1279:    s1 c@ s2 c@ - ?dup 
 1280:    if
 1281:      unloop exit
 1282:    then
 1283:    s1 char+ s2 char+
 1284:  loop
 1285:  2drop
 1286:  u1 u2 - ;
 1287: @end example
 1288: Here it is clear from the start that @code{s1} has a different value
 1289: in every loop iteration.
 1290: 
 1291: @node Implementation,  , Programming Style, gforth locals
 1292: @subsubsection Implementation
 1293: 
 1294: GNU Forth uses an extra locals stack. The most compelling reason for
 1295: this is that the return stack is not float-aligned; using an extra stack
 1296: also eliminates the problems and restrictions of using the return stack
 1297: as locals stack. Like the other stacks, the locals stack grows toward
 1298: lower addresses. A few primitives allow an efficient implementation:
 1299: 
 1300: doc-@local#
 1301: doc-f@local#
 1302: doc-laddr#
 1303: doc-lp+!#
 1304: doc-lp!
 1305: doc->l
 1306: doc-f>l
 1307: 
 1308: In addition to these primitives, some specializations of these
 1309: primitives for commonly occurring inline arguments are provided for
 1310: efficiency reasons, e.g., @code{@@local0} as specialization of
 1311: @code{@@local#} for the inline argument 0. The following compiling words
 1312: compile the right specialized version, or the general version, as
 1313: appropriate:
 1314: 
 1315: doc-compile-@@local
 1316: doc-compile-f@@local
 1317: doc-compile-lp+!
 1318: 
 1319: Combinations of conditional branches and @code{lp+!#} like
 1320: @code{?branch-lp+!#} (the locals pointer is only changed if the branch
 1321: is taken) are provided for efficiency and correctness in loops.
 1322: 
 1323: A special area in the dictionary space is reserved for keeping the
 1324: local variable names. @code{@{} switches the dictionary pointer to this
 1325: area and @code{@}} switches it back and generates the locals
 1326: initializing code. @code{W:} etc.@ are normal defining words. This
 1327: special area is cleared at the start of every colon definition.
 1328: 
 1329: A special feature of GNU Forths dictionary is used to implement the
 1330: definition of locals without type specifiers: every wordlist (aka
 1331: vocabulary) has its own methods for searching
 1332: etc. (@pxref{Wordlists}). For the present purpose we defined a wordlist
 1333: with a special search method: When it is searched for a word, it
 1334: actually creates that word using @code{W:}. @code{@{} changes the search
 1335: order to first search the wordlist containing @code{@}}, @code{W:} etc.,
 1336: and then the wordlist for defining locals without type specifiers.
 1337: 
 1338: The lifetime rules support a stack discipline within a colon
 1339: definition: The lifetime of a local is either nested with other locals
 1340: lifetimes or it does not overlap them.
 1341: 
 1342: At @code{BEGIN}, @code{IF}, and @code{AHEAD} no code for locals stack
 1343: pointer manipulation is generated. Between control structure words
 1344: locals definitions can push locals onto the locals stack. @code{AGAIN}
 1345: is the simplest of the other three control flow words. It has to
 1346: restore the locals stack depth of the corresponding @code{BEGIN}
 1347: before branching. The code looks like this:
 1348: @format
 1349: @code{lp+!#} current-locals-size @minus{} dest-locals-size
 1350: @code{branch} <begin>
 1351: @end format
 1352: 
 1353: @code{UNTIL} is a little more complicated: If it branches back, it
 1354: must adjust the stack just like @code{AGAIN}. But if it falls through,
 1355: the locals stack must not be changed. The compiler generates the
 1356: following code:
 1357: @format
 1358: @code{?branch-lp+!#} <begin> current-locals-size @minus{} dest-locals-size
 1359: @end format
 1360: The locals stack pointer is only adjusted if the branch is taken.
 1361: 
 1362: @code{THEN} can produce somewhat inefficient code:
 1363: @format
 1364: @code{lp+!#} current-locals-size @minus{} orig-locals-size
 1365: <orig target>:
 1366: @code{lp+!#} orig-locals-size @minus{} new-locals-size
 1367: @end format
 1368: The second @code{lp+!#} adjusts the locals stack pointer from the
 1369: level at the @var{orig} point to the level after the @code{THEN}. The
 1370: first @code{lp+!#} adjusts the locals stack pointer from the current
 1371: level to the level at the orig point, so the complete effect is an
 1372: adjustment from the current level to the right level after the
 1373: @code{THEN}.
 1374: 
 1375: In a conventional Forth implementation a dest control-flow stack entry
 1376: is just the target address and an orig entry is just the address to be
 1377: patched. Our locals implementation adds a wordlist to every orig or dest
 1378: item. It is the list of locals visible (or assumed visible) at the point
 1379: described by the entry. Our implementation also adds a tag to identify
 1380: the kind of entry, in particular to differentiate between live and dead
 1381: (reachable and unreachable) orig entries.
 1382: 
 1383: A few unusual operations have to be performed on locals wordlists:
 1384: 
 1385: doc-common-list
 1386: doc-sub-list?
 1387: doc-list-size
 1388: 
 1389: Several features of our locals wordlist implementation make these
 1390: operations easy to implement: The locals wordlists are organised as
 1391: linked lists; the tails of these lists are shared, if the lists
 1392: contain some of the same locals; and the address of a name is greater
 1393: than the address of the names behind it in the list.
 1394: 
 1395: Another important implementation detail is the variable
 1396: @code{dead-code}. It is used by @code{BEGIN} and @code{THEN} to
 1397: determine if they can be reached directly or only through the branch
 1398: that they resolve. @code{dead-code} is set by @code{UNREACHABLE},
 1399: @code{AHEAD}, @code{EXIT} etc., and cleared at the start of a colon
 1400: definition, by @code{BEGIN} and usually by @code{THEN}.
 1401: 
 1402: Counted loops are similar to other loops in most respects, but
 1403: @code{LEAVE} requires special attention: It performs basically the same
 1404: service as @code{AHEAD}, but it does not create a control-flow stack
 1405: entry. Therefore the information has to be stored elsewhere;
 1406: traditionally, the information was stored in the target fields of the
 1407: branches created by the @code{LEAVE}s, by organizing these fields into a
 1408: linked list. Unfortunately, this clever trick does not provide enough
 1409: space for storing our extended control flow information. Therefore, we
 1410: introduce another stack, the leave stack. It contains the control-flow
 1411: stack entries for all unresolved @code{LEAVE}s.
 1412: 
 1413: Local names are kept until the end of the colon definition, even if
 1414: they are no longer visible in any control-flow path. In a few cases
 1415: this may lead to increased space needs for the locals name area, but
 1416: usually less than reclaiming this space would cost in code size.
 1417: 
 1418: 
 1419: @node ANS Forth locals,  , gforth locals, Locals
 1420: @subsection ANS Forth locals
 1421: 
 1422: The ANS Forth locals wordset does not define a syntax for locals, but
 1423: words that make it possible to define various syntaxes. One of the
 1424: possible syntaxes is a subset of the syntax we used in the gforth locals
 1425: wordset, i.e.:
 1426: 
 1427: @example
 1428: @{ local1 local2 ... -- comment @}
 1429: @end example
 1430: or
 1431: @example
 1432: @{ local1 local2 ... @}
 1433: @end example
 1434: 
 1435: The order of the locals corresponds to the order in a stack comment. The
 1436: restrictions are:
 1437: 
 1438: @itemize @bullet
 1439: @item
 1440: Locals can only be cell-sized values (no type specifers are allowed).
 1441: @item
 1442: Locals can be defined only outside control structures.
 1443: @item
 1444: Locals can interfere with explicit usage of the return stack. For the
 1445: exact (and long) rules, see the standard. If you don't use return stack
 1446: accessing words in a definition using locals, you will we all right. The
 1447: purpose of this rule is to make locals implementation on the return
 1448: stack easier.
 1449: @item
 1450: The whole definition must be in one line.
 1451: @end itemize
 1452: 
 1453: Locals defined in this way behave like @code{VALUE}s
 1454: (@xref{Values}). I.e., they are initialized from the stack. Using their
 1455: name produces their value. Their value can be changed using @code{TO}.
 1456: 
 1457: Since this syntax is supported by gforth directly, you need not do
 1458: anything to use it. If you want to port a program using this syntax to
 1459: another ANS Forth system, use @file{anslocal.fs} to implement the syntax
 1460: on the other system.
 1461: 
 1462: Note that a syntax shown in the standard, section A.13 looks
 1463: similar, but is quite different in having the order of locals
 1464: reversed. Beware!
 1465: 
 1466: The ANS Forth locals wordset itself consists of the following word
 1467: 
 1468: doc-(local)
 1469: 
 1470: The ANS Forth locals extension wordset defines a syntax, but it is so
 1471: awful that we strongly recommend not to use it. We have implemented this
 1472: syntax to make porting to gforth easy, but do not document it here. The
 1473: problem with this syntax is that the locals are defined in an order
 1474: reversed with respect to the standard stack comment notation, making
 1475: programs harder to read, and easier to misread and miswrite. The only
 1476: merit of this syntax is that it is easy to implement using the ANS Forth
 1477: locals wordset.
 1478: 
 1479: @node Defining Words, Wordlists, Locals, Words
 1480: @section Defining Words
 1481: 
 1482: @node Values,  , Defining Words, Defining Words
 1483: @subsection Values
 1484: 
 1485: @node Wordlists, Files, Defining Words, Words
 1486: @section Wordlists
 1487: 
 1488: @node Files, Blocks, Wordlists, Words
 1489: @section Files
 1490: 
 1491: @node Blocks, Other I/O, Files, Words
 1492: @section Blocks
 1493: 
 1494: @node Other I/O, Programming Tools, Blocks, Words
 1495: @section Other I/O
 1496: 
 1497: @node Programming Tools, Threading Words, Other I/O, Words
 1498: @section Programming Tools
 1499: 
 1500: @menu
 1501: * Debugging::                   Simple and quick.
 1502: * Assertions::                  Making your programs self-checking.
 1503: @end menu
 1504: 
 1505: @node Debugging, Assertions, Programming Tools, Programming Tools
 1506: @subsection Debugging
 1507: 
 1508: The simple debugging aids provided in @file{debugging.fs}
 1509: are meant to support a different style of debugging than the
 1510: tracing/stepping debuggers used in languages with long turn-around
 1511: times.
 1512: 
 1513: A much better (faster) way in fast-compilig languages is to add
 1514: printing code at well-selected places, let the program run, look at
 1515: the output, see where things went wrong, add more printing code, etc.,
 1516: until the bug is found.
 1517: 
 1518: The word @code{~~} is easy to insert. It just prints debugging
 1519: information (by default the source location and the stack contents). It
 1520: is also easy to remove (@kbd{C-x ~} in the Emacs Forth mode to
 1521: query-replace them with nothing). The deferred words
 1522: @code{printdebugdata} and @code{printdebugline} control the output of
 1523: @code{~~}. The default source location output format works well with
 1524: Emacs' compilation mode, so you can step through the program at the
 1525: source level using @kbd{C-x `} (the advantage over a stepping debugger
 1526: is that you can step in any direction and you know where the crash has
 1527: happened or where the strange data has occurred).
 1528: 
 1529: Note that the default actions clobber the contents of the pictured
 1530: numeric output string, so you should not use @code{~~}, e.g., between
 1531: @code{<#} and @code{#>}.
 1532: 
 1533: doc-~~
 1534: doc-printdebugdata
 1535: doc-printdebugline
 1536: 
 1537: @node Assertions,  , Debugging, Programming Tools
 1538: @subsection Assertions
 1539: 
 1540: It is a good idea to make your programs self-checking, in particular, if
 1541: you use an assumption (e.g., that a certain field of a data structure is
 1542: never zero) that may become wrong during maintenance. GForth supports
 1543: assertions for this purpose. They are used like this:
 1544: 
 1545: @example
 1546: assert( @var{flag} )
 1547: @end example
 1548: 
 1549: The code between @code{assert(} and @code{)} should compute a flag, that
 1550: should be true if everything is alright and false otherwise. It should
 1551: not change anything else on the stack. The overall stack effect of the
 1552: assertion is @code{( -- )}. E.g.
 1553: 
 1554: @example
 1555: assert( 1 1 + 2 = ) \ what we learn in school
 1556: assert( dup 0<> ) \ assert that the top of stack is not zero
 1557: assert( false ) \ this code should not be reached
 1558: @end example
 1559: 
 1560: The need for assertions is different at different times. During
 1561: debugging, we want more checking, in production we sometimes care more
 1562: for speed. Therefore, assertions can be turned off, i.e., the assertion
 1563: becomes a comment. Depending on the importance of an assertion and the
 1564: time it takes to check it, you may want to turn off some assertions and
 1565: keep others turned on. GForth provides several levels of assertions for
 1566: this purpose:
 1567: 
 1568: doc-assert0(
 1569: doc-assert1(
 1570: doc-assert2(
 1571: doc-assert3(
 1572: doc-assert(
 1573: doc-)
 1574: 
 1575: @code{Assert(} is the same as @code{assert1(}. The variable
 1576: @code{assert-level} specifies the highest assertions that are turned
 1577: on. I.e., at the default @code{assert-level} of one, @code{assert0(} and
 1578: @code{assert1(} assertions perform checking, while @code{assert2(} and
 1579: @code{assert3(} assertions are treated as comments.
 1580: 
 1581: Note that the @code{assert-level} is evaluated at compile-time, not at
 1582: run-time. I.e., you cannot turn assertions on or off at run-time, you
 1583: have to set the @code{assert-level} appropriately before compiling a
 1584: piece of code. You can compile several pieces of code at several
 1585: @code{assert-level}s (e.g., a trusted library at level 1 and newly
 1586: written code at level 3).
 1587: 
 1588: doc-assert-level
 1589: 
 1590: If an assertion fails, a message compatible with Emacs' compilation mode
 1591: is produced and the execution is aborted (currently with @code{ABORT"}.
 1592: If there is interest, we will introduce a special throw code. But if you
 1593: intend to @code{catch} a specific condition, using @code{throw} is
 1594: probably more appropriate than an assertion).
 1595: 
 1596: @node Threading Words,  , Programming Tools, Words
 1597: @section Threading Words
 1598: 
 1599: These words provide access to code addresses and other threading stuff
 1600: in gforth (and, possibly, other interpretive Forths). It more or less
 1601: abstracts away the differences between direct and indirect threading
 1602: (and, for direct threading, the machine dependences). However, at
 1603: present this wordset is still inclomplete. It is also pretty low-level;
 1604: some day it will hopefully be made unnecessary by an internals words set
 1605: that abstracts implementation details away completely.
 1606: 
 1607: doc->code-address
 1608: doc->does-code
 1609: doc-code-address!
 1610: doc-does-code!
 1611: doc-does-handler!
 1612: doc-/does-handler
 1613: 
 1614: @node ANS conformance, Model, Words, Top
 1615: @chapter ANS conformance
 1616: 
 1617: @node Model, Emacs and GForth, ANS conformance, Top
 1618: @chapter Model
 1619: 
 1620: @node Emacs and GForth, Internals, Model, Top
 1621: @chapter Emacs and GForth
 1622: 
 1623: GForth comes with @file{gforth.el}, an improved version of
 1624: @file{forth.el} by Goran Rydqvist (icluded in the TILE package). The
 1625: improvements are a better (but still not perfect) handling of
 1626: indentation. I have also added comment paragraph filling (@kbd{M-q}),
 1627: commenting (@kbd{C-x \}) and uncommenting (@kbd{C-x |}) regions and
 1628: removing debugging tracers (@kbd{C-x ~}). I left the stuff I do not use
 1629: alone, even though some of it only makes sense for TILE. To get a
 1630: description of these features, enter Forth mode and type @kbd{C-h m}.
 1631: 
 1632: In addition, GForth supports Emacs quite well: The source code locations
 1633: given in error messages, debugging output (from @code{~~}) and failed
 1634: assertion messages are in the right format for Emacs' compilation mode
 1635: (@pxref{Compilation, , Running Compilations under Emacs, emacs, Emacs
 1636: Manual}) so the source location corresponding to an error or other
 1637: message is only a few keystrokes away (@kbd{C-x `} for the next error,
 1638: @kbd{C-c C-c} for the error under the cursor).
 1639: 
 1640: Also, if you @code{include} @file{etags.fs}, a new @file{TAGS} file
 1641: (@pxref{Tags, , Tags Tables, emacs, Emacs Manual}) will be produced that
 1642: contains the definitions of all words defined afterwards. You can then
 1643: find the source for a word using @kbd{M-.}. Note that emacs can use
 1644: several tags files at the same time (e.g., one for the gforth sources
 1645: and one for your program).
 1646: 
 1647: To get all these benefits, add the following lines to your @file{.emacs}
 1648: file:
 1649: 
 1650: @example
 1651: (autoload 'forth-mode "gforth.el")
 1652: (setq auto-mode-alist (cons '("\\.fs\\'" . forth-mode) auto-mode-alist))
 1653: @end example
 1654: 
 1655: @node Internals, Bugs, Emacs and GForth, Top
 1656: @chapter Internals
 1657: 
 1658: Reading this section is not necessary for programming with gforth. It
 1659: should be helpful for finding your way in the gforth sources.
 1660: 
 1661: @menu
 1662: * Portability::                 
 1663: * Threading::                   
 1664: * Primitives::                  
 1665: * System Architecture::         
 1666: @end menu
 1667: 
 1668: @node Portability, Threading, Internals, Internals
 1669: @section Portability
 1670: 
 1671: One of the main goals of the effort is availability across a wide range
 1672: of personal machines. fig-Forth, and, to a lesser extent, F83, achieved
 1673: this goal by manually coding the engine in assembly language for several
 1674: then-popular processors. This approach is very labor-intensive and the
 1675: results are short-lived due to progress in computer architecture.
 1676: 
 1677: Others have avoided this problem by coding in C, e.g., Mitch Bradley
 1678: (cforth), Mikael Patel (TILE) and Dirk Zoller (pfe). This approach is
 1679: particularly popular for UNIX-based Forths due to the large variety of
 1680: architectures of UNIX machines. Unfortunately an implementation in C
 1681: does not mix well with the goals of efficiency and with using
 1682: traditional techniques: Indirect or direct threading cannot be expressed
 1683: in C, and switch threading, the fastest technique available in C, is
 1684: significantly slower. Another problem with C is that it's very
 1685: cumbersome to express double integer arithmetic.
 1686: 
 1687: Fortunately, there is a portable language that does not have these
 1688: limitations: GNU C, the version of C processed by the GNU C compiler
 1689: (@pxref{C Extensions, , Extensions to the C Language Family, gcc.info,
 1690: GNU C Manual}). Its labels as values feature (@pxref{Labels as Values, ,
 1691: Labels as Values, gcc.info, GNU C Manual}) makes direct and indirect
 1692: threading possible, its @code{long long} type (@pxref{Long Long, ,
 1693: Double-Word Integers, gcc.info, GNU C Manual}) corresponds to Forths
 1694: double numbers. GNU C is available for free on all important (and many
 1695: unimportant) UNIX machines, VMS, 80386s running MS-DOS, the Amiga, and
 1696: the Atari ST, so a Forth written in GNU C can run on all these
 1697: machines@footnote{Due to Apple's look-and-feel lawsuit it is not
 1698: available on the Mac (@pxref{Boycott, , Protect Your Freedom---Fight
 1699: ``Look And Feel'', gcc.info, GNU C Manual}).}.
 1700: 
 1701: Writing in a portable language has the reputation of producing code that
 1702: is slower than assembly. For our Forth engine we repeatedly looked at
 1703: the code produced by the compiler and eliminated most compiler-induced
 1704: inefficiencies by appropriate changes in the source-code.
 1705: 
 1706: However, register allocation cannot be portably influenced by the
 1707: programmer, leading to some inefficiencies on register-starved
 1708: machines. We use explicit register declarations (@pxref{Explicit Reg
 1709: Vars, , Variables in Specified Registers, gcc.info, GNU C Manual}) to
 1710: improve the speed on some machines. They are turned on by using the
 1711: @code{gcc} switch @code{-DFORCE_REG}. Unfortunately, this feature not
 1712: only depends on the machine, but also on the compiler version: On some
 1713: machines some compiler versions produce incorrect code when certain
 1714: explicit register declarations are used. So by default
 1715: @code{-DFORCE_REG} is not used.
 1716: 
 1717: @node Threading, Primitives, Portability, Internals
 1718: @section Threading
 1719: 
 1720: GNU C's labels as values extension (available since @code{gcc-2.0},
 1721: @pxref{Labels as Values, , Labels as Values, gcc.info, GNU C Manual})
 1722: makes it possible to take the address of @var{label} by writing
 1723: @code{&&@var{label}}.  This address can then be used in a statement like
 1724: @code{goto *@var{address}}. I.e., @code{goto *&&x} is the same as
 1725: @code{goto x}.
 1726: 
 1727: With this feature an indirect threaded NEXT looks like:
 1728: @example
 1729: cfa = *ip++;
 1730: ca = *cfa;
 1731: goto *ca;
 1732: @end example
 1733: For those unfamiliar with the names: @code{ip} is the Forth instruction
 1734: pointer; the @code{cfa} (code-field address) corresponds to ANS Forths
 1735: execution token and points to the code field of the next word to be
 1736: executed; The @code{ca} (code address) fetched from there points to some
 1737: executable code, e.g., a primitive or the colon definition handler
 1738: @code{docol}.
 1739: 
 1740: Direct threading is even simpler:
 1741: @example
 1742: ca = *ip++;
 1743: goto *ca;
 1744: @end example
 1745: 
 1746: Of course we have packaged the whole thing neatly in macros called
 1747: @code{NEXT} and @code{NEXT1} (the part of NEXT after fetching the cfa).
 1748: 
 1749: @menu
 1750: * Scheduling::                  
 1751: * Direct or Indirect Threaded?::  
 1752: * DOES>::                       
 1753: @end menu
 1754: 
 1755: @node Scheduling, Direct or Indirect Threaded?, Threading, Threading
 1756: @subsection Scheduling
 1757: 
 1758: There is a little complication: Pipelined and superscalar processors,
 1759: i.e., RISC and some modern CISC machines can process independent
 1760: instructions while waiting for the results of an instruction. The
 1761: compiler usually reorders (schedules) the instructions in a way that
 1762: achieves good usage of these delay slots. However, on our first tries
 1763: the compiler did not do well on scheduling primitives. E.g., for
 1764: @code{+} implemented as
 1765: @example
 1766: n=sp[0]+sp[1];
 1767: sp++;
 1768: sp[0]=n;
 1769: NEXT;
 1770: @end example
 1771: the NEXT comes strictly after the other code, i.e., there is nearly no
 1772: scheduling. After a little thought the problem becomes clear: The
 1773: compiler cannot know that sp and ip point to different addresses (and
 1774: the version of @code{gcc} we used would not know it even if it was
 1775: possible), so it could not move the load of the cfa above the store to
 1776: the TOS. Indeed the pointers could be the same, if code on or very near
 1777: the top of stack were executed. In the interest of speed we chose to
 1778: forbid this probably unused ``feature'' and helped the compiler in
 1779: scheduling: NEXT is divided into the loading part (@code{NEXT_P1}) and
 1780: the goto part (@code{NEXT_P2}). @code{+} now looks like:
 1781: @example
 1782: n=sp[0]+sp[1];
 1783: sp++;
 1784: NEXT_P1;
 1785: sp[0]=n;
 1786: NEXT_P2;
 1787: @end example
 1788: This can be scheduled optimally by the compiler.
 1789: 
 1790: This division can be turned off with the switch @code{-DCISC_NEXT}. This
 1791: switch is on by default on machines that do not profit from scheduling
 1792: (e.g., the 80386), in order to preserve registers.
 1793: 
 1794: @node Direct or Indirect Threaded?, DOES>, Scheduling, Threading
 1795: @subsection Direct or Indirect Threaded?
 1796: 
 1797: Both! After packaging the nasty details in macro definitions we
 1798: realized that we could switch between direct and indirect threading by
 1799: simply setting a compilation flag (@code{-DDIRECT_THREADED}) and
 1800: defining a few machine-specific macros for the direct-threading case.
 1801: On the Forth level we also offer access words that hide the
 1802: differences between the threading methods (@pxref{Threading Words}).
 1803: 
 1804: Indirect threading is implemented completely
 1805: machine-independently. Direct threading needs routines for creating
 1806: jumps to the executable code (e.g. to docol or dodoes). These routines
 1807: are inherently machine-dependent, but they do not amount to many source
 1808: lines. I.e., even porting direct threading to a new machine is a small
 1809: effort.
 1810: 
 1811: @node DOES>,  , Direct or Indirect Threaded?, Threading
 1812: @subsection DOES>
 1813: One of the most complex parts of a Forth engine is @code{dodoes}, i.e.,
 1814: the chunk of code executed by every word defined by a
 1815: @code{CREATE}...@code{DOES>} pair. The main problem here is: How to find
 1816: the Forth code to be executed, i.e. the code after the @code{DOES>} (the
 1817: DOES-code)? There are two solutions:
 1818: 
 1819: In fig-Forth the code field points directly to the dodoes and the
 1820: DOES-code address is stored in the cell after the code address
 1821: (i.e. at cfa cell+). It may seem that this solution is illegal in the
 1822: Forth-79 and all later standards, because in fig-Forth this address
 1823: lies in the body (which is illegal in these standards). However, by
 1824: making the code field larger for all words this solution becomes legal
 1825: again. We use this approach for the indirect threaded version. Leaving
 1826: a cell unused in most words is a bit wasteful, but on the machines we
 1827: are targetting this is hardly a problem. The other reason for having a
 1828: code field size of two cells is to avoid having different image files
 1829: for direct and indirect threaded systems (@pxref{System Architecture}).
 1830: 
 1831: The other approach is that the code field points or jumps to the cell
 1832: after @code{DOES}. In this variant there is a jump to @code{dodoes} at
 1833: this address. @code{dodoes} can then get the DOES-code address by
 1834: computing the code address, i.e., the address of the jump to dodoes,
 1835: and add the length of that jump field. A variant of this is to have a
 1836: call to @code{dodoes} after the @code{DOES>}; then the return address
 1837: (which can be found in the return register on RISCs) is the DOES-code
 1838: address. Since the two cells available in the code field are usually
 1839: used up by the jump to the code address in direct threading, we use
 1840: this approach for direct threading. We did not want to add another
 1841: cell to the code field.
 1842: 
 1843: @node Primitives, System Architecture, Threading, Internals
 1844: @section Primitives
 1845: 
 1846: @menu
 1847: * Automatic Generation::        
 1848: * TOS Optimization::            
 1849: * Produced code::               
 1850: @end menu
 1851: 
 1852: @node Automatic Generation, TOS Optimization, Primitives, Primitives
 1853: @subsection Automatic Generation
 1854: 
 1855: Since the primitives are implemented in a portable language, there is no
 1856: longer any need to minimize the number of primitives. On the contrary,
 1857: having many primitives is an advantage: speed. In order to reduce the
 1858: number of errors in primitives and to make programming them easier, we
 1859: provide a tool, the primitive generator (@file{prims2x.fs}), that
 1860: automatically generates most (and sometimes all) of the C code for a
 1861: primitive from the stack effect notation.  The source for a primitive
 1862: has the following form:
 1863: 
 1864: @format
 1865: @var{Forth-name}	@var{stack-effect}	@var{category}	[@var{pronounc.}]
 1866: [@code{""}@var{glossary entry}@code{""}]
 1867: @var{C code}
 1868: [@code{:}
 1869: @var{Forth code}]
 1870: @end format
 1871: 
 1872: The items in brackets are optional. The category and glossary fields
 1873: are there for generating the documentation, the Forth code is there
 1874: for manual implementations on machines without GNU C. E.g., the source
 1875: for the primitive @code{+} is:
 1876: @example
 1877: +    n1 n2 -- n    core    plus
 1878: n = n1+n2;
 1879: @end example
 1880: 
 1881: This looks like a specification, but in fact @code{n = n1+n2} is C
 1882: code. Our primitive generation tool extracts a lot of information from
 1883: the stack effect notations@footnote{We use a one-stack notation, even
 1884: though we have separate data and floating-point stacks; The separate
 1885: notation can be generated easily from the unified notation.}: The number
 1886: of items popped from and pushed on the stack, their type, and by what
 1887: name they are referred to in the C code. It then generates a C code
 1888: prelude and postlude for each primitive. The final C code for @code{+}
 1889: looks like this:
 1890: 
 1891: @example
 1892: I_plus:	/* + ( n1 n2 -- n ) */  /* label, stack effect */
 1893: /*  */                          /* documentation */
 1894: @{
 1895: DEF_CA                          /* definition of variable ca (indirect threading) */
 1896: Cell n1;                        /* definitions of variables */
 1897: Cell n2;
 1898: Cell n;
 1899: n1 = (Cell) sp[1];              /* input */
 1900: n2 = (Cell) TOS;
 1901: sp += 1;                        /* stack adjustment */
 1902: NAME("+")                       /* debugging output (with -DDEBUG) */
 1903: @{
 1904: n = n1+n2;                      /* C code taken from the source */
 1905: @}
 1906: NEXT_P1;                        /* NEXT part 1 */
 1907: TOS = (Cell)n;                  /* output */
 1908: NEXT_P2;                        /* NEXT part 2 */
 1909: @}
 1910: @end example
 1911: 
 1912: This looks long and inefficient, but the GNU C compiler optimizes quite
 1913: well and produces optimal code for @code{+} on, e.g., the R3000 and the
 1914: HP RISC machines: Defining the @code{n}s does not produce any code, and
 1915: using them as intermediate storage also adds no cost.
 1916: 
 1917: There are also other optimizations, that are not illustrated by this
 1918: example: Assignments between simple variables are usually for free (copy
 1919: propagation). If one of the stack items is not used by the primitive
 1920: (e.g.  in @code{drop}), the compiler eliminates the load from the stack
 1921: (dead code elimination). On the other hand, there are some things that
 1922: the compiler does not do, therefore they are performed by
 1923: @file{prims2x.fs}: The compiler does not optimize code away that stores
 1924: a stack item to the place where it just came from (e.g., @code{over}).
 1925: 
 1926: While programming a primitive is usually easy, there are a few cases
 1927: where the programmer has to take the actions of the generator into
 1928: account, most notably @code{?dup}, but also words that do not (always)
 1929: fall through to NEXT.
 1930: 
 1931: @node TOS Optimization, Produced code, Automatic Generation, Primitives
 1932: @subsection TOS Optimization
 1933: 
 1934: An important optimization for stack machine emulators, e.g., Forth
 1935: engines, is keeping  one or more of the top stack items in
 1936: registers.  If a word has the stack effect @var{in1}...@var{inx} @code{--}
 1937: @var{out1}...@var{outy}, keeping the top @var{n} items in registers
 1938: @itemize
 1939: @item
 1940: is better than keeping @var{n-1} items, if @var{x>=n} and @var{y>=n},
 1941: due to fewer loads from and stores to the stack.
 1942: @item is slower than keeping @var{n-1} items, if @var{x<>y} and @var{x<n} and
 1943: @var{y<n}, due to additional moves between registers.
 1944: @end itemize
 1945: 
 1946: In particular, keeping one item in a register is never a disadvantage,
 1947: if there are enough registers. Keeping two items in registers is a
 1948: disadvantage for frequent words like @code{?branch}, constants,
 1949: variables, literals and @code{i}. Therefore our generator only produces
 1950: code that keeps zero or one items in registers. The generated C code
 1951: covers both cases; the selection between these alternatives is made at
 1952: C-compile time using the switch @code{-DUSE_TOS}. @code{TOS} in the C
 1953: code for @code{+} is just a simple variable name in the one-item case,
 1954: otherwise it is a macro that expands into @code{sp[0]}. Note that the
 1955: GNU C compiler tries to keep simple variables like @code{TOS} in
 1956: registers, and it usually succeeds, if there are enough registers.
 1957: 
 1958: The primitive generator performs the TOS optimization for the
 1959: floating-point stack, too (@code{-DUSE_FTOS}). For floating-point
 1960: operations the benefit of this optimization is even larger:
 1961: floating-point operations take quite long on most processors, but can be
 1962: performed in parallel with other operations as long as their results are
 1963: not used. If the FP-TOS is kept in a register, this works. If
 1964: it is kept on the stack, i.e., in memory, the store into memory has to
 1965: wait for the result of the floating-point operation, lengthening the
 1966: execution time of the primitive considerably.
 1967: 
 1968: The TOS optimization makes the automatic generation of primitives a
 1969: bit more complicated. Just replacing all occurrences of @code{sp[0]} by
 1970: @code{TOS} is not sufficient. There are some special cases to
 1971: consider:
 1972: @itemize
 1973: @item In the case of @code{dup ( w -- w w )} the generator must not
 1974: eliminate the store to the original location of the item on the stack,
 1975: if the TOS optimization is turned on.
 1976: @item Primitives with stack effects of the form @code{--}
 1977: @var{out1}...@var{outy} must store the TOS to the stack at the start.
 1978: Likewise, primitives with the stack effect @var{in1}...@var{inx} @code{--}
 1979: must load the TOS from the stack at the end. But for the null stack
 1980: effect @code{--} no stores or loads should be generated.
 1981: @end itemize
 1982: 
 1983: @node Produced code,  , TOS Optimization, Primitives
 1984: @subsection Produced code
 1985: 
 1986: To see what assembly code is produced for the primitives on your machine
 1987: with your compiler and your flag settings, type @code{make engine.s} and
 1988: look at the resulting file @file{engine.s}.
 1989: 
 1990: @node System Architecture,  , Primitives, Internals
 1991: @section System Architecture
 1992: 
 1993: Our Forth system consists not only of primitives, but also of
 1994: definitions written in Forth. Since the Forth compiler itself belongs
 1995: to those definitions, it is not possible to start the system with the
 1996: primitives and the Forth source alone. Therefore we provide the Forth
 1997: code as an image file in nearly executable form. At the start of the
 1998: system a C routine loads the image file into memory, sets up the
 1999: memory (stacks etc.) according to information in the image file, and
 2000: starts executing Forth code.
 2001: 
 2002: The image file format is a compromise between the goals of making it
 2003: easy to generate image files and making them portable. The easiest way
 2004: to generate an image file is to just generate a memory dump. However,
 2005: this kind of image file cannot be used on a different machine, or on
 2006: the next version of the engine on the same machine, it even might not
 2007: work with the same engine compiled by a different version of the C
 2008: compiler. We would like to have as few versions of the image file as
 2009: possible, because we do not want to distribute many versions of the
 2010: same image file, and to make it easy for the users to use their image
 2011: files on many machines. We currently need to create a different image
 2012: file for machines with different cell sizes and different byte order
 2013: (little- or big-endian)@footnote{We consider adding information to the
 2014: image file that enables the loader to change the byte order.}.
 2015: 
 2016: Forth code that is going to end up in a portable image file has to
 2017: comply to some restrictions: addresses have to be stored in memory with
 2018: special words (@code{A!}, @code{A,}, etc.) in order to make the code
 2019: relocatable. Cells, floats, etc., have to be stored at the natural
 2020: alignment boundaries@footnote{E.g., store floats (8 bytes) at an address
 2021: dividable by~8. This happens automatically in our system when you use
 2022: the ANS Forth alignment words.}, in order to avoid alignment faults on
 2023: machines with stricter alignment. The image file is produced by a
 2024: metacompiler (@file{cross.fs}).
 2025: 
 2026: So, unlike the image file of Mitch Bradleys @code{cforth}, our image
 2027: file is not directly executable, but has to undergo some manipulations
 2028: during loading. Address relocation is performed at image load-time, not
 2029: at run-time. The loader also has to replace tokens standing for
 2030: primitive calls with the appropriate code-field addresses (or code
 2031: addresses in the case of direct threading).
 2032: 
 2033: @node Bugs, Pedigree, Internals, Top
 2034: @chapter Bugs
 2035: 
 2036: @node Pedigree, Word Index, Bugs, Top
 2037: @chapter Pedigree
 2038: 
 2039: @node Word Index, Node Index, Pedigree, Top
 2040: @chapter Word Index
 2041: 
 2042: @node Node Index,  , Word Index, Top
 2043: @chapter Node Index
 2044: 
 2045: @contents
 2046: @bye
 2047: 

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>