Annotation of gforth/gforth.ds, revision 1.7

1.1       anton       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.)
1.4       anton       4: @setfilename gforth.info
1.1       anton       5: @settitle GNU Forth Manual
1.4       anton       6: @comment @setchapternewpage odd
1.1       anton       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:      
1.4       anton      18: @ignore
1.1       anton      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:      
1.4       anton      24: @end ignore
1.1       anton      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
1.4       anton      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
1.1       anton      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: 
1.4       anton     256: @node Words, ANS conformance, Invocation, Top
1.1       anton     257: @chapter Forth Words
                    258: 
                    259: @menu
1.4       anton     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::             
1.1       anton     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: 
1.4       anton     281: @format
1.1       anton     282: @var{word}     @var{Stack effect}   @var{wordset}   @var{pronunciation}
1.4       anton     283: @end format
1.1       anton     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: 
1.4       anton     318: The type of a stack item is specified by the character(s) the name
                    319: starts with:
1.1       anton     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: 
1.4       anton     356: @node Arithmetic, Stack Manipulation, Notation, Words
1.1       anton     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
1.4       anton     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
1.1       anton     375: 
1.4       anton     376: @node Single precision, Bitwise operations, Arithmetic, Arithmetic
1.1       anton     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: 
1.4       anton     389: @node Bitwise operations, Mixed precision, Single precision, Arithmetic
1.1       anton     390: @subsection Bitwise operations
                    391: doc-and
                    392: doc-or
                    393: doc-xor
                    394: doc-invert
                    395: doc-2*
                    396: doc-2/
                    397: 
1.4       anton     398: @node Mixed precision, Double precision, Bitwise operations, Arithmetic
1.1       anton     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: 
1.4       anton     410: @node Double precision, Floating Point, Mixed precision, Arithmetic
1.1       anton     411: @subsection Double precision
                    412: doc-d+
                    413: doc-d-
                    414: doc-dnegate
                    415: doc-dabs
                    416: doc-dmin
                    417: doc-dmax
                    418: 
1.4       anton     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
1.6       anton     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}.
1.4       anton     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
1.6       anton     452: doc-falog
1.4       anton     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
1.1       anton     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
1.4       anton     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
1.1       anton     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: 
1.4       anton     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
1.1       anton     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: 
1.4       anton     520: @node Floating point stack, Return stack, Data stack, Stack Manipulation
1.1       anton     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: 
1.4       anton     530: @node Return stack, Locals stack, Floating point stack, Stack Manipulation
1.1       anton     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: 
1.4       anton     541: @node Locals stack, Stack pointer manipulation, Return stack, Stack Manipulation
1.1       anton     542: @subsection Locals stack
                    543: 
1.4       anton     544: @node Stack pointer manipulation,  , Locals stack, Stack Manipulation
1.1       anton     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: 
1.4       anton     555: @node Memory access, Control Structures, Stack Manipulation, Words
1.1       anton     556: @section Memory access
                    557: 
1.4       anton     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
1.1       anton     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: 
1.4       anton     581: @node Address arithmetic, Memory block access, Stack-Memory transfers, Memory access
1.1       anton     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: 
1.4       anton     630: @node Memory block access,  , Address arithmetic, Memory access
1.1       anton     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: 
1.4       anton     644: @node Control Structures, Locals, Memory access, Words
1.1       anton     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: 
1.4       anton     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
1.1       anton     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: 
1.4       anton     680: You can use @code{THEN} instead of @code{ENDIF}. Indeed, @code{THEN} is
1.1       anton     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
1.4       anton     707:   @dots{}
1.1       anton     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: 
1.4       anton     716: @node Simple Loops, Counted Loops, Selection, Control Structures
1.1       anton     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: 
1.4       anton     748: @node Counted Loops, Arbitrary control structures, Simple Loops, Control Structures
1.1       anton     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: 
1.2       anton     790: @code{4 0 ?DO  i .  2 +LOOP}   prints @code{0 2}
1.1       anton     791: 
1.2       anton     792: @code{4 1 ?DO  i .  2 +LOOP}   prints @code{1 3}
1.1       anton     793: 
                    794: The behaviour of @code{@var{n} +LOOP} is peculiar when @var{n} is negative:
                    795: 
1.2       anton     796: @code{-1 0 ?DO  i .  -1 +LOOP}  prints @code{0 -1}
1.1       anton     797: 
1.2       anton     798: @code{ 0 0 ?DO  i .  -1 +LOOP}  prints nothing
1.1       anton     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: 
1.7     ! pazsan    804: @code{-2 0 ?DO  i .  -1 S+LOOP}  prints @code{0 -1}
1.1       anton     805: 
1.7     ! pazsan    806: @code{-1 0 ?DO  i .  -1 S+LOOP}  prints @code{0}
1.1       anton     807: 
1.7     ! pazsan    808: @code{ 0 0 ?DO  i .  -1 S+LOOP}  prints nothing
1.1       anton     809: 
1.2       anton     810: The loop is terminated when the border between @var{limit@minus{}sgn(n)} and
1.1       anton     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: 
1.4       anton     835: @node Arbitrary control structures, Calls and returns, Counted Loops, Control Structures
1.2       anton     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: 
1.3       anton     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
1.2       anton     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: 
1.3       anton     865: doc-else
                    866: doc-while
                    867: doc-repeat
1.2       anton     868: 
                    869: Counted loop words constitute a separate group of words:
                    870: 
1.3       anton     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
1.2       anton     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
1.3       anton     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
1.7     ! pazsan    888: resolved (by using one of the loop-ending words or @code{DONE}).
1.2       anton     889: 
                    890: Another group of control structure words are
                    891: 
1.3       anton     892: doc-case
                    893: doc-endcase
                    894: doc-of
                    895: doc-endof
1.2       anton     896: 
                    897: @i{case-sys} and @i{of-sys} cannot be processed using @code{cs-pick} and
                    898: @code{cs-roll}.
                    899: 
1.3       anton     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: 
1.4       anton     943: @node Calls and returns, Exception Handling, Arbitrary control structures, Control Structures
1.3       anton     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: 
1.4       anton     957: @node Exception Handling,  , Calls and returns, Control Structures
1.3       anton     958: @subsection Exception Handling
                    959: 
                    960: doc-catch
                    961: doc-throw
                    962: 
1.4       anton     963: @node Locals, Defining Words, Control Structures, Words
1.1       anton     964: @section Locals
                    965: 
1.2       anton     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
1.4       anton     973: * gforth locals::               
                    974: * ANS Forth locals::            
1.2       anton     975: @end menu
                    976: 
1.4       anton     977: @node gforth locals, ANS Forth locals, Locals, Locals
1.2       anton     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: 
1.7     ! pazsan   1041: GNU Forth allows defining locals everywhere in a colon definition. This
        !          1042: poses the following questions:
1.2       anton    1043: 
1.4       anton    1044: @menu
                   1045: * Where are locals visible by name?::  
                   1046: * How long do locals live? ::   
                   1047: * Programming Style::           
                   1048: * Implementation::              
                   1049: @end menu
                   1050: 
                   1051: @node Where are locals visible by name?, How long do locals live?, gforth locals, gforth locals
1.2       anton    1052: @subsubsection Where are locals visible by name?
                   1053: 
                   1054: Basically, the answer is that locals are visible where you would expect
                   1055: it in block-structured languages, and sometimes a little longer. If you
                   1056: want to restrict the scope of a local, enclose its definition in
                   1057: @code{SCOPE}...@code{ENDSCOPE}.
                   1058: 
                   1059: doc-scope
                   1060: doc-endscope
                   1061: 
                   1062: These words behave like control structure words, so you can use them
                   1063: with @code{CS-PICK} and @code{CS-ROLL} to restrict the scope in
                   1064: arbitrary ways.
                   1065: 
                   1066: If you want a more exact answer to the visibility question, here's the
                   1067: basic principle: A local is visible in all places that can only be
                   1068: reached through the definition of the local@footnote{In compiler
                   1069: construction terminology, all places dominated by the definition of the
                   1070: local.}. In other words, it is not visible in places that can be reached
                   1071: without going through the definition of the local. E.g., locals defined
                   1072: in @code{IF}...@code{ENDIF} are visible until the @code{ENDIF}, locals
                   1073: defined in @code{BEGIN}...@code{UNTIL} are visible after the
                   1074: @code{UNTIL} (until, e.g., a subsequent @code{ENDSCOPE}).
                   1075: 
                   1076: The reasoning behind this solution is: We want to have the locals
                   1077: visible as long as it is meaningful. The user can always make the
                   1078: visibility shorter by using explicit scoping. In a place that can
                   1079: only be reached through the definition of a local, the meaning of a
                   1080: local name is clear. In other places it is not: How is the local
                   1081: initialized at the control flow path that does not contain the
                   1082: definition? Which local is meant, if the same name is defined twice in
                   1083: two independent control flow paths?
                   1084: 
                   1085: This should be enough detail for nearly all users, so you can skip the
                   1086: rest of this section. If you relly must know all the gory details and
                   1087: options, read on.
                   1088: 
                   1089: In order to implement this rule, the compiler has to know which places
                   1090: are unreachable. It knows this automatically after @code{AHEAD},
                   1091: @code{AGAIN}, @code{EXIT} and @code{LEAVE}; in other cases (e.g., after
                   1092: most @code{THROW}s), you can use the word @code{UNREACHABLE} to tell the
                   1093: compiler that the control flow never reaches that place. If
                   1094: @code{UNREACHABLE} is not used where it could, the only consequence is
                   1095: that the visibility of some locals is more limited than the rule above
                   1096: says. If @code{UNREACHABLE} is used where it should not (i.e., if you
                   1097: lie to the compiler), buggy code will be produced.
                   1098: 
                   1099: Another problem with this rule is that at @code{BEGIN}, the compiler
1.3       anton    1100: does not know which locals will be visible on the incoming
                   1101: back-edge. All problems discussed in the following are due to this
                   1102: ignorance of the compiler (we discuss the problems using @code{BEGIN}
                   1103: loops as examples; the discussion also applies to @code{?DO} and other
1.2       anton    1104: loops). Perhaps the most insidious example is:
                   1105: @example
                   1106: AHEAD
                   1107: BEGIN
                   1108:   x
                   1109: [ 1 CS-ROLL ] THEN
1.4       anton    1110:   @{ x @}
1.2       anton    1111:   ...
                   1112: UNTIL
                   1113: @end example
                   1114: 
                   1115: This should be legal according to the visibility rule. The use of
                   1116: @code{x} can only be reached through the definition; but that appears
                   1117: textually below the use.
                   1118: 
                   1119: From this example it is clear that the visibility rules cannot be fully
                   1120: implemented without major headaches. Our implementation treats common
                   1121: cases as advertised and the exceptions are treated in a safe way: The
                   1122: compiler makes a reasonable guess about the locals visible after a
                   1123: @code{BEGIN}; if it is too pessimistic, the
                   1124: user will get a spurious error about the local not being defined; if the
                   1125: compiler is too optimistic, it will notice this later and issue a
                   1126: warning. In the case above the compiler would complain about @code{x}
                   1127: being undefined at its use. You can see from the obscure examples in
                   1128: this section that it takes quite unusual control structures to get the
                   1129: compiler into trouble, and even then it will often do fine.
                   1130: 
                   1131: If the @code{BEGIN} is reachable from above, the most optimistic guess
                   1132: is that all locals visible before the @code{BEGIN} will also be
                   1133: visible after the @code{BEGIN}. This guess is valid for all loops that
                   1134: are entered only through the @code{BEGIN}, in particular, for normal
                   1135: @code{BEGIN}...@code{WHILE}...@code{REPEAT} and
                   1136: @code{BEGIN}...@code{UNTIL} loops and it is implemented in our
                   1137: compiler. When the branch to the @code{BEGIN} is finally generated by
                   1138: @code{AGAIN} or @code{UNTIL}, the compiler checks the guess and
                   1139: warns the user if it was too optimisitic:
                   1140: @example
                   1141: IF
1.4       anton    1142:   @{ x @}
1.2       anton    1143: BEGIN
                   1144:   \ x ? 
                   1145: [ 1 cs-roll ] THEN
                   1146:   ...
                   1147: UNTIL
                   1148: @end example
                   1149: 
                   1150: Here, @code{x} lives only until the @code{BEGIN}, but the compiler
                   1151: optimistically assumes that it lives until the @code{THEN}. It notices
                   1152: this difference when it compiles the @code{UNTIL} and issues a
                   1153: warning. The user can avoid the warning, and make sure that @code{x}
                   1154: is not used in the wrong area by using explicit scoping:
                   1155: @example
                   1156: IF
                   1157:   SCOPE
1.4       anton    1158:   @{ x @}
1.2       anton    1159:   ENDSCOPE
                   1160: BEGIN
                   1161: [ 1 cs-roll ] THEN
                   1162:   ...
                   1163: UNTIL
                   1164: @end example
                   1165: 
                   1166: Since the guess is optimistic, there will be no spurious error messages
                   1167: about undefined locals.
                   1168: 
                   1169: If the @code{BEGIN} is not reachable from above (e.g., after
                   1170: @code{AHEAD} or @code{EXIT}), the compiler cannot even make an
                   1171: optimistic guess, as the locals visible after the @code{BEGIN} may be
                   1172: defined later. Therefore, the compiler assumes that no locals are
                   1173: visible after the @code{BEGIN}. However, the useer can use
                   1174: @code{ASSUME-LIVE} to make the compiler assume that the same locals are
                   1175: visible at the BEGIN as at the point where the item was created.
                   1176: 
                   1177: doc-assume-live
                   1178: 
                   1179: E.g.,
                   1180: @example
1.4       anton    1181: @{ x @}
1.2       anton    1182: AHEAD
                   1183: ASSUME-LIVE
                   1184: BEGIN
                   1185:   x
                   1186: [ 1 CS-ROLL ] THEN
                   1187:   ...
                   1188: UNTIL
                   1189: @end example
                   1190: 
                   1191: Other cases where the locals are defined before the @code{BEGIN} can be
                   1192: handled by inserting an appropriate @code{CS-ROLL} before the
                   1193: @code{ASSUME-LIVE} (and changing the control-flow stack manipulation
                   1194: behind the @code{ASSUME-LIVE}).
                   1195: 
                   1196: Cases where locals are defined after the @code{BEGIN} (but should be
                   1197: visible immediately after the @code{BEGIN}) can only be handled by
                   1198: rearranging the loop. E.g., the ``most insidious'' example above can be
                   1199: arranged into:
                   1200: @example
                   1201: BEGIN
1.4       anton    1202:   @{ x @}
1.2       anton    1203:   ... 0=
                   1204: WHILE
                   1205:   x
                   1206: REPEAT
                   1207: @end example
                   1208: 
1.4       anton    1209: @node How long do locals live?, Programming Style, Where are locals visible by name?, gforth locals
1.2       anton    1210: @subsubsection How long do locals live?
                   1211: 
                   1212: The right answer for the lifetime question would be: A local lives at
                   1213: least as long as it can be accessed. For a value-flavoured local this
                   1214: means: until the end of its visibility. However, a variable-flavoured
                   1215: local could be accessed through its address far beyond its visibility
                   1216: scope. Ultimately, this would mean that such locals would have to be
                   1217: garbage collected. Since this entails un-Forth-like implementation
                   1218: complexities, I adopted the same cowardly solution as some other
                   1219: languages (e.g., C): The local lives only as long as it is visible;
                   1220: afterwards its address is invalid (and programs that access it
                   1221: afterwards are erroneous).
                   1222: 
1.4       anton    1223: @node Programming Style, Implementation, How long do locals live?, gforth locals
1.2       anton    1224: @subsubsection Programming Style
                   1225: 
                   1226: The freedom to define locals anywhere has the potential to change
                   1227: programming styles dramatically. In particular, the need to use the
                   1228: return stack for intermediate storage vanishes. Moreover, all stack
                   1229: manipulations (except @code{PICK}s and @code{ROLL}s with run-time
                   1230: determined arguments) can be eliminated: If the stack items are in the
                   1231: wrong order, just write a locals definition for all of them; then
                   1232: write the items in the order you want.
                   1233: 
                   1234: This seems a little far-fetched and eliminating stack manipulations is
1.4       anton    1235: unlikely to become a conscious programming objective. Still, the number
                   1236: of stack manipulations will be reduced dramatically if local variables
                   1237: are used liberally (e.g., compare @code{max} in @ref{gforth locals} with
                   1238: a traditional implementation of @code{max}).
1.2       anton    1239: 
                   1240: This shows one potential benefit of locals: making Forth programs more
                   1241: readable. Of course, this benefit will only be realized if the
                   1242: programmers continue to honour the principle of factoring instead of
                   1243: using the added latitude to make the words longer.
                   1244: 
                   1245: Using @code{TO} can and should be avoided.  Without @code{TO},
                   1246: every value-flavoured local has only a single assignment and many
                   1247: advantages of functional languages apply to Forth. I.e., programs are
                   1248: easier to analyse, to optimize and to read: It is clear from the
                   1249: definition what the local stands for, it does not turn into something
                   1250: different later.
                   1251: 
                   1252: E.g., a definition using @code{TO} might look like this:
                   1253: @example
                   1254: : strcmp @{ addr1 u1 addr2 u2 -- n @}
                   1255:  u1 u2 min 0
                   1256:  ?do
                   1257:    addr1 c@ addr2 c@ - ?dup
                   1258:    if
                   1259:      unloop exit
                   1260:    then
                   1261:    addr1 char+ TO addr1
                   1262:    addr2 char+ TO addr2
                   1263:  loop
                   1264:  u1 u2 - ;
                   1265: @end example
                   1266: Here, @code{TO} is used to update @code{addr1} and @code{addr2} at
                   1267: every loop iteration. @code{strcmp} is a typical example of the
                   1268: readability problems of using @code{TO}. When you start reading
                   1269: @code{strcmp}, you think that @code{addr1} refers to the start of the
                   1270: string. Only near the end of the loop you realize that it is something
                   1271: else.
                   1272: 
                   1273: This can be avoided by defining two locals at the start of the loop that
                   1274: are initialized with the right value for the current iteration.
                   1275: @example
                   1276: : strcmp @{ addr1 u1 addr2 u2 -- n @}
                   1277:  addr1 addr2
                   1278:  u1 u2 min 0 
                   1279:  ?do @{ s1 s2 @}
                   1280:    s1 c@ s2 c@ - ?dup 
                   1281:    if
                   1282:      unloop exit
                   1283:    then
                   1284:    s1 char+ s2 char+
                   1285:  loop
                   1286:  2drop
                   1287:  u1 u2 - ;
                   1288: @end example
                   1289: Here it is clear from the start that @code{s1} has a different value
                   1290: in every loop iteration.
                   1291: 
1.4       anton    1292: @node Implementation,  , Programming Style, gforth locals
1.2       anton    1293: @subsubsection Implementation
                   1294: 
                   1295: GNU Forth uses an extra locals stack. The most compelling reason for
                   1296: this is that the return stack is not float-aligned; using an extra stack
                   1297: also eliminates the problems and restrictions of using the return stack
                   1298: as locals stack. Like the other stacks, the locals stack grows toward
                   1299: lower addresses. A few primitives allow an efficient implementation:
                   1300: 
                   1301: doc-@local#
                   1302: doc-f@local#
                   1303: doc-laddr#
                   1304: doc-lp+!#
                   1305: doc-lp!
                   1306: doc->l
                   1307: doc-f>l
                   1308: 
                   1309: In addition to these primitives, some specializations of these
                   1310: primitives for commonly occurring inline arguments are provided for
                   1311: efficiency reasons, e.g., @code{@@local0} as specialization of
                   1312: @code{@@local#} for the inline argument 0. The following compiling words
                   1313: compile the right specialized version, or the general version, as
                   1314: appropriate:
                   1315: 
                   1316: doc-compile-@@local
                   1317: doc-compile-f@@local
                   1318: doc-compile-lp+!
                   1319: 
                   1320: Combinations of conditional branches and @code{lp+!#} like
                   1321: @code{?branch-lp+!#} (the locals pointer is only changed if the branch
                   1322: is taken) are provided for efficiency and correctness in loops.
                   1323: 
                   1324: A special area in the dictionary space is reserved for keeping the
                   1325: local variable names. @code{@{} switches the dictionary pointer to this
                   1326: area and @code{@}} switches it back and generates the locals
                   1327: initializing code. @code{W:} etc.@ are normal defining words. This
                   1328: special area is cleared at the start of every colon definition.
                   1329: 
                   1330: A special feature of GNU Forths dictionary is used to implement the
                   1331: definition of locals without type specifiers: every wordlist (aka
                   1332: vocabulary) has its own methods for searching
1.4       anton    1333: etc. (@pxref{Wordlists}). For the present purpose we defined a wordlist
1.2       anton    1334: with a special search method: When it is searched for a word, it
                   1335: actually creates that word using @code{W:}. @code{@{} changes the search
                   1336: order to first search the wordlist containing @code{@}}, @code{W:} etc.,
                   1337: and then the wordlist for defining locals without type specifiers.
                   1338: 
                   1339: The lifetime rules support a stack discipline within a colon
                   1340: definition: The lifetime of a local is either nested with other locals
                   1341: lifetimes or it does not overlap them.
                   1342: 
                   1343: At @code{BEGIN}, @code{IF}, and @code{AHEAD} no code for locals stack
                   1344: pointer manipulation is generated. Between control structure words
                   1345: locals definitions can push locals onto the locals stack. @code{AGAIN}
                   1346: is the simplest of the other three control flow words. It has to
                   1347: restore the locals stack depth of the corresponding @code{BEGIN}
                   1348: before branching. The code looks like this:
                   1349: @format
                   1350: @code{lp+!#} current-locals-size @minus{} dest-locals-size
                   1351: @code{branch} <begin>
                   1352: @end format
                   1353: 
                   1354: @code{UNTIL} is a little more complicated: If it branches back, it
                   1355: must adjust the stack just like @code{AGAIN}. But if it falls through,
                   1356: the locals stack must not be changed. The compiler generates the
                   1357: following code:
                   1358: @format
                   1359: @code{?branch-lp+!#} <begin> current-locals-size @minus{} dest-locals-size
                   1360: @end format
                   1361: The locals stack pointer is only adjusted if the branch is taken.
                   1362: 
                   1363: @code{THEN} can produce somewhat inefficient code:
                   1364: @format
                   1365: @code{lp+!#} current-locals-size @minus{} orig-locals-size
                   1366: <orig target>:
                   1367: @code{lp+!#} orig-locals-size @minus{} new-locals-size
                   1368: @end format
                   1369: The second @code{lp+!#} adjusts the locals stack pointer from the
1.4       anton    1370: level at the @var{orig} point to the level after the @code{THEN}. The
1.2       anton    1371: first @code{lp+!#} adjusts the locals stack pointer from the current
                   1372: level to the level at the orig point, so the complete effect is an
                   1373: adjustment from the current level to the right level after the
                   1374: @code{THEN}.
                   1375: 
                   1376: In a conventional Forth implementation a dest control-flow stack entry
                   1377: is just the target address and an orig entry is just the address to be
                   1378: patched. Our locals implementation adds a wordlist to every orig or dest
                   1379: item. It is the list of locals visible (or assumed visible) at the point
                   1380: described by the entry. Our implementation also adds a tag to identify
                   1381: the kind of entry, in particular to differentiate between live and dead
                   1382: (reachable and unreachable) orig entries.
                   1383: 
                   1384: A few unusual operations have to be performed on locals wordlists:
                   1385: 
                   1386: doc-common-list
                   1387: doc-sub-list?
                   1388: doc-list-size
                   1389: 
                   1390: Several features of our locals wordlist implementation make these
                   1391: operations easy to implement: The locals wordlists are organised as
                   1392: linked lists; the tails of these lists are shared, if the lists
                   1393: contain some of the same locals; and the address of a name is greater
                   1394: than the address of the names behind it in the list.
                   1395: 
                   1396: Another important implementation detail is the variable
                   1397: @code{dead-code}. It is used by @code{BEGIN} and @code{THEN} to
                   1398: determine if they can be reached directly or only through the branch
                   1399: that they resolve. @code{dead-code} is set by @code{UNREACHABLE},
                   1400: @code{AHEAD}, @code{EXIT} etc., and cleared at the start of a colon
                   1401: definition, by @code{BEGIN} and usually by @code{THEN}.
                   1402: 
                   1403: Counted loops are similar to other loops in most respects, but
                   1404: @code{LEAVE} requires special attention: It performs basically the same
                   1405: service as @code{AHEAD}, but it does not create a control-flow stack
                   1406: entry. Therefore the information has to be stored elsewhere;
                   1407: traditionally, the information was stored in the target fields of the
                   1408: branches created by the @code{LEAVE}s, by organizing these fields into a
                   1409: linked list. Unfortunately, this clever trick does not provide enough
                   1410: space for storing our extended control flow information. Therefore, we
                   1411: introduce another stack, the leave stack. It contains the control-flow
                   1412: stack entries for all unresolved @code{LEAVE}s.
                   1413: 
                   1414: Local names are kept until the end of the colon definition, even if
                   1415: they are no longer visible in any control-flow path. In a few cases
                   1416: this may lead to increased space needs for the locals name area, but
                   1417: usually less than reclaiming this space would cost in code size.
                   1418: 
                   1419: 
1.4       anton    1420: @node ANS Forth locals,  , gforth locals, Locals
1.2       anton    1421: @subsection ANS Forth locals
                   1422: 
                   1423: The ANS Forth locals wordset does not define a syntax for locals, but
                   1424: words that make it possible to define various syntaxes. One of the
                   1425: possible syntaxes is a subset of the syntax we used in the gforth locals
                   1426: wordset, i.e.:
                   1427: 
                   1428: @example
                   1429: @{ local1 local2 ... -- comment @}
                   1430: @end example
                   1431: or
                   1432: @example
                   1433: @{ local1 local2 ... @}
                   1434: @end example
                   1435: 
                   1436: The order of the locals corresponds to the order in a stack comment. The
                   1437: restrictions are:
1.1       anton    1438: 
1.2       anton    1439: @itemize @bullet
                   1440: @item
                   1441: Locals can only be cell-sized values (no type specifers are allowed).
                   1442: @item
                   1443: Locals can be defined only outside control structures.
                   1444: @item
                   1445: Locals can interfere with explicit usage of the return stack. For the
                   1446: exact (and long) rules, see the standard. If you don't use return stack
                   1447: accessing words in a definition using locals, you will we all right. The
                   1448: purpose of this rule is to make locals implementation on the return
                   1449: stack easier.
                   1450: @item
                   1451: The whole definition must be in one line.
                   1452: @end itemize
                   1453: 
                   1454: Locals defined in this way behave like @code{VALUE}s
1.4       anton    1455: (@xref{Values}). I.e., they are initialized from the stack. Using their
1.2       anton    1456: name produces their value. Their value can be changed using @code{TO}.
                   1457: 
                   1458: Since this syntax is supported by gforth directly, you need not do
                   1459: anything to use it. If you want to port a program using this syntax to
                   1460: another ANS Forth system, use @file{anslocal.fs} to implement the syntax
                   1461: on the other system.
                   1462: 
                   1463: Note that a syntax shown in the standard, section A.13 looks
                   1464: similar, but is quite different in having the order of locals
                   1465: reversed. Beware!
                   1466: 
                   1467: The ANS Forth locals wordset itself consists of the following word
                   1468: 
                   1469: doc-(local)
                   1470: 
                   1471: The ANS Forth locals extension wordset defines a syntax, but it is so
                   1472: awful that we strongly recommend not to use it. We have implemented this
                   1473: syntax to make porting to gforth easy, but do not document it here. The
                   1474: problem with this syntax is that the locals are defined in an order
                   1475: reversed with respect to the standard stack comment notation, making
                   1476: programs harder to read, and easier to misread and miswrite. The only
                   1477: merit of this syntax is that it is easy to implement using the ANS Forth
                   1478: locals wordset.
1.3       anton    1479: 
1.4       anton    1480: @node Defining Words, Wordlists, Locals, Words
                   1481: @section Defining Words
                   1482: 
                   1483: @node Values,  , Defining Words, Defining Words
                   1484: @subsection Values
                   1485: 
                   1486: @node Wordlists, Files, Defining Words, Words
                   1487: @section Wordlists
                   1488: 
                   1489: @node Files, Blocks, Wordlists, Words
                   1490: @section Files
                   1491: 
                   1492: @node Blocks, Other I/O, Files, Words
                   1493: @section Blocks
                   1494: 
                   1495: @node Other I/O, Programming Tools, Blocks, Words
                   1496: @section Other I/O
                   1497: 
                   1498: @node Programming Tools, Threading Words, Other I/O, Words
                   1499: @section Programming Tools
                   1500: 
1.5       anton    1501: @menu
                   1502: * Debugging::                   Simple and quick.
                   1503: * Assertions::                  Making your programs self-checking.
                   1504: @end menu
                   1505: 
                   1506: @node Debugging, Assertions, Programming Tools, Programming Tools
1.4       anton    1507: @subsection Debugging
                   1508: 
                   1509: The simple debugging aids provided in @file{debugging.fs}
                   1510: are meant to support a different style of debugging than the
                   1511: tracing/stepping debuggers used in languages with long turn-around
                   1512: times.
                   1513: 
                   1514: A much better (faster) way in fast-compilig languages is to add
                   1515: printing code at well-selected places, let the program run, look at
                   1516: the output, see where things went wrong, add more printing code, etc.,
                   1517: until the bug is found.
                   1518: 
                   1519: The word @code{~~} is easy to insert. It just prints debugging
                   1520: information (by default the source location and the stack contents). It
                   1521: is also easy to remove (@kbd{C-x ~} in the Emacs Forth mode to
                   1522: query-replace them with nothing). The deferred words
                   1523: @code{printdebugdata} and @code{printdebugline} control the output of
                   1524: @code{~~}. The default source location output format works well with
                   1525: Emacs' compilation mode, so you can step through the program at the
1.5       anton    1526: source level using @kbd{C-x `} (the advantage over a stepping debugger
                   1527: is that you can step in any direction and you know where the crash has
                   1528: happened or where the strange data has occurred).
1.4       anton    1529: 
                   1530: Note that the default actions clobber the contents of the pictured
                   1531: numeric output string, so you should not use @code{~~}, e.g., between
                   1532: @code{<#} and @code{#>}.
                   1533: 
                   1534: doc-~~
                   1535: doc-printdebugdata
                   1536: doc-printdebugline
                   1537: 
1.5       anton    1538: @node Assertions,  , Debugging, Programming Tools
1.4       anton    1539: @subsection Assertions
                   1540: 
1.5       anton    1541: It is a good idea to make your programs self-checking, in particular, if
                   1542: you use an assumption (e.g., that a certain field of a data structure is
                   1543: never zero) that may become wrong during maintenance. GForth supports
                   1544: assertions for this purpose. They are used like this:
                   1545: 
                   1546: @example
                   1547: assert( @var{flag} )
                   1548: @end example
                   1549: 
                   1550: The code between @code{assert(} and @code{)} should compute a flag, that
                   1551: should be true if everything is alright and false otherwise. It should
                   1552: not change anything else on the stack. The overall stack effect of the
                   1553: assertion is @code{( -- )}. E.g.
                   1554: 
                   1555: @example
                   1556: assert( 1 1 + 2 = ) \ what we learn in school
                   1557: assert( dup 0<> ) \ assert that the top of stack is not zero
                   1558: assert( false ) \ this code should not be reached
                   1559: @end example
                   1560: 
                   1561: The need for assertions is different at different times. During
                   1562: debugging, we want more checking, in production we sometimes care more
                   1563: for speed. Therefore, assertions can be turned off, i.e., the assertion
                   1564: becomes a comment. Depending on the importance of an assertion and the
                   1565: time it takes to check it, you may want to turn off some assertions and
                   1566: keep others turned on. GForth provides several levels of assertions for
                   1567: this purpose:
                   1568: 
                   1569: doc-assert0(
                   1570: doc-assert1(
                   1571: doc-assert2(
                   1572: doc-assert3(
                   1573: doc-assert(
                   1574: doc-)
                   1575: 
                   1576: @code{Assert(} is the same as @code{assert1(}. The variable
                   1577: @code{assert-level} specifies the highest assertions that are turned
                   1578: on. I.e., at the default @code{assert-level} of one, @code{assert0(} and
                   1579: @code{assert1(} assertions perform checking, while @code{assert2(} and
                   1580: @code{assert3(} assertions are treated as comments.
                   1581: 
                   1582: Note that the @code{assert-level} is evaluated at compile-time, not at
                   1583: run-time. I.e., you cannot turn assertions on or off at run-time, you
                   1584: have to set the @code{assert-level} appropriately before compiling a
                   1585: piece of code. You can compile several pieces of code at several
                   1586: @code{assert-level}s (e.g., a trusted library at level 1 and newly
                   1587: written code at level 3).
                   1588: 
                   1589: doc-assert-level
                   1590: 
                   1591: If an assertion fails, a message compatible with Emacs' compilation mode
                   1592: is produced and the execution is aborted (currently with @code{ABORT"}.
                   1593: If there is interest, we will introduce a special throw code. But if you
                   1594: intend to @code{catch} a specific condition, using @code{throw} is
                   1595: probably more appropriate than an assertion).
                   1596: 
1.4       anton    1597: @node Threading Words,  , Programming Tools, Words
                   1598: @section Threading Words
                   1599: 
                   1600: These words provide access to code addresses and other threading stuff
                   1601: in gforth (and, possibly, other interpretive Forths). It more or less
                   1602: abstracts away the differences between direct and indirect threading
                   1603: (and, for direct threading, the machine dependences). However, at
                   1604: present this wordset is still inclomplete. It is also pretty low-level;
                   1605: some day it will hopefully be made unnecessary by an internals words set
                   1606: that abstracts implementation details away completely.
                   1607: 
                   1608: doc->code-address
                   1609: doc->does-code
                   1610: doc-code-address!
                   1611: doc-does-code!
                   1612: doc-does-handler!
                   1613: doc-/does-handler
                   1614: 
                   1615: @node ANS conformance, Model, Words, Top
                   1616: @chapter ANS conformance
                   1617: 
                   1618: @node Model, Emacs and GForth, ANS conformance, Top
                   1619: @chapter Model
                   1620: 
                   1621: @node Emacs and GForth, Internals, Model, Top
                   1622: @chapter Emacs and GForth
                   1623: 
                   1624: GForth comes with @file{gforth.el}, an improved version of
                   1625: @file{forth.el} by Goran Rydqvist (icluded in the TILE package). The
                   1626: improvements are a better (but still not perfect) handling of
                   1627: indentation. I have also added comment paragraph filling (@kbd{M-q}),
                   1628: commenting (@kbd{C-x \}) and uncommenting (@kbd{C-x |}) regions and
                   1629: removing debugging tracers (@kbd{C-x ~}). I left the stuff I do not use
                   1630: alone, even though some of it only makes sense for TILE. To get a
                   1631: description of these features, enter Forth mode and type @kbd{C-h m}.
                   1632: 
                   1633: In addition, GForth supports Emacs quite well: The source code locations
                   1634: given in error messages, debugging output (from @code{~~}) and failed
                   1635: assertion messages are in the right format for Emacs' compilation mode
                   1636: (@pxref{Compilation, , Running Compilations under Emacs, emacs, Emacs
                   1637: Manual}) so the source location corresponding to an error or other
                   1638: message is only a few keystrokes away (@kbd{C-x `} for the next error,
                   1639: @kbd{C-c C-c} for the error under the cursor).
                   1640: 
                   1641: Also, if you @code{include} @file{etags.fs}, a new @file{TAGS} file
                   1642: (@pxref{Tags, , Tags Tables, emacs, Emacs Manual}) will be produced that
                   1643: contains the definitions of all words defined afterwards. You can then
                   1644: find the source for a word using @kbd{M-.}. Note that emacs can use
                   1645: several tags files at the same time (e.g., one for the gforth sources
                   1646: and one for your program).
                   1647: 
                   1648: To get all these benefits, add the following lines to your @file{.emacs}
                   1649: file:
                   1650: 
                   1651: @example
                   1652: (autoload 'forth-mode "gforth.el")
                   1653: (setq auto-mode-alist (cons '("\\.fs\\'" . forth-mode) auto-mode-alist))
                   1654: @end example
                   1655: 
                   1656: @node Internals, Bugs, Emacs and GForth, Top
1.3       anton    1657: @chapter Internals
                   1658: 
                   1659: Reading this section is not necessary for programming with gforth. It
                   1660: should be helpful for finding your way in the gforth sources.
                   1661: 
1.4       anton    1662: @menu
                   1663: * Portability::                 
                   1664: * Threading::                   
                   1665: * Primitives::                  
                   1666: * System Architecture::         
                   1667: @end menu
                   1668: 
                   1669: @node Portability, Threading, Internals, Internals
1.3       anton    1670: @section Portability
                   1671: 
                   1672: One of the main goals of the effort is availability across a wide range
                   1673: of personal machines. fig-Forth, and, to a lesser extent, F83, achieved
                   1674: this goal by manually coding the engine in assembly language for several
                   1675: then-popular processors. This approach is very labor-intensive and the
                   1676: results are short-lived due to progress in computer architecture.
                   1677: 
                   1678: Others have avoided this problem by coding in C, e.g., Mitch Bradley
                   1679: (cforth), Mikael Patel (TILE) and Dirk Zoller (pfe). This approach is
                   1680: particularly popular for UNIX-based Forths due to the large variety of
                   1681: architectures of UNIX machines. Unfortunately an implementation in C
                   1682: does not mix well with the goals of efficiency and with using
                   1683: traditional techniques: Indirect or direct threading cannot be expressed
                   1684: in C, and switch threading, the fastest technique available in C, is
                   1685: significantly slower. Another problem with C is that it's very
                   1686: cumbersome to express double integer arithmetic.
                   1687: 
                   1688: Fortunately, there is a portable language that does not have these
                   1689: limitations: GNU C, the version of C processed by the GNU C compiler
                   1690: (@pxref{C Extensions, , Extensions to the C Language Family, gcc.info,
                   1691: GNU C Manual}). Its labels as values feature (@pxref{Labels as Values, ,
                   1692: Labels as Values, gcc.info, GNU C Manual}) makes direct and indirect
                   1693: threading possible, its @code{long long} type (@pxref{Long Long, ,
                   1694: Double-Word Integers, gcc.info, GNU C Manual}) corresponds to Forths
                   1695: double numbers. GNU C is available for free on all important (and many
                   1696: unimportant) UNIX machines, VMS, 80386s running MS-DOS, the Amiga, and
                   1697: the Atari ST, so a Forth written in GNU C can run on all these
                   1698: machines@footnote{Due to Apple's look-and-feel lawsuit it is not
1.5       anton    1699: available on the Mac (@pxref{Boycott, , Protect Your Freedom---Fight
1.3       anton    1700: ``Look And Feel'', gcc.info, GNU C Manual}).}.
                   1701: 
                   1702: Writing in a portable language has the reputation of producing code that
                   1703: is slower than assembly. For our Forth engine we repeatedly looked at
                   1704: the code produced by the compiler and eliminated most compiler-induced
                   1705: inefficiencies by appropriate changes in the source-code.
                   1706: 
                   1707: However, register allocation cannot be portably influenced by the
                   1708: programmer, leading to some inefficiencies on register-starved
                   1709: machines. We use explicit register declarations (@pxref{Explicit Reg
                   1710: Vars, , Variables in Specified Registers, gcc.info, GNU C Manual}) to
                   1711: improve the speed on some machines. They are turned on by using the
                   1712: @code{gcc} switch @code{-DFORCE_REG}. Unfortunately, this feature not
                   1713: only depends on the machine, but also on the compiler version: On some
                   1714: machines some compiler versions produce incorrect code when certain
                   1715: explicit register declarations are used. So by default
                   1716: @code{-DFORCE_REG} is not used.
                   1717: 
1.4       anton    1718: @node Threading, Primitives, Portability, Internals
1.3       anton    1719: @section Threading
                   1720: 
                   1721: GNU C's labels as values extension (available since @code{gcc-2.0},
                   1722: @pxref{Labels as Values, , Labels as Values, gcc.info, GNU C Manual})
                   1723: makes it possible to take the address of @var{label} by writing
                   1724: @code{&&@var{label}}.  This address can then be used in a statement like
                   1725: @code{goto *@var{address}}. I.e., @code{goto *&&x} is the same as
                   1726: @code{goto x}.
                   1727: 
                   1728: With this feature an indirect threaded NEXT looks like:
                   1729: @example
                   1730: cfa = *ip++;
                   1731: ca = *cfa;
                   1732: goto *ca;
                   1733: @end example
                   1734: For those unfamiliar with the names: @code{ip} is the Forth instruction
                   1735: pointer; the @code{cfa} (code-field address) corresponds to ANS Forths
                   1736: execution token and points to the code field of the next word to be
                   1737: executed; The @code{ca} (code address) fetched from there points to some
                   1738: executable code, e.g., a primitive or the colon definition handler
                   1739: @code{docol}.
                   1740: 
                   1741: Direct threading is even simpler:
                   1742: @example
                   1743: ca = *ip++;
                   1744: goto *ca;
                   1745: @end example
                   1746: 
                   1747: Of course we have packaged the whole thing neatly in macros called
                   1748: @code{NEXT} and @code{NEXT1} (the part of NEXT after fetching the cfa).
                   1749: 
1.4       anton    1750: @menu
                   1751: * Scheduling::                  
                   1752: * Direct or Indirect Threaded?::  
                   1753: * DOES>::                       
                   1754: @end menu
                   1755: 
                   1756: @node Scheduling, Direct or Indirect Threaded?, Threading, Threading
1.3       anton    1757: @subsection Scheduling
                   1758: 
                   1759: There is a little complication: Pipelined and superscalar processors,
                   1760: i.e., RISC and some modern CISC machines can process independent
                   1761: instructions while waiting for the results of an instruction. The
                   1762: compiler usually reorders (schedules) the instructions in a way that
                   1763: achieves good usage of these delay slots. However, on our first tries
                   1764: the compiler did not do well on scheduling primitives. E.g., for
                   1765: @code{+} implemented as
                   1766: @example
                   1767: n=sp[0]+sp[1];
                   1768: sp++;
                   1769: sp[0]=n;
                   1770: NEXT;
                   1771: @end example
                   1772: the NEXT comes strictly after the other code, i.e., there is nearly no
                   1773: scheduling. After a little thought the problem becomes clear: The
                   1774: compiler cannot know that sp and ip point to different addresses (and
1.4       anton    1775: the version of @code{gcc} we used would not know it even if it was
                   1776: possible), so it could not move the load of the cfa above the store to
                   1777: the TOS. Indeed the pointers could be the same, if code on or very near
                   1778: the top of stack were executed. In the interest of speed we chose to
                   1779: forbid this probably unused ``feature'' and helped the compiler in
                   1780: scheduling: NEXT is divided into the loading part (@code{NEXT_P1}) and
                   1781: the goto part (@code{NEXT_P2}). @code{+} now looks like:
1.3       anton    1782: @example
                   1783: n=sp[0]+sp[1];
                   1784: sp++;
                   1785: NEXT_P1;
                   1786: sp[0]=n;
                   1787: NEXT_P2;
                   1788: @end example
1.4       anton    1789: This can be scheduled optimally by the compiler.
1.3       anton    1790: 
                   1791: This division can be turned off with the switch @code{-DCISC_NEXT}. This
                   1792: switch is on by default on machines that do not profit from scheduling
                   1793: (e.g., the 80386), in order to preserve registers.
                   1794: 
1.4       anton    1795: @node Direct or Indirect Threaded?, DOES>, Scheduling, Threading
1.3       anton    1796: @subsection Direct or Indirect Threaded?
                   1797: 
                   1798: Both! After packaging the nasty details in macro definitions we
                   1799: realized that we could switch between direct and indirect threading by
                   1800: simply setting a compilation flag (@code{-DDIRECT_THREADED}) and
                   1801: defining a few machine-specific macros for the direct-threading case.
                   1802: On the Forth level we also offer access words that hide the
                   1803: differences between the threading methods (@pxref{Threading Words}).
                   1804: 
                   1805: Indirect threading is implemented completely
                   1806: machine-independently. Direct threading needs routines for creating
                   1807: jumps to the executable code (e.g. to docol or dodoes). These routines
                   1808: are inherently machine-dependent, but they do not amount to many source
                   1809: lines. I.e., even porting direct threading to a new machine is a small
                   1810: effort.
                   1811: 
1.4       anton    1812: @node DOES>,  , Direct or Indirect Threaded?, Threading
1.3       anton    1813: @subsection DOES>
                   1814: One of the most complex parts of a Forth engine is @code{dodoes}, i.e.,
                   1815: the chunk of code executed by every word defined by a
                   1816: @code{CREATE}...@code{DOES>} pair. The main problem here is: How to find
                   1817: the Forth code to be executed, i.e. the code after the @code{DOES>} (the
                   1818: DOES-code)? There are two solutions:
                   1819: 
                   1820: In fig-Forth the code field points directly to the dodoes and the
                   1821: DOES-code address is stored in the cell after the code address
                   1822: (i.e. at cfa cell+). It may seem that this solution is illegal in the
                   1823: Forth-79 and all later standards, because in fig-Forth this address
                   1824: lies in the body (which is illegal in these standards). However, by
                   1825: making the code field larger for all words this solution becomes legal
                   1826: again. We use this approach for the indirect threaded version. Leaving
                   1827: a cell unused in most words is a bit wasteful, but on the machines we
                   1828: are targetting this is hardly a problem. The other reason for having a
                   1829: code field size of two cells is to avoid having different image files
1.4       anton    1830: for direct and indirect threaded systems (@pxref{System Architecture}).
1.3       anton    1831: 
                   1832: The other approach is that the code field points or jumps to the cell
                   1833: after @code{DOES}. In this variant there is a jump to @code{dodoes} at
                   1834: this address. @code{dodoes} can then get the DOES-code address by
                   1835: computing the code address, i.e., the address of the jump to dodoes,
                   1836: and add the length of that jump field. A variant of this is to have a
                   1837: call to @code{dodoes} after the @code{DOES>}; then the return address
                   1838: (which can be found in the return register on RISCs) is the DOES-code
                   1839: address. Since the two cells available in the code field are usually
                   1840: used up by the jump to the code address in direct threading, we use
                   1841: this approach for direct threading. We did not want to add another
                   1842: cell to the code field.
                   1843: 
1.4       anton    1844: @node Primitives, System Architecture, Threading, Internals
1.3       anton    1845: @section Primitives
                   1846: 
1.4       anton    1847: @menu
                   1848: * Automatic Generation::        
                   1849: * TOS Optimization::            
                   1850: * Produced code::               
                   1851: @end menu
                   1852: 
                   1853: @node Automatic Generation, TOS Optimization, Primitives, Primitives
1.3       anton    1854: @subsection Automatic Generation
                   1855: 
                   1856: Since the primitives are implemented in a portable language, there is no
                   1857: longer any need to minimize the number of primitives. On the contrary,
                   1858: having many primitives is an advantage: speed. In order to reduce the
                   1859: number of errors in primitives and to make programming them easier, we
                   1860: provide a tool, the primitive generator (@file{prims2x.fs}), that
                   1861: automatically generates most (and sometimes all) of the C code for a
                   1862: primitive from the stack effect notation.  The source for a primitive
                   1863: has the following form:
                   1864: 
                   1865: @format
                   1866: @var{Forth-name}       @var{stack-effect}      @var{category}  [@var{pronounc.}]
                   1867: [@code{""}@var{glossary entry}@code{""}]
                   1868: @var{C code}
                   1869: [@code{:}
                   1870: @var{Forth code}]
                   1871: @end format
                   1872: 
                   1873: The items in brackets are optional. The category and glossary fields
                   1874: are there for generating the documentation, the Forth code is there
                   1875: for manual implementations on machines without GNU C. E.g., the source
                   1876: for the primitive @code{+} is:
                   1877: @example
                   1878: +    n1 n2 -- n    core    plus
                   1879: n = n1+n2;
                   1880: @end example
                   1881: 
                   1882: This looks like a specification, but in fact @code{n = n1+n2} is C
                   1883: code. Our primitive generation tool extracts a lot of information from
                   1884: the stack effect notations@footnote{We use a one-stack notation, even
                   1885: though we have separate data and floating-point stacks; The separate
                   1886: notation can be generated easily from the unified notation.}: The number
                   1887: of items popped from and pushed on the stack, their type, and by what
                   1888: name they are referred to in the C code. It then generates a C code
                   1889: prelude and postlude for each primitive. The final C code for @code{+}
                   1890: looks like this:
                   1891: 
                   1892: @example
                   1893: I_plus:        /* + ( n1 n2 -- n ) */  /* label, stack effect */
                   1894: /*  */                          /* documentation */
1.4       anton    1895: @{
1.3       anton    1896: DEF_CA                          /* definition of variable ca (indirect threading) */
                   1897: Cell n1;                        /* definitions of variables */
                   1898: Cell n2;
                   1899: Cell n;
                   1900: n1 = (Cell) sp[1];              /* input */
                   1901: n2 = (Cell) TOS;
                   1902: sp += 1;                        /* stack adjustment */
                   1903: NAME("+")                       /* debugging output (with -DDEBUG) */
1.4       anton    1904: @{
1.3       anton    1905: n = n1+n2;                      /* C code taken from the source */
1.4       anton    1906: @}
1.3       anton    1907: NEXT_P1;                        /* NEXT part 1 */
                   1908: TOS = (Cell)n;                  /* output */
                   1909: NEXT_P2;                        /* NEXT part 2 */
1.4       anton    1910: @}
1.3       anton    1911: @end example
                   1912: 
                   1913: This looks long and inefficient, but the GNU C compiler optimizes quite
                   1914: well and produces optimal code for @code{+} on, e.g., the R3000 and the
                   1915: HP RISC machines: Defining the @code{n}s does not produce any code, and
                   1916: using them as intermediate storage also adds no cost.
                   1917: 
                   1918: There are also other optimizations, that are not illustrated by this
                   1919: example: Assignments between simple variables are usually for free (copy
                   1920: propagation). If one of the stack items is not used by the primitive
                   1921: (e.g.  in @code{drop}), the compiler eliminates the load from the stack
                   1922: (dead code elimination). On the other hand, there are some things that
                   1923: the compiler does not do, therefore they are performed by
                   1924: @file{prims2x.fs}: The compiler does not optimize code away that stores
                   1925: a stack item to the place where it just came from (e.g., @code{over}).
                   1926: 
                   1927: While programming a primitive is usually easy, there are a few cases
                   1928: where the programmer has to take the actions of the generator into
                   1929: account, most notably @code{?dup}, but also words that do not (always)
                   1930: fall through to NEXT.
                   1931: 
1.4       anton    1932: @node TOS Optimization, Produced code, Automatic Generation, Primitives
1.3       anton    1933: @subsection TOS Optimization
                   1934: 
                   1935: An important optimization for stack machine emulators, e.g., Forth
                   1936: engines, is keeping  one or more of the top stack items in
1.4       anton    1937: registers.  If a word has the stack effect @var{in1}...@var{inx} @code{--}
                   1938: @var{out1}...@var{outy}, keeping the top @var{n} items in registers
1.3       anton    1939: @itemize
                   1940: @item
                   1941: is better than keeping @var{n-1} items, if @var{x>=n} and @var{y>=n},
                   1942: due to fewer loads from and stores to the stack.
                   1943: @item is slower than keeping @var{n-1} items, if @var{x<>y} and @var{x<n} and
                   1944: @var{y<n}, due to additional moves between registers.
                   1945: @end itemize
                   1946: 
                   1947: In particular, keeping one item in a register is never a disadvantage,
                   1948: if there are enough registers. Keeping two items in registers is a
                   1949: disadvantage for frequent words like @code{?branch}, constants,
                   1950: variables, literals and @code{i}. Therefore our generator only produces
                   1951: code that keeps zero or one items in registers. The generated C code
                   1952: covers both cases; the selection between these alternatives is made at
                   1953: C-compile time using the switch @code{-DUSE_TOS}. @code{TOS} in the C
                   1954: code for @code{+} is just a simple variable name in the one-item case,
                   1955: otherwise it is a macro that expands into @code{sp[0]}. Note that the
                   1956: GNU C compiler tries to keep simple variables like @code{TOS} in
                   1957: registers, and it usually succeeds, if there are enough registers.
                   1958: 
                   1959: The primitive generator performs the TOS optimization for the
                   1960: floating-point stack, too (@code{-DUSE_FTOS}). For floating-point
                   1961: operations the benefit of this optimization is even larger:
                   1962: floating-point operations take quite long on most processors, but can be
                   1963: performed in parallel with other operations as long as their results are
                   1964: not used. If the FP-TOS is kept in a register, this works. If
                   1965: it is kept on the stack, i.e., in memory, the store into memory has to
                   1966: wait for the result of the floating-point operation, lengthening the
                   1967: execution time of the primitive considerably.
                   1968: 
                   1969: The TOS optimization makes the automatic generation of primitives a
                   1970: bit more complicated. Just replacing all occurrences of @code{sp[0]} by
                   1971: @code{TOS} is not sufficient. There are some special cases to
                   1972: consider:
                   1973: @itemize
                   1974: @item In the case of @code{dup ( w -- w w )} the generator must not
                   1975: eliminate the store to the original location of the item on the stack,
                   1976: if the TOS optimization is turned on.
1.4       anton    1977: @item Primitives with stack effects of the form @code{--}
                   1978: @var{out1}...@var{outy} must store the TOS to the stack at the start.
                   1979: Likewise, primitives with the stack effect @var{in1}...@var{inx} @code{--}
1.3       anton    1980: must load the TOS from the stack at the end. But for the null stack
                   1981: effect @code{--} no stores or loads should be generated.
                   1982: @end itemize
                   1983: 
1.4       anton    1984: @node Produced code,  , TOS Optimization, Primitives
1.3       anton    1985: @subsection Produced code
                   1986: 
                   1987: To see what assembly code is produced for the primitives on your machine
                   1988: with your compiler and your flag settings, type @code{make engine.s} and
1.4       anton    1989: look at the resulting file @file{engine.s}.
1.3       anton    1990: 
1.4       anton    1991: @node System Architecture,  , Primitives, Internals
1.3       anton    1992: @section System Architecture
                   1993: 
                   1994: Our Forth system consists not only of primitives, but also of
                   1995: definitions written in Forth. Since the Forth compiler itself belongs
                   1996: to those definitions, it is not possible to start the system with the
                   1997: primitives and the Forth source alone. Therefore we provide the Forth
                   1998: code as an image file in nearly executable form. At the start of the
                   1999: system a C routine loads the image file into memory, sets up the
                   2000: memory (stacks etc.) according to information in the image file, and
                   2001: starts executing Forth code.
                   2002: 
                   2003: The image file format is a compromise between the goals of making it
                   2004: easy to generate image files and making them portable. The easiest way
                   2005: to generate an image file is to just generate a memory dump. However,
                   2006: this kind of image file cannot be used on a different machine, or on
                   2007: the next version of the engine on the same machine, it even might not
                   2008: work with the same engine compiled by a different version of the C
                   2009: compiler. We would like to have as few versions of the image file as
                   2010: possible, because we do not want to distribute many versions of the
                   2011: same image file, and to make it easy for the users to use their image
                   2012: files on many machines. We currently need to create a different image
                   2013: file for machines with different cell sizes and different byte order
                   2014: (little- or big-endian)@footnote{We consider adding information to the
                   2015: image file that enables the loader to change the byte order.}.
                   2016: 
                   2017: Forth code that is going to end up in a portable image file has to
1.4       anton    2018: comply to some restrictions: addresses have to be stored in memory with
                   2019: special words (@code{A!}, @code{A,}, etc.) in order to make the code
                   2020: relocatable. Cells, floats, etc., have to be stored at the natural
                   2021: alignment boundaries@footnote{E.g., store floats (8 bytes) at an address
                   2022: dividable by~8. This happens automatically in our system when you use
                   2023: the ANS Forth alignment words.}, in order to avoid alignment faults on
                   2024: machines with stricter alignment. The image file is produced by a
                   2025: metacompiler (@file{cross.fs}).
1.3       anton    2026: 
                   2027: So, unlike the image file of Mitch Bradleys @code{cforth}, our image
                   2028: file is not directly executable, but has to undergo some manipulations
                   2029: during loading. Address relocation is performed at image load-time, not
                   2030: at run-time. The loader also has to replace tokens standing for
                   2031: primitive calls with the appropriate code-field addresses (or code
                   2032: addresses in the case of direct threading).
1.4       anton    2033: 
                   2034: @node Bugs, Pedigree, Internals, Top
                   2035: @chapter Bugs
                   2036: 
                   2037: @node Pedigree, Word Index, Bugs, Top
                   2038: @chapter Pedigree
                   2039: 
                   2040: @node Word Index, Node Index, Pedigree, Top
                   2041: @chapter Word Index
                   2042: 
                   2043: @node Node Index,  , Word Index, Top
                   2044: @chapter Node Index
1.1       anton    2045: 
                   2046: @contents
                   2047: @bye
                   2048: 

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