File:  [gforth] / gforth / Attic / gforth.ds
Revision 1.37: download - view: text, annotated - select for diffs
Mon Sep 30 13:16:09 1996 UTC (23 years, 2 months ago) by anton
Branches: MAIN
CVS tags: HEAD
Documentation changes
Building in a dir different from the srcdir now works
a few bug fixes

    1: \input texinfo   @c -*-texinfo-*-
    2: @comment The source is gforth.ds, from which gforth.texi is generated
    3: @comment %**start of header (This is for running Texinfo on a region.)
    4: @setfilename gforth.info
    5: @settitle Gforth Manual
    6: @comment @setchapternewpage odd
    7: @comment %**end of header (This is for running Texinfo on a region.)
    8: 
    9: @ifinfo
   10: This file documents Gforth 0.2
   11: 
   12: Copyright @copyright{} 1995,1996 Free Software Foundation, Inc.
   13: 
   14:      Permission is granted to make and distribute verbatim copies of
   15:      this manual provided the copyright notice and this permission notice
   16:      are preserved on all copies.
   17:      
   18: @ignore
   19:      Permission is granted to process this file through TeX and print the
   20:      results, provided the printed document carries a copying permission
   21:      notice identical to this one except for the removal of this paragraph
   22:      (this paragraph not being relevant to the printed manual).
   23:      
   24: @end ignore
   25:      Permission is granted to copy and distribute modified versions of this
   26:      manual under the conditions for verbatim copying, provided also that the
   27:      sections entitled "Distribution" and "General Public License" are
   28:      included exactly as in the original, and provided that the entire
   29:      resulting derived work is distributed under the terms of a permission
   30:      notice identical to this one.
   31:      
   32:      Permission is granted to copy and distribute translations of this manual
   33:      into another language, under the above conditions for modified versions,
   34:      except that the sections entitled "Distribution" and "General Public
   35:      License" may be included in a translation approved by the author instead
   36:      of in the original English.
   37: @end ifinfo
   38: 
   39: @finalout
   40: @titlepage
   41: @sp 10
   42: @center @titlefont{Gforth Manual}
   43: @sp 2
   44: @center for version 0.2
   45: @sp 2
   46: @center Anton Ertl
   47: @center Bernd Paysan
   48: @sp 3
   49: @center This manual is under construction
   50: 
   51: @comment  The following two commands start the copyright page.
   52: @page
   53: @vskip 0pt plus 1filll
   54: Copyright @copyright{} 1995,1996 Free Software Foundation, Inc.
   55: 
   56: @comment !! Published by ... or You can get a copy of this manual ...
   57: 
   58:      Permission is granted to make and distribute verbatim copies of
   59:      this manual provided the copyright notice and this permission notice
   60:      are preserved on all copies.
   61:      
   62:      Permission is granted to copy and distribute modified versions of this
   63:      manual under the conditions for verbatim copying, provided also that the
   64:      sections entitled "Distribution" and "General Public License" are
   65:      included exactly as in the original, and provided that the entire
   66:      resulting derived work is distributed under the terms of a permission
   67:      notice identical to this one.
   68:      
   69:      Permission is granted to copy and distribute translations of this manual
   70:      into another language, under the above conditions for modified versions,
   71:      except that the sections entitled "Distribution" and "General Public
   72:      License" may be included in a translation approved by the author instead
   73:      of in the original English.
   74: @end titlepage
   75: 
   76: 
   77: @node Top, License, (dir), (dir)
   78: @ifinfo
   79: Gforth is a free implementation of ANS Forth available on many
   80: personal machines. This manual corresponds to version 0.2.
   81: @end ifinfo
   82: 
   83: @menu
   84: * License::                     
   85: * Goals::                       About the Gforth Project
   86: * Other Books::                 Things you might want to read
   87: * Invocation::                  Starting Gforth
   88: * Words::                       Forth words available in Gforth
   89: * ANS conformance::             Implementation-defined options etc.
   90: * Model::                       The abstract machine of Gforth
   91: * Integrating Gforth::          Forth as scripting language for applications.
   92: * Emacs and Gforth::            The Gforth Mode
   93: * Internals::                   Implementation details
   94: * Bugs::                        How to report them
   95: * Origin::                      Authors and ancestors of Gforth
   96: * Word Index::                  An item for each Forth word
   97: * Node Index::                  An item for each node
   98: @end menu
   99: 
  100: @node License, Goals, Top, Top
  101: @unnumbered GNU GENERAL PUBLIC LICENSE
  102: @center Version 2, June 1991
  103: 
  104: @display
  105: Copyright @copyright{} 1989, 1991 Free Software Foundation, Inc.
  106: 675 Mass Ave, Cambridge, MA 02139, USA
  107: 
  108: Everyone is permitted to copy and distribute verbatim copies
  109: of this license document, but changing it is not allowed.
  110: @end display
  111: 
  112: @unnumberedsec Preamble
  113: 
  114:   The licenses for most software are designed to take away your
  115: freedom to share and change it.  By contrast, the GNU General Public
  116: License is intended to guarantee your freedom to share and change free
  117: software---to make sure the software is free for all its users.  This
  118: General Public License applies to most of the Free Software
  119: Foundation's software and to any other program whose authors commit to
  120: using it.  (Some other Free Software Foundation software is covered by
  121: the GNU Library General Public License instead.)  You can apply it to
  122: your programs, too.
  123: 
  124:   When we speak of free software, we are referring to freedom, not
  125: price.  Our General Public Licenses are designed to make sure that you
  126: have the freedom to distribute copies of free software (and charge for
  127: this service if you wish), that you receive source code or can get it
  128: if you want it, that you can change the software or use pieces of it
  129: in new free programs; and that you know you can do these things.
  130: 
  131:   To protect your rights, we need to make restrictions that forbid
  132: anyone to deny you these rights or to ask you to surrender the rights.
  133: These restrictions translate to certain responsibilities for you if you
  134: distribute copies of the software, or if you modify it.
  135: 
  136:   For example, if you distribute copies of such a program, whether
  137: gratis or for a fee, you must give the recipients all the rights that
  138: you have.  You must make sure that they, too, receive or can get the
  139: source code.  And you must show them these terms so they know their
  140: rights.
  141: 
  142:   We protect your rights with two steps: (1) copyright the software, and
  143: (2) offer you this license which gives you legal permission to copy,
  144: distribute and/or modify the software.
  145: 
  146:   Also, for each author's protection and ours, we want to make certain
  147: that everyone understands that there is no warranty for this free
  148: software.  If the software is modified by someone else and passed on, we
  149: want its recipients to know that what they have is not the original, so
  150: that any problems introduced by others will not reflect on the original
  151: authors' reputations.
  152: 
  153:   Finally, any free program is threatened constantly by software
  154: patents.  We wish to avoid the danger that redistributors of a free
  155: program will individually obtain patent licenses, in effect making the
  156: program proprietary.  To prevent this, we have made it clear that any
  157: patent must be licensed for everyone's free use or not licensed at all.
  158: 
  159:   The precise terms and conditions for copying, distribution and
  160: modification follow.
  161: 
  162: @iftex
  163: @unnumberedsec TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
  164: @end iftex
  165: @ifinfo
  166: @center TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
  167: @end ifinfo
  168: 
  169: @enumerate 0
  170: @item
  171: This License applies to any program or other work which contains
  172: a notice placed by the copyright holder saying it may be distributed
  173: under the terms of this General Public License.  The ``Program'', below,
  174: refers to any such program or work, and a ``work based on the Program''
  175: means either the Program or any derivative work under copyright law:
  176: that is to say, a work containing the Program or a portion of it,
  177: either verbatim or with modifications and/or translated into another
  178: language.  (Hereinafter, translation is included without limitation in
  179: the term ``modification''.)  Each licensee is addressed as ``you''.
  180: 
  181: Activities other than copying, distribution and modification are not
  182: covered by this License; they are outside its scope.  The act of
  183: running the Program is not restricted, and the output from the Program
  184: is covered only if its contents constitute a work based on the
  185: Program (independent of having been made by running the Program).
  186: Whether that is true depends on what the Program does.
  187: 
  188: @item
  189: You may copy and distribute verbatim copies of the Program's
  190: source code as you receive it, in any medium, provided that you
  191: conspicuously and appropriately publish on each copy an appropriate
  192: copyright notice and disclaimer of warranty; keep intact all the
  193: notices that refer to this License and to the absence of any warranty;
  194: and give any other recipients of the Program a copy of this License
  195: along with the Program.
  196: 
  197: You may charge a fee for the physical act of transferring a copy, and
  198: you may at your option offer warranty protection in exchange for a fee.
  199: 
  200: @item
  201: You may modify your copy or copies of the Program or any portion
  202: of it, thus forming a work based on the Program, and copy and
  203: distribute such modifications or work under the terms of Section 1
  204: above, provided that you also meet all of these conditions:
  205: 
  206: @enumerate a
  207: @item
  208: You must cause the modified files to carry prominent notices
  209: stating that you changed the files and the date of any change.
  210: 
  211: @item
  212: You must cause any work that you distribute or publish, that in
  213: whole or in part contains or is derived from the Program or any
  214: part thereof, to be licensed as a whole at no charge to all third
  215: parties under the terms of this License.
  216: 
  217: @item
  218: If the modified program normally reads commands interactively
  219: when run, you must cause it, when started running for such
  220: interactive use in the most ordinary way, to print or display an
  221: announcement including an appropriate copyright notice and a
  222: notice that there is no warranty (or else, saying that you provide
  223: a warranty) and that users may redistribute the program under
  224: these conditions, and telling the user how to view a copy of this
  225: License.  (Exception: if the Program itself is interactive but
  226: does not normally print such an announcement, your work based on
  227: the Program is not required to print an announcement.)
  228: @end enumerate
  229: 
  230: These requirements apply to the modified work as a whole.  If
  231: identifiable sections of that work are not derived from the Program,
  232: and can be reasonably considered independent and separate works in
  233: themselves, then this License, and its terms, do not apply to those
  234: sections when you distribute them as separate works.  But when you
  235: distribute the same sections as part of a whole which is a work based
  236: on the Program, the distribution of the whole must be on the terms of
  237: this License, whose permissions for other licensees extend to the
  238: entire whole, and thus to each and every part regardless of who wrote it.
  239: 
  240: Thus, it is not the intent of this section to claim rights or contest
  241: your rights to work written entirely by you; rather, the intent is to
  242: exercise the right to control the distribution of derivative or
  243: collective works based on the Program.
  244: 
  245: In addition, mere aggregation of another work not based on the Program
  246: with the Program (or with a work based on the Program) on a volume of
  247: a storage or distribution medium does not bring the other work under
  248: the scope of this License.
  249: 
  250: @item
  251: You may copy and distribute the Program (or a work based on it,
  252: under Section 2) in object code or executable form under the terms of
  253: Sections 1 and 2 above provided that you also do one of the following:
  254: 
  255: @enumerate a
  256: @item
  257: Accompany it with the complete corresponding machine-readable
  258: source code, which must be distributed under the terms of Sections
  259: 1 and 2 above on a medium customarily used for software interchange; or,
  260: 
  261: @item
  262: Accompany it with a written offer, valid for at least three
  263: years, to give any third party, for a charge no more than your
  264: cost of physically performing source distribution, a complete
  265: machine-readable copy of the corresponding source code, to be
  266: distributed under the terms of Sections 1 and 2 above on a medium
  267: customarily used for software interchange; or,
  268: 
  269: @item
  270: Accompany it with the information you received as to the offer
  271: to distribute corresponding source code.  (This alternative is
  272: allowed only for noncommercial distribution and only if you
  273: received the program in object code or executable form with such
  274: an offer, in accord with Subsection b above.)
  275: @end enumerate
  276: 
  277: The source code for a work means the preferred form of the work for
  278: making modifications to it.  For an executable work, complete source
  279: code means all the source code for all modules it contains, plus any
  280: associated interface definition files, plus the scripts used to
  281: control compilation and installation of the executable.  However, as a
  282: special exception, the source code distributed need not include
  283: anything that is normally distributed (in either source or binary
  284: form) with the major components (compiler, kernel, and so on) of the
  285: operating system on which the executable runs, unless that component
  286: itself accompanies the executable.
  287: 
  288: If distribution of executable or object code is made by offering
  289: access to copy from a designated place, then offering equivalent
  290: access to copy the source code from the same place counts as
  291: distribution of the source code, even though third parties are not
  292: compelled to copy the source along with the object code.
  293: 
  294: @item
  295: You may not copy, modify, sublicense, or distribute the Program
  296: except as expressly provided under this License.  Any attempt
  297: otherwise to copy, modify, sublicense or distribute the Program is
  298: void, and will automatically terminate your rights under this License.
  299: However, parties who have received copies, or rights, from you under
  300: this License will not have their licenses terminated so long as such
  301: parties remain in full compliance.
  302: 
  303: @item
  304: You are not required to accept this License, since you have not
  305: signed it.  However, nothing else grants you permission to modify or
  306: distribute the Program or its derivative works.  These actions are
  307: prohibited by law if you do not accept this License.  Therefore, by
  308: modifying or distributing the Program (or any work based on the
  309: Program), you indicate your acceptance of this License to do so, and
  310: all its terms and conditions for copying, distributing or modifying
  311: the Program or works based on it.
  312: 
  313: @item
  314: Each time you redistribute the Program (or any work based on the
  315: Program), the recipient automatically receives a license from the
  316: original licensor to copy, distribute or modify the Program subject to
  317: these terms and conditions.  You may not impose any further
  318: restrictions on the recipients' exercise of the rights granted herein.
  319: You are not responsible for enforcing compliance by third parties to
  320: this License.
  321: 
  322: @item
  323: If, as a consequence of a court judgment or allegation of patent
  324: infringement or for any other reason (not limited to patent issues),
  325: conditions are imposed on you (whether by court order, agreement or
  326: otherwise) that contradict the conditions of this License, they do not
  327: excuse you from the conditions of this License.  If you cannot
  328: distribute so as to satisfy simultaneously your obligations under this
  329: License and any other pertinent obligations, then as a consequence you
  330: may not distribute the Program at all.  For example, if a patent
  331: license would not permit royalty-free redistribution of the Program by
  332: all those who receive copies directly or indirectly through you, then
  333: the only way you could satisfy both it and this License would be to
  334: refrain entirely from distribution of the Program.
  335: 
  336: If any portion of this section is held invalid or unenforceable under
  337: any particular circumstance, the balance of the section is intended to
  338: apply and the section as a whole is intended to apply in other
  339: circumstances.
  340: 
  341: It is not the purpose of this section to induce you to infringe any
  342: patents or other property right claims or to contest validity of any
  343: such claims; this section has the sole purpose of protecting the
  344: integrity of the free software distribution system, which is
  345: implemented by public license practices.  Many people have made
  346: generous contributions to the wide range of software distributed
  347: through that system in reliance on consistent application of that
  348: system; it is up to the author/donor to decide if he or she is willing
  349: to distribute software through any other system and a licensee cannot
  350: impose that choice.
  351: 
  352: This section is intended to make thoroughly clear what is believed to
  353: be a consequence of the rest of this License.
  354: 
  355: @item
  356: If the distribution and/or use of the Program is restricted in
  357: certain countries either by patents or by copyrighted interfaces, the
  358: original copyright holder who places the Program under this License
  359: may add an explicit geographical distribution limitation excluding
  360: those countries, so that distribution is permitted only in or among
  361: countries not thus excluded.  In such case, this License incorporates
  362: the limitation as if written in the body of this License.
  363: 
  364: @item
  365: The Free Software Foundation may publish revised and/or new versions
  366: of the General Public License from time to time.  Such new versions will
  367: be similar in spirit to the present version, but may differ in detail to
  368: address new problems or concerns.
  369: 
  370: Each version is given a distinguishing version number.  If the Program
  371: specifies a version number of this License which applies to it and ``any
  372: later version'', you have the option of following the terms and conditions
  373: either of that version or of any later version published by the Free
  374: Software Foundation.  If the Program does not specify a version number of
  375: this License, you may choose any version ever published by the Free Software
  376: Foundation.
  377: 
  378: @item
  379: If you wish to incorporate parts of the Program into other free
  380: programs whose distribution conditions are different, write to the author
  381: to ask for permission.  For software which is copyrighted by the Free
  382: Software Foundation, write to the Free Software Foundation; we sometimes
  383: make exceptions for this.  Our decision will be guided by the two goals
  384: of preserving the free status of all derivatives of our free software and
  385: of promoting the sharing and reuse of software generally.
  386: 
  387: @iftex
  388: @heading NO WARRANTY
  389: @end iftex
  390: @ifinfo
  391: @center NO WARRANTY
  392: @end ifinfo
  393: 
  394: @item
  395: BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
  396: FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
  397: OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
  398: PROVIDE THE PROGRAM ``AS IS'' WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
  399: OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  400: MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
  401: TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
  402: PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
  403: REPAIR OR CORRECTION.
  404: 
  405: @item
  406: IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
  407: WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
  408: REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
  409: INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
  410: OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
  411: TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
  412: YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
  413: PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
  414: POSSIBILITY OF SUCH DAMAGES.
  415: @end enumerate
  416: 
  417: @iftex
  418: @heading END OF TERMS AND CONDITIONS
  419: @end iftex
  420: @ifinfo
  421: @center END OF TERMS AND CONDITIONS
  422: @end ifinfo
  423: 
  424: @page
  425: @unnumberedsec How to Apply These Terms to Your New Programs
  426: 
  427:   If you develop a new program, and you want it to be of the greatest
  428: possible use to the public, the best way to achieve this is to make it
  429: free software which everyone can redistribute and change under these terms.
  430: 
  431:   To do so, attach the following notices to the program.  It is safest
  432: to attach them to the start of each source file to most effectively
  433: convey the exclusion of warranty; and each file should have at least
  434: the ``copyright'' line and a pointer to where the full notice is found.
  435: 
  436: @smallexample
  437: @var{one line to give the program's name and a brief idea of what it does.}
  438: Copyright (C) 19@var{yy}  @var{name of author}
  439: 
  440: This program is free software; you can redistribute it and/or modify 
  441: it under the terms of the GNU General Public License as published by 
  442: the Free Software Foundation; either version 2 of the License, or 
  443: (at your option) any later version.
  444: 
  445: This program is distributed in the hope that it will be useful,
  446: but WITHOUT ANY WARRANTY; without even the implied warranty of
  447: MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  448: GNU General Public License for more details.
  449: 
  450: You should have received a copy of the GNU General Public License
  451: along with this program; if not, write to the Free Software
  452: Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  453: @end smallexample
  454: 
  455: Also add information on how to contact you by electronic and paper mail.
  456: 
  457: If the program is interactive, make it output a short notice like this
  458: when it starts in an interactive mode:
  459: 
  460: @smallexample
  461: Gnomovision version 69, Copyright (C) 19@var{yy} @var{name of author}
  462: Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
  463: type `show w'.  
  464: This is free software, and you are welcome to redistribute it 
  465: under certain conditions; type `show c' for details.
  466: @end smallexample
  467: 
  468: The hypothetical commands @samp{show w} and @samp{show c} should show
  469: the appropriate parts of the General Public License.  Of course, the
  470: commands you use may be called something other than @samp{show w} and
  471: @samp{show c}; they could even be mouse-clicks or menu items---whatever
  472: suits your program.
  473: 
  474: You should also get your employer (if you work as a programmer) or your
  475: school, if any, to sign a ``copyright disclaimer'' for the program, if
  476: necessary.  Here is a sample; alter the names:
  477: 
  478: @smallexample
  479: Yoyodyne, Inc., hereby disclaims all copyright interest in the program
  480: `Gnomovision' (which makes passes at compilers) written by James Hacker.
  481: 
  482: @var{signature of Ty Coon}, 1 April 1989
  483: Ty Coon, President of Vice
  484: @end smallexample
  485: 
  486: This General Public License does not permit incorporating your program into
  487: proprietary programs.  If your program is a subroutine library, you may
  488: consider it more useful to permit linking proprietary applications with the
  489: library.  If this is what you want to do, use the GNU Library General
  490: Public License instead of this License.
  491: 
  492: @iftex
  493: @node    Preface
  494: @comment node-name,     next,           previous, up
  495: @unnumbered Preface
  496: @cindex Preface
  497: This manual documents Gforth. The reader is expected to know
  498: Forth. This manual is primarily a reference manual. @xref{Other Books}
  499: for introductory material.
  500: @end iftex
  501: 
  502: @node    Goals, Other Books, License, Top
  503: @comment node-name,     next,           previous, up
  504: @chapter Goals of Gforth
  505: @cindex Goals
  506: The goal of the Gforth Project is to develop a standard model for
  507: ANSI Forth. This can be split into several subgoals:
  508: 
  509: @itemize @bullet
  510: @item
  511: Gforth should conform to the ANSI Forth standard.
  512: @item
  513: It should be a model, i.e. it should define all the
  514: implementation-dependent things.
  515: @item
  516: It should become standard, i.e. widely accepted and used. This goal
  517: is the most difficult one.
  518: @end itemize
  519: 
  520: To achieve these goals Gforth should be
  521: @itemize @bullet
  522: @item
  523: Similar to previous models (fig-Forth, F83)
  524: @item
  525: Powerful. It should provide for all the things that are considered
  526: necessary today and even some that are not yet considered necessary.
  527: @item
  528: Efficient. It should not get the reputation of being exceptionally
  529: slow.
  530: @item
  531: Free.
  532: @item
  533: Available on many machines/easy to port.
  534: @end itemize
  535: 
  536: Have we achieved these goals? Gforth conforms to the ANS Forth
  537: standard. It may be considered a model, but we have not yet documented
  538: which parts of the model are stable and which parts we are likely to
  539: change. It certainly has not yet become a de facto standard. It has some
  540: similarities and some differences to previous models. It has some
  541: powerful features, but not yet everything that we envisioned. We
  542: certainly have achieved our execution speed goals (@pxref{Performance}).
  543: It is free and available on many machines.
  544: 
  545: @node Other Books, Invocation, Goals, Top
  546: @chapter Other books on ANS Forth
  547: 
  548: As the standard is relatively new, there are not many books out yet. It
  549: is not recommended to learn Forth by using Gforth and a book that is
  550: not written for ANS Forth, as you will not know your mistakes from the
  551: deviations of the book.
  552: 
  553: There is, of course, the standard, the definite reference if you want to
  554: write ANS Forth programs. It is available in printed form from the
  555: National Standards Institute Sales Department (Tel.: USA (212) 642-4900;
  556: Fax.: USA (212) 302-1286) as document @cite{X3.215-1994} for about $200. You
  557: can also get it from Global Engineering Documents (Tel.: USA (800)
  558: 854-7179; Fax.: (303) 843-9880) for about $300.
  559: 
  560: @cite{dpANS6}, the last draft of the standard, which was then submitted to ANSI
  561: for publication is available electronically and for free in some MS Word
  562: format, and it has been converted to HTML. Some pointers to these
  563: versions can be found through
  564: @*@file{http://www.complang.tuwien.ac.at/projects/forth.html}.
  565: 
  566: @cite{Forth: The new model} by Jack Woehr (Prentice-Hall, 1993) is an
  567: introductory book based on a draft version of the standard. It does not
  568: cover the whole standard. It also contains interesting background
  569: information (Jack Woehr was in the ANS Forth Technical Committe). It is
  570: not appropriate for complete newbies, but programmers experienced in
  571: other languages should find it ok.
  572: 
  573: @node Invocation, Words, Other Books, Top
  574: @chapter Invocation
  575: 
  576: You will usually just say @code{gforth}. In many other cases the default
  577: Gforth image will be invoked like this:
  578: 
  579: @example
  580: gforth [files] [-e forth-code]
  581: @end example
  582: 
  583: executing the contents of the files and the Forth code in the order they
  584: are given.
  585: 
  586: In general, the command line looks like this:
  587: 
  588: @example
  589: gforth [initialization options] [image-specific options]
  590: @end example
  591: 
  592: The initialization options must come before the rest of the command
  593: line. They are:
  594: 
  595: @table @code
  596: @item --image-file @var{file}
  597: @item -i @var{file}
  598: Loads the Forth image @var{file} instead of the default
  599: @file{gforth.fi}.
  600: 
  601: @item --path @var{path}
  602: @item -p @var{path}
  603: Uses @var{path} for searching the image file and Forth source code
  604: files instead of the default in the environment variable
  605: @code{GFORTHPATH} or the path specified at installation time (typically
  606: @file{/usr/local/lib/gforth:.}). A path is given as a @code{:}-separated
  607: list.
  608: 
  609: @item --dictionary-size @var{size}
  610: @item -m @var{size}
  611: Allocate @var{size} space for the Forth dictionary space instead of
  612: using the default specified in the image (typically 256K). The
  613: @var{size} specification consists of an integer and a unit (e.g.,
  614: @code{4M}). The unit can be one of @code{b} (bytes), @code{e} (element
  615: size, in this case Cells), @code{k} (kilobytes), and @code{M}
  616: (Megabytes). If no unit is specified, @code{e} is used.
  617: 
  618: @item --data-stack-size @var{size}
  619: @item -d @var{size}
  620: Allocate @var{size} space for the data stack instead of using the
  621: default specified in the image (typically 16K).
  622: 
  623: @item --return-stack-size @var{size}
  624: @item -r @var{size}
  625: Allocate @var{size} space for the return stack instead of using the
  626: default specified in the image (typically 16K).
  627: 
  628: @item --fp-stack-size @var{size}
  629: @item -f @var{size}
  630: Allocate @var{size} space for the floating point stack instead of
  631: using the default specified in the image (typically 16K). In this case
  632: the unit specifier @code{e} refers to floating point numbers.
  633: 
  634: @item --locals-stack-size @var{size}
  635: @item -l @var{size}
  636: Allocate @var{size} space for the locals stack instead of using the
  637: default specified in the image (typically 16K).
  638: 
  639: @end table
  640: 
  641: As explained above, the image-specific command-line arguments for the
  642: default image @file{gforth.fi} consist of a sequence of filenames and
  643: @code{-e @var{forth-code}} options that are interpreted in the seqence
  644: in which they are given. The @code{-e @var{forth-code}} or
  645: @code{--evaluate @var{forth-code}} option evaluates the forth
  646: code. This option takes only one argument; if you want to evaluate more
  647: Forth words, you have to quote them or use several @code{-e}s. To exit
  648: after processing the command line (instead of entering interactive mode)
  649: append @code{-e bye} to the command line.
  650: 
  651: If you have several versions of Gforth installed, @code{gforth} will
  652: invoke the version that was installed last. @code{gforth-@var{version}}
  653: invokes a specific version. You may want to use the option
  654: @code{--path}, if your environment contains the variable
  655: @code{GFORTHPATH}.
  656: 
  657: Not yet implemented:
  658: On startup the system first executes the system initialization file
  659: (unless the option @code{--no-init-file} is given; note that the system
  660: resulting from using this option may not be ANS Forth conformant). Then
  661: the user initialization file @file{.gforth.fs} is executed, unless the
  662: option @code{--no-rc} is given; this file is first searched in @file{.},
  663: then in @file{~}, then in the normal path (see above).
  664: 
  665: @node Words, ANS conformance, Invocation, Top
  666: @chapter Forth Words
  667: 
  668: @menu
  669: * Notation::                    
  670: * Arithmetic::                  
  671: * Stack Manipulation::          
  672: * Memory access::               
  673: * Control Structures::          
  674: * Locals::                      
  675: * Defining Words::              
  676: * Tokens for Words::            
  677: * Wordlists::                   
  678: * Files::                       
  679: * Blocks::                      
  680: * Other I/O::                   
  681: * Programming Tools::           
  682: * Assembler and Code words::    
  683: * Threading Words::             
  684: @end menu
  685: 
  686: @node Notation, Arithmetic, Words, Words
  687: @section Notation
  688: 
  689: The Forth words are described in this section in the glossary notation
  690: that has become a de-facto standard for Forth texts, i.e.
  691: 
  692: @format
  693: @var{word}     @var{Stack effect}   @var{wordset}   @var{pronunciation}
  694: @end format
  695: @var{Description}
  696: 
  697: @table @var
  698: @item word
  699: The name of the word. BTW, Gforth is case insensitive, so you can
  700: type the words in in lower case (However, @pxref{core-idef}).
  701: 
  702: @item Stack effect
  703: The stack effect is written in the notation @code{@var{before} --
  704: @var{after}}, where @var{before} and @var{after} describe the top of
  705: stack entries before and after the execution of the word. The rest of
  706: the stack is not touched by the word. The top of stack is rightmost,
  707: i.e., a stack sequence is written as it is typed in. Note that Gforth
  708: uses a separate floating point stack, but a unified stack
  709: notation. Also, return stack effects are not shown in @var{stack
  710: effect}, but in @var{Description}. The name of a stack item describes
  711: the type and/or the function of the item. See below for a discussion of
  712: the types.
  713: 
  714: All words have two stack effects: A compile-time stack effect and a
  715: run-time stack effect. The compile-time stack-effect of most words is
  716: @var{ -- }. If the compile-time stack-effect of a word deviates from
  717: this standard behaviour, or the word does other unusual things at
  718: compile time, both stack effects are shown; otherwise only the run-time
  719: stack effect is shown.
  720: 
  721: @item pronunciation
  722: How the word is pronounced
  723: 
  724: @item wordset
  725: The ANS Forth standard is divided into several wordsets. A standard
  726: system need not support all of them. So, the fewer wordsets your program
  727: uses the more portable it will be in theory. However, we suspect that
  728: most ANS Forth systems on personal machines will feature all
  729: wordsets. Words that are not defined in the ANS standard have
  730: @code{gforth} or @code{gforth-internal} as wordset. @code{gforth}
  731: describes words that will work in future releases of Gforth;
  732: @code{gforth-internal} words are more volatile. Environmental query
  733: strings are also displayed like words; you can recognize them by the
  734: @code{environment} in the wordset field.
  735: 
  736: @item Description
  737: A description of the behaviour of the word.
  738: @end table
  739: 
  740: The type of a stack item is specified by the character(s) the name
  741: starts with:
  742: 
  743: @table @code
  744: @item f
  745: Bool, i.e. @code{false} or @code{true}.
  746: @item c
  747: Char
  748: @item w
  749: Cell, can contain an integer or an address
  750: @item n
  751: signed integer
  752: @item u
  753: unsigned integer
  754: @item d
  755: double sized signed integer
  756: @item ud
  757: double sized unsigned integer
  758: @item r
  759: Float (on the FP stack)
  760: @item a_
  761: Cell-aligned address
  762: @item c_
  763: Char-aligned address (note that a Char may have two bytes in Windows NT)
  764: @item f_
  765: Float-aligned address
  766: @item df_
  767: Address aligned for IEEE double precision float
  768: @item sf_
  769: Address aligned for IEEE single precision float
  770: @item xt
  771: Execution token, same size as Cell
  772: @item wid
  773: Wordlist ID, same size as Cell
  774: @item f83name
  775: Pointer to a name structure
  776: @item "
  777: string in the input stream (not the stack). The terminating character is
  778: a blank by default. If it is not a blank, it is shown in @code{<>}
  779: quotes.
  780: 
  781: @end table
  782: 
  783: @node Arithmetic, Stack Manipulation, Notation, Words
  784: @section Arithmetic
  785: Forth arithmetic is not checked, i.e., you will not hear about integer
  786: overflow on addition or multiplication, you may hear about division by
  787: zero if you are lucky. The operator is written after the operands, but
  788: the operands are still in the original order. I.e., the infix @code{2-1}
  789: corresponds to @code{2 1 -}. Forth offers a variety of division
  790: operators. If you perform division with potentially negative operands,
  791: you do not want to use @code{/} or @code{/mod} with its undefined
  792: behaviour, but rather @code{fm/mod} or @code{sm/mod} (probably the
  793: former, @pxref{Mixed precision}).
  794: 
  795: @menu
  796: * Single precision::            
  797: * Bitwise operations::          
  798: * Mixed precision::             operations with single and double-cell integers
  799: * Double precision::            Double-cell integer arithmetic
  800: * Floating Point::              
  801: @end menu
  802: 
  803: @node Single precision, Bitwise operations, Arithmetic, Arithmetic
  804: @subsection Single precision
  805: doc-+
  806: doc--
  807: doc-*
  808: doc-/
  809: doc-mod
  810: doc-/mod
  811: doc-negate
  812: doc-abs
  813: doc-min
  814: doc-max
  815: 
  816: @node Bitwise operations, Mixed precision, Single precision, Arithmetic
  817: @subsection Bitwise operations
  818: doc-and
  819: doc-or
  820: doc-xor
  821: doc-invert
  822: doc-2*
  823: doc-2/
  824: 
  825: @node Mixed precision, Double precision, Bitwise operations, Arithmetic
  826: @subsection Mixed precision
  827: doc-m+
  828: doc-*/
  829: doc-*/mod
  830: doc-m*
  831: doc-um*
  832: doc-m*/
  833: doc-um/mod
  834: doc-fm/mod
  835: doc-sm/rem
  836: 
  837: @node Double precision, Floating Point, Mixed precision, Arithmetic
  838: @subsection Double precision
  839: 
  840: The outer (aka text) interpreter converts numbers containing a dot into
  841: a double precision number. Note that only numbers with the dot as last
  842: character are standard-conforming.
  843: 
  844: doc-d+
  845: doc-d-
  846: doc-dnegate
  847: doc-dabs
  848: doc-dmin
  849: doc-dmax
  850: 
  851: @node Floating Point,  , Double precision, Arithmetic
  852: @subsection Floating Point
  853: 
  854: The format of floating point numbers recognized by the outer (aka text)
  855: interpreter is: a signed decimal number, possibly containing a decimal
  856: point (@code{.}), followed by @code{E} or @code{e}, optionally followed
  857: by a signed integer (the exponent). E.g., @code{1e} ist the same as
  858: @code{+1.0e+0}. Note that a number without @code{e}
  859: is not interpreted as floating-point number, but as double (if the
  860: number contains a @code{.}) or single precision integer. Also,
  861: conversions between string and floating point numbers always use base
  862: 10, irrespective of the value of @code{BASE}. If @code{BASE} contains a
  863: value greater then 14, the @code{E} may be interpreted as digit and the
  864: number will be interpreted as integer, unless it has a signed exponent
  865: (both @code{+} and @code{-} are allowed as signs).
  866: 
  867: Angles in floating point operations are given in radians (a full circle
  868: has 2 pi radians). Note, that Gforth has a separate floating point
  869: stack, but we use the unified notation.
  870: 
  871: Floating point numbers have a number of unpleasant surprises for the
  872: unwary (e.g., floating point addition is not associative) and even a few
  873: for the wary. You should not use them unless you know what you are doing
  874: or you don't care that the results you get are totally bogus. If you
  875: want to learn about the problems of floating point numbers (and how to
  876: avoid them), you might start with @cite{David Goldberg, What Every
  877: Computer Scientist Should Know About Floating-Point Arithmetic, ACM
  878: Computing Surveys 23(1):5@minus{}48, March 1991}.
  879: 
  880: doc-f+
  881: doc-f-
  882: doc-f*
  883: doc-f/
  884: doc-fnegate
  885: doc-fabs
  886: doc-fmax
  887: doc-fmin
  888: doc-floor
  889: doc-fround
  890: doc-f**
  891: doc-fsqrt
  892: doc-fexp
  893: doc-fexpm1
  894: doc-fln
  895: doc-flnp1
  896: doc-flog
  897: doc-falog
  898: doc-fsin
  899: doc-fcos
  900: doc-fsincos
  901: doc-ftan
  902: doc-fasin
  903: doc-facos
  904: doc-fatan
  905: doc-fatan2
  906: doc-fsinh
  907: doc-fcosh
  908: doc-ftanh
  909: doc-fasinh
  910: doc-facosh
  911: doc-fatanh
  912: 
  913: @node Stack Manipulation, Memory access, Arithmetic, Words
  914: @section Stack Manipulation
  915: 
  916: Gforth has a data stack (aka parameter stack) for characters, cells,
  917: addresses, and double cells, a floating point stack for floating point
  918: numbers, a return stack for storing the return addresses of colon
  919: definitions and other data, and a locals stack for storing local
  920: variables. Note that while every sane Forth has a separate floating
  921: point stack, this is not strictly required; an ANS Forth system could
  922: theoretically keep floating point numbers on the data stack. As an
  923: additional difficulty, you don't know how many cells a floating point
  924: number takes. It is reportedly possible to write words in a way that
  925: they work also for a unified stack model, but we do not recommend trying
  926: it. Instead, just say that your program has an environmental dependency
  927: on a separate FP stack.
  928: 
  929: Also, a Forth system is allowed to keep the local variables on the
  930: return stack. This is reasonable, as local variables usually eliminate
  931: the need to use the return stack explicitly. So, if you want to produce
  932: a standard complying program and if you are using local variables in a
  933: word, forget about return stack manipulations in that word (see the
  934: standard document for the exact rules).
  935: 
  936: @menu
  937: * Data stack::                  
  938: * Floating point stack::        
  939: * Return stack::                
  940: * Locals stack::                
  941: * Stack pointer manipulation::  
  942: @end menu
  943: 
  944: @node Data stack, Floating point stack, Stack Manipulation, Stack Manipulation
  945: @subsection Data stack
  946: doc-drop
  947: doc-nip
  948: doc-dup
  949: doc-over
  950: doc-tuck
  951: doc-swap
  952: doc-rot
  953: doc--rot
  954: doc-?dup
  955: doc-pick
  956: doc-roll
  957: doc-2drop
  958: doc-2nip
  959: doc-2dup
  960: doc-2over
  961: doc-2tuck
  962: doc-2swap
  963: doc-2rot
  964: 
  965: @node Floating point stack, Return stack, Data stack, Stack Manipulation
  966: @subsection Floating point stack
  967: doc-fdrop
  968: doc-fnip
  969: doc-fdup
  970: doc-fover
  971: doc-ftuck
  972: doc-fswap
  973: doc-frot
  974: 
  975: @node Return stack, Locals stack, Floating point stack, Stack Manipulation
  976: @subsection Return stack
  977: doc->r
  978: doc-r>
  979: doc-r@
  980: doc-rdrop
  981: doc-2>r
  982: doc-2r>
  983: doc-2r@
  984: doc-2rdrop
  985: 
  986: @node Locals stack, Stack pointer manipulation, Return stack, Stack Manipulation
  987: @subsection Locals stack
  988: 
  989: @node Stack pointer manipulation,  , Locals stack, Stack Manipulation
  990: @subsection Stack pointer manipulation
  991: doc-sp@
  992: doc-sp!
  993: doc-fp@
  994: doc-fp!
  995: doc-rp@
  996: doc-rp!
  997: doc-lp@
  998: doc-lp!
  999: 
 1000: @node Memory access, Control Structures, Stack Manipulation, Words
 1001: @section Memory access
 1002: 
 1003: @menu
 1004: * Stack-Memory transfers::      
 1005: * Address arithmetic::          
 1006: * Memory block access::         
 1007: @end menu
 1008: 
 1009: @node Stack-Memory transfers, Address arithmetic, Memory access, Memory access
 1010: @subsection Stack-Memory transfers
 1011: 
 1012: doc-@
 1013: doc-!
 1014: doc-+!
 1015: doc-c@
 1016: doc-c!
 1017: doc-2@
 1018: doc-2!
 1019: doc-f@
 1020: doc-f!
 1021: doc-sf@
 1022: doc-sf!
 1023: doc-df@
 1024: doc-df!
 1025: 
 1026: @node Address arithmetic, Memory block access, Stack-Memory transfers, Memory access
 1027: @subsection Address arithmetic
 1028: 
 1029: ANS Forth does not specify the sizes of the data types. Instead, it
 1030: offers a number of words for computing sizes and doing address
 1031: arithmetic. Basically, address arithmetic is performed in terms of
 1032: address units (aus); on most systems the address unit is one byte. Note
 1033: that a character may have more than one au, so @code{chars} is no noop
 1034: (on systems where it is a noop, it compiles to nothing).
 1035: 
 1036: ANS Forth also defines words for aligning addresses for specific
 1037: addresses. Many computers require that accesses to specific data types
 1038: must only occur at specific addresses; e.g., that cells may only be
 1039: accessed at addresses divisible by 4. Even if a machine allows unaligned
 1040: accesses, it can usually perform aligned accesses faster. 
 1041: 
 1042: For the performance-conscious: alignment operations are usually only
 1043: necessary during the definition of a data structure, not during the
 1044: (more frequent) accesses to it.
 1045: 
 1046: ANS Forth defines no words for character-aligning addresses. This is not
 1047: an oversight, but reflects the fact that addresses that are not
 1048: char-aligned have no use in the standard and therefore will not be
 1049: created.
 1050: 
 1051: The standard guarantees that addresses returned by @code{CREATE}d words
 1052: are cell-aligned; in addition, Gforth guarantees that these addresses
 1053: are aligned for all purposes.
 1054: 
 1055: Note that the standard defines a word @code{char}, which has nothing to
 1056: do with address arithmetic.
 1057: 
 1058: doc-chars
 1059: doc-char+
 1060: doc-cells
 1061: doc-cell+
 1062: doc-align
 1063: doc-aligned
 1064: doc-floats
 1065: doc-float+
 1066: doc-falign
 1067: doc-faligned
 1068: doc-sfloats
 1069: doc-sfloat+
 1070: doc-sfalign
 1071: doc-sfaligned
 1072: doc-dfloats
 1073: doc-dfloat+
 1074: doc-dfalign
 1075: doc-dfaligned
 1076: doc-maxalign
 1077: doc-maxaligned
 1078: doc-cfalign
 1079: doc-cfaligned
 1080: doc-address-unit-bits
 1081: 
 1082: @node Memory block access,  , Address arithmetic, Memory access
 1083: @subsection Memory block access
 1084: 
 1085: doc-move
 1086: doc-erase
 1087: 
 1088: While the previous words work on address units, the rest works on
 1089: characters.
 1090: 
 1091: doc-cmove
 1092: doc-cmove>
 1093: doc-fill
 1094: doc-blank
 1095: 
 1096: @node Control Structures, Locals, Memory access, Words
 1097: @section Control Structures
 1098: 
 1099: Control structures in Forth cannot be used in interpret state, only in
 1100: compile state, i.e., in a colon definition. We do not like this
 1101: limitation, but have not seen a satisfying way around it yet, although
 1102: many schemes have been proposed.
 1103: 
 1104: @menu
 1105: * Selection::                   
 1106: * Simple Loops::                
 1107: * Counted Loops::               
 1108: * Arbitrary control structures::  
 1109: * Calls and returns::           
 1110: * Exception Handling::          
 1111: @end menu
 1112: 
 1113: @node Selection, Simple Loops, Control Structures, Control Structures
 1114: @subsection Selection
 1115: 
 1116: @example
 1117: @var{flag}
 1118: IF
 1119:   @var{code}
 1120: ENDIF
 1121: @end example
 1122: or
 1123: @example
 1124: @var{flag}
 1125: IF
 1126:   @var{code1}
 1127: ELSE
 1128:   @var{code2}
 1129: ENDIF
 1130: @end example
 1131: 
 1132: You can use @code{THEN} instead of @code{ENDIF}. Indeed, @code{THEN} is
 1133: standard, and @code{ENDIF} is not, although it is quite popular. We
 1134: recommend using @code{ENDIF}, because it is less confusing for people
 1135: who also know other languages (and is not prone to reinforcing negative
 1136: prejudices against Forth in these people). Adding @code{ENDIF} to a
 1137: system that only supplies @code{THEN} is simple:
 1138: @example
 1139: : endif   POSTPONE then ; immediate
 1140: @end example
 1141: 
 1142: [According to @cite{Webster's New Encyclopedic Dictionary}, @dfn{then
 1143: (adv.)}  has the following meanings:
 1144: @quotation
 1145: ... 2b: following next after in order ... 3d: as a necessary consequence
 1146: (if you were there, then you saw them).
 1147: @end quotation
 1148: Forth's @code{THEN} has the meaning 2b, whereas @code{THEN} in Pascal
 1149: and many other programming languages has the meaning 3d.]
 1150: 
 1151: Gforth also provides the words @code{?dup-if} and @code{?dup-0=-if}, so
 1152: you can avoid using @code{?dup}. Using these alternatives is also more
 1153: efficient than using @code{?dup}. Definitions in plain standard Forth
 1154: for @code{ENDIF}, @code{?DUP-IF} and @code{?DUP-0=-IF} are provided in
 1155: @file{compat/control.fs}.
 1156: 
 1157: @example
 1158: @var{n}
 1159: CASE
 1160:   @var{n1} OF @var{code1} ENDOF
 1161:   @var{n2} OF @var{code2} ENDOF
 1162:   @dots{}
 1163: ENDCASE
 1164: @end example
 1165: 
 1166: Executes the first @var{codei}, where the @var{ni} is equal to
 1167: @var{n}. A default case can be added by simply writing the code after
 1168: the last @code{ENDOF}. It may use @var{n}, which is on top of the stack,
 1169: but must not consume it.
 1170: 
 1171: @node Simple Loops, Counted Loops, Selection, Control Structures
 1172: @subsection Simple Loops
 1173: 
 1174: @example
 1175: BEGIN
 1176:   @var{code1}
 1177:   @var{flag}
 1178: WHILE
 1179:   @var{code2}
 1180: REPEAT
 1181: @end example
 1182: 
 1183: @var{code1} is executed and @var{flag} is computed. If it is true,
 1184: @var{code2} is executed and the loop is restarted; If @var{flag} is false, execution continues after the @code{REPEAT}.
 1185: 
 1186: @example
 1187: BEGIN
 1188:   @var{code}
 1189:   @var{flag}
 1190: UNTIL
 1191: @end example
 1192: 
 1193: @var{code} is executed. The loop is restarted if @code{flag} is false.
 1194: 
 1195: @example
 1196: BEGIN
 1197:   @var{code}
 1198: AGAIN
 1199: @end example
 1200: 
 1201: This is an endless loop.
 1202: 
 1203: @node Counted Loops, Arbitrary control structures, Simple Loops, Control Structures
 1204: @subsection Counted Loops
 1205: 
 1206: The basic counted loop is:
 1207: @example
 1208: @var{limit} @var{start}
 1209: ?DO
 1210:   @var{body}
 1211: LOOP
 1212: @end example
 1213: 
 1214: This performs one iteration for every integer, starting from @var{start}
 1215: and up to, but excluding @var{limit}. The counter, aka index, can be
 1216: accessed with @code{i}. E.g., the loop
 1217: @example
 1218: 10 0 ?DO
 1219:   i .
 1220: LOOP
 1221: @end example
 1222: prints
 1223: @example
 1224: 0 1 2 3 4 5 6 7 8 9
 1225: @end example
 1226: The index of the innermost loop can be accessed with @code{i}, the index
 1227: of the next loop with @code{j}, and the index of the third loop with
 1228: @code{k}.
 1229: 
 1230: The loop control data are kept on the return stack, so there are some
 1231: restrictions on mixing return stack accesses and counted loop
 1232: words. E.g., if you put values on the return stack outside the loop, you
 1233: cannot read them inside the loop. If you put values on the return stack
 1234: within a loop, you have to remove them before the end of the loop and
 1235: before accessing the index of the loop.
 1236: 
 1237: There are several variations on the counted loop:
 1238: 
 1239: @code{LEAVE} leaves the innermost counted loop immediately.
 1240: 
 1241: If @var{start} is greater than @var{limit}, a @code{?DO} loop is entered
 1242: (and @code{LOOP} iterates until they become equal by wrap-around
 1243: arithmetic). This behaviour is usually not what you want. Therefore,
 1244: Gforth offers @code{+DO} and @code{U+DO} (as replacements for
 1245: @code{?DO}), which do not enter the loop if @var{start} is greater than
 1246: @var{limit}; @code{+DO} is for signed loop parameters, @code{U+DO} for
 1247: unsigned loop parameters.
 1248: 
 1249: @code{LOOP} can be replaced with @code{@var{n} +LOOP}; this updates the
 1250: index by @var{n} instead of by 1. The loop is terminated when the border
 1251: between @var{limit-1} and @var{limit} is crossed. E.g.:
 1252: 
 1253: @code{4 0 +DO  i .  2 +LOOP}   prints @code{0 2}
 1254: 
 1255: @code{4 1 +DO  i .  2 +LOOP}   prints @code{1 3}
 1256: 
 1257: The behaviour of @code{@var{n} +LOOP} is peculiar when @var{n} is negative:
 1258: 
 1259: @code{-1 0 ?DO  i .  -1 +LOOP}  prints @code{0 -1}
 1260: 
 1261: @code{ 0 0 ?DO  i .  -1 +LOOP}  prints nothing
 1262: 
 1263: Therefore we recommend avoiding @code{@var{n} +LOOP} with negative
 1264: @var{n}. One alternative is @code{@var{u} -LOOP}, which reduces the
 1265: index by @var{u} each iteration. The loop is terminated when the border
 1266: between @var{limit+1} and @var{limit} is crossed. Gforth also provides
 1267: @code{-DO} and @code{U-DO} for down-counting loops. E.g.:
 1268: 
 1269: @code{-2 0 -DO  i .  1 -LOOP}  prints @code{0 -1}
 1270: 
 1271: @code{-1 0 -DO  i .  1 -LOOP}  prints @code{0}
 1272: 
 1273: @code{ 0 0 -DO  i .  1 -LOOP}  prints nothing
 1274: 
 1275: Unfortunately, @code{+DO}, @code{U+DO}, @code{-DO}, @code{U-DO} and
 1276: @code{-LOOP} are not in the ANS Forth standard. However, an
 1277: implementation for these words that uses only standard words is provided
 1278: in @file{compat/loops.fs}.
 1279: 
 1280: @code{?DO} can also be replaced by @code{DO}. @code{DO} always enters
 1281: the loop, independent of the loop parameters. Do not use @code{DO}, even
 1282: if you know that the loop is entered in any case. Such knowledge tends
 1283: to become invalid during maintenance of a program, and then the
 1284: @code{DO} will make trouble.
 1285: 
 1286: @code{UNLOOP} is used to prepare for an abnormal loop exit, e.g., via
 1287: @code{EXIT}. @code{UNLOOP} removes the loop control parameters from the
 1288: return stack so @code{EXIT} can get to its return address.
 1289: 
 1290: Another counted loop is
 1291: @example
 1292: @var{n}
 1293: FOR
 1294:   @var{body}
 1295: NEXT
 1296: @end example
 1297: This is the preferred loop of native code compiler writers who are too
 1298: lazy to optimize @code{?DO} loops properly. In Gforth, this loop
 1299: iterates @var{n+1} times; @code{i} produces values starting with @var{n}
 1300: and ending with 0. Other Forth systems may behave differently, even if
 1301: they support @code{FOR} loops. To avoid problems, don't use @code{FOR}
 1302: loops.
 1303: 
 1304: @node Arbitrary control structures, Calls and returns, Counted Loops, Control Structures
 1305: @subsection Arbitrary control structures
 1306: 
 1307: ANS Forth permits and supports using control structures in a non-nested
 1308: way. Information about incomplete control structures is stored on the
 1309: control-flow stack. This stack may be implemented on the Forth data
 1310: stack, and this is what we have done in Gforth.
 1311: 
 1312: An @i{orig} entry represents an unresolved forward branch, a @i{dest}
 1313: entry represents a backward branch target. A few words are the basis for
 1314: building any control structure possible (except control structures that
 1315: need storage, like calls, coroutines, and backtracking).
 1316: 
 1317: doc-if
 1318: doc-ahead
 1319: doc-then
 1320: doc-begin
 1321: doc-until
 1322: doc-again
 1323: doc-cs-pick
 1324: doc-cs-roll
 1325: 
 1326: On many systems control-flow stack items take one word, in Gforth they
 1327: currently take three (this may change in the future). Therefore it is a
 1328: really good idea to manipulate the control flow stack with
 1329: @code{cs-pick} and @code{cs-roll}, not with data stack manipulation
 1330: words.
 1331: 
 1332: Some standard control structure words are built from these words:
 1333: 
 1334: doc-else
 1335: doc-while
 1336: doc-repeat
 1337: 
 1338: Gforth adds some more control-structure words:
 1339: 
 1340: doc-endif
 1341: doc-?dup-if
 1342: doc-?dup-0=-if
 1343: 
 1344: Counted loop words constitute a separate group of words:
 1345: 
 1346: doc-?do
 1347: doc-+do
 1348: doc-u+do
 1349: doc--do
 1350: doc-u-do
 1351: doc-do
 1352: doc-for
 1353: doc-loop
 1354: doc-+loop
 1355: doc--loop
 1356: doc-next
 1357: doc-leave
 1358: doc-?leave
 1359: doc-unloop
 1360: doc-done
 1361: 
 1362: The standard does not allow using @code{cs-pick} and @code{cs-roll} on
 1363: @i{do-sys}. Our system allows it, but it's your job to ensure that for
 1364: every @code{?DO} etc. there is exactly one @code{UNLOOP} on any path
 1365: through the definition (@code{LOOP} etc. compile an @code{UNLOOP} on the
 1366: fall-through path). Also, you have to ensure that all @code{LEAVE}s are
 1367: resolved (by using one of the loop-ending words or @code{DONE}).
 1368: 
 1369: Another group of control structure words are
 1370: 
 1371: doc-case
 1372: doc-endcase
 1373: doc-of
 1374: doc-endof
 1375: 
 1376: @i{case-sys} and @i{of-sys} cannot be processed using @code{cs-pick} and
 1377: @code{cs-roll}.
 1378: 
 1379: @subsubsection Programming Style
 1380: 
 1381: In order to ensure readability we recommend that you do not create
 1382: arbitrary control structures directly, but define new control structure
 1383: words for the control structure you want and use these words in your
 1384: program.
 1385: 
 1386: E.g., instead of writing
 1387: 
 1388: @example
 1389: begin
 1390:   ...
 1391: if [ 1 cs-roll ]
 1392:   ...
 1393: again then
 1394: @end example
 1395: 
 1396: we recommend defining control structure words, e.g.,
 1397: 
 1398: @example
 1399: : while ( dest -- orig dest )
 1400:  POSTPONE if
 1401:  1 cs-roll ; immediate
 1402: 
 1403: : repeat ( orig dest -- )
 1404:  POSTPONE again
 1405:  POSTPONE then ; immediate
 1406: @end example
 1407: 
 1408: and then using these to create the control structure:
 1409: 
 1410: @example
 1411: begin
 1412:   ...
 1413: while
 1414:   ...
 1415: repeat
 1416: @end example
 1417: 
 1418: That's much easier to read, isn't it? Of course, @code{REPEAT} and
 1419: @code{WHILE} are predefined, so in this example it would not be
 1420: necessary to define them.
 1421: 
 1422: @node Calls and returns, Exception Handling, Arbitrary control structures, Control Structures
 1423: @subsection Calls and returns
 1424: 
 1425: A definition can be called simply be writing the name of the
 1426: definition. When the end of the definition is reached, it returns. An
 1427: earlier return can be forced using
 1428: 
 1429: doc-exit
 1430: 
 1431: Don't forget to clean up the return stack and @code{UNLOOP} any
 1432: outstanding @code{?DO}...@code{LOOP}s before @code{EXIT}ing. The
 1433: primitive compiled by @code{EXIT} is
 1434: 
 1435: doc-;s
 1436: 
 1437: @node Exception Handling,  , Calls and returns, Control Structures
 1438: @subsection Exception Handling
 1439: 
 1440: doc-catch
 1441: doc-throw
 1442: 
 1443: @node Locals, Defining Words, Control Structures, Words
 1444: @section Locals
 1445: 
 1446: Local variables can make Forth programming more enjoyable and Forth
 1447: programs easier to read. Unfortunately, the locals of ANS Forth are
 1448: laden with restrictions. Therefore, we provide not only the ANS Forth
 1449: locals wordset, but also our own, more powerful locals wordset (we
 1450: implemented the ANS Forth locals wordset through our locals wordset).
 1451: 
 1452: The ideas in this section have also been published in the paper
 1453: @cite{Automatic Scoping of Local Variables} by M. Anton Ertl, presented
 1454: at EuroForth '94; it is available at
 1455: @*@file{http://www.complang.tuwien.ac.at/papers/ertl94l.ps.gz}.
 1456: 
 1457: @menu
 1458: * Gforth locals::               
 1459: * ANS Forth locals::            
 1460: @end menu
 1461: 
 1462: @node Gforth locals, ANS Forth locals, Locals, Locals
 1463: @subsection Gforth locals
 1464: 
 1465: Locals can be defined with
 1466: 
 1467: @example
 1468: @{ local1 local2 ... -- comment @}
 1469: @end example
 1470: or
 1471: @example
 1472: @{ local1 local2 ... @}
 1473: @end example
 1474: 
 1475: E.g.,
 1476: @example
 1477: : max @{ n1 n2 -- n3 @}
 1478:  n1 n2 > if
 1479:    n1
 1480:  else
 1481:    n2
 1482:  endif ;
 1483: @end example
 1484: 
 1485: The similarity of locals definitions with stack comments is intended. A
 1486: locals definition often replaces the stack comment of a word. The order
 1487: of the locals corresponds to the order in a stack comment and everything
 1488: after the @code{--} is really a comment.
 1489: 
 1490: This similarity has one disadvantage: It is too easy to confuse locals
 1491: declarations with stack comments, causing bugs and making them hard to
 1492: find. However, this problem can be avoided by appropriate coding
 1493: conventions: Do not use both notations in the same program. If you do,
 1494: they should be distinguished using additional means, e.g. by position.
 1495: 
 1496: The name of the local may be preceded by a type specifier, e.g.,
 1497: @code{F:} for a floating point value:
 1498: 
 1499: @example
 1500: : CX* @{ F: Ar F: Ai F: Br F: Bi -- Cr Ci @}
 1501: \ complex multiplication
 1502:  Ar Br f* Ai Bi f* f-
 1503:  Ar Bi f* Ai Br f* f+ ;
 1504: @end example
 1505: 
 1506: Gforth currently supports cells (@code{W:}, @code{W^}), doubles
 1507: (@code{D:}, @code{D^}), floats (@code{F:}, @code{F^}) and characters
 1508: (@code{C:}, @code{C^}) in two flavours: a value-flavoured local (defined
 1509: with @code{W:}, @code{D:} etc.) produces its value and can be changed
 1510: with @code{TO}. A variable-flavoured local (defined with @code{W^} etc.)
 1511: produces its address (which becomes invalid when the variable's scope is
 1512: left). E.g., the standard word @code{emit} can be defined in therms of
 1513: @code{type} like this:
 1514: 
 1515: @example
 1516: : emit @{ C^ char* -- @}
 1517:     char* 1 type ;
 1518: @end example
 1519: 
 1520: A local without type specifier is a @code{W:} local. Both flavours of
 1521: locals are initialized with values from the data or FP stack.
 1522: 
 1523: Currently there is no way to define locals with user-defined data
 1524: structures, but we are working on it.
 1525: 
 1526: Gforth allows defining locals everywhere in a colon definition. This
 1527: poses the following questions:
 1528: 
 1529: @menu
 1530: * Where are locals visible by name?::  
 1531: * How long do locals live?::    
 1532: * Programming Style::           
 1533: * Implementation::              
 1534: @end menu
 1535: 
 1536: @node Where are locals visible by name?, How long do locals live?, Gforth locals, Gforth locals
 1537: @subsubsection Where are locals visible by name?
 1538: 
 1539: Basically, the answer is that locals are visible where you would expect
 1540: it in block-structured languages, and sometimes a little longer. If you
 1541: want to restrict the scope of a local, enclose its definition in
 1542: @code{SCOPE}...@code{ENDSCOPE}.
 1543: 
 1544: doc-scope
 1545: doc-endscope
 1546: 
 1547: These words behave like control structure words, so you can use them
 1548: with @code{CS-PICK} and @code{CS-ROLL} to restrict the scope in
 1549: arbitrary ways.
 1550: 
 1551: If you want a more exact answer to the visibility question, here's the
 1552: basic principle: A local is visible in all places that can only be
 1553: reached through the definition of the local@footnote{In compiler
 1554: construction terminology, all places dominated by the definition of the
 1555: local.}. In other words, it is not visible in places that can be reached
 1556: without going through the definition of the local. E.g., locals defined
 1557: in @code{IF}...@code{ENDIF} are visible until the @code{ENDIF}, locals
 1558: defined in @code{BEGIN}...@code{UNTIL} are visible after the
 1559: @code{UNTIL} (until, e.g., a subsequent @code{ENDSCOPE}).
 1560: 
 1561: The reasoning behind this solution is: We want to have the locals
 1562: visible as long as it is meaningful. The user can always make the
 1563: visibility shorter by using explicit scoping. In a place that can
 1564: only be reached through the definition of a local, the meaning of a
 1565: local name is clear. In other places it is not: How is the local
 1566: initialized at the control flow path that does not contain the
 1567: definition? Which local is meant, if the same name is defined twice in
 1568: two independent control flow paths?
 1569: 
 1570: This should be enough detail for nearly all users, so you can skip the
 1571: rest of this section. If you relly must know all the gory details and
 1572: options, read on.
 1573: 
 1574: In order to implement this rule, the compiler has to know which places
 1575: are unreachable. It knows this automatically after @code{AHEAD},
 1576: @code{AGAIN}, @code{EXIT} and @code{LEAVE}; in other cases (e.g., after
 1577: most @code{THROW}s), you can use the word @code{UNREACHABLE} to tell the
 1578: compiler that the control flow never reaches that place. If
 1579: @code{UNREACHABLE} is not used where it could, the only consequence is
 1580: that the visibility of some locals is more limited than the rule above
 1581: says. If @code{UNREACHABLE} is used where it should not (i.e., if you
 1582: lie to the compiler), buggy code will be produced.
 1583: 
 1584: Another problem with this rule is that at @code{BEGIN}, the compiler
 1585: does not know which locals will be visible on the incoming
 1586: back-edge. All problems discussed in the following are due to this
 1587: ignorance of the compiler (we discuss the problems using @code{BEGIN}
 1588: loops as examples; the discussion also applies to @code{?DO} and other
 1589: loops). Perhaps the most insidious example is:
 1590: @example
 1591: AHEAD
 1592: BEGIN
 1593:   x
 1594: [ 1 CS-ROLL ] THEN
 1595:   @{ x @}
 1596:   ...
 1597: UNTIL
 1598: @end example
 1599: 
 1600: This should be legal according to the visibility rule. The use of
 1601: @code{x} can only be reached through the definition; but that appears
 1602: textually below the use.
 1603: 
 1604: From this example it is clear that the visibility rules cannot be fully
 1605: implemented without major headaches. Our implementation treats common
 1606: cases as advertised and the exceptions are treated in a safe way: The
 1607: compiler makes a reasonable guess about the locals visible after a
 1608: @code{BEGIN}; if it is too pessimistic, the
 1609: user will get a spurious error about the local not being defined; if the
 1610: compiler is too optimistic, it will notice this later and issue a
 1611: warning. In the case above the compiler would complain about @code{x}
 1612: being undefined at its use. You can see from the obscure examples in
 1613: this section that it takes quite unusual control structures to get the
 1614: compiler into trouble, and even then it will often do fine.
 1615: 
 1616: If the @code{BEGIN} is reachable from above, the most optimistic guess
 1617: is that all locals visible before the @code{BEGIN} will also be
 1618: visible after the @code{BEGIN}. This guess is valid for all loops that
 1619: are entered only through the @code{BEGIN}, in particular, for normal
 1620: @code{BEGIN}...@code{WHILE}...@code{REPEAT} and
 1621: @code{BEGIN}...@code{UNTIL} loops and it is implemented in our
 1622: compiler. When the branch to the @code{BEGIN} is finally generated by
 1623: @code{AGAIN} or @code{UNTIL}, the compiler checks the guess and
 1624: warns the user if it was too optimisitic:
 1625: @example
 1626: IF
 1627:   @{ x @}
 1628: BEGIN
 1629:   \ x ? 
 1630: [ 1 cs-roll ] THEN
 1631:   ...
 1632: UNTIL
 1633: @end example
 1634: 
 1635: Here, @code{x} lives only until the @code{BEGIN}, but the compiler
 1636: optimistically assumes that it lives until the @code{THEN}. It notices
 1637: this difference when it compiles the @code{UNTIL} and issues a
 1638: warning. The user can avoid the warning, and make sure that @code{x}
 1639: is not used in the wrong area by using explicit scoping:
 1640: @example
 1641: IF
 1642:   SCOPE
 1643:   @{ x @}
 1644:   ENDSCOPE
 1645: BEGIN
 1646: [ 1 cs-roll ] THEN
 1647:   ...
 1648: UNTIL
 1649: @end example
 1650: 
 1651: Since the guess is optimistic, there will be no spurious error messages
 1652: about undefined locals.
 1653: 
 1654: If the @code{BEGIN} is not reachable from above (e.g., after
 1655: @code{AHEAD} or @code{EXIT}), the compiler cannot even make an
 1656: optimistic guess, as the locals visible after the @code{BEGIN} may be
 1657: defined later. Therefore, the compiler assumes that no locals are
 1658: visible after the @code{BEGIN}. However, the user can use
 1659: @code{ASSUME-LIVE} to make the compiler assume that the same locals are
 1660: visible at the BEGIN as at the point where the top control-flow stack
 1661: item was created.
 1662: 
 1663: doc-assume-live
 1664: 
 1665: E.g.,
 1666: @example
 1667: @{ x @}
 1668: AHEAD
 1669: ASSUME-LIVE
 1670: BEGIN
 1671:   x
 1672: [ 1 CS-ROLL ] THEN
 1673:   ...
 1674: UNTIL
 1675: @end example
 1676: 
 1677: Other cases where the locals are defined before the @code{BEGIN} can be
 1678: handled by inserting an appropriate @code{CS-ROLL} before the
 1679: @code{ASSUME-LIVE} (and changing the control-flow stack manipulation
 1680: behind the @code{ASSUME-LIVE}).
 1681: 
 1682: Cases where locals are defined after the @code{BEGIN} (but should be
 1683: visible immediately after the @code{BEGIN}) can only be handled by
 1684: rearranging the loop. E.g., the ``most insidious'' example above can be
 1685: arranged into:
 1686: @example
 1687: BEGIN
 1688:   @{ x @}
 1689:   ... 0=
 1690: WHILE
 1691:   x
 1692: REPEAT
 1693: @end example
 1694: 
 1695: @node How long do locals live?, Programming Style, Where are locals visible by name?, Gforth locals
 1696: @subsubsection How long do locals live?
 1697: 
 1698: The right answer for the lifetime question would be: A local lives at
 1699: least as long as it can be accessed. For a value-flavoured local this
 1700: means: until the end of its visibility. However, a variable-flavoured
 1701: local could be accessed through its address far beyond its visibility
 1702: scope. Ultimately, this would mean that such locals would have to be
 1703: garbage collected. Since this entails un-Forth-like implementation
 1704: complexities, I adopted the same cowardly solution as some other
 1705: languages (e.g., C): The local lives only as long as it is visible;
 1706: afterwards its address is invalid (and programs that access it
 1707: afterwards are erroneous).
 1708: 
 1709: @node Programming Style, Implementation, How long do locals live?, Gforth locals
 1710: @subsubsection Programming Style
 1711: 
 1712: The freedom to define locals anywhere has the potential to change
 1713: programming styles dramatically. In particular, the need to use the
 1714: return stack for intermediate storage vanishes. Moreover, all stack
 1715: manipulations (except @code{PICK}s and @code{ROLL}s with run-time
 1716: determined arguments) can be eliminated: If the stack items are in the
 1717: wrong order, just write a locals definition for all of them; then
 1718: write the items in the order you want.
 1719: 
 1720: This seems a little far-fetched and eliminating stack manipulations is
 1721: unlikely to become a conscious programming objective. Still, the number
 1722: of stack manipulations will be reduced dramatically if local variables
 1723: are used liberally (e.g., compare @code{max} in @ref{Gforth locals} with
 1724: a traditional implementation of @code{max}).
 1725: 
 1726: This shows one potential benefit of locals: making Forth programs more
 1727: readable. Of course, this benefit will only be realized if the
 1728: programmers continue to honour the principle of factoring instead of
 1729: using the added latitude to make the words longer.
 1730: 
 1731: Using @code{TO} can and should be avoided.  Without @code{TO},
 1732: every value-flavoured local has only a single assignment and many
 1733: advantages of functional languages apply to Forth. I.e., programs are
 1734: easier to analyse, to optimize and to read: It is clear from the
 1735: definition what the local stands for, it does not turn into something
 1736: different later.
 1737: 
 1738: E.g., a definition using @code{TO} might look like this:
 1739: @example
 1740: : strcmp @{ addr1 u1 addr2 u2 -- n @}
 1741:  u1 u2 min 0
 1742:  ?do
 1743:    addr1 c@@ addr2 c@@ -
 1744:    ?dup-if
 1745:      unloop exit
 1746:    then
 1747:    addr1 char+ TO addr1
 1748:    addr2 char+ TO addr2
 1749:  loop
 1750:  u1 u2 - ;
 1751: @end example
 1752: Here, @code{TO} is used to update @code{addr1} and @code{addr2} at
 1753: every loop iteration. @code{strcmp} is a typical example of the
 1754: readability problems of using @code{TO}. When you start reading
 1755: @code{strcmp}, you think that @code{addr1} refers to the start of the
 1756: string. Only near the end of the loop you realize that it is something
 1757: else.
 1758: 
 1759: This can be avoided by defining two locals at the start of the loop that
 1760: are initialized with the right value for the current iteration.
 1761: @example
 1762: : strcmp @{ addr1 u1 addr2 u2 -- n @}
 1763:  addr1 addr2
 1764:  u1 u2 min 0 
 1765:  ?do @{ s1 s2 @}
 1766:    s1 c@@ s2 c@@ -
 1767:    ?dup-if
 1768:      unloop exit
 1769:    then
 1770:    s1 char+ s2 char+
 1771:  loop
 1772:  2drop
 1773:  u1 u2 - ;
 1774: @end example
 1775: Here it is clear from the start that @code{s1} has a different value
 1776: in every loop iteration.
 1777: 
 1778: @node Implementation,  , Programming Style, Gforth locals
 1779: @subsubsection Implementation
 1780: 
 1781: Gforth uses an extra locals stack. The most compelling reason for
 1782: this is that the return stack is not float-aligned; using an extra stack
 1783: also eliminates the problems and restrictions of using the return stack
 1784: as locals stack. Like the other stacks, the locals stack grows toward
 1785: lower addresses. A few primitives allow an efficient implementation:
 1786: 
 1787: doc-@local#
 1788: doc-f@local#
 1789: doc-laddr#
 1790: doc-lp+!#
 1791: doc-lp!
 1792: doc->l
 1793: doc-f>l
 1794: 
 1795: In addition to these primitives, some specializations of these
 1796: primitives for commonly occurring inline arguments are provided for
 1797: efficiency reasons, e.g., @code{@@local0} as specialization of
 1798: @code{@@local#} for the inline argument 0. The following compiling words
 1799: compile the right specialized version, or the general version, as
 1800: appropriate:
 1801: 
 1802: doc-compile-@local
 1803: doc-compile-f@local
 1804: doc-compile-lp+!
 1805: 
 1806: Combinations of conditional branches and @code{lp+!#} like
 1807: @code{?branch-lp+!#} (the locals pointer is only changed if the branch
 1808: is taken) are provided for efficiency and correctness in loops.
 1809: 
 1810: A special area in the dictionary space is reserved for keeping the
 1811: local variable names. @code{@{} switches the dictionary pointer to this
 1812: area and @code{@}} switches it back and generates the locals
 1813: initializing code. @code{W:} etc.@ are normal defining words. This
 1814: special area is cleared at the start of every colon definition.
 1815: 
 1816: A special feature of Gforth's dictionary is used to implement the
 1817: definition of locals without type specifiers: every wordlist (aka
 1818: vocabulary) has its own methods for searching
 1819: etc. (@pxref{Wordlists}). For the present purpose we defined a wordlist
 1820: with a special search method: When it is searched for a word, it
 1821: actually creates that word using @code{W:}. @code{@{} changes the search
 1822: order to first search the wordlist containing @code{@}}, @code{W:} etc.,
 1823: and then the wordlist for defining locals without type specifiers.
 1824: 
 1825: The lifetime rules support a stack discipline within a colon
 1826: definition: The lifetime of a local is either nested with other locals
 1827: lifetimes or it does not overlap them.
 1828: 
 1829: At @code{BEGIN}, @code{IF}, and @code{AHEAD} no code for locals stack
 1830: pointer manipulation is generated. Between control structure words
 1831: locals definitions can push locals onto the locals stack. @code{AGAIN}
 1832: is the simplest of the other three control flow words. It has to
 1833: restore the locals stack depth of the corresponding @code{BEGIN}
 1834: before branching. The code looks like this:
 1835: @format
 1836: @code{lp+!#} current-locals-size @minus{} dest-locals-size
 1837: @code{branch} <begin>
 1838: @end format
 1839: 
 1840: @code{UNTIL} is a little more complicated: If it branches back, it
 1841: must adjust the stack just like @code{AGAIN}. But if it falls through,
 1842: the locals stack must not be changed. The compiler generates the
 1843: following code:
 1844: @format
 1845: @code{?branch-lp+!#} <begin> current-locals-size @minus{} dest-locals-size
 1846: @end format
 1847: The locals stack pointer is only adjusted if the branch is taken.
 1848: 
 1849: @code{THEN} can produce somewhat inefficient code:
 1850: @format
 1851: @code{lp+!#} current-locals-size @minus{} orig-locals-size
 1852: <orig target>:
 1853: @code{lp+!#} orig-locals-size @minus{} new-locals-size
 1854: @end format
 1855: The second @code{lp+!#} adjusts the locals stack pointer from the
 1856: level at the @var{orig} point to the level after the @code{THEN}. The
 1857: first @code{lp+!#} adjusts the locals stack pointer from the current
 1858: level to the level at the orig point, so the complete effect is an
 1859: adjustment from the current level to the right level after the
 1860: @code{THEN}.
 1861: 
 1862: In a conventional Forth implementation a dest control-flow stack entry
 1863: is just the target address and an orig entry is just the address to be
 1864: patched. Our locals implementation adds a wordlist to every orig or dest
 1865: item. It is the list of locals visible (or assumed visible) at the point
 1866: described by the entry. Our implementation also adds a tag to identify
 1867: the kind of entry, in particular to differentiate between live and dead
 1868: (reachable and unreachable) orig entries.
 1869: 
 1870: A few unusual operations have to be performed on locals wordlists:
 1871: 
 1872: doc-common-list
 1873: doc-sub-list?
 1874: doc-list-size
 1875: 
 1876: Several features of our locals wordlist implementation make these
 1877: operations easy to implement: The locals wordlists are organised as
 1878: linked lists; the tails of these lists are shared, if the lists
 1879: contain some of the same locals; and the address of a name is greater
 1880: than the address of the names behind it in the list.
 1881: 
 1882: Another important implementation detail is the variable
 1883: @code{dead-code}. It is used by @code{BEGIN} and @code{THEN} to
 1884: determine if they can be reached directly or only through the branch
 1885: that they resolve. @code{dead-code} is set by @code{UNREACHABLE},
 1886: @code{AHEAD}, @code{EXIT} etc., and cleared at the start of a colon
 1887: definition, by @code{BEGIN} and usually by @code{THEN}.
 1888: 
 1889: Counted loops are similar to other loops in most respects, but
 1890: @code{LEAVE} requires special attention: It performs basically the same
 1891: service as @code{AHEAD}, but it does not create a control-flow stack
 1892: entry. Therefore the information has to be stored elsewhere;
 1893: traditionally, the information was stored in the target fields of the
 1894: branches created by the @code{LEAVE}s, by organizing these fields into a
 1895: linked list. Unfortunately, this clever trick does not provide enough
 1896: space for storing our extended control flow information. Therefore, we
 1897: introduce another stack, the leave stack. It contains the control-flow
 1898: stack entries for all unresolved @code{LEAVE}s.
 1899: 
 1900: Local names are kept until the end of the colon definition, even if
 1901: they are no longer visible in any control-flow path. In a few cases
 1902: this may lead to increased space needs for the locals name area, but
 1903: usually less than reclaiming this space would cost in code size.
 1904: 
 1905: 
 1906: @node ANS Forth locals,  , Gforth locals, Locals
 1907: @subsection ANS Forth locals
 1908: 
 1909: The ANS Forth locals wordset does not define a syntax for locals, but
 1910: words that make it possible to define various syntaxes. One of the
 1911: possible syntaxes is a subset of the syntax we used in the Gforth locals
 1912: wordset, i.e.:
 1913: 
 1914: @example
 1915: @{ local1 local2 ... -- comment @}
 1916: @end example
 1917: or
 1918: @example
 1919: @{ local1 local2 ... @}
 1920: @end example
 1921: 
 1922: The order of the locals corresponds to the order in a stack comment. The
 1923: restrictions are:
 1924: 
 1925: @itemize @bullet
 1926: @item
 1927: Locals can only be cell-sized values (no type specifiers are allowed).
 1928: @item
 1929: Locals can be defined only outside control structures.
 1930: @item
 1931: Locals can interfere with explicit usage of the return stack. For the
 1932: exact (and long) rules, see the standard. If you don't use return stack
 1933: accessing words in a definition using locals, you will be all right. The
 1934: purpose of this rule is to make locals implementation on the return
 1935: stack easier.
 1936: @item
 1937: The whole definition must be in one line.
 1938: @end itemize
 1939: 
 1940: Locals defined in this way behave like @code{VALUE}s (@xref{Simple
 1941: Defining Words}). I.e., they are initialized from the stack. Using their
 1942: name produces their value. Their value can be changed using @code{TO}.
 1943: 
 1944: Since this syntax is supported by Gforth directly, you need not do
 1945: anything to use it. If you want to port a program using this syntax to
 1946: another ANS Forth system, use @file{compat/anslocal.fs} to implement the
 1947: syntax on the other system.
 1948: 
 1949: Note that a syntax shown in the standard, section A.13 looks
 1950: similar, but is quite different in having the order of locals
 1951: reversed. Beware!
 1952: 
 1953: The ANS Forth locals wordset itself consists of the following word
 1954: 
 1955: doc-(local)
 1956: 
 1957: The ANS Forth locals extension wordset defines a syntax, but it is so
 1958: awful that we strongly recommend not to use it. We have implemented this
 1959: syntax to make porting to Gforth easy, but do not document it here. The
 1960: problem with this syntax is that the locals are defined in an order
 1961: reversed with respect to the standard stack comment notation, making
 1962: programs harder to read, and easier to misread and miswrite. The only
 1963: merit of this syntax is that it is easy to implement using the ANS Forth
 1964: locals wordset.
 1965: 
 1966: @node Defining Words, Tokens for Words, Locals, Words
 1967: @section Defining Words
 1968: 
 1969: @menu
 1970: * Simple Defining Words::       
 1971: * Colon Definitions::           
 1972: * User-defined Defining Words::  
 1973: * Supplying names::             
 1974: * Interpretation and Compilation Semantics::  
 1975: @end menu
 1976: 
 1977: @node Simple Defining Words, Colon Definitions, Defining Words, Defining Words
 1978: @subsection Simple Defining Words
 1979: 
 1980: doc-constant
 1981: doc-2constant
 1982: doc-fconstant
 1983: doc-variable
 1984: doc-2variable
 1985: doc-fvariable
 1986: doc-create
 1987: doc-user
 1988: doc-value
 1989: doc-to
 1990: doc-defer
 1991: doc-is
 1992: 
 1993: @node Colon Definitions, User-defined Defining Words, Simple Defining Words, Defining Words
 1994: @subsection Colon Definitions
 1995: 
 1996: @example
 1997: : name ( ... -- ... )
 1998:     word1 word2 word3 ;
 1999: @end example
 2000: 
 2001: creates a word called @code{name}, that, upon execution, executes
 2002: @code{word1 word2 word3}. @code{name} is a @dfn{(colon) definition}.
 2003: 
 2004: The explanation above is somewhat superficial. @xref{Interpretation and
 2005: Compilation Semantics} for an in-depth discussion of some of the issues
 2006: involved.
 2007: 
 2008: doc-:
 2009: doc-;
 2010: 
 2011: @node User-defined Defining Words, Supplying names, Colon Definitions, Defining Words
 2012: @subsection User-defined Defining Words
 2013: 
 2014: You can create new defining words simply by wrapping defining-time code
 2015: around existing defining words and putting the sequence in a colon
 2016: definition.
 2017: 
 2018: If you want the words defined with your defining words to behave
 2019: differently from words defined with standard defining words, you can
 2020: write your defining word like this:
 2021: 
 2022: @example
 2023: : def-word ( "name" -- )
 2024:     Create @var{code1}
 2025: DOES> ( ... -- ... )
 2026:     @var{code2} ;
 2027: 
 2028: def-word name
 2029: @end example
 2030: 
 2031: Technically, this fragment defines a defining word @code{def-word}, and
 2032: a word @code{name}; when you execute @code{name}, the address of the
 2033: body of @code{name} is put on the data stack and @var{code2} is executed
 2034: (the address of the body of @code{name} is the address @code{HERE}
 2035: returns immediately after the @code{CREATE}).
 2036: 
 2037: In other words, if you make the following definitions:
 2038: 
 2039: @example
 2040: : def-word1 ( "name" -- )
 2041:     Create @var{code1} ;
 2042: 
 2043: : action1 ( ... -- ... )
 2044:     @var{code2} ;
 2045: 
 2046: def-word name1
 2047: @end example
 2048: 
 2049: Using @code{name1 action1} is equivalent to using @code{name}.
 2050: 
 2051: E.g., you can implement @code{Constant} in this way:
 2052: 
 2053: @example
 2054: : constant ( w "name" -- )
 2055:     create ,
 2056: DOES> ( -- w )
 2057:     @@ ;
 2058: @end example
 2059: 
 2060: When you create a constant with @code{5 constant five}, first a new word
 2061: @code{five} is created, then the value 5 is laid down in the body of
 2062: @code{five} with @code{,}. When @code{five} is invoked, the address of
 2063: the body is put on the stack, and @code{@@} retrieves the value 5.
 2064: 
 2065: In the example above the stack comment after the @code{DOES>} specifies
 2066: the stack effect of the defined words, not the stack effect of the
 2067: following code (the following code expects the address of the body on
 2068: the top of stack, which is not reflected in the stack comment). This is
 2069: the convention that I use and recommend (it clashes a bit with using
 2070: locals declarations for stack effect specification, though).
 2071: 
 2072: @subsubsection Applications of @code{CREATE..DOES>}
 2073: 
 2074: You may wonder how to use this feature. Here are some usage patterns:
 2075: 
 2076: When you see a sequence of code occurring several times, and you can
 2077: identify a meaning, you will factor it out as a colon definition. When
 2078: you see similar colon definitions, you can factor them using
 2079: @code{CREATE..DOES>}. E.g., an assembler usually defines several words
 2080: that look very similar:
 2081: @example
 2082: : ori, ( reg-taget reg-source n -- )
 2083:     0 asm-reg-reg-imm ;
 2084: : andi, ( reg-taget reg-source n -- )
 2085:     1 asm-reg-reg-imm ;
 2086: @end example
 2087: 
 2088: This could be factored with:
 2089: @example
 2090: : reg-reg-imm ( op-code -- )
 2091:     create ,
 2092: DOES> ( reg-taget reg-source n -- )
 2093:     @@ asm-reg-reg-imm ;
 2094: 
 2095: 0 reg-reg-imm ori,
 2096: 1 reg-reg-imm andi,
 2097: @end example
 2098: 
 2099: Another view of @code{CREATE..DOES>} is to consider it as a crude way to
 2100: supply a part of the parameters for a word (known as @dfn{currying} in
 2101: the functional language community). E.g., @code{+} needs two
 2102: parameters. Creating versions of @code{+} with one parameter fixed can
 2103: be done like this:
 2104: @example
 2105: : curry+ ( n1 -- )
 2106:     create ,
 2107: DOES> ( n2 -- n1+n2 )
 2108:     @@ + ;
 2109: 
 2110:  3 curry+ 3+
 2111: -2 curry+ 2-
 2112: @end example
 2113: 
 2114: @subsubsection The gory details of @code{CREATE..DOES>}
 2115: 
 2116: doc-does>
 2117: 
 2118: This means that you need not use @code{CREATE} and @code{DOES>} in the
 2119: same definition; E.g., you can put the @code{DOES>}-part in a separate
 2120: definition. This allows us to, e.g., select among different DOES>-parts:
 2121: @example
 2122: : does1 
 2123: DOES> ( ... -- ... )
 2124:     ... ;
 2125: 
 2126: : does2
 2127: DOES> ( ... -- ... )
 2128:     ... ;
 2129: 
 2130: : def-word ( ... -- ... )
 2131:     create ...
 2132:     IF
 2133:        does1
 2134:     ELSE
 2135:        does2
 2136:     ENDIF ;
 2137: @end example
 2138: 
 2139: In a standard program you can apply a @code{DOES>}-part only if the last
 2140: word was defined with @code{CREATE}. In Gforth, the @code{DOES>}-part
 2141: will override the behaviour of the last word defined in any case. In a
 2142: standard program, you can use @code{DOES>} only in a colon
 2143: definition. In Gforth, you can also use it in interpretation state, in a
 2144: kind of one-shot mode:
 2145: @example
 2146: CREATE name ( ... -- ... )
 2147:   @var{initialization}
 2148: DOES>
 2149:   @var{code} ;
 2150: @end example
 2151: This is equivalwent to the standard
 2152: @example
 2153: :noname
 2154: DOES>
 2155:     @var{code} ;
 2156: CREATE name EXECUTE ( ... -- ... )
 2157:     @var{initialization}
 2158: @end example
 2159: 
 2160: You can get the address of the body of a word with
 2161: 
 2162: doc->body
 2163: 
 2164: @node Supplying names, Interpretation and Compilation Semantics, User-defined Defining Words, Defining Words
 2165: @subsection Supplying names for the defined words
 2166: 
 2167: By default, defining words take the names for the defined words from the
 2168: input stream. Sometimes you want to supply the name from a string. You
 2169: can do this with
 2170: 
 2171: doc-nextname
 2172: 
 2173: E.g.,
 2174: 
 2175: @example
 2176: s" foo" nextname create
 2177: @end example
 2178: is equivalent to
 2179: @example
 2180: create foo
 2181: @end example
 2182: 
 2183: Sometimes you want to define a word without a name. You can do this with
 2184: 
 2185: doc-noname
 2186: 
 2187: To make any use of the newly defined word, you need its execution
 2188: token. You can get it with
 2189: 
 2190: doc-lastxt
 2191: 
 2192: E.g., you can initialize a deferred word with an anonymous colon
 2193: definition:
 2194: @example
 2195: Defer deferred
 2196: noname : ( ... -- ... )
 2197:   ... ;
 2198: lastxt IS deferred
 2199: @end example
 2200: 
 2201: @code{lastxt} also works when the last word was not defined as
 2202: @code{noname}. 
 2203: 
 2204: The standard has also recognized the need for anonymous words and
 2205: provides
 2206: 
 2207: doc-:noname
 2208: 
 2209: This leaves the execution token for the word on the stack after the
 2210: closing @code{;}. You can rewrite the last example with @code{:noname}:
 2211: @example
 2212: Defer deferred
 2213: :noname ( ... -- ... )
 2214:   ... ;
 2215: IS deferred
 2216: @end example
 2217: 
 2218: @node Interpretation and Compilation Semantics,  , Supplying names, Defining Words
 2219: @subsection Interpretation and Compilation Semantics
 2220: 
 2221: The @dfn{interpretation semantics} of a word are what the text
 2222: interpreter does when it encounters the word in interpret state. It also
 2223: appears in some other contexts, e.g., the execution token returned by
 2224: @code{' @var{word}} identifies the interpretation semantics of
 2225: @var{word} (in other words, @code{' @var{word} execute} is equivalent to
 2226: interpret-state text interpretation of @code{@var{word}}).
 2227: 
 2228: The @dfn{compilation semantics} of a word are what the text interpreter
 2229: does when it encounters the word in compile state. It also appears in
 2230: other contexts, e.g, @code{POSTPONE @var{word}} compiles@footnote{In
 2231: standard terminology, ``appends to the current definition''.} the
 2232: compilation semantics of @var{word}.
 2233: 
 2234: The standard also talks about @dfn{execution semantics}. They are used
 2235: only for defining the interpretation and compilation semantics of many
 2236: words. By default, the interpretation semantics of a word are to
 2237: @code{execute} its execution semantics, and the compilation semantics of
 2238: a word are to @code{compile,} its execution semantics.@footnote{In
 2239: standard terminology: The default interpretation semantics are its
 2240: execution semantics; the default compilation semantics are to append its
 2241: execution semantics to the execution semantics of the current
 2242: definition.}
 2243: 
 2244: You can change the compilation semantics into @code{execute}ing the
 2245: execution semantics with
 2246: 
 2247: doc-immediate
 2248: 
 2249: You can remove the interpretation semantics of a word with
 2250: 
 2251: doc-compile-only
 2252: doc-restrict
 2253: 
 2254: Note that ticking (@code{'}) compile-only words gives an error
 2255: (``Interpreting a compile-only word'').
 2256: 
 2257: Gforth also allows you to define words with arbitrary combinations of
 2258: interpretation and compilation semantics.
 2259: 
 2260: doc-interpret/compile:
 2261: 
 2262: This feature was introduced for implementing @code{TO} and @code{S"}. I
 2263: recommend that you do not define such words, as cute as they may be:
 2264: they make it hard to get at both parts of the word in some contexts.
 2265: E.g., assume you want to get an execution token for the compilation
 2266: part. Instead, define two words, one that embodies the interpretation
 2267: part, and one that embodies the compilation part.
 2268: 
 2269: There is, however, a potentially useful application of this feature:
 2270: Providing differing implementations for the default semantics. While
 2271: this introduces redundancy and is therefore usually a bad idea, a
 2272: performance improvement may be worth the trouble. E.g., consider the
 2273: word @code{foobar}:
 2274: 
 2275: @example
 2276: : foobar
 2277:     foo bar ;
 2278: @end example
 2279: 
 2280: Let us assume that @code{foobar} is called so frequently that the
 2281: calling overhead would take a significant amount of the run-time. We can
 2282: optimize it with @code{interpret/compile:}:
 2283: 
 2284: @example
 2285: :noname
 2286:    foo bar ;
 2287: :noname
 2288:    POSTPONE foo POSTPONE bar ;
 2289: interpret/compile: foobar
 2290: @end example
 2291: 
 2292: This definition has the same interpretation semantics and essentially
 2293: the same compilation semantics as the simple definition of
 2294: @code{foobar}, but the implementation of the compilation semantics is
 2295: more efficient with respect to run-time.
 2296: 
 2297: Some people try to use state-smart words to emulate the feature provided
 2298: by @code{interpret/compile:} (words are state-smart if they check
 2299: @code{STATE} during execution). E.g., they would try to code
 2300: @code{foobar} like this:
 2301: 
 2302: @example
 2303: : foobar
 2304:   STATE @@
 2305:   IF ( compilation state )
 2306:     POSTPONE foo POSTPONE bar
 2307:   ELSE
 2308:     foo bar
 2309:   ENDIF ; immediate
 2310: @end example
 2311: 
 2312: While this works if @code{foobar} is processed only by the text
 2313: interpreter, it does not work in other contexts (like @code{'} or
 2314: @code{POSTPONE}). E.g., @code{' foobar} will produce an execution token
 2315: for a state-smart word, not for the interpretation semantics of the
 2316: original @code{foobar}; when you execute this execution token (directly
 2317: with @code{EXECUTE} or indirectly through @code{COMPILE,}) in compile
 2318: state, the result will not be what you expected (i.e., it will not
 2319: perform @code{foo bar}). State-smart words are a bad idea. Simply don't
 2320: write them!
 2321: 
 2322: It is also possible to write defining words that define words with
 2323: arbitrary combinations of interpretation and compilation semantics (or,
 2324: preferably, arbitrary combinations of implementations of the default
 2325: semantics). In general, this looks like:
 2326: 
 2327: @example
 2328: : def-word
 2329:     create-interpret/compile
 2330:     @var{code1}
 2331: interpretation>
 2332:     @var{code2}
 2333: <interpretation
 2334: compilation>
 2335:     @var{code3}
 2336: <compilation ;
 2337: @end example
 2338: 
 2339: For a @var{word} defined with @code{def-word}, the interpretation
 2340: semantics are to push the address of the body of @var{word} and perform
 2341: @var{code2}, and the compilation semantics are to push the address of
 2342: the body of @var{word} and perform @var{code3}. E.g., @code{constant}
 2343: can also be defined like this:
 2344: 
 2345: @example
 2346: : constant ( n "name" -- )
 2347:     create-interpret/compile
 2348:     ,
 2349: interpretation> ( -- n )
 2350:     @@
 2351: <interpretation
 2352: compilation> ( compilation. -- ; run-time. -- n )
 2353:     @@ postpone literal
 2354: <compilation ;
 2355: @end example
 2356: 
 2357: doc-create-interpret/compile
 2358: doc-interpretation>
 2359: doc-<interpretation
 2360: doc-compilation>
 2361: doc-<compilation
 2362: 
 2363: Note that words defined with @code{interpret/compile:} and
 2364: @code{create-interpret/compile} have an extended header structure that
 2365: differs from other words; however, unless you try to access them with
 2366: plain address arithmetic, you should not notice this. Words for
 2367: accessing the header structure usually know how to deal with this; e.g.,
 2368: @code{' word >body} also gives you the body of a word created with
 2369: @code{create-interpret/compile}.
 2370: 
 2371: @node Tokens for Words, Wordlists, Defining Words, Words
 2372: @section Tokens for Words
 2373: 
 2374: This chapter describes the creation and use of tokens that represent
 2375: words on the stack (and in data space).
 2376: 
 2377: Named words have interpretation and compilation semantics. Unnamed words
 2378: just have execution semantics.
 2379: 
 2380: An @dfn{execution token} represents the execution semantics of an
 2381: unnamed word. An execution token occupies one cell. As explained in
 2382: section @ref{Supplying names}, the execution token of the last words
 2383: defined can be produced with
 2384: 
 2385: short-lastxt
 2386: 
 2387: You can perform the semantics represented by an execution token with
 2388: doc-execute
 2389: You can compile the word with
 2390: doc-compile,
 2391: 
 2392: In Gforth, the abstract data type @emph{execution token} is implemented
 2393: as CFA (code field address).
 2394: 
 2395: The interpretation semantics of a named word are also represented by an
 2396: execution token. You can get it with
 2397: 
 2398: doc-[']
 2399: doc-'
 2400: 
 2401: For literals, you use @code{'} in interpreted code and @code{[']} in
 2402: compiled code. Gforth's @code{'} and @code{[']} behave somewhat unusual
 2403: by complaining about compile-only words. To get an execution token for a
 2404: compiling word @var{X}, use @code{COMP' @var{X} drop} or @code{[COMP']
 2405: @var{X} drop}.
 2406: 
 2407: The compilation semantics are represented by a @dfn{compilation token}
 2408: consisting of two cells: @var{w xt}. The top cell @var{xt} is an
 2409: execution token. The compilation semantics represented by the
 2410: compilation token can be performed with @code{execute}, which consumes
 2411: the whole compilation token, with an additional stack effect determined
 2412: by the represented compilation semantics.
 2413: 
 2414: doc-[comp']
 2415: doc-comp'
 2416: 
 2417: At present, the @var{w} part of a compilation token is an execution
 2418: token, and the @var{xt} part represents either @code{execute} or
 2419: @code{compile,}. However, don't rely on that kowledge, unless necessary;
 2420: we may introduce unusual compilation tokens in the future (e.g.,
 2421: compilation tokens representing the compilation semantics of literals).
 2422: 
 2423: Named words are also represented by the @dfn{name token}. The abstract
 2424: data type @emph{name token} is implemented as NFA (name field address).
 2425: 
 2426: doc-find-name
 2427: doc-name>int
 2428: doc-name?int
 2429: doc-name>comp
 2430: doc-name>string
 2431: 
 2432: @node Wordlists, Files, Tokens for Words, Words
 2433: @section Wordlists
 2434: 
 2435: @node Files, Blocks, Wordlists, Words
 2436: @section Files
 2437: 
 2438: @node Blocks, Other I/O, Files, Words
 2439: @section Blocks
 2440: 
 2441: @node Other I/O, Programming Tools, Blocks, Words
 2442: @section Other I/O
 2443: 
 2444: @node Programming Tools, Assembler and Code words, Other I/O, Words
 2445: @section Programming Tools
 2446: 
 2447: @menu
 2448: * Debugging::                   Simple and quick.
 2449: * Assertions::                  Making your programs self-checking.
 2450: @end menu
 2451: 
 2452: @node Debugging, Assertions, Programming Tools, Programming Tools
 2453: @subsection Debugging
 2454: 
 2455: The simple debugging aids provided in @file{debugging.fs}
 2456: are meant to support a different style of debugging than the
 2457: tracing/stepping debuggers used in languages with long turn-around
 2458: times.
 2459: 
 2460: A much better (faster) way in fast-compilig languages is to add
 2461: printing code at well-selected places, let the program run, look at
 2462: the output, see where things went wrong, add more printing code, etc.,
 2463: until the bug is found.
 2464: 
 2465: The word @code{~~} is easy to insert. It just prints debugging
 2466: information (by default the source location and the stack contents). It
 2467: is also easy to remove (@kbd{C-x ~} in the Emacs Forth mode to
 2468: query-replace them with nothing). The deferred words
 2469: @code{printdebugdata} and @code{printdebugline} control the output of
 2470: @code{~~}. The default source location output format works well with
 2471: Emacs' compilation mode, so you can step through the program at the
 2472: source level using @kbd{C-x `} (the advantage over a stepping debugger
 2473: is that you can step in any direction and you know where the crash has
 2474: happened or where the strange data has occurred).
 2475: 
 2476: Note that the default actions clobber the contents of the pictured
 2477: numeric output string, so you should not use @code{~~}, e.g., between
 2478: @code{<#} and @code{#>}.
 2479: 
 2480: doc-~~
 2481: doc-printdebugdata
 2482: doc-printdebugline
 2483: 
 2484: @node Assertions,  , Debugging, Programming Tools
 2485: @subsection Assertions
 2486: 
 2487: It is a good idea to make your programs self-checking, in particular, if
 2488: you use an assumption (e.g., that a certain field of a data structure is
 2489: never zero) that may become wrong during maintenance. Gforth supports
 2490: assertions for this purpose. They are used like this:
 2491: 
 2492: @example
 2493: assert( @var{flag} )
 2494: @end example
 2495: 
 2496: The code between @code{assert(} and @code{)} should compute a flag, that
 2497: should be true if everything is alright and false otherwise. It should
 2498: not change anything else on the stack. The overall stack effect of the
 2499: assertion is @code{( -- )}. E.g.
 2500: 
 2501: @example
 2502: assert( 1 1 + 2 = ) \ what we learn in school
 2503: assert( dup 0<> ) \ assert that the top of stack is not zero
 2504: assert( false ) \ this code should not be reached
 2505: @end example
 2506: 
 2507: The need for assertions is different at different times. During
 2508: debugging, we want more checking, in production we sometimes care more
 2509: for speed. Therefore, assertions can be turned off, i.e., the assertion
 2510: becomes a comment. Depending on the importance of an assertion and the
 2511: time it takes to check it, you may want to turn off some assertions and
 2512: keep others turned on. Gforth provides several levels of assertions for
 2513: this purpose:
 2514: 
 2515: doc-assert0(
 2516: doc-assert1(
 2517: doc-assert2(
 2518: doc-assert3(
 2519: doc-assert(
 2520: doc-)
 2521: 
 2522: @code{Assert(} is the same as @code{assert1(}. The variable
 2523: @code{assert-level} specifies the highest assertions that are turned
 2524: on. I.e., at the default @code{assert-level} of one, @code{assert0(} and
 2525: @code{assert1(} assertions perform checking, while @code{assert2(} and
 2526: @code{assert3(} assertions are treated as comments.
 2527: 
 2528: Note that the @code{assert-level} is evaluated at compile-time, not at
 2529: run-time. I.e., you cannot turn assertions on or off at run-time, you
 2530: have to set the @code{assert-level} appropriately before compiling a
 2531: piece of code. You can compile several pieces of code at several
 2532: @code{assert-level}s (e.g., a trusted library at level 1 and newly
 2533: written code at level 3).
 2534: 
 2535: doc-assert-level
 2536: 
 2537: If an assertion fails, a message compatible with Emacs' compilation mode
 2538: is produced and the execution is aborted (currently with @code{ABORT"}.
 2539: If there is interest, we will introduce a special throw code. But if you
 2540: intend to @code{catch} a specific condition, using @code{throw} is
 2541: probably more appropriate than an assertion).
 2542: 
 2543: @node Assembler and Code words, Threading Words, Programming Tools, Words
 2544: @section Assembler and Code words
 2545: 
 2546: Gforth provides some words for defining primitives (words written in
 2547: machine code), and for defining the the machine-code equivalent of
 2548: @code{DOES>}-based defining words. However, the machine-independent
 2549: nature of Gforth poses a few problems: First of all. Gforth runs on
 2550: several architectures, so it can provide no standard assembler. What's
 2551: worse is that the register allocation not only depends on the processor,
 2552: but also on the @code{gcc} version and options used.
 2553: 
 2554: The words that Gforth offers encapsulate some system dependences (e.g., the
 2555: header structure), so a system-independent assembler may be used in
 2556: Gforth. If you do not have an assembler, you can compile machine code
 2557: directly with @code{,} and @code{c,}.
 2558: 
 2559: doc-assembler
 2560: doc-code
 2561: doc-end-code
 2562: doc-;code
 2563: doc-flush-icache
 2564: 
 2565: If @code{flush-icache} does not work correctly, @code{code} words
 2566: etc. will not work (reliably), either.
 2567: 
 2568: These words are rarely used. Therefore they reside in @code{code.fs},
 2569: which is usually not loaded (except @code{flush-icache}, which is always
 2570: present). You can load them with @code{require code.fs}.
 2571: 
 2572: In the assembly code you will want to refer to the inner interpreter's
 2573: registers (e.g., the data stack pointer) and you may want to use other
 2574: registers for temporary storage. Unfortunately, the register allocation
 2575: is installation-dependent.
 2576: 
 2577: The easiest solution is to use explicit register declarations
 2578: (@pxref{Explicit Reg Vars, , Variables in Specified Registers, gcc.info,
 2579: GNU C Manual}) for all of the inner interpreter's registers: You have to
 2580: compile Gforth with @code{-DFORCE_REG} (configure option
 2581: @code{--enable-force-reg}) and the appropriate declarations must be
 2582: present in the @code{machine.h} file (see @code{mips.h} for an example;
 2583: you can find a full list of all declarable register symbols with
 2584: @code{grep register engine.c}). If you give explicit registers to all
 2585: variables that are declared at the beginning of @code{engine()}, you
 2586: should be able to use the other caller-saved registers for temporary
 2587: storage. Alternatively, you can use the @code{gcc} option
 2588: @code{-ffixed-REG} (@pxref{Code Gen Options, , Options for Code
 2589: Generation Conventions, gcc.info, GNU C Manual}) to reserve a register
 2590: (however, this restriction on register allocation may slow Gforth
 2591: significantly).
 2592: 
 2593: If this solution is not viable (e.g., because @code{gcc} does not allow
 2594: you to explicitly declare all the registers you need), you have to find
 2595: out by looking at the code where the inner interpreter's registers
 2596: reside and which registers can be used for temporary storage. You can
 2597: get an assembly listing of the engine's code with @code{make engine.s}.
 2598: 
 2599: In any case, it is good practice to abstract your assembly code from the
 2600: actual register allocation. E.g., if the data stack pointer resides in
 2601: register @code{$17}, create an alias for this register called @code{sp},
 2602: and use that in your assembly code.
 2603: 
 2604: Another option for implementing normal and defining words efficiently
 2605: is: adding the wanted functionality to the source of Gforth. For normal
 2606: words you just have to edit @file{primitives} (@pxref{Automatic
 2607: Generation}), defining words (equivalent to @code{;CODE} words, for fast
 2608: defined words) may require changes in @file{engine.c}, @file{kernal.fs},
 2609: @file{prims2x.fs}, and possibly @file{cross.fs}.
 2610: 
 2611: 
 2612: @node Threading Words,  , Assembler and Code words, Words
 2613: @section Threading Words
 2614: 
 2615: These words provide access to code addresses and other threading stuff
 2616: in Gforth (and, possibly, other interpretive Forths). It more or less
 2617: abstracts away the differences between direct and indirect threading
 2618: (and, for direct threading, the machine dependences). However, at
 2619: present this wordset is still inclomplete. It is also pretty low-level;
 2620: some day it will hopefully be made unnecessary by an internals words set
 2621: that abstracts implementation details away completely.
 2622: 
 2623: doc->code-address
 2624: doc->does-code
 2625: doc-code-address!
 2626: doc-does-code!
 2627: doc-does-handler!
 2628: doc-/does-handler
 2629: 
 2630: The code addresses produced by various defining words are produced by
 2631: the following words:
 2632: 
 2633: doc-docol:
 2634: doc-docon:
 2635: doc-dovar:
 2636: doc-douser:
 2637: doc-dodefer:
 2638: doc-dofield:
 2639: 
 2640: You can recognize words defined by a @code{CREATE}...@code{DOES>} word
 2641: with @code{>DOES-CODE}. If the word was defined in that way, the value
 2642: returned is different from 0 and identifies the @code{DOES>} used by the
 2643: defining word.
 2644: 
 2645: @node ANS conformance, Model, Words, Top
 2646: @chapter ANS conformance
 2647: 
 2648: To the best of our knowledge, Gforth is an
 2649: 
 2650: ANS Forth System
 2651: @itemize @bullet
 2652: @item providing the Core Extensions word set
 2653: @item providing the Block word set
 2654: @item providing the Block Extensions word set
 2655: @item providing the Double-Number word set
 2656: @item providing the Double-Number Extensions word set
 2657: @item providing the Exception word set
 2658: @item providing the Exception Extensions word set
 2659: @item providing the Facility word set
 2660: @item providing @code{MS} and @code{TIME&DATE} from the Facility Extensions word set
 2661: @item providing the File Access word set
 2662: @item providing the File Access Extensions word set
 2663: @item providing the Floating-Point word set
 2664: @item providing the Floating-Point Extensions word set
 2665: @item providing the Locals word set
 2666: @item providing the Locals Extensions word set
 2667: @item providing the Memory-Allocation word set
 2668: @item providing the Memory-Allocation Extensions word set (that one's easy)
 2669: @item providing the Programming-Tools word set
 2670: @item providing @code{;CODE}, @code{AHEAD}, @code{ASSEMBLER}, @code{BYE}, @code{CODE}, @code{CS-PICK}, @code{CS-ROLL}, @code{STATE}, @code{[ELSE]}, @code{[IF]}, @code{[THEN]} from the Programming-Tools Extensions word set
 2671: @item providing the Search-Order word set
 2672: @item providing the Search-Order Extensions word set
 2673: @item providing the String word set
 2674: @item providing the String Extensions word set (another easy one)
 2675: @end itemize
 2676: 
 2677: In addition, ANS Forth systems are required to document certain
 2678: implementation choices. This chapter tries to meet these
 2679: requirements. In many cases it gives a way to ask the system for the
 2680: information instead of providing the information directly, in
 2681: particular, if the information depends on the processor, the operating
 2682: system or the installation options chosen, or if they are likely to
 2683: change during the maintenance of Gforth.
 2684: 
 2685: @comment The framework for the rest has been taken from pfe.
 2686: 
 2687: @menu
 2688: * The Core Words::              
 2689: * The optional Block word set::  
 2690: * The optional Double Number word set::  
 2691: * The optional Exception word set::  
 2692: * The optional Facility word set::  
 2693: * The optional File-Access word set::  
 2694: * The optional Floating-Point word set::  
 2695: * The optional Locals word set::  
 2696: * The optional Memory-Allocation word set::  
 2697: * The optional Programming-Tools word set::  
 2698: * The optional Search-Order word set::  
 2699: @end menu
 2700: 
 2701: 
 2702: @c =====================================================================
 2703: @node The Core Words, The optional Block word set, ANS conformance, ANS conformance
 2704: @comment  node-name,  next,  previous,  up
 2705: @section The Core Words
 2706: @c =====================================================================
 2707: 
 2708: @menu
 2709: * core-idef::                   Implementation Defined Options                   
 2710: * core-ambcond::                Ambiguous Conditions                
 2711: * core-other::                  Other System Documentation                  
 2712: @end menu
 2713: 
 2714: @c ---------------------------------------------------------------------
 2715: @node core-idef, core-ambcond, The Core Words, The Core Words
 2716: @subsection Implementation Defined Options
 2717: @c ---------------------------------------------------------------------
 2718: 
 2719: @table @i
 2720: 
 2721: @item (Cell) aligned addresses:
 2722: processor-dependent. Gforth's alignment words perform natural alignment
 2723: (e.g., an address aligned for a datum of size 8 is divisible by
 2724: 8). Unaligned accesses usually result in a @code{-23 THROW}.
 2725: 
 2726: @item @code{EMIT} and non-graphic characters:
 2727: The character is output using the C library function (actually, macro)
 2728: @code{putc}.
 2729: 
 2730: @item character editing of @code{ACCEPT} and @code{EXPECT}:
 2731: This is modeled on the GNU readline library (@pxref{Readline
 2732: Interaction, , Command Line Editing, readline, The GNU Readline
 2733: Library}) with Emacs-like key bindings. @kbd{Tab} deviates a little by
 2734: producing a full word completion every time you type it (instead of
 2735: producing the common prefix of all completions).
 2736: 
 2737: @item character set:
 2738: The character set of your computer and display device. Gforth is
 2739: 8-bit-clean (but some other component in your system may make trouble).
 2740: 
 2741: @item Character-aligned address requirements:
 2742: installation-dependent. Currently a character is represented by a C
 2743: @code{unsigned char}; in the future we might switch to @code{wchar_t}
 2744: (Comments on that requested).
 2745: 
 2746: @item character-set extensions and matching of names:
 2747: Any character except the ASCII NUL charcter can be used in a
 2748: name. Matching is case-insensitive (except in @code{TABLE}s. The
 2749: matching is performed using the C function @code{strncasecmp}, whose
 2750: function is probably influenced by the locale. E.g., the @code{C} locale
 2751: does not know about accents and umlauts, so they are matched
 2752: case-sensitively in that locale. For portability reasons it is best to
 2753: write programs such that they work in the @code{C} locale. Then one can
 2754: use libraries written by a Polish programmer (who might use words
 2755: containing ISO Latin-2 encoded characters) and by a French programmer
 2756: (ISO Latin-1) in the same program (of course, @code{WORDS} will produce
 2757: funny results for some of the words (which ones, depends on the font you
 2758: are using)). Also, the locale you prefer may not be available in other
 2759: operating systems. Hopefully, Unicode will solve these problems one day.
 2760: 
 2761: @item conditions under which control characters match a space delimiter:
 2762: If @code{WORD} is called with the space character as a delimiter, all
 2763: white-space characters (as identified by the C macro @code{isspace()})
 2764: are delimiters. @code{PARSE}, on the other hand, treats space like other
 2765: delimiters. @code{PARSE-WORD} treats space like @code{WORD}, but behaves
 2766: like @code{PARSE} otherwise. @code{(NAME)}, which is used by the outer
 2767: interpreter (aka text interpreter) by default, treats all white-space
 2768: characters as delimiters.
 2769: 
 2770: @item format of the control flow stack:
 2771: The data stack is used as control flow stack. The size of a control flow
 2772: stack item in cells is given by the constant @code{cs-item-size}. At the
 2773: time of this writing, an item consists of a (pointer to a) locals list
 2774: (third), an address in the code (second), and a tag for identifying the
 2775: item (TOS). The following tags are used: @code{defstart},
 2776: @code{live-orig}, @code{dead-orig}, @code{dest}, @code{do-dest},
 2777: @code{scopestart}.
 2778: 
 2779: @item conversion of digits > 35
 2780: The characters @code{[\]^_'} are the digits with the decimal value
 2781: 36@minus{}41. There is no way to input many of the larger digits.
 2782: 
 2783: @item display after input terminates in @code{ACCEPT} and @code{EXPECT}:
 2784: The cursor is moved to the end of the entered string. If the input is
 2785: terminated using the @kbd{Return} key, a space is typed.
 2786: 
 2787: @item exception abort sequence of @code{ABORT"}:
 2788: The error string is stored into the variable @code{"error} and a
 2789: @code{-2 throw} is performed.
 2790: 
 2791: @item input line terminator:
 2792: For interactive input, @kbd{C-m} (CR) and @kbd{C-j} (LF) terminate
 2793: lines. One of these characters is typically produced when you type the
 2794: @kbd{Enter} or @kbd{Return} key.
 2795: 
 2796: @item maximum size of a counted string:
 2797: @code{s" /counted-string" environment? drop .}. Currently 255 characters
 2798: on all ports, but this may change.
 2799: 
 2800: @item maximum size of a parsed string:
 2801: Given by the constant @code{/line}. Currently 255 characters.
 2802: 
 2803: @item maximum size of a definition name, in characters:
 2804: 31
 2805: 
 2806: @item maximum string length for @code{ENVIRONMENT?}, in characters:
 2807: 31
 2808: 
 2809: @item method of selecting the user input device:
 2810: The user input device is the standard input. There is currently no way to
 2811: change it from within Gforth. However, the input can typically be
 2812: redirected in the command line that starts Gforth.
 2813: 
 2814: @item method of selecting the user output device:
 2815: @code{EMIT} and @code{TYPE} output to the file-id stored in the value
 2816: @code{outfile-id} (@code{stdout} by default). Gforth uses buffered
 2817: output, so output on a terminal does not become visible before the next
 2818: newline or buffer overflow. Output on non-terminals is invisible until
 2819: the buffer overflows.
 2820: 
 2821: @item methods of dictionary compilation:
 2822: What are we expected to document here?
 2823: 
 2824: @item number of bits in one address unit:
 2825: @code{s" address-units-bits" environment? drop .}. 8 in all current
 2826: ports.
 2827: 
 2828: @item number representation and arithmetic:
 2829: Processor-dependent. Binary two's complement on all current ports.
 2830: 
 2831: @item ranges for integer types:
 2832: Installation-dependent. Make environmental queries for @code{MAX-N},
 2833: @code{MAX-U}, @code{MAX-D} and @code{MAX-UD}. The lower bounds for
 2834: unsigned (and positive) types is 0. The lower bound for signed types on
 2835: two's complement and one's complement machines machines can be computed
 2836: by adding 1 to the upper bound.
 2837: 
 2838: @item read-only data space regions:
 2839: The whole Forth data space is writable.
 2840: 
 2841: @item size of buffer at @code{WORD}:
 2842: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
 2843: shared with the pictured numeric output string. If overwriting
 2844: @code{PAD} is acceptable, it is as large as the remaining dictionary
 2845: space, although only as much can be sensibly used as fits in a counted
 2846: string.
 2847: 
 2848: @item size of one cell in address units:
 2849: @code{1 cells .}.
 2850: 
 2851: @item size of one character in address units:
 2852: @code{1 chars .}. 1 on all current ports.
 2853: 
 2854: @item size of the keyboard terminal buffer:
 2855: Varies. You can determine the size at a specific time using @code{lp@@
 2856: tib - .}. It is shared with the locals stack and TIBs of files that
 2857: include the current file. You can change the amount of space for TIBs
 2858: and locals stack at Gforth startup with the command line option
 2859: @code{-l}.
 2860: 
 2861: @item size of the pictured numeric output buffer:
 2862: @code{PAD HERE - .}. 104 characters on 32-bit machines. The buffer is
 2863: shared with @code{WORD}.
 2864: 
 2865: @item size of the scratch area returned by @code{PAD}:
 2866: The remainder of dictionary space. You can even use the unused part of
 2867: the data stack space. The current size can be computed with @code{sp@@
 2868: pad - .}.
 2869: 
 2870: @item system case-sensitivity characteristics:
 2871: Dictionary searches are case insensitive (except in
 2872: @code{TABLE}s). However, as explained above under @i{character-set
 2873: extensions}, the matching for non-ASCII characters is determined by the
 2874: locale you are using. In the default @code{C} locale all non-ASCII
 2875: characters are matched case-sensitively.
 2876: 
 2877: @item system prompt:
 2878: @code{ ok} in interpret state, @code{ compiled} in compile state.
 2879: 
 2880: @item division rounding:
 2881: installation dependent. @code{s" floored" environment? drop .}. We leave
 2882: the choice to @code{gcc} (what to use for @code{/}) and to you (whether to use
 2883: @code{fm/mod}, @code{sm/rem} or simply @code{/}).
 2884: 
 2885: @item values of @code{STATE} when true:
 2886: -1.
 2887: 
 2888: @item values returned after arithmetic overflow:
 2889: On two's complement machines, arithmetic is performed modulo
 2890: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
 2891: arithmetic (with appropriate mapping for signed types). Division by zero
 2892: typically results in a @code{-55 throw} (Floating-point unidentified
 2893: fault), although a @code{-10 throw} (divide by zero) would be more
 2894: appropriate.
 2895: 
 2896: @item whether the current definition can be found after @t{DOES>}:
 2897: No.
 2898: 
 2899: @end table
 2900: 
 2901: @c ---------------------------------------------------------------------
 2902: @node core-ambcond, core-other, core-idef, The Core Words
 2903: @subsection Ambiguous conditions
 2904: @c ---------------------------------------------------------------------
 2905: 
 2906: @table @i
 2907: 
 2908: @item a name is neither a word nor a number:
 2909: @code{-13 throw} (Undefined word). Actually, @code{-13 bounce}, which
 2910: preserves the data and FP stack, so you don't lose more work than
 2911: necessary.
 2912: 
 2913: @item a definition name exceeds the maximum length allowed:
 2914: @code{-19 throw} (Word name too long)
 2915: 
 2916: @item addressing a region not inside the various data spaces of the forth system:
 2917: The stacks, code space and name space are accessible. Machine code space is
 2918: typically readable. Accessing other addresses gives results dependent on
 2919: the operating system. On decent systems: @code{-9 throw} (Invalid memory
 2920: address).
 2921: 
 2922: @item argument type incompatible with parameter:
 2923: This is usually not caught. Some words perform checks, e.g., the control
 2924: flow words, and issue a @code{ABORT"} or @code{-12 THROW} (Argument type
 2925: mismatch).
 2926: 
 2927: @item attempting to obtain the execution token of a word with undefined execution semantics:
 2928: @code{-14 throw} (Interpreting a compile-only word). In some cases, you
 2929: get an execution token for @code{compile-only-error} (which performs a
 2930: @code{-14 throw} when executed).
 2931: 
 2932: @item dividing by zero:
 2933: typically results in a @code{-55 throw} (floating point unidentified
 2934: fault), although a @code{-10 throw} (divide by zero) would be more
 2935: appropriate.
 2936: 
 2937: @item insufficient data stack or return stack space:
 2938: Not checked. This typically results in mysterious illegal memory
 2939: accesses, producing @code{-9 throw} (Invalid memory address) or
 2940: @code{-23 throw} (Address alignment exception).
 2941: 
 2942: @item insufficient space for loop control parameters:
 2943: like other return stack overflows.
 2944: 
 2945: @item insufficient space in the dictionary:
 2946: Not checked. Similar results as stack overflows. However, typically the
 2947: error appears at a different place when one inserts or removes code.
 2948: 
 2949: @item interpreting a word with undefined interpretation semantics:
 2950: For some words, we defined interpretation semantics. For the others:
 2951: @code{-14 throw} (Interpreting a compile-only word).
 2952: 
 2953: @item modifying the contents of the input buffer or a string literal:
 2954: These are located in writable memory and can be modified.
 2955: 
 2956: @item overflow of the pictured numeric output string:
 2957: Not checked.
 2958: 
 2959: @item parsed string overflow:
 2960: @code{PARSE} cannot overflow. @code{WORD} does not check for overflow.
 2961: 
 2962: @item producing a result out of range:
 2963: On two's complement machines, arithmetic is performed modulo
 2964: 2**bits-per-cell for single arithmetic and 4**bits-per-cell for double
 2965: arithmetic (with appropriate mapping for signed types). Division by zero
 2966: typically results in a @code{-55 throw} (floatingpoint unidentified
 2967: fault), although a @code{-10 throw} (divide by zero) would be more
 2968: appropriate. @code{convert} and @code{>number} currently overflow
 2969: silently.
 2970: 
 2971: @item reading from an empty data or return stack:
 2972: The data stack is checked by the outer (aka text) interpreter after
 2973: every word executed. If it has underflowed, a @code{-4 throw} (Stack
 2974: underflow) is performed. Apart from that, the stacks are not checked and
 2975: underflows can result in similar behaviour as overflows (of adjacent
 2976: stacks).
 2977: 
 2978: @item unexpected end of the input buffer, resulting in an attempt to use a zero-length string as a name:
 2979: @code{Create} and its descendants perform a @code{-16 throw} (Attempt to
 2980: use zero-length string as a name). Words like @code{'} probably will not
 2981: find what they search. Note that it is possible to create zero-length
 2982: names with @code{nextname} (should it not?).
 2983: 
 2984: @item @code{>IN} greater than input buffer:
 2985: The next invocation of a parsing word returns a string wih length 0.
 2986: 
 2987: @item @code{RECURSE} appears after @code{DOES>}:
 2988: Compiles a recursive call to the defining word, not to the defined word.
 2989: 
 2990: @item argument input source different than current input source for @code{RESTORE-INPUT}:
 2991: @code{-12 THROW}. Note that, once an input file is closed (e.g., because
 2992: the end of the file was reached), its source-id may be
 2993: reused. Therefore, restoring an input source specification referencing a
 2994: closed file may lead to unpredictable results instead of a @code{-12
 2995: THROW}.
 2996: 
 2997: In the future, Gforth may be able to restore input source specifications
 2998: from other than the current input soruce.
 2999: 
 3000: @item data space containing definitions gets de-allocated:
 3001: Deallocation with @code{allot} is not checked. This typically resuls in
 3002: memory access faults or execution of illegal instructions.
 3003: 
 3004: @item data space read/write with incorrect alignment:
 3005: Processor-dependent. Typically results in a @code{-23 throw} (Address
 3006: alignment exception). Under Linux on a 486 or later processor with
 3007: alignment turned on, incorrect alignment results in a @code{-9 throw}
 3008: (Invalid memory address). There are reportedly some processors with
 3009: alignment restrictions that do not report them.
 3010: 
 3011: @item data space pointer not properly aligned, @code{,}, @code{C,}:
 3012: Like other alignment errors.
 3013: 
 3014: @item less than u+2 stack items (@code{PICK} and @code{ROLL}):
 3015: Not checked. May cause an illegal memory access.
 3016: 
 3017: @item loop control parameters not available:
 3018: Not checked. The counted loop words simply assume that the top of return
 3019: stack items are loop control parameters and behave accordingly.
 3020: 
 3021: @item most recent definition does not have a name (@code{IMMEDIATE}):
 3022: @code{abort" last word was headerless"}.
 3023: 
 3024: @item name not defined by @code{VALUE} used by @code{TO}:
 3025: @code{-32 throw} (Invalid name argument) (unless name was defined by
 3026: @code{CONSTANT}; then it just changes the constant).
 3027: 
 3028: @item name not found (@code{'}, @code{POSTPONE}, @code{[']}, @code{[COMPILE]}):
 3029: @code{-13 throw} (Undefined word)
 3030: 
 3031: @item parameters are not of the same type (@code{DO}, @code{?DO}, @code{WITHIN}):
 3032: Gforth behaves as if they were of the same type. I.e., you can predict
 3033: the behaviour by interpreting all parameters as, e.g., signed.
 3034: 
 3035: @item @code{POSTPONE} or @code{[COMPILE]} applied to @code{TO}:
 3036: Assume @code{: X POSTPONE TO ; IMMEDIATE}. @code{X} performs the
 3037: compilation semantics of @code{TO}.
 3038: 
 3039: @item String longer than a counted string returned by @code{WORD}:
 3040: Not checked. The string will be ok, but the count will, of course,
 3041: contain only the least significant bits of the length.
 3042: 
 3043: @item u greater than or equal to the number of bits in a cell (@code{LSHIFT}, @code{RSHIFT}):
 3044: Processor-dependent. Typical behaviours are returning 0 and using only
 3045: the low bits of the shift count.
 3046: 
 3047: @item word not defined via @code{CREATE}:
 3048: @code{>BODY} produces the PFA of the word no matter how it was defined.
 3049: 
 3050: @code{DOES>} changes the execution semantics of the last defined word no
 3051: matter how it was defined. E.g., @code{CONSTANT DOES>} is equivalent to
 3052: @code{CREATE , DOES>}.
 3053: 
 3054: @item words improperly used outside @code{<#} and @code{#>}:
 3055: Not checked. As usual, you can expect memory faults.
 3056: 
 3057: @end table
 3058: 
 3059: 
 3060: @c ---------------------------------------------------------------------
 3061: @node core-other,  , core-ambcond, The Core Words
 3062: @subsection Other system documentation
 3063: @c ---------------------------------------------------------------------
 3064: 
 3065: @table @i
 3066: 
 3067: @item nonstandard words using @code{PAD}:
 3068: None.
 3069: 
 3070: @item operator's terminal facilities available:
 3071: After processing the command line, Gforth goes into interactive mode,
 3072: and you can give commands to Gforth interactively. The actual facilities
 3073: available depend on how you invoke Gforth.
 3074: 
 3075: @item program data space available:
 3076: @code{sp@@ here - .} gives the space remaining for dictionary and data
 3077: stack together.
 3078: 
 3079: @item return stack space available:
 3080: By default 16 KBytes. The default can be overridden with the @code{-r}
 3081: switch (@pxref{Invocation}) when Gforth starts up.
 3082: 
 3083: @item stack space available:
 3084: @code{sp@@ here - .} gives the space remaining for dictionary and data
 3085: stack together.
 3086: 
 3087: @item system dictionary space required, in address units:
 3088: Type @code{here forthstart - .} after startup. At the time of this
 3089: writing, this gives 70108 (bytes) on a 32-bit system.
 3090: @end table
 3091: 
 3092: 
 3093: @c =====================================================================
 3094: @node The optional Block word set, The optional Double Number word set, The Core Words, ANS conformance
 3095: @section The optional Block word set
 3096: @c =====================================================================
 3097: 
 3098: @menu
 3099: * block-idef::                  Implementation Defined Options                  
 3100: * block-ambcond::               Ambiguous Conditions               
 3101: * block-other::                 Other System Documentation                 
 3102: @end menu
 3103: 
 3104: 
 3105: @c ---------------------------------------------------------------------
 3106: @node block-idef, block-ambcond, The optional Block word set, The optional Block word set
 3107: @subsection Implementation Defined Options
 3108: @c ---------------------------------------------------------------------
 3109: 
 3110: @table @i
 3111: 
 3112: @item the format for display by @code{LIST}:
 3113: First the screen number is displayed, then 16 lines of 64 characters,
 3114: each line preceded by the line number.
 3115: 
 3116: @item the length of a line affected by @code{\}:
 3117: 64 characters.
 3118: @end table
 3119: 
 3120: 
 3121: @c ---------------------------------------------------------------------
 3122: @node block-ambcond, block-other, block-idef, The optional Block word set
 3123: @subsection Ambiguous conditions
 3124: @c ---------------------------------------------------------------------
 3125: 
 3126: @table @i
 3127: 
 3128: @item correct block read was not possible:
 3129: Typically results in a @code{throw} of some OS-derived value (between
 3130: -512 and -2048). If the blocks file was just not long enough, blanks are
 3131: supplied for the missing portion.
 3132: 
 3133: @item I/O exception in block transfer:
 3134: Typically results in a @code{throw} of some OS-derived value (between
 3135: -512 and -2048).
 3136: 
 3137: @item invalid block number:
 3138: @code{-35 throw} (Invalid block number)
 3139: 
 3140: @item a program directly alters the contents of @code{BLK}:
 3141: The input stream is switched to that other block, at the same
 3142: position. If the storing to @code{BLK} happens when interpreting
 3143: non-block input, the system will get quite confused when the block ends.
 3144: 
 3145: @item no current block buffer for @code{UPDATE}:
 3146: @code{UPDATE} has no effect.
 3147: 
 3148: @end table
 3149: 
 3150: 
 3151: @c ---------------------------------------------------------------------
 3152: @node block-other,  , block-ambcond, The optional Block word set
 3153: @subsection Other system documentation
 3154: @c ---------------------------------------------------------------------
 3155: 
 3156: @table @i
 3157: 
 3158: @item any restrictions a multiprogramming system places on the use of buffer addresses:
 3159: No restrictions (yet).
 3160: 
 3161: @item the number of blocks available for source and data:
 3162: depends on your disk space.
 3163: 
 3164: @end table
 3165: 
 3166: 
 3167: @c =====================================================================
 3168: @node The optional Double Number word set, The optional Exception word set, The optional Block word set, ANS conformance
 3169: @section The optional Double Number word set
 3170: @c =====================================================================
 3171: 
 3172: @menu
 3173: * double-ambcond::              Ambiguous Conditions              
 3174: @end menu
 3175: 
 3176: 
 3177: @c ---------------------------------------------------------------------
 3178: @node double-ambcond,  , The optional Double Number word set, The optional Double Number word set
 3179: @subsection Ambiguous conditions
 3180: @c ---------------------------------------------------------------------
 3181: 
 3182: @table @i
 3183: 
 3184: @item @var{d} outside of range of @var{n} in @code{D>S}:
 3185: The least significant cell of @var{d} is produced.
 3186: 
 3187: @end table
 3188: 
 3189: 
 3190: @c =====================================================================
 3191: @node The optional Exception word set, The optional Facility word set, The optional Double Number word set, ANS conformance
 3192: @section The optional Exception word set
 3193: @c =====================================================================
 3194: 
 3195: @menu
 3196: * exception-idef::              Implementation Defined Options              
 3197: @end menu
 3198: 
 3199: 
 3200: @c ---------------------------------------------------------------------
 3201: @node exception-idef,  , The optional Exception word set, The optional Exception word set
 3202: @subsection Implementation Defined Options
 3203: @c ---------------------------------------------------------------------
 3204: 
 3205: @table @i
 3206: @item @code{THROW}-codes used in the system:
 3207: The codes -256@minus{}-511 are used for reporting signals (see
 3208: @file{errore.fs}). The codes -512@minus{}-2047 are used for OS errors
 3209: (for file and memory allocation operations). The mapping from OS error
 3210: numbers to throw code is -512@minus{}@code{errno}. One side effect of
 3211: this mapping is that undefined OS errors produce a message with a
 3212: strange number; e.g., @code{-1000 THROW} results in @code{Unknown error
 3213: 488} on my system.
 3214: @end table
 3215: 
 3216: @c =====================================================================
 3217: @node The optional Facility word set, The optional File-Access word set, The optional Exception word set, ANS conformance
 3218: @section The optional Facility word set
 3219: @c =====================================================================
 3220: 
 3221: @menu
 3222: * facility-idef::               Implementation Defined Options               
 3223: * facility-ambcond::            Ambiguous Conditions            
 3224: @end menu
 3225: 
 3226: 
 3227: @c ---------------------------------------------------------------------
 3228: @node facility-idef, facility-ambcond, The optional Facility word set, The optional Facility word set
 3229: @subsection Implementation Defined Options
 3230: @c ---------------------------------------------------------------------
 3231: 
 3232: @table @i
 3233: 
 3234: @item encoding of keyboard events (@code{EKEY}):
 3235: Not yet implemeted.
 3236: 
 3237: @item duration of a system clock tick
 3238: System dependent. With respect to @code{MS}, the time is specified in
 3239: microseconds. How well the OS and the hardware implement this, is
 3240: another question.
 3241: 
 3242: @item repeatability to be expected from the execution of @code{MS}:
 3243: System dependent. On Unix, a lot depends on load. If the system is
 3244: lightly loaded, and the delay is short enough that Gforth does not get
 3245: swapped out, the performance should be acceptable. Under MS-DOS and
 3246: other single-tasking systems, it should be good.
 3247: 
 3248: @end table
 3249: 
 3250: 
 3251: @c ---------------------------------------------------------------------
 3252: @node facility-ambcond,  , facility-idef, The optional Facility word set
 3253: @subsection Ambiguous conditions
 3254: @c ---------------------------------------------------------------------
 3255: 
 3256: @table @i
 3257: 
 3258: @item @code{AT-XY} can't be performed on user output device:
 3259: Largely terminal dependant. No range checks are done on the arguments.
 3260: No errors are reported. You may see some garbage appearing, you may see
 3261: simply nothing happen.
 3262: 
 3263: @end table
 3264: 
 3265: 
 3266: @c =====================================================================
 3267: @node The optional File-Access word set, The optional Floating-Point word set, The optional Facility word set, ANS conformance
 3268: @section The optional File-Access word set
 3269: @c =====================================================================
 3270: 
 3271: @menu
 3272: * file-idef::                   Implementation Defined Options                   
 3273: * file-ambcond::                Ambiguous Conditions                
 3274: @end menu
 3275: 
 3276: 
 3277: @c ---------------------------------------------------------------------
 3278: @node file-idef, file-ambcond, The optional File-Access word set, The optional File-Access word set
 3279: @subsection Implementation Defined Options
 3280: @c ---------------------------------------------------------------------
 3281: 
 3282: @table @i
 3283: 
 3284: @item File access methods used:
 3285: @code{R/O}, @code{R/W} and @code{BIN} work as you would
 3286: expect. @code{W/O} translates into the C file opening mode @code{w} (or
 3287: @code{wb}): The file is cleared, if it exists, and created, if it does
 3288: not (both with @code{open-file} and @code{create-file}).  Under Unix
 3289: @code{create-file} creates a file with 666 permissions modified by your
 3290: umask.
 3291: 
 3292: @item file exceptions:
 3293: The file words do not raise exceptions (except, perhaps, memory access
 3294: faults when you pass illegal addresses or file-ids).
 3295: 
 3296: @item file line terminator:
 3297: System-dependent. Gforth uses C's newline character as line
 3298: terminator. What the actual character code(s) of this are is
 3299: system-dependent.
 3300: 
 3301: @item file name format
 3302: System dependent. Gforth just uses the file name format of your OS.
 3303: 
 3304: @item information returned by @code{FILE-STATUS}:
 3305: @code{FILE-STATUS} returns the most powerful file access mode allowed
 3306: for the file: Either @code{R/O}, @code{W/O} or @code{R/W}. If the file
 3307: cannot be accessed, @code{R/O BIN} is returned. @code{BIN} is applicable
 3308: along with the retured mode.
 3309: 
 3310: @item input file state after an exception when including source:
 3311: All files that are left via the exception are closed.
 3312: 
 3313: @item @var{ior} values and meaning:
 3314: The @var{ior}s returned by the file and memory allocation words are
 3315: intended as throw codes. They typically are in the range
 3316: -512@minus{}-2047 of OS errors.  The mapping from OS error numbers to
 3317: @var{ior}s is -512@minus{}@var{errno}.
 3318: 
 3319: @item maximum depth of file input nesting:
 3320: limited by the amount of return stack, locals/TIB stack, and the number
 3321: of open files available. This should not give you troubles.
 3322: 
 3323: @item maximum size of input line:
 3324: @code{/line}. Currently 255.
 3325: 
 3326: @item methods of mapping block ranges to files:
 3327: By default, blocks are accessed in the file @file{blocks.fb} in the
 3328: current working directory. The file can be switched with @code{USE}.
 3329: 
 3330: @item number of string buffers provided by @code{S"}:
 3331: 1
 3332: 
 3333: @item size of string buffer used by @code{S"}:
 3334: @code{/line}. currently 255.
 3335: 
 3336: @end table
 3337: 
 3338: @c ---------------------------------------------------------------------
 3339: @node file-ambcond,  , file-idef, The optional File-Access word set
 3340: @subsection Ambiguous conditions
 3341: @c ---------------------------------------------------------------------
 3342: 
 3343: @table @i
 3344: 
 3345: @item attempting to position a file outside it's boundaries:
 3346: @code{REPOSITION-FILE} is performed as usual: Afterwards,
 3347: @code{FILE-POSITION} returns the value given to @code{REPOSITION-FILE}.
 3348: 
 3349: @item attempting to read from file positions not yet written:
 3350: End-of-file, i.e., zero characters are read and no error is reported.
 3351: 
 3352: @item @var{file-id} is invalid (@code{INCLUDE-FILE}):
 3353: An appropriate exception may be thrown, but a memory fault or other
 3354: problem is more probable.
 3355: 
 3356: @item I/O exception reading or closing @var{file-id} (@code{include-file}, @code{included}):
 3357: The @var{ior} produced by the operation, that discovered the problem, is
 3358: thrown.
 3359: 
 3360: @item named file cannot be opened (@code{included}):
 3361: The @var{ior} produced by @code{open-file} is thrown.
 3362: 
 3363: @item requesting an unmapped block number:
 3364: There are no unmapped legal block numbers. On some operating systems,
 3365: writing a block with a large number may overflow the file system and
 3366: have an error message as consequence.
 3367: 
 3368: @item using @code{source-id} when @code{blk} is non-zero:
 3369: @code{source-id} performs its function. Typically it will give the id of
 3370: the source which loaded the block. (Better ideas?)
 3371: 
 3372: @end table
 3373: 
 3374: 
 3375: @c =====================================================================
 3376: @node  The optional Floating-Point word set, The optional Locals word set, The optional File-Access word set, ANS conformance
 3377: @section The optional Floating-Point word set
 3378: @c =====================================================================
 3379: 
 3380: @menu
 3381: * floating-idef::               Implementation Defined Options
 3382: * floating-ambcond::            Ambiguous Conditions            
 3383: @end menu
 3384: 
 3385: 
 3386: @c ---------------------------------------------------------------------
 3387: @node floating-idef, floating-ambcond, The optional Floating-Point word set, The optional Floating-Point word set
 3388: @subsection Implementation Defined Options
 3389: @c ---------------------------------------------------------------------
 3390: 
 3391: @table @i
 3392: 
 3393: @item format and range of floating point numbers:
 3394: System-dependent; the @code{double} type of C.
 3395: 
 3396: @item results of @code{REPRESENT} when @var{float} is out of range:
 3397: System dependent; @code{REPRESENT} is implemented using the C library
 3398: function @code{ecvt()} and inherits its behaviour in this respect.
 3399: 
 3400: @item rounding or truncation of floating-point numbers:
 3401: System dependent; the rounding behaviour is inherited from the hosting C
 3402: compiler. IEEE-FP-based (i.e., most) systems by default round to
 3403: nearest, and break ties by rounding to even (i.e., such that the last
 3404: bit of the mantissa is 0).
 3405: 
 3406: @item size of floating-point stack:
 3407: @code{s" FLOATING-STACK" environment? drop .}. Can be changed at startup
 3408: with the command-line option @code{-f}.
 3409: 
 3410: @item width of floating-point stack:
 3411: @code{1 floats}.
 3412: 
 3413: @end table
 3414: 
 3415: 
 3416: @c ---------------------------------------------------------------------
 3417: @node floating-ambcond,  , floating-idef, The optional Floating-Point word set
 3418: @subsection Ambiguous conditions
 3419: @c ---------------------------------------------------------------------
 3420: 
 3421: @table @i
 3422: 
 3423: @item @code{df@@} or @code{df!} used with an address that is not double-float  aligned:
 3424: System-dependent. Typically results in a @code{-23 THROW} like other
 3425: alignment violations.
 3426: 
 3427: @item @code{f@@} or @code{f!} used with an address that is not float  aligned:
 3428: System-dependent. Typically results in a @code{-23 THROW} like other
 3429: alignment violations.
 3430: 
 3431: @item Floating-point result out of range:
 3432: System-dependent. Can result in a @code{-55 THROW} (Floating-point
 3433: unidentified fault), or can produce a special value representing, e.g.,
 3434: Infinity.
 3435: 
 3436: @item @code{sf@@} or @code{sf!} used with an address that is not single-float  aligned:
 3437: System-dependent. Typically results in an alignment fault like other
 3438: alignment violations.
 3439: 
 3440: @item BASE is not decimal (@code{REPRESENT}, @code{F.}, @code{FE.}, @code{FS.}):
 3441: The floating-point number is converted into decimal nonetheless.
 3442: 
 3443: @item Both arguments are equal to zero (@code{FATAN2}):
 3444: System-dependent. @code{FATAN2} is implemented using the C library
 3445: function @code{atan2()}.
 3446: 
 3447: @item Using ftan on an argument @var{r1} where cos(@var{r1}) is zero:
 3448: System-dependent. Anyway, typically the cos of @var{r1} will not be zero
 3449: because of small errors and the tan will be a very large (or very small)
 3450: but finite number.
 3451: 
 3452: @item @var{d} cannot be presented precisely as a float in @code{D>F}:
 3453: The result is rounded to the nearest float.
 3454: 
 3455: @item dividing by zero:
 3456: @code{-55 throw} (Floating-point unidentified fault)
 3457: 
 3458: @item exponent too big for conversion (@code{DF!}, @code{DF@@}, @code{SF!}, @code{SF@@}):
 3459: System dependent. On IEEE-FP based systems the number is converted into
 3460: an infinity.
 3461: 
 3462: @item @var{float}<1 (@code{facosh}):
 3463: @code{-55 throw} (Floating-point unidentified fault)
 3464: 
 3465: @item @var{float}=<-1 (@code{flnp1}):
 3466: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
 3467: negative infinity is typically produced for @var{float}=-1.
 3468: 
 3469: @item @var{float}=<0 (@code{fln}, @code{flog}):
 3470: @code{-55 throw} (Floating-point unidentified fault). On IEEE-FP systems
 3471: negative infinity is typically produced for @var{float}=0.
 3472: 
 3473: @item @var{float}<0 (@code{fasinh}, @code{fsqrt}):
 3474: @code{-55 throw} (Floating-point unidentified fault). @code{fasinh}
 3475: produces values for these inputs on my Linux box (Bug in the C library?)
 3476: 
 3477: @item |@var{float}|>1 (@code{facos}, @code{fasin}, @code{fatanh}):
 3478: @code{-55 throw} (Floating-point unidentified fault).
 3479: 
 3480: @item integer part of float cannot be represented by @var{d} in @code{f>d}:
 3481: @code{-55 throw} (Floating-point unidentified fault).
 3482: 
 3483: @item string larger than pictured numeric output area (@code{f.}, @code{fe.}, @code{fs.}):
 3484: This does not happen.
 3485: @end table
 3486: 
 3487: 
 3488: 
 3489: @c =====================================================================
 3490: @node  The optional Locals word set, The optional Memory-Allocation word set, The optional Floating-Point word set, ANS conformance
 3491: @section The optional Locals word set
 3492: @c =====================================================================
 3493: 
 3494: @menu
 3495: * locals-idef::                 Implementation Defined Options                 
 3496: * locals-ambcond::              Ambiguous Conditions              
 3497: @end menu
 3498: 
 3499: 
 3500: @c ---------------------------------------------------------------------
 3501: @node locals-idef, locals-ambcond, The optional Locals word set, The optional Locals word set
 3502: @subsection Implementation Defined Options
 3503: @c ---------------------------------------------------------------------
 3504: 
 3505: @table @i
 3506: 
 3507: @item maximum number of locals in a definition:
 3508: @code{s" #locals" environment? drop .}. Currently 15. This is a lower
 3509: bound, e.g., on a 32-bit machine there can be 41 locals of up to 8
 3510: characters. The number of locals in a definition is bounded by the size
 3511: of locals-buffer, which contains the names of the locals.
 3512: 
 3513: @end table
 3514: 
 3515: 
 3516: @c ---------------------------------------------------------------------
 3517: @node locals-ambcond,  , locals-idef, The optional Locals word set
 3518: @subsection Ambiguous conditions
 3519: @c ---------------------------------------------------------------------
 3520: 
 3521: @table @i
 3522: 
 3523: @item executing a named local in interpretation state:
 3524: @code{-14 throw} (Interpreting a compile-only word).
 3525: 
 3526: @item @var{name} not defined by @code{VALUE} or @code{(LOCAL)} (@code{TO}):
 3527: @code{-32 throw} (Invalid name argument)
 3528: 
 3529: @end table
 3530: 
 3531: 
 3532: @c =====================================================================
 3533: @node  The optional Memory-Allocation word set, The optional Programming-Tools word set, The optional Locals word set, ANS conformance
 3534: @section The optional Memory-Allocation word set
 3535: @c =====================================================================
 3536: 
 3537: @menu
 3538: * memory-idef::                 Implementation Defined Options                 
 3539: @end menu
 3540: 
 3541: 
 3542: @c ---------------------------------------------------------------------
 3543: @node memory-idef,  , The optional Memory-Allocation word set, The optional Memory-Allocation word set
 3544: @subsection Implementation Defined Options
 3545: @c ---------------------------------------------------------------------
 3546: 
 3547: @table @i
 3548: 
 3549: @item values and meaning of @var{ior}:
 3550: The @var{ior}s returned by the file and memory allocation words are
 3551: intended as throw codes. They typically are in the range
 3552: -512@minus{}-2047 of OS errors.  The mapping from OS error numbers to
 3553: @var{ior}s is -512@minus{}@var{errno}.
 3554: 
 3555: @end table
 3556: 
 3557: @c =====================================================================
 3558: @node  The optional Programming-Tools word set, The optional Search-Order word set, The optional Memory-Allocation word set, ANS conformance
 3559: @section The optional Programming-Tools word set
 3560: @c =====================================================================
 3561: 
 3562: @menu
 3563: * programming-idef::            Implementation Defined Options            
 3564: * programming-ambcond::         Ambiguous Conditions         
 3565: @end menu
 3566: 
 3567: 
 3568: @c ---------------------------------------------------------------------
 3569: @node programming-idef, programming-ambcond, The optional Programming-Tools word set, The optional Programming-Tools word set
 3570: @subsection Implementation Defined Options
 3571: @c ---------------------------------------------------------------------
 3572: 
 3573: @table @i
 3574: 
 3575: @item ending sequence for input following @code{;code} and @code{code}:
 3576: @code{end-code}
 3577: 
 3578: @item manner of processing input following @code{;code} and @code{code}:
 3579: The @code{assembler} vocabulary is pushed on the search order stack, and
 3580: the input is processed by the text interpreter, (starting) in interpret
 3581: state.
 3582: 
 3583: @item search order capability for @code{EDITOR} and @code{ASSEMBLER}:
 3584: The ANS Forth search order word set.
 3585: 
 3586: @item source and format of display by @code{SEE}:
 3587: The source for @code{see} is the intermediate code used by the inner
 3588: interpreter.  The current @code{see} tries to output Forth source code
 3589: as well as possible.
 3590: 
 3591: @end table
 3592: 
 3593: @c ---------------------------------------------------------------------
 3594: @node programming-ambcond,  , programming-idef, The optional Programming-Tools word set
 3595: @subsection Ambiguous conditions
 3596: @c ---------------------------------------------------------------------
 3597: 
 3598: @table @i
 3599: 
 3600: @item deleting the compilation wordlist (@code{FORGET}):
 3601: Not implemented (yet).
 3602: 
 3603: @item fewer than @var{u}+1 items on the control flow stack (@code{CS-PICK}, @code{CS-ROLL}):
 3604: This typically results in an @code{abort"} with a descriptive error
 3605: message (may change into a @code{-22 throw} (Control structure mismatch)
 3606: in the future). You may also get a memory access error. If you are
 3607: unlucky, this ambiguous condition is not caught.
 3608: 
 3609: @item @var{name} can't be found (@code{forget}):
 3610: Not implemented (yet).
 3611: 
 3612: @item @var{name} not defined via @code{CREATE}:
 3613: @code{;code} behaves like @code{DOES>} in this respect, i.e., it changes
 3614: the execution semantics of the last defined word no matter how it was
 3615: defined.
 3616: 
 3617: @item @code{POSTPONE} applied to @code{[IF]}:
 3618: After defining @code{: X POSTPONE [IF] ; IMMEDIATE}. @code{X} is
 3619: equivalent to @code{[IF]}.
 3620: 
 3621: @item reaching the end of the input source before matching @code{[ELSE]} or @code{[THEN]}:
 3622: Continue in the same state of conditional compilation in the next outer
 3623: input source. Currently there is no warning to the user about this.
 3624: 
 3625: @item removing a needed definition (@code{FORGET}):
 3626: Not implemented (yet).
 3627: 
 3628: @end table
 3629: 
 3630: 
 3631: @c =====================================================================
 3632: @node  The optional Search-Order word set,  , The optional Programming-Tools word set, ANS conformance
 3633: @section The optional Search-Order word set
 3634: @c =====================================================================
 3635: 
 3636: @menu
 3637: * search-idef::                 Implementation Defined Options                 
 3638: * search-ambcond::              Ambiguous Conditions              
 3639: @end menu
 3640: 
 3641: 
 3642: @c ---------------------------------------------------------------------
 3643: @node search-idef, search-ambcond, The optional Search-Order word set, The optional Search-Order word set
 3644: @subsection Implementation Defined Options
 3645: @c ---------------------------------------------------------------------
 3646: 
 3647: @table @i
 3648: 
 3649: @item maximum number of word lists in search order:
 3650: @code{s" wordlists" environment? drop .}. Currently 16.
 3651: 
 3652: @item minimum search order:
 3653: @code{root root}.
 3654: 
 3655: @end table
 3656: 
 3657: @c ---------------------------------------------------------------------
 3658: @node search-ambcond,  , search-idef, The optional Search-Order word set
 3659: @subsection Ambiguous conditions
 3660: @c ---------------------------------------------------------------------
 3661: 
 3662: @table @i
 3663: 
 3664: @item changing the compilation wordlist (during compilation):
 3665: The word is entered into the wordlist that was the compilation wordlist
 3666: at the start of the definition. Any changes to the name field (e.g.,
 3667: @code{immediate}) or the code field (e.g., when executing @code{DOES>})
 3668: are applied to the latest defined word (as reported by @code{last} or
 3669: @code{lastxt}), if possible, irrespective of the compilation wordlist.
 3670: 
 3671: @item search order empty (@code{previous}):
 3672: @code{abort" Vocstack empty"}.
 3673: 
 3674: @item too many word lists in search order (@code{also}):
 3675: @code{abort" Vocstack full"}.
 3676: 
 3677: @end table
 3678: 
 3679: @node Model, Integrating Gforth, ANS conformance, Top
 3680: @chapter Model
 3681: 
 3682: This chapter has yet to be written. It will contain information, on
 3683: which internal structures you can rely.
 3684: 
 3685: @node Integrating Gforth, Emacs and Gforth, Model, Top
 3686: @chapter Integrating Gforth into C programs
 3687: 
 3688: This is not yet implemented.
 3689: 
 3690: Several people like to use Forth as scripting language for applications
 3691: that are otherwise written in C, C++, or some other language.
 3692: 
 3693: The Forth system ATLAST provides facilities for embedding it into
 3694: applications; unfortunately it has several disadvantages: most
 3695: importantly, it is not based on ANS Forth, and it is apparently dead
 3696: (i.e., not developed further and not supported). The facilities
 3697: provided by Gforth in this area are inspired by ATLASTs facilities, so
 3698: making the switch should not be hard.
 3699: 
 3700: We also tried to design the interface such that it can easily be
 3701: implemented by other Forth systems, so that we may one day arrive at a
 3702: standardized interface. Such a standard interface would allow you to
 3703: replace the Forth system without having to rewrite C code.
 3704: 
 3705: You embed the Gforth interpreter by linking with the library
 3706: @code{libgforth.a} (give the compiler the option @code{-lgforth}).  All
 3707: global symbols in this library that belong to the interface, have the
 3708: prefix @code{forth_}. (Global symbols that are used internally have the
 3709: prefix @code{gforth_}).
 3710: 
 3711: You can include the declarations of Forth types and the functions and
 3712: variables of the interface with @code{#include <forth.h>}.
 3713: 
 3714: Types.
 3715: 
 3716: Variables.
 3717: 
 3718: Data and FP Stack pointer. Area sizes.
 3719: 
 3720: functions.
 3721: 
 3722: forth_init(imagefile)
 3723: forth_evaluate(string) exceptions?
 3724: forth_goto(address) (or forth_execute(xt)?)
 3725: forth_continue() (a corountining mechanism)
 3726: 
 3727: Adding primitives.
 3728: 
 3729: No checking.
 3730: 
 3731: Signals?
 3732: 
 3733: Accessing the Stacks
 3734: 
 3735: @node Emacs and Gforth, Internals, Integrating Gforth, Top
 3736: @chapter Emacs and Gforth
 3737: 
 3738: Gforth comes with @file{gforth.el}, an improved version of
 3739: @file{forth.el} by Goran Rydqvist (included in the TILE package). The
 3740: improvements are a better (but still not perfect) handling of
 3741: indentation. I have also added comment paragraph filling (@kbd{M-q}),
 3742: commenting (@kbd{C-x \}) and uncommenting (@kbd{C-u C-x \}) regions and
 3743: removing debugging tracers (@kbd{C-x ~}, @pxref{Debugging}). I left the
 3744: stuff I do not use alone, even though some of it only makes sense for
 3745: TILE. To get a description of these features, enter Forth mode and type
 3746: @kbd{C-h m}.
 3747: 
 3748: In addition, Gforth supports Emacs quite well: The source code locations
 3749: given in error messages, debugging output (from @code{~~}) and failed
 3750: assertion messages are in the right format for Emacs' compilation mode
 3751: (@pxref{Compilation, , Running Compilations under Emacs, emacs, Emacs
 3752: Manual}) so the source location corresponding to an error or other
 3753: message is only a few keystrokes away (@kbd{C-x `} for the next error,
 3754: @kbd{C-c C-c} for the error under the cursor).
 3755: 
 3756: Also, if you @code{include} @file{etags.fs}, a new @file{TAGS} file
 3757: (@pxref{Tags, , Tags Tables, emacs, Emacs Manual}) will be produced that
 3758: contains the definitions of all words defined afterwards. You can then
 3759: find the source for a word using @kbd{M-.}. Note that emacs can use
 3760: several tags files at the same time (e.g., one for the Gforth sources
 3761: and one for your program, @pxref{Select Tags Table,,Selecting a Tags
 3762: Table,emacs, Emacs Manual}). The TAGS file for the preloaded words is
 3763: @file{$(datadir)/gforth/$(VERSION)/TAGS} (e.g.,
 3764: @file{/usr/local/share/gforth/0.2.0/TAGS}).
 3765: 
 3766: To get all these benefits, add the following lines to your @file{.emacs}
 3767: file:
 3768: 
 3769: @example
 3770: (autoload 'forth-mode "gforth.el")
 3771: (setq auto-mode-alist (cons '("\\.fs\\'" . forth-mode) auto-mode-alist))
 3772: @end example
 3773: 
 3774: @node Internals, Bugs, Emacs and Gforth, Top
 3775: @chapter Internals
 3776: 
 3777: Reading this section is not necessary for programming with Gforth. It
 3778: should be helpful for finding your way in the Gforth sources.
 3779: 
 3780: The ideas in this section have also been published in the papers
 3781: @cite{ANS fig/GNU/??? Forth} (in German) by Bernd Paysan, presented at
 3782: the Forth-Tagung '93 and @cite{A Portable Forth Engine} by M. Anton
 3783: Ertl, presented at EuroForth '93; the latter is available at
 3784: @*@file{http://www.complang.tuwien.ac.at/papers/ertl93.ps.Z}.
 3785: 
 3786: @menu
 3787: * Portability::                 
 3788: * Threading::                   
 3789: * Primitives::                  
 3790: * System Architecture::         
 3791: * Performance::                 
 3792: @end menu
 3793: 
 3794: @node Portability, Threading, Internals, Internals
 3795: @section Portability
 3796: 
 3797: One of the main goals of the effort is availability across a wide range
 3798: of personal machines. fig-Forth, and, to a lesser extent, F83, achieved
 3799: this goal by manually coding the engine in assembly language for several
 3800: then-popular processors. This approach is very labor-intensive and the
 3801: results are short-lived due to progress in computer architecture.
 3802: 
 3803: Others have avoided this problem by coding in C, e.g., Mitch Bradley
 3804: (cforth), Mikael Patel (TILE) and Dirk Zoller (pfe). This approach is
 3805: particularly popular for UNIX-based Forths due to the large variety of
 3806: architectures of UNIX machines. Unfortunately an implementation in C
 3807: does not mix well with the goals of efficiency and with using
 3808: traditional techniques: Indirect or direct threading cannot be expressed
 3809: in C, and switch threading, the fastest technique available in C, is
 3810: significantly slower. Another problem with C is that it's very
 3811: cumbersome to express double integer arithmetic.
 3812: 
 3813: Fortunately, there is a portable language that does not have these
 3814: limitations: GNU C, the version of C processed by the GNU C compiler
 3815: (@pxref{C Extensions, , Extensions to the C Language Family, gcc.info,
 3816: GNU C Manual}). Its labels as values feature (@pxref{Labels as Values, ,
 3817: Labels as Values, gcc.info, GNU C Manual}) makes direct and indirect
 3818: threading possible, its @code{long long} type (@pxref{Long Long, ,
 3819: Double-Word Integers, gcc.info, GNU C Manual}) corresponds to Forth's
 3820: double numbers@footnote{Unfortunately, long longs are not implemented
 3821: properly on all machines (e.g., on alpha-osf1, long longs are only 64
 3822: bits, the same size as longs (and pointers), but they should be twice as
 3823: long according to @ref{Long Long, , Double-Word Integers, gcc.info, GNU
 3824: C Manual}). So, we had to implement doubles in C after all. Still, on
 3825: most machines we can use long longs and achieve better performance than
 3826: with the emulation package.}. GNU C is available for free on all
 3827: important (and many unimportant) UNIX machines, VMS, 80386s running
 3828: MS-DOS, the Amiga, and the Atari ST, so a Forth written in GNU C can run
 3829: on all these machines.
 3830: 
 3831: Writing in a portable language has the reputation of producing code that
 3832: is slower than assembly. For our Forth engine we repeatedly looked at
 3833: the code produced by the compiler and eliminated most compiler-induced
 3834: inefficiencies by appropriate changes in the source-code.
 3835: 
 3836: However, register allocation cannot be portably influenced by the
 3837: programmer, leading to some inefficiencies on register-starved
 3838: machines. We use explicit register declarations (@pxref{Explicit Reg
 3839: Vars, , Variables in Specified Registers, gcc.info, GNU C Manual}) to
 3840: improve the speed on some machines. They are turned on by using the
 3841: @code{gcc} switch @code{-DFORCE_REG}. Unfortunately, this feature not
 3842: only depends on the machine, but also on the compiler version: On some
 3843: machines some compiler versions produce incorrect code when certain
 3844: explicit register declarations are used. So by default
 3845: @code{-DFORCE_REG} is not used.
 3846: 
 3847: @node Threading, Primitives, Portability, Internals
 3848: @section Threading
 3849: 
 3850: GNU C's labels as values extension (available since @code{gcc-2.0},
 3851: @pxref{Labels as Values, , Labels as Values, gcc.info, GNU C Manual})
 3852: makes it possible to take the address of @var{label} by writing
 3853: @code{&&@var{label}}.  This address can then be used in a statement like
 3854: @code{goto *@var{address}}. I.e., @code{goto *&&x} is the same as
 3855: @code{goto x}.
 3856: 
 3857: With this feature an indirect threaded NEXT looks like:
 3858: @example
 3859: cfa = *ip++;
 3860: ca = *cfa;
 3861: goto *ca;
 3862: @end example
 3863: For those unfamiliar with the names: @code{ip} is the Forth instruction
 3864: pointer; the @code{cfa} (code-field address) corresponds to ANS Forths
 3865: execution token and points to the code field of the next word to be
 3866: executed; The @code{ca} (code address) fetched from there points to some
 3867: executable code, e.g., a primitive or the colon definition handler
 3868: @code{docol}.
 3869: 
 3870: Direct threading is even simpler:
 3871: @example
 3872: ca = *ip++;
 3873: goto *ca;
 3874: @end example
 3875: 
 3876: Of course we have packaged the whole thing neatly in macros called
 3877: @code{NEXT} and @code{NEXT1} (the part of NEXT after fetching the cfa).
 3878: 
 3879: @menu
 3880: * Scheduling::                  
 3881: * Direct or Indirect Threaded?::  
 3882: * DOES>::                       
 3883: @end menu
 3884: 
 3885: @node Scheduling, Direct or Indirect Threaded?, Threading, Threading
 3886: @subsection Scheduling
 3887: 
 3888: There is a little complication: Pipelined and superscalar processors,
 3889: i.e., RISC and some modern CISC machines can process independent
 3890: instructions while waiting for the results of an instruction. The
 3891: compiler usually reorders (schedules) the instructions in a way that
 3892: achieves good usage of these delay slots. However, on our first tries
 3893: the compiler did not do well on scheduling primitives. E.g., for
 3894: @code{+} implemented as
 3895: @example
 3896: n=sp[0]+sp[1];
 3897: sp++;
 3898: sp[0]=n;
 3899: NEXT;
 3900: @end example
 3901: the NEXT comes strictly after the other code, i.e., there is nearly no
 3902: scheduling. After a little thought the problem becomes clear: The
 3903: compiler cannot know that sp and ip point to different addresses (and
 3904: the version of @code{gcc} we used would not know it even if it was
 3905: possible), so it could not move the load of the cfa above the store to
 3906: the TOS. Indeed the pointers could be the same, if code on or very near
 3907: the top of stack were executed. In the interest of speed we chose to
 3908: forbid this probably unused ``feature'' and helped the compiler in
 3909: scheduling: NEXT is divided into the loading part (@code{NEXT_P1}) and
 3910: the goto part (@code{NEXT_P2}). @code{+} now looks like:
 3911: @example
 3912: n=sp[0]+sp[1];
 3913: sp++;
 3914: NEXT_P1;
 3915: sp[0]=n;
 3916: NEXT_P2;
 3917: @end example
 3918: This can be scheduled optimally by the compiler.
 3919: 
 3920: This division can be turned off with the switch @code{-DCISC_NEXT}. This
 3921: switch is on by default on machines that do not profit from scheduling
 3922: (e.g., the 80386), in order to preserve registers.
 3923: 
 3924: @node Direct or Indirect Threaded?, DOES>, Scheduling, Threading
 3925: @subsection Direct or Indirect Threaded?
 3926: 
 3927: Both! After packaging the nasty details in macro definitions we
 3928: realized that we could switch between direct and indirect threading by
 3929: simply setting a compilation flag (@code{-DDIRECT_THREADED}) and
 3930: defining a few machine-specific macros for the direct-threading case.
 3931: On the Forth level we also offer access words that hide the
 3932: differences between the threading methods (@pxref{Threading Words}).
 3933: 
 3934: Indirect threading is implemented completely
 3935: machine-independently. Direct threading needs routines for creating
 3936: jumps to the executable code (e.g. to docol or dodoes). These routines
 3937: are inherently machine-dependent, but they do not amount to many source
 3938: lines. I.e., even porting direct threading to a new machine is a small
 3939: effort.
 3940: 
 3941: @node DOES>,  , Direct or Indirect Threaded?, Threading
 3942: @subsection DOES>
 3943: One of the most complex parts of a Forth engine is @code{dodoes}, i.e.,
 3944: the chunk of code executed by every word defined by a
 3945: @code{CREATE}...@code{DOES>} pair. The main problem here is: How to find
 3946: the Forth code to be executed, i.e. the code after the @code{DOES>} (the
 3947: DOES-code)? There are two solutions:
 3948: 
 3949: In fig-Forth the code field points directly to the dodoes and the
 3950: DOES-code address is stored in the cell after the code address
 3951: (i.e. at cfa cell+). It may seem that this solution is illegal in the
 3952: Forth-79 and all later standards, because in fig-Forth this address
 3953: lies in the body (which is illegal in these standards). However, by
 3954: making the code field larger for all words this solution becomes legal
 3955: again. We use this approach for the indirect threaded version. Leaving
 3956: a cell unused in most words is a bit wasteful, but on the machines we
 3957: are targetting this is hardly a problem. The other reason for having a
 3958: code field size of two cells is to avoid having different image files
 3959: for direct and indirect threaded systems (@pxref{System Architecture}).
 3960: 
 3961: The other approach is that the code field points or jumps to the cell
 3962: after @code{DOES}. In this variant there is a jump to @code{dodoes} at
 3963: this address. @code{dodoes} can then get the DOES-code address by
 3964: computing the code address, i.e., the address of the jump to dodoes,
 3965: and add the length of that jump field. A variant of this is to have a
 3966: call to @code{dodoes} after the @code{DOES>}; then the return address
 3967: (which can be found in the return register on RISCs) is the DOES-code
 3968: address. Since the two cells available in the code field are usually
 3969: used up by the jump to the code address in direct threading, we use
 3970: this approach for direct threading. We did not want to add another
 3971: cell to the code field.
 3972: 
 3973: @node Primitives, System Architecture, Threading, Internals
 3974: @section Primitives
 3975: 
 3976: @menu
 3977: * Automatic Generation::        
 3978: * TOS Optimization::            
 3979: * Produced code::               
 3980: @end menu
 3981: 
 3982: @node Automatic Generation, TOS Optimization, Primitives, Primitives
 3983: @subsection Automatic Generation
 3984: 
 3985: Since the primitives are implemented in a portable language, there is no
 3986: longer any need to minimize the number of primitives. On the contrary,
 3987: having many primitives is an advantage: speed. In order to reduce the
 3988: number of errors in primitives and to make programming them easier, we
 3989: provide a tool, the primitive generator (@file{prims2x.fs}), that
 3990: automatically generates most (and sometimes all) of the C code for a
 3991: primitive from the stack effect notation.  The source for a primitive
 3992: has the following form:
 3993: 
 3994: @format
 3995: @var{Forth-name}	@var{stack-effect}	@var{category}	[@var{pronounc.}]
 3996: [@code{""}@var{glossary entry}@code{""}]
 3997: @var{C code}
 3998: [@code{:}
 3999: @var{Forth code}]
 4000: @end format
 4001: 
 4002: The items in brackets are optional. The category and glossary fields
 4003: are there for generating the documentation, the Forth code is there
 4004: for manual implementations on machines without GNU C. E.g., the source
 4005: for the primitive @code{+} is:
 4006: @example
 4007: +    n1 n2 -- n    core    plus
 4008: n = n1+n2;
 4009: @end example
 4010: 
 4011: This looks like a specification, but in fact @code{n = n1+n2} is C
 4012: code. Our primitive generation tool extracts a lot of information from
 4013: the stack effect notations@footnote{We use a one-stack notation, even
 4014: though we have separate data and floating-point stacks; The separate
 4015: notation can be generated easily from the unified notation.}: The number
 4016: of items popped from and pushed on the stack, their type, and by what
 4017: name they are referred to in the C code. It then generates a C code
 4018: prelude and postlude for each primitive. The final C code for @code{+}
 4019: looks like this:
 4020: 
 4021: @example
 4022: I_plus:	/* + ( n1 n2 -- n ) */  /* label, stack effect */
 4023: /*  */                          /* documentation */
 4024: @{
 4025: DEF_CA                          /* definition of variable ca (indirect threading) */
 4026: Cell n1;                        /* definitions of variables */
 4027: Cell n2;
 4028: Cell n;
 4029: n1 = (Cell) sp[1];              /* input */
 4030: n2 = (Cell) TOS;
 4031: sp += 1;                        /* stack adjustment */
 4032: NAME("+")                       /* debugging output (with -DDEBUG) */
 4033: @{
 4034: n = n1+n2;                      /* C code taken from the source */
 4035: @}
 4036: NEXT_P1;                        /* NEXT part 1 */
 4037: TOS = (Cell)n;                  /* output */
 4038: NEXT_P2;                        /* NEXT part 2 */
 4039: @}
 4040: @end example
 4041: 
 4042: This looks long and inefficient, but the GNU C compiler optimizes quite
 4043: well and produces optimal code for @code{+} on, e.g., the R3000 and the
 4044: HP RISC machines: Defining the @code{n}s does not produce any code, and
 4045: using them as intermediate storage also adds no cost.
 4046: 
 4047: There are also other optimizations, that are not illustrated by this
 4048: example: Assignments between simple variables are usually for free (copy
 4049: propagation). If one of the stack items is not used by the primitive
 4050: (e.g.  in @code{drop}), the compiler eliminates the load from the stack
 4051: (dead code elimination). On the other hand, there are some things that
 4052: the compiler does not do, therefore they are performed by
 4053: @file{prims2x.fs}: The compiler does not optimize code away that stores
 4054: a stack item to the place where it just came from (e.g., @code{over}).
 4055: 
 4056: While programming a primitive is usually easy, there are a few cases
 4057: where the programmer has to take the actions of the generator into
 4058: account, most notably @code{?dup}, but also words that do not (always)
 4059: fall through to NEXT.
 4060: 
 4061: @node TOS Optimization, Produced code, Automatic Generation, Primitives
 4062: @subsection TOS Optimization
 4063: 
 4064: An important optimization for stack machine emulators, e.g., Forth
 4065: engines, is keeping  one or more of the top stack items in
 4066: registers.  If a word has the stack effect @var{in1}...@var{inx} @code{--}
 4067: @var{out1}...@var{outy}, keeping the top @var{n} items in registers
 4068: @itemize @bullet
 4069: @item
 4070: is better than keeping @var{n-1} items, if @var{x>=n} and @var{y>=n},
 4071: due to fewer loads from and stores to the stack.
 4072: @item is slower than keeping @var{n-1} items, if @var{x<>y} and @var{x<n} and
 4073: @var{y<n}, due to additional moves between registers.
 4074: @end itemize
 4075: 
 4076: In particular, keeping one item in a register is never a disadvantage,
 4077: if there are enough registers. Keeping two items in registers is a
 4078: disadvantage for frequent words like @code{?branch}, constants,
 4079: variables, literals and @code{i}. Therefore our generator only produces
 4080: code that keeps zero or one items in registers. The generated C code
 4081: covers both cases; the selection between these alternatives is made at
 4082: C-compile time using the switch @code{-DUSE_TOS}. @code{TOS} in the C
 4083: code for @code{+} is just a simple variable name in the one-item case,
 4084: otherwise it is a macro that expands into @code{sp[0]}. Note that the
 4085: GNU C compiler tries to keep simple variables like @code{TOS} in
 4086: registers, and it usually succeeds, if there are enough registers.
 4087: 
 4088: The primitive generator performs the TOS optimization for the
 4089: floating-point stack, too (@code{-DUSE_FTOS}). For floating-point
 4090: operations the benefit of this optimization is even larger:
 4091: floating-point operations take quite long on most processors, but can be
 4092: performed in parallel with other operations as long as their results are
 4093: not used. If the FP-TOS is kept in a register, this works. If
 4094: it is kept on the stack, i.e., in memory, the store into memory has to
 4095: wait for the result of the floating-point operation, lengthening the
 4096: execution time of the primitive considerably.
 4097: 
 4098: The TOS optimization makes the automatic generation of primitives a
 4099: bit more complicated. Just replacing all occurrences of @code{sp[0]} by
 4100: @code{TOS} is not sufficient. There are some special cases to
 4101: consider:
 4102: @itemize @bullet
 4103: @item In the case of @code{dup ( w -- w w )} the generator must not
 4104: eliminate the store to the original location of the item on the stack,
 4105: if the TOS optimization is turned on.
 4106: @item Primitives with stack effects of the form @code{--}
 4107: @var{out1}...@var{outy} must store the TOS to the stack at the start.
 4108: Likewise, primitives with the stack effect @var{in1}...@var{inx} @code{--}
 4109: must load the TOS from the stack at the end. But for the null stack
 4110: effect @code{--} no stores or loads should be generated.
 4111: @end itemize
 4112: 
 4113: @node Produced code,  , TOS Optimization, Primitives
 4114: @subsection Produced code
 4115: 
 4116: To see what assembly code is produced for the primitives on your machine
 4117: with your compiler and your flag settings, type @code{make engine.s} and
 4118: look at the resulting file @file{engine.s}.
 4119: 
 4120: @node System Architecture, Performance, Primitives, Internals
 4121: @section System Architecture
 4122: 
 4123: Our Forth system consists not only of primitives, but also of
 4124: definitions written in Forth. Since the Forth compiler itself belongs
 4125: to those definitions, it is not possible to start the system with the
 4126: primitives and the Forth source alone. Therefore we provide the Forth
 4127: code as an image file in nearly executable form. At the start of the
 4128: system a C routine loads the image file into memory, sets up the
 4129: memory (stacks etc.) according to information in the image file, and
 4130: starts executing Forth code.
 4131: 
 4132: The image file format is a compromise between the goals of making it
 4133: easy to generate image files and making them portable. The easiest way
 4134: to generate an image file is to just generate a memory dump. However,
 4135: this kind of image file cannot be used on a different machine, or on
 4136: the next version of the engine on the same machine, it even might not
 4137: work with the same engine compiled by a different version of the C
 4138: compiler. We would like to have as few versions of the image file as
 4139: possible, because we do not want to distribute many versions of the
 4140: same image file, and to make it easy for the users to use their image
 4141: files on many machines. We currently need to create a different image
 4142: file for machines with different cell sizes and different byte order
 4143: (little- or big-endian)@footnote{We are considering adding information to the
 4144: image file that enables the loader to change the byte order.}.
 4145: 
 4146: Forth code that is going to end up in a portable image file has to
 4147: comply to some restrictions: addresses have to be stored in memory with
 4148: special words (@code{A!}, @code{A,}, etc.) in order to make the code
 4149: relocatable. Cells, floats, etc., have to be stored at the natural
 4150: alignment boundaries@footnote{E.g., store floats (8 bytes) at an address
 4151: dividable by~8. This happens automatically in our system when you use
 4152: the ANS Forth alignment words.}, in order to avoid alignment faults on
 4153: machines with stricter alignment. The image file is produced by a
 4154: metacompiler (@file{cross.fs}).
 4155: 
 4156: So, unlike the image file of Mitch Bradleys @code{cforth}, our image
 4157: file is not directly executable, but has to undergo some manipulations
 4158: during loading. Address relocation is performed at image load-time, not
 4159: at run-time. The loader also has to replace tokens standing for
 4160: primitive calls with the appropriate code-field addresses (or code
 4161: addresses in the case of direct threading).
 4162: 
 4163: @node  Performance,  , System Architecture, Internals
 4164: @section Performance
 4165: 
 4166: On RISCs the Gforth engine is very close to optimal; i.e., it is usually
 4167: impossible to write a significantly faster engine.
 4168: 
 4169: On register-starved machines like the 386 architecture processors
 4170: improvements are possible, because @code{gcc} does not utilize the
 4171: registers as well as a human, even with explicit register declarations;
 4172: e.g., Bernd Beuster wrote a Forth system fragment in assembly language
 4173: and hand-tuned it for the 486; this system is 1.19 times faster on the
 4174: Sieve benchmark on a 486DX2/66 than Gforth compiled with
 4175: @code{gcc-2.6.3} with @code{-DFORCE_REG}.
 4176: 
 4177: However, this potential advantage of assembly language implementations
 4178: is not necessarily realized in complete Forth systems: We compared
 4179: Gforth (direct threaded, compiled with @code{gcc-2.6.3} and
 4180: @code{-DFORCE_REG}) with Win32Forth 1.2093, LMI's NT Forth (Beta, May
 4181: 1994) and Eforth (with and without peephole (aka pinhole) optimization
 4182: of the threaded code); all these systems were written in assembly
 4183: language. We also compared Gforth with three systems written in C:
 4184: PFE-0.9.14 (compiled with @code{gcc-2.6.3} with the default
 4185: configuration for Linux: @code{-O2 -fomit-frame-pointer -DUSE_REGS
 4186: -DUNROLL_NEXT}), ThisForth Beta (compiled with gcc-2.6.3 -O3
 4187: -fomit-frame-pointer; ThisForth employs peephole optimization of the
 4188: threaded code) and TILE (compiled with @code{make opt}). We benchmarked
 4189: Gforth, PFE, ThisForth and TILE on a 486DX2/66 under Linux. Kenneth
 4190: O'Heskin kindly provided the results for Win32Forth and NT Forth on a
 4191: 486DX2/66 with similar memory performance under Windows NT. Marcel
 4192: Hendrix ported Eforth to Linux, then extended it to run the benchmarks,
 4193: added the peephole optimizer, ran the benchmarks and reported the
 4194: results.
 4195:  
 4196: We used four small benchmarks: the ubiquitous Sieve; bubble-sorting and
 4197: matrix multiplication come from the Stanford integer benchmarks and have
 4198: been translated into Forth by Martin Fraeman; we used the versions
 4199: included in the TILE Forth package, but with bigger data set sizes; and
 4200: a recursive Fibonacci number computation for benchmarking calling
 4201: performance. The following table shows the time taken for the benchmarks
 4202: scaled by the time taken by Gforth (in other words, it shows the speedup
 4203: factor that Gforth achieved over the other systems).
 4204: 
 4205: @example
 4206: relative      Win32-    NT       eforth       This-
 4207:   time  Gforth Forth Forth eforth  +opt   PFE Forth  TILE
 4208: sieve     1.00  1.39  1.14   1.39  0.85  1.58  3.18  8.58
 4209: bubble    1.00  1.31  1.41   1.48  0.88  1.50        3.88
 4210: matmul    1.00  1.47  1.35   1.46  1.16  1.58        4.09
 4211: fib       1.00  1.52  1.34   1.22  1.13  1.74  2.99  4.30
 4212: @end example
 4213: 
 4214: You may find the good performance of Gforth compared with the systems
 4215: written in assembly language quite surprising. One important reason for
 4216: the disappointing performance of these systems is probably that they are
 4217: not written optimally for the 486 (e.g., they use the @code{lods}
 4218: instruction). In addition, Win32Forth uses a comfortable, but costly
 4219: method for relocating the Forth image: like @code{cforth}, it computes
 4220: the actual addresses at run time, resulting in two address computations
 4221: per NEXT (@pxref{System Architecture}).
 4222: 
 4223: Only Eforth with the peephole optimizer performs comparable to
 4224: Gforth. The speedups achieved with peephole optimization of threaded
 4225: code are quite remarkable. Adding a peephole optimizer to Gforth should
 4226: cause similar speedups.
 4227: 
 4228: The speedup of Gforth over PFE, ThisForth and TILE can be easily
 4229: explained with the self-imposed restriction to standard C, which makes
 4230: efficient threading impossible (however, the measured implementation of
 4231: PFE uses a GNU C extension: @ref{Global Reg Vars, , Defining Global
 4232: Register Variables, gcc.info, GNU C Manual}).  Moreover, current C
 4233: compilers have a hard time optimizing other aspects of the ThisForth
 4234: and the TILE source.
 4235: 
 4236: Note that the performance of Gforth on 386 architecture processors
 4237: varies widely with the version of @code{gcc} used. E.g., @code{gcc-2.5.8}
 4238: failed to allocate any of the virtual machine registers into real
 4239: machine registers by itself and would not work correctly with explicit
 4240: register declarations, giving a 1.3 times slower engine (on a 486DX2/66
 4241: running the Sieve) than the one measured above.
 4242: 
 4243: In @cite{Translating Forth to Efficient C} by M. Anton Ertl and Martin
 4244: Maierhofer (presented at EuroForth '95), an indirect threaded version of
 4245: Gforth is compared with Win32Forth, NT Forth, PFE, and ThisForth; that
 4246: version of Gforth is 2\%@minus{}8\% slower on a 486 than the version
 4247: used here. The paper available at
 4248: @*@file{http://www.complang.tuwien.ac.at/papers/ertl&maierhofer95.ps.gz};
 4249: it also contains numbers for some native code systems. You can find
 4250: numbers for Gforth on various machines in @file{Benchres}.
 4251: 
 4252: @node Bugs, Origin, Internals, Top
 4253: @chapter Bugs
 4254: 
 4255: Known bugs are described in the file BUGS in the Gforth distribution.
 4256: 
 4257: If you find a bug, please send a bug report to
 4258: @code{bug-gforth@@gnu.ai.mit.edu}. A bug report should
 4259: describe the Gforth version used (it is announced at the start of an
 4260: interactive Gforth session), the machine and operating system (on Unix
 4261: systems you can use @code{uname -a} to produce this information), the
 4262: installation options (send the @code{config.status} file), and a
 4263: complete list of changes you (or your installer) have made to the Gforth
 4264: sources (if any); it should contain a program (or a sequence of keyboard
 4265: commands) that reproduces the bug and a description of what you think
 4266: constitutes the buggy behaviour.
 4267: 
 4268: For a thorough guide on reporting bugs read @ref{Bug Reporting, , How
 4269: to Report Bugs, gcc.info, GNU C Manual}.
 4270: 
 4271: 
 4272: @node Origin, Word Index, Bugs, Top
 4273: @chapter Authors and Ancestors of Gforth
 4274: 
 4275: @section Authors and Contributors
 4276: 
 4277: The Gforth project was started in mid-1992 by Bernd Paysan and Anton
 4278: Ertl. The third major author was Jens Wilke.  Lennart Benschop (who was
 4279: one of Gforth's first users, in mid-1993) and Stuart Ramsden inspired us
 4280: with their continuous feedback. Lennart Benshop contributed
 4281: @file{glosgen.fs}, while Stuart Ramsden has been working on automatic
 4282: support for calling C libraries. Helpful comments also came from Paul
 4283: Kleinrubatscher, Christian Pirker, Dirk Zoller, Marcel Hendrix, John
 4284: Wavrik, Stott Bolton and Marc de Groot.
 4285: 
 4286: Gforth also owes a lot to the authors of the tools we used (GCC, CVS,
 4287: and autoconf, among others), and to the creators of the Internet: Gforth
 4288: was developed across the Internet, and its authors have not met
 4289: physically yet.
 4290: 
 4291: @section Pedigree
 4292: 
 4293: Gforth descends from BigForth (1993) and fig-Forth. Gforth and PFE (by
 4294: Dirk Zoller) will cross-fertilize each other. Of course, a significant
 4295: part of the design of Gforth was prescribed by ANS Forth.
 4296: 
 4297: Bernd Paysan wrote BigForth, a descendent from TurboForth, an unreleased
 4298: 32 bit native code version of VolksForth for the Atari ST, written
 4299: mostly by Dietrich Weineck.
 4300: 
 4301: VolksForth descends from F83. It was written by Klaus Schleisiek, Bernd
 4302: Pennemann, Georg Rehfeld and Dietrich Weineck for the C64 (called
 4303: UltraForth there) in the mid-80s and ported to the Atari ST in 1986.
 4304: 
 4305: Henry Laxen and Mike Perry wrote F83 as a model implementation of the
 4306: Forth-83 standard. !! Pedigree? When?
 4307: 
 4308: A team led by Bill Ragsdale implemented fig-Forth on many processors in
 4309: 1979. Robert Selzer and Bill Ragsdale developed the original
 4310: implementation of fig-Forth for the 6502 based on microForth.
 4311: 
 4312: The principal architect of microForth was Dean Sanderson. microForth was
 4313: FORTH, Inc.'s first off-the-shelf product. It was developped in 1976 for
 4314: the 1802, and subsequently implemented on the 8080, the 6800 and the
 4315: Z80.
 4316: 
 4317: All earlier Forth systems were custom-made, usually by Charles Moore,
 4318: who discovered (as he puts it) Forth during the late 60s. The first full
 4319: Forth existed in 1971.
 4320: 
 4321: A part of the information in this section comes from @cite{The Evolution
 4322: of Forth} by Elizabeth D. Rather, Donald R. Colburn and Charles
 4323: H. Moore, presented at the HOPL-II conference and preprinted in SIGPLAN
 4324: Notices 28(3), 1993.  You can find more historical and genealogical
 4325: information about Forth there.
 4326: 
 4327: @node Word Index, Node Index, Origin, Top
 4328: @chapter Word Index
 4329: 
 4330: This index is as incomplete as the manual. Each word is listed with
 4331: stack effect and wordset.
 4332: 
 4333: @printindex fn
 4334: 
 4335: @node Node Index,  , Word Index, Top
 4336: @chapter Node Index
 4337: 
 4338: This index is even less complete than the manual.
 4339: 
 4340: @contents
 4341: @bye
 4342: 

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