Diff for /gforth/doc/vmgen.texi between versions 1.1 and 1.30

version 1.1, 2002/05/16 09:07:29 version 1.30, 2007/12/31 18:56:18
Line 1 Line 1
   \input texinfo    @c -*-texinfo-*-
   @comment %**start of header
   @setfilename vmgen.info
 @include version.texi  @include version.texi
   @settitle Vmgen (Gforth @value{VERSION})
   @c @syncodeindex pg cp
   @comment %**end of header
   This manual is for Vmgen
   (version @value{VERSION}, @value{UPDATED}),
   the virtual machine interpreter generator
   Copyright @copyright{} 2002,2003,2005 Free Software Foundation, Inc.
   Permission is granted to copy, distribute and/or modify this document
   under the terms of the GNU Free Documentation License, Version 1.2 or
   any later version published by the Free Software Foundation; with no
   Invariant Sections, with the Front-Cover texts being ``A GNU Manual,''
   and with the Back-Cover Texts as in (a) below.  A copy of the
   license is included in the section entitled ``GNU Free Documentation
   (a) The FSF's Back-Cover Text is: ``You have freedom to copy and modify
   this GNU Manual, like GNU software.  Copies published by the Free
   Software Foundation raise funds for GNU development.''
   @end quotation
   @end copying
   @dircategory Software development
   * Vmgen: (vmgen).               Virtual machine interpreter generator
   @end direntry
   @title Vmgen
   @subtitle for Gforth version @value{VERSION}, @value{UPDATED}
   @author M. Anton Ertl (@email{anton@@mips.complang.tuwien.ac.at})
   @vskip 0pt plus 1filll
   @end titlepage
   @node Top, Introduction, (dir), (dir)
   @top Vmgen
   @end ifnottex
   * Introduction::                What can Vmgen do for you?
   * Why interpreters?::           Advantages and disadvantages
   * Concepts::                    VM interpreter background
   * Invoking Vmgen::              
   * Example::                     
   * Input File Format::           
   * Error messages::              reported by Vmgen
   * Using the generated code::    
   * Hints::                       VM archictecture, efficiency
   * The future::                  
   * Changes::                     from earlier versions
   * Contact::                     Bug reporting etc.
   * Copying This Manual::         Manual License
   * Index::                       
    --- The Detailed Node Listing ---
   * Front end and VM interpreter::  Modularizing an interpretive system
   * Data handling::               Stacks, registers, immediate arguments
   * Dispatch::                    From one VM instruction to the next
   * Example overview::            
   * Using profiling to create superinstructions::  
   Input File Format
   * Input File Grammar::          
   * Simple instructions::         
   * Superinstructions::           
   * Store Optimization::          
   * Register Machines::           How to define register VM instructions
   Input File Grammar
   * Eval escapes::                what follows \E
   Simple instructions
   * Explicit stack access::       If the C code accesses a stack pointer
   * C Code Macros::               Macros recognized by Vmgen
   * C Code restrictions::         Vmgen makes assumptions about C code
   * Stack growth direction::      is configurable per stack
   Using the generated code
   * VM engine::                   Executing VM code
   * VM instruction table::        
   * VM code generation::          Creating VM code (in the front-end)
   * Peephole optimization::       Creating VM superinstructions
   * VM disassembler::             for debugging the front end
   * VM profiler::                 for finding worthwhile superinstructions
   * Floating point::              and stacks
   Copying This Manual
   * GNU Free Documentation License::  License for copying this manual.
   @end detailmenu
   @end menu
 @c @ifnottex  @c @ifnottex
 This file documents vmgen (Gforth @value{VERSION}).  @c This file documents Vmgen (Gforth @value{VERSION}).
 @section Introduction  @c ************************************************************
   @node Introduction, Why interpreters?, Top, Top
   @chapter Introduction
 Vmgen is a tool for writing efficient interpreters.  It takes a simple  Vmgen is a tool for writing efficient interpreters.  It takes a simple
 virtual machine description and generates efficient C code for dealing  virtual machine description and generates efficient C code for dealing
Line 12  it).  The run-time efficiency of the res Line 133  it).  The run-time efficiency of the res
 within a factor of 10 of machine code produced by an optimizing  within a factor of 10 of machine code produced by an optimizing
 compiler.  compiler.
 The interpreter design strategy supported by vmgen is to divide the  The interpreter design strategy supported by Vmgen is to divide the
 interpreter into two parts:  interpreter into two parts:
 @itemize @bullet  @itemize @bullet
Line 29  machine code. Line 150  machine code.
 @end itemize  @end itemize
 Such a division is usually used in interpreters, for modularity as well  Such a division is usually used in interpreters, for modularity as well
 as for efficiency reasons.  The virtual machine code is typically passed  as for efficiency.  The virtual machine code is typically passed between
 between front end and virtual machine interpreter in memory, like in a  front end and virtual machine interpreter in memory, like in a
 load-and-go compiler; this avoids the complexity and time cost of  load-and-go compiler; this avoids the complexity and time cost of
 writing the code to a file and reading it again.  writing the code to a file and reading it again.
Line 39  A @emph{virtual machine} (VM) represents Line 160  A @emph{virtual machine} (VM) represents
 machine code.  Control flow occurs through VM branch instructions, like  machine code.  Control flow occurs through VM branch instructions, like
 in a real machine.  in a real machine.
 In this setup, vmgen can generate most of the code dealing with virtual  @cindex functionality features overview
   In this setup, Vmgen can generate most of the code dealing with virtual
 machine instructions from a simple description of the virtual machine  machine instructions from a simple description of the virtual machine
 instructions (@pxref...), in particular:  instructions (@pxref{Input File Format}), in particular:
 @table @emph  @table @strong
 @item VM instruction execution  @item VM instruction execution
Line 59  typically provide other means for debugg Line 181  typically provide other means for debugg
 source level.  source level.
 @item VM code profiling  @item VM code profiling
 Useful for optimizing the VM insterpreter with superinstructions  Useful for optimizing the VM interpreter with superinstructions
 (@pxref...).  (@pxref{VM profiler}).
 @end table  @end table
 VMgen supports efficient interpreters though various optimizations, in  To create parts of the interpretive system that do not deal with VM
   instructions, you have to use other tools (e.g., @command{bison}) and/or
   hand-code them.
   @cindex efficiency features overview
   Vmgen supports efficient interpreters though various optimizations, in
 particular  particular
 @itemize  @itemize @bullet
 @item Threaded code  @item Threaded code
Line 81  Replicating VM (super)instructions for b Line 209  Replicating VM (super)instructions for b
 @end itemize  @end itemize
 As a result, vmgen-based interpreters are only about an order of  @cindex speed for JVM
 magintude slower than native code from an optimizing C compiler on small  As a result, Vmgen-based interpreters are only about an order of
   magnitude slower than native code from an optimizing C compiler on small
 benchmarks; on large benchmarks, which spend more time in the run-time  benchmarks; on large benchmarks, which spend more time in the run-time
 system, the slowdown is often less (e.g., the slowdown over the best JVM  system, the slowdown is often less (e.g., the slowdown of a
 JIT compiler we measured is only a factor of 2-3 for large benchmarks  Vmgen-generated JVM interpreter over the best JVM JIT compiler we
 (and some other JITs were slower than our interpreter).  measured is only a factor of 2-3 for large benchmarks; some other JITs
   and all other interpreters we looked at were slower than our
 VMs are usually designed as stack machines (passing data between VM  VMs are usually designed as stack machines (passing data between VM
 instructions on a stack), and vmgen supports such designs especially  instructions on a stack), and Vmgen supports such designs especially
 well; however, you can also use vmgen for implementing a register VM and  well; however, you can also use Vmgen for implementing a register VM
 still benefit from most of the advantages offered by vmgen.  (@pxref{Register Machines}) and still benefit from most of the advantages
   offered by Vmgen.
   There are many potential uses of the instruction descriptions that are
   not implemented at the moment, but we are open for feature requests, and
   we will consider new features if someone asks for them; so the feature
   list above is not exhaustive.
   @c *********************************************************************
   @node Why interpreters?, Concepts, Introduction, Top
   @chapter Why interpreters?
   @cindex interpreters, advantages
   @cindex advantages of interpreters
   @cindex advantages of vmgen
   Interpreters are a popular language implementation technique because
   they combine all three of the following advantages:
   @itemize @bullet
   @item Ease of implementation
   @item Portability
   @item Fast edit-compile-run cycle
   @end itemize
   Vmgen makes it even easier to implement interpreters.
   @cindex speed of interpreters
   The main disadvantage of interpreters is their run-time speed.  However,
   there are huge differences between different interpreters in this area:
   the slowdown over optimized C code on programs consisting of simple
   operations is typically a factor of 10 for the more efficient
   interpreters, and a factor of 1000 for the less efficient ones (the
   slowdown for programs executing complex operations is less, because the
   time spent in libraries for executing complex operations is the same in
   all implementation strategies).
   Vmgen supports techniques for building efficient interpreters.
   @c ********************************************************************
   @node Concepts, Invoking Vmgen, Why interpreters?, Top
   @chapter Concepts
   * Front end and VM interpreter::  Modularizing an interpretive system
   * Data handling::               Stacks, registers, immediate arguments
   * Dispatch::                    From one VM instruction to the next
   @end menu
   @c --------------------------------------------------------------------
   @node Front end and VM interpreter, Data handling, Concepts, Concepts
   @section Front end and VM interpreter
   @cindex modularization of interpreters
   @cindex front-end
   Interpretive systems are typically divided into a @emph{front end} that
   parses the input language and produces an intermediate representation
   for the program, and an interpreter that executes the intermediate
   representation of the program.
   @cindex virtual machine
   @cindex VM
   @cindex VM instruction
   @cindex instruction, VM
   @cindex VM branch instruction
   @cindex branch instruction, VM
   @cindex VM register
   @cindex register, VM
   @cindex opcode, VM instruction
   @cindex immediate argument, VM instruction
   For efficient interpreters the intermediate representation of choice is
   virtual machine code (rather than, e.g., an abstract syntax tree).
   @emph{Virtual machine} (VM) code consists of VM instructions arranged
   sequentially in memory; they are executed in sequence by the VM
   interpreter, but VM branch instructions can change the control flow and
   are used for implementing control structures.  The conceptual similarity
   to real machine code results in the name @emph{virtual machine}.
   Various terms similar to terms for real machines are used; e.g., there
   are @emph{VM registers} (like the instruction pointer and stack
   pointer(s)), and the VM instruction consists of an @emph{opcode} and
   @emph{immediate arguments}.
   In this framework, Vmgen supports building the VM interpreter and any
   other component dealing with VM instructions.  It does not have any
   support for the front end, apart from VM code generation support.  The
   front end can be implemented with classical compiler front-end
   techniques, supported by tools like @command{flex} and @command{bison}.
   The intermediate representation is usually just internal to the
   interpreter, but some systems also support saving it to a file, either
   as an image file, or in a full-blown linkable file format (e.g., JVM).
   Vmgen currently has no special support for such features, but the
   information in the instruction descriptions can be helpful, and we are
   open to feature requests and suggestions.
   @c --------------------------------------------------------------------
   @node Data handling, Dispatch, Front end and VM interpreter, Concepts
   @section Data handling
   @cindex stack machine
   @cindex register machine
   Most VMs use one or more stacks for passing temporary data between VM
   instructions.  Another option is to use a register machine architecture
   for the virtual machine; we believe that using a stack architecture is
   usually both simpler and faster.
   however, this option is slower or
   significantly more complex to implement than a stack machine architecture.
   Vmgen has special support and optimizations for stack VMs, making their
   implementation easy and efficient.
   You can also implement a register VM with Vmgen (@pxref{Register
   Machines}), and you will still profit from most Vmgen features.
   @cindex stack item size
   @cindex size, stack items
   Stack items all have the same size, so they typically will be as wide as
   an integer, pointer, or floating-point value.  Vmgen supports treating
   two consecutive stack items as a single value, but anything larger is
   best kept in some other memory area (e.g., the heap), with pointers to
   the data on the stack.
   @cindex instruction stream
   @cindex immediate arguments
   Another source of data is immediate arguments VM instructions (in the VM
   instruction stream).  The VM instruction stream is handled similar to a
   stack in Vmgen.
   @cindex garbage collection
   @cindex reference counting
   Vmgen has no built-in support for, nor restrictions against
   @emph{garbage collection}.  If you need garbage collection, you need to
   provide it in your run-time libraries.  Using @emph{reference counting}
   is probably harder, but might be possible (contact us if you are
   @c reference counting might be possible by including counting code in 
   @c the conversion macros.
   @c --------------------------------------------------------------------
   @node Dispatch,  , Data handling, Concepts
   @section Dispatch
   @cindex Dispatch of VM instructions
   @cindex main interpreter loop
   Understanding this section is probably not necessary for using Vmgen,
   but it may help.  You may want to skip it now, and read it if you find statements about dispatch methods confusing.
   After executing one VM instruction, the VM interpreter has to dispatch
   the next VM instruction (Vmgen calls the dispatch routine @samp{NEXT}).
   Vmgen supports two methods of dispatch:
   @table @strong
   @item switch dispatch
   @cindex switch dispatch
   In this method the VM interpreter contains a giant @code{switch}
   statement, with one @code{case} for each VM instruction.  The VM
   instruction opcodes are represented by integers (e.g., produced by an
   @code{enum}) in the VM code, and dispatch occurs by loading the next
   opcode, @code{switch}ing on it, and continuing at the appropriate
   @code{case}; after executing the VM instruction, the VM interpreter
   jumps back to the dispatch code.
   @item threaded code
   @cindex threaded code
   This method represents a VM instruction opcode by the address of the
   start of the machine code fragment for executing the VM instruction.
   Dispatch consists of loading this address, jumping to it, and
   incrementing the VM instruction pointer.  Typically the threaded-code
   dispatch code is appended directly to the code for executing the VM
   instruction.  Threaded code cannot be implemented in ANSI C, but it can
   be implemented using GNU C's labels-as-values extension (@pxref{Labels
   as Values, , Labels as Values, gcc.info, GNU C Manual}).
   @c call threading
   @end table
   Threaded code can be twice as fast as switch dispatch, depending on the
   interpreter, the benchmark, and the machine.
   @c *************************************************************
   @node Invoking Vmgen, Example, Concepts, Top
   @chapter Invoking Vmgen
   @cindex Invoking Vmgen
   The usual way to invoke Vmgen is as follows:
   vmgen @var{inputfile}
   @end example
   Here @var{inputfile} is the VM instruction description file, which
   usually ends in @file{.vmg}.  The output filenames are made by taking
   the basename of @file{inputfile} (i.e., the output files will be created
   in the current working directory) and replacing @file{.vmg} with
   @file{-vm.i}, @file{-disasm.i}, @file{-gen.i}, @file{-labels.i},
   @file{-profile.i}, and @file{-peephole.i}.  E.g., @command{vmgen
   hack/foo.vmg} will create @file{foo-vm.i}, @file{foo-disasm.i},
   @file{foo-gen.i}, @file{foo-labels.i}, @file{foo-profile.i} and
   The command-line options supported by Vmgen are
   @table @option
   @cindex -h, command-line option
   @cindex --help, command-line option
   @item --help
   @itemx -h
   Print a message about the command-line options
   @cindex -v, command-line option
   @cindex --version, command-line option
   @item --version
   @itemx -v
   Print version and exit
   @end table
   @c ****************************************************************
   @node Example, Input File Format, Invoking Vmgen, Top
   @chapter Example
   @cindex example of a Vmgen-based interpreter
   * Example overview::            
   * Using profiling to create superinstructions::  
   @end menu
   @c --------------------------------------------------------------------
   @node Example overview, Using profiling to create superinstructions, Example, Example
   @section Example overview
   @cindex example overview
   @cindex @file{vmgen-ex}
   @cindex @file{vmgen-ex2}
   There are two versions of the same example for using Vmgen:
   @file{vmgen-ex} and @file{vmgen-ex2} (you can also see Gforth as
   example, but it uses additional (undocumented) features, and also
   differs in some other respects).  The example implements @emph{mini}, a
   tiny Modula-2-like language with a small JavaVM-like virtual machine.
   The difference between the examples is that @file{vmgen-ex} uses many
   casts, and @file{vmgen-ex2} tries to avoids most casts and uses unions
   instead.  In the rest of this manual we usually mention just files in
   @file{vmgen-ex}; if you want to use unions, use the equivalent file in
   @cindex unions example
   @cindex casts example
   The files provided with each example are:
   @cindex example files
   disasm.c           wrapper file
   engine.c           wrapper file
   peephole.c         wrapper file
   profile.c          wrapper file
   mini-inst.vmg      simple VM instructions
   mini-super.vmg     superinstructions (empty at first)
   mini.h             common declarations
   mini.l             scanner
   mini.y             front end (parser, VM code generator)
   support.c          main() and other support functions
   fib.mini           example mini program
   simple.mini        example mini program
   test.mini          example mini program (tests everything)
   test.out           test.mini output
   stat.awk           script for aggregating profile information
   peephole-blacklist list of instructions not allowed in superinstructions
   seq2rule.awk       script for creating superinstructions
   @end example
   For your own interpreter, you would typically copy the following files
   and change little, if anything:
   @cindex wrapper files
   disasm.c           wrapper file
   engine.c           wrapper file
   peephole.c         wrapper file
   profile.c          wrapper file
   stat.awk           script for aggregating profile information
   seq2rule.awk       script for creating superinstructions
   @end example
   You would typically change much in or replace the following files:
   mini-inst.vmg      simple VM instructions
   mini.h             common declarations
   mini.l             scanner
   mini.y             front end (parser, VM code generator)
   support.c          main() and other support functions
   peephole-blacklist list of instructions not allowed in superinstructions
   @end example
   You can build the example by @code{cd}ing into the example's directory,
   and then typing @code{make}; you can check that it works with @code{make
   check}.  You can run run mini programs like this:
   ./mini fib.mini
   @end example
   To learn about the options, type @code{./mini -h}.
   @c --------------------------------------------------------------------
   @node Using profiling to create superinstructions,  , Example overview, Example
   @section Using profiling to create superinstructions
   @cindex profiling example
   @cindex superinstructions example
   I have not added rules for this in the @file{Makefile} (there are many
   options for selecting superinstructions, and I did not want to hardcode
   one into the @file{Makefile}), but there are some supporting scripts, and
   here's an example:
   Suppose you want to use @file{fib.mini} and @file{test.mini} as training
   programs, you get the profiles like this:
   make fib.prof test.prof #takes a few seconds
   @end example
   You can aggregate these profiles with @file{stat.awk}:
   awk -f stat.awk fib.prof test.prof
   @end example
   The result contains lines like:
         2      16        36910041 loadlocal lit
   @end example
   This means that the sequence @code{loadlocal lit} statically occurs a
   total of 16 times in 2 profiles, with a dynamic execution count of
   The numbers can be used in various ways to select superinstructions.
   E.g., if you just want to select all sequences with a dynamic
   execution count exceeding 10000, you would use the following pipeline:
   awk -f stat.awk fib.prof test.prof|
   awk '$3>=10000'|                #select sequences
   fgrep -v -f peephole-blacklist| #eliminate wrong instructions
   awk -f seq2rule.awk|  #transform sequences into superinstruction rules
   sort -k 3 >mini-super.vmg       #sort sequences
   @end example
   The file @file{peephole-blacklist} contains all instructions that
   directly access a stack or stack pointer (for mini: @code{call},
   @code{return}); the sort step is necessary to ensure that prefixes
   precede larger superinstructions.
   Now you can create a version of mini with superinstructions by just
   saying @samp{make}
   @c ***************************************************************
   @node Input File Format, Error messages, Example, Top
   @chapter Input File Format
   @cindex input file format
   @cindex format, input file
   Vmgen takes as input a file containing specifications of virtual machine
   instructions.  This file usually has a name ending in @file{.vmg}.
   Most examples are taken from the example in @file{vmgen-ex}.
   * Input File Grammar::          
   * Simple instructions::         
   * Superinstructions::           
   * Store Optimization::          
   * Register Machines::           How to define register VM instructions
   @end menu
   @c --------------------------------------------------------------------
   @node Input File Grammar, Simple instructions, Input File Format, Input File Format
   @section Input File Grammar
   @cindex grammar, input file
   @cindex input file grammar
   The grammar is in EBNF format, with @code{@var{a}|@var{b}} meaning
   ``@var{a} or @var{b}'', @code{@{@var{c}@}} meaning 0 or more repetitions
   of @var{c} and @code{[@var{d}]} meaning 0 or 1 repetitions of @var{d}.
   @cindex free-format, not
   @cindex newlines, significance in syntax
   Vmgen input is not free-format, so you have to take care where you put
   newlines (and, in a few cases, white space).
   description: @{instruction|comment|eval-escape|c-escape@}
   instruction: simple-inst|superinst
   simple-inst: ident '(' stack-effect ')' newline c-code newline newline
   stack-effect: @{ident@} '--' @{ident@}
   super-inst: ident '=' ident @{ident@}  
   comment:      '\ '  text newline
   eval-escape:  '\E ' text newline
   c-escape:     '\C ' text newline
   @end example
   @c \+ \- \g \f \c
   Note that the @code{\}s in this grammar are meant literally, not as
   C-style encodings for non-printable characters.
   There are two ways to delimit the C code in @code{simple-inst}:
   @itemize @bullet
   If you start it with a @samp{@{} at the start of a line (i.e., not even
   white space before it), you have to end it with a @samp{@}} at the start
   of a line (followed by a newline).  In this case you may have empty
   lines within the C code (typically used between variable definitions and
   You do not start it with @samp{@{}.  Then the C code ends at the first
   empty line, so you cannot have empty lines within this code.
   @end itemize
   The text in @code{comment}, @code{eval-escape} and @code{c-escape} must
   not contain a newline.  @code{Ident} must conform to the usual
   conventions of C identifiers (otherwise the C compiler would choke on
   the Vmgen output), except that idents in @code{stack-effect} may have a
   stack prefix (for stack prefix syntax, @pxref{Eval escapes}).
   @cindex C escape
   @cindex @code{\C}
   @cindex conditional compilation of Vmgen output
   The @code{c-escape} passes the text through to each output file (without
   the @samp{\C}).  This is useful mainly for conditional compilation
   (i.e., you write @samp{\C #if ...} etc.).
   @cindex sync lines
   @cindex @code{#line}
   In addition to the syntax given in the grammer, Vmgen also processes
   sync lines (lines starting with @samp{#line}), as produced by @samp{m4
   -s} (@pxref{Invoking m4, , Invoking m4, m4.info, GNU m4}) and similar
   tools.  This allows associating C compiler error messages with the
   original source of the C code.
   Vmgen understands a few extensions beyond the grammar given here, but
   these extensions are only useful for building Gforth.  You can find a
   description of the format used for Gforth in @file{prim}.
   * Eval escapes::                what follows \E
   @end menu
   @node Eval escapes,  , Input File Grammar, Input File Grammar
   @subsection Eval escapes
   @cindex escape to Forth
   @cindex eval escape
   @cindex @code{\E}
   @c woanders?
   The text in @code{eval-escape} is Forth code that is evaluated when
   Vmgen reads the line.  You will normally use this feature to define
   stacks and types.
   If you do not know (and do not want to learn) Forth, you can build the
   text according to the following grammar; these rules are normally all
   Forth you need for using Vmgen:
   text: stack-decl|type-prefix-decl|stack-prefix-decl|set-flag
   stack-decl: 'stack ' ident ident ident
       's" ' string '" ' ('single'|'double') ident 'type-prefix' ident
   stack-prefix-decl:  ident 'stack-prefix' string
   set-flag: ('store-optimization'|'include-skipped-insts') ('on'|'off')
   @end example
   Note that the syntax of this code is not checked thoroughly (there are
   many other Forth program fragments that could be written in an
   A stack prefix can contain letters, digits, or @samp{:}, and may start
   with an @samp{#}; e.g., in Gforth the return stack has the stack prefix
   @samp{R:}.  This restriction is not checked during the stack prefix
   definition, but it is enforced by the parsing rules for stack items
   If you know Forth, the stack effects of the non-standard words involved
   @findex stack
   @findex type-prefix
   @findex single
   @findex double
   @findex stack-prefix
   @findex store-optimization
   stack                 ( "name" "pointer" "type" -- )
                         ( name execution: -- stack )
   type-prefix           ( addr u item-size stack "prefix" -- )
   single                ( -- item-size )
   double                ( -- item-size )
   stack-prefix          ( stack "prefix" -- )
   store-optimization    ( -- addr )
   include-skipped-insts ( -- addr )
   @end example
   An @var{item-size} takes three cells on the stack.
   @c --------------------------------------------------------------------
   @node Simple instructions, Superinstructions, Input File Grammar, Input File Format
   @section Simple instructions
   @cindex simple VM instruction
   @cindex instruction, simple VM
   We will use the following simple VM instruction description as example:
   sub ( i1 i2 -- i )
   i = i1-i2;
   @end example
   The first line specifies the name of the VM instruction (@code{sub}) and
   its stack effect (@code{i1 i2 -- i}).  The rest of the description is
   just plain C code.
   @cindex stack effect
   @cindex effect, stack
   The stack effect specifies that @code{sub} pulls two integers from the
   data stack and puts them in the C variables @code{i1} and @code{i2}
   (with the rightmost item (@code{i2}) taken from the top of stack;
   intuition: if you push @code{i1}, then @code{i2} on the stack, the
   resulting stack picture is @code{i1 i2}) and later pushes one integer
   (@code{i}) on the data stack (the rightmost item is on the top
   @cindex prefix, type
   @cindex type prefix
   @cindex default stack of a type prefix
   How do we know the type and stack of the stack items?  Vmgen uses
   prefixes, similar to Fortran; in contrast to Fortran, you have to
   define the prefix first:
   \E s" Cell"   single data-stack type-prefix i
   @end example
   This defines the prefix @code{i} to refer to the type @code{Cell}
   (defined as @code{long} in @file{mini.h}) and, by default, to the
   @code{data-stack}.  It also specifies that this type takes one stack
   item (@code{single}).  The type prefix is part of the variable name.
   @cindex stack definition
   @cindex defining a stack
   Before we can use @code{data-stack} in this way, we have to define it:
   \E stack data-stack sp Cell
   @end example
   @c !! use something other than Cell
   @cindex stack basic type
   @cindex basic type of a stack
   @cindex type of a stack, basic
   This line defines the stack @code{data-stack}, which uses the stack
   pointer @code{sp}, and each item has the basic type @code{Cell}; other
   types have to fit into one or two @code{Cell}s (depending on whether the
   type is @code{single} or @code{double} wide), and are cast from and to
   Cells on accessing the @code{data-stack} with type cast macros
   (@pxref{VM engine}).  By default, stacks grow towards lower addresses in
   Vmgen-erated interpreters (@pxref{Stack growth direction}).
   @cindex stack prefix
   @cindex prefix, stack
   We can override the default stack of a stack item by using a stack
   prefix.  E.g., consider the following instruction:
   lit ( #i -- i )
   @end example
   The VM instruction @code{lit} takes the item @code{i} from the
   instruction stream (indicated by the prefix @code{#}), and pushes it on
   the (default) data stack.  The stack prefix is not part of the variable
   name.  Stack prefixes are defined like this:
   \E inst-stream stack-prefix #
   \E data-stack  stack-prefix S:
   @end example
   This definition defines that the stack prefix @code{#} specifies the
   ``stack'' @code{inst-stream}.  Since the instruction stream behaves a
   little differently than an ordinary stack, it is predefined, and you do
   not need to define it.
   @cindex instruction stream
   The instruction stream contains instructions and their immediate
   arguments, so specifying that an argument comes from the instruction
   stream indicates an immediate argument.  Of course, instruction stream
   arguments can only appear to the left of @code{--} in the stack effect.
   If there are multiple instruction stream arguments, the leftmost is the
   first one (just as the intuition suggests).
   * Explicit stack access::       If the C code accesses a stack pointer
   * C Code Macros::               Macros recognized by Vmgen
   * C Code restrictions::         Vmgen makes assumptions about C code
   * Stack growth direction::      is configurable per stack
   @end menu
   @c --------------------------------------------------------------------
   @node  Explicit stack access, C Code Macros, Simple instructions, Simple instructions
   @subsection Explicit stack access
   @cindex stack access, explicit
   @cindex Stack pointer access
   @cindex explicit stack access
   Not all stack effects can be specified using the stack effect
   specifications above.  For VM instructions that have other stack
   effects, you can specify them explicitly by accessing the stack
   pointer in the C code; however, you have to notify Vmgen of such
   explicit stack accesses, otherwise Vmgens optimizations could conflict
   with your explicit stack accesses.
   You notify Vmgen by putting @code{...} with the appropriate stack
   prefix into the stack comment.  Then the VM instruction will first
   take the other stack items specified in the stack effect into C
   variables, then make sure that all other stack items for that stack
   are in memory, and that the stack pointer for the stack points to the
   top-of-stack (by default, unless you change the stack access
   transformation: @pxref{Stack growth direction}).
   The general rule is: If you mention a stack pointer in the C code of a
   VM instruction, you should put a @code{...} for that stack in the stack
   Consider this example:
   return ( #iadjust S:... target afp i1 -- i2 )
   sp = (Cell *)(((char *)sp)+iadjust);
   fp = afp;
   @end example
   First the variables @code{target afp i1} are popped off the stack,
   then the stack pointer @code{sp} is set correctly for the new stack
   depth, then the C code changes the stack depth and does other things,
   and finally @code{i2} is pushed on the stack with the new depth.
   The position of the @code{...} within the stack effect does not
   matter.  You can use several @code{...}s, for different stacks, and
   also several for the same stack (that has no additional effect).  If
   you use @code{...} without a stack prefix, this specifies all the
   stacks except the instruction stream.
   You cannot use @code{...} for the instruction stream, but that is not
   necessary: At the start of the C code, @code{IP} points to the start
   of the next VM instruction (i.e., right beyond the end of the current
   VM instruction), and you can change the instruction pointer with
   @code{SET_IP} (@pxref{VM engine}).
   @c --------------------------------------------------------------------
   @node C Code Macros, C Code restrictions, Explicit stack access, Simple instructions
   @subsection C Code Macros
   @cindex macros recognized by Vmgen
   @cindex basic block, VM level
   Vmgen recognizes the following strings in the C code part of simple
   @table @code
   @item SET_IP
   @findex SET_IP
   As far as Vmgen is concerned, a VM instruction containing this ends a VM
   basic block (used in profiling to delimit profiled sequences).  On the C
   level, this also sets the instruction pointer.
   @item SUPER_END
   @findex SUPER_END
   This ends a basic block (for profiling), even if the instruction
   contains no @code{SET_IP}.
   @item INST_TAIL;
   @findex INST_TAIL;
   Vmgen replaces @samp{INST_TAIL;} with code for ending a VM instruction and
   dispatching the next VM instruction.  Even without a @samp{INST_TAIL;} this
   happens automatically when control reaches the end of the C code.  If
   you want to have this in the middle of the C code, you need to use
   @samp{INST_TAIL;}.  A typical example is a conditional VM branch:
   if (branch_condition) @{
     SET_IP(target); INST_TAIL;
   /* implicit tail follows here */
   @end example
   In this example, @samp{INST_TAIL;} is not strictly necessary, because there
   is another one implicitly after the if-statement, but using it improves
   branch prediction accuracy slightly and allows other optimizations.
   This indicates that the implicit tail at the end of the VM instruction
   dispatches the sequentially next VM instruction even if there is a
   @code{SET_IP} in the VM instruction.  This enables an optimization that
   is not yet implemented in the vmgen-ex code (but in Gforth).  The
   typical application is in conditional VM branches:
   if (branch_condition) @{
     SET_IP(target); INST_TAIL; /* now this INST_TAIL is necessary */
   @end example
   @item VM_JUMP
   @findex VM_JUMP
   @code{VM_JUMP(target)} is equivalent to @code{goto *(target)}, but
   allows Vmgen to do dynamic superinstructions and replication.  You
   still need to say @code{SUPER_END}.  Also, the goto only happens at
   the end (wherever the VM_JUMP is).  Essentially, this just suppresses
   much of the ordinary dispatch mechanism.
   @end table
   Note that Vmgen is not smart about C-level tokenization, comments,
   strings, or conditional compilation, so it will interpret even a
   commented-out SUPER_END as ending a basic block (or, e.g.,
   @samp{RESET_IP;} as @samp{SET_IP;}).  Conversely, Vmgen requires the literal
   presence of these strings; Vmgen will not see them if they are hiding in
   a C preprocessor macro.
   @c --------------------------------------------------------------------
   @node C Code restrictions, Stack growth direction, C Code Macros, Simple instructions
   @subsection C Code restrictions
   @cindex C code restrictions
   @cindex restrictions on C code
   @cindex assumptions about C code
   @cindex accessing stack (pointer)
   @cindex stack pointer, access
   @cindex instruction pointer, access
   Vmgen generates code and performs some optimizations under the
   assumption that the user-supplied C code does not access the stack
   pointers or stack items, and that accesses to the instruction pointer
   only occur through special macros.  In general you should heed these
   restrictions.  However, if you need to break these restrictions, read
   the following.
   Accessing a stack or stack pointer directly can be a problem for several
   @cindex stack caching, restriction on C code
   @cindex superinstructions, restrictions on components
   @itemize @bullet
   Vmgen optionally supports caching the top-of-stack item in a local
   variable (that is allocated to a register).  This is the most frequent
   source of trouble.  You can deal with it either by not using
   top-of-stack caching (slowdown factor 1-1.4, depending on machine), or
   by inserting flushing code (e.g., @samp{IF_spTOS(sp[...] = spTOS);}) at
   the start and reloading code (e.g., @samp{IF_spTOS(spTOS = sp[0])}) at
   the end of problematic C code.  Vmgen inserts a stack pointer update
   before the start of the user-supplied C code, so the flushing code has
   to use an index that corrects for that.  In the future, this flushing
   may be done automatically by mentioning a special string in the C code.
   @c sometimes flushing and/or reloading unnecessary
   The Vmgen-erated code loads the stack items from stack-pointer-indexed
   memory into variables before the user-supplied C code, and stores them
   from variables to stack-pointer-indexed memory afterwards.  If you do
   any writes to the stack through its stack pointer in your C code, it
   will not affect the variables, and your write may be overwritten by the
   stores after the C code.  Similarly, a read from a stack using a stack
   pointer will not reflect computations of stack items in the same VM
   Superinstructions keep stack items in variables across the whole
   superinstruction.  So you should not include VM instructions, that
   access a stack or stack pointer, as components of superinstructions
   (@pxref{VM profiler}).
   @end itemize
   You should access the instruction pointer only through its special
   macros (@samp{IP}, @samp{SET_IP}, @samp{IPTOS}); this ensure that these
   macros can be implemented in several ways for best performance.
   @samp{IP} points to the next instruction, and @samp{IPTOS} is its
   @c --------------------------------------------------------------------
   @node Stack growth direction,  , C Code restrictions, Simple instructions
   @subsection Stack growth direction
   @cindex stack growth direction
   @cindex @code{stack-access-transform}
   By default, the stacks grow towards lower addresses.  You can change
   this for a stack by setting the @code{stack-access-transform} field of
   the stack to an xt @code{( itemnum -- index )} that performs the
   appropriate index transformation.
   E.g., if you want to let @code{data-stack} grow towards higher
   addresses, with the stack pointer always pointing just beyond the
   top-of-stack, use this right after defining @code{data-stack}:
   \E : sp-access-transform ( itemnum -- index ) negate 1- ;
   \E ' sp-access-transform ' data-stack >body stack-access-transform !
   @end example
   This means that @code{sp-access-transform} will be used to generate
   indexes for accessing @code{data-stack}.  The definition of
   @code{sp-access-transform} above transforms n into -n-1, e.g, 1 into -2.
   This will access the 0th data-stack element (top-of-stack) at sp[-1],
   the 1st at sp[-2], etc., which is the typical way upward-growing
   stacks are used.  If you need a different transform and do not know
   enough Forth to program it, let me know.
   @c --------------------------------------------------------------------
   @node Superinstructions, Store Optimization, Simple instructions, Input File Format
   @section Superinstructions
   @cindex superinstructions, defining
   @cindex defining superinstructions
   Note: don't invest too much work in (static) superinstructions; a future
   version of Vmgen will support dynamic superinstructions (see Ian
   Piumarta and Fabio Riccardi, @cite{Optimizing Direct Threaded Code by
   Selective Inlining}, PLDI'98), and static superinstructions have much
   less benefit in that context (preliminary results indicate only a factor
   1.1 speedup).
   Here is an example of a superinstruction definition:
   lit_sub = lit sub
   @end example
   @code{lit_sub} is the name of the superinstruction, and @code{lit} and
   @code{sub} are its components.  This superinstruction performs the same
   action as the sequence @code{lit} and @code{sub}.  It is generated
   automatically by the VM code generation functions whenever that sequence
   occurs, so if you want to use this superinstruction, you just need to
   add this definition (and even that can be partially automatized,
   @pxref{VM profiler}).
   @cindex prefixes of superinstructions
   Vmgen requires that the component instructions are simple instructions
   defined before superinstructions using the components.  Currently, Vmgen
   also requires that all the subsequences at the start of a
   superinstruction (prefixes) must be defined as superinstruction before
   the superinstruction.  I.e., if you want to define a superinstruction
   foo4 = load add sub mul
   @end example
   you first have to define @code{load}, @code{add}, @code{sub} and
   @code{mul}, plus
   foo2 = load add
   foo3 = load add sub
   @end example
   Here, @code{sumof4} is the longest prefix of @code{sumof5}, and @code{sumof3}
   is the longest prefix of @code{sumof4}.
   Note that Vmgen assumes that only the code it generates accesses stack
   pointers, the instruction pointer, and various stack items, and it
   performs optimizations based on this assumption.  Therefore, VM
   instructions where your C code changes the instruction pointer should
   only be used as last component; a VM instruction where your C code
   accesses a stack pointer should not be used as component at all.  Vmgen
   does not check these restrictions, they just result in bugs in your
   @cindex include-skipped-insts
   The Vmgen flag @code{include-skipped-insts} influences superinstruction
   code generation.  Currently there is no support in the peephole
   optimizer for both variations, so leave this flag alone for now.
   @c -------------------------------------------------------------------
   @node  Store Optimization, Register Machines, Superinstructions, Input File Format
   @section Store Optimization
   @cindex store optimization
   @cindex optimization, stack stores
   @cindex stack stores, optimization
   @cindex eliminating stack stores
   This minor optimization (0.6\%--0.8\% reduction in executed instructions
   for Gforth) puts additional requirements on the instruction descriptions
   and is therefore disabled by default.
   What does it do?  Consider an instruction like
   dup ( n -- n n )
   @end example
   For simplicity, also assume that we are not caching the top-of-stack in
   a register.  Now, the C code for dup first loads @code{n} from the
   stack, and then stores it twice to the stack, one time to the address
   where it came from; that time is unnecessary, but gcc does not optimize
   it away, so vmgen can do it instead (if you turn on the store
   Vmgen uses the stack item's name to determine if the stack item contains
   the same value as it did at the start.  Therefore, if you use the store
   optimization, you have to ensure that stack items that have the same
   name on input and output also have the same value, and are not changed
   in the C code you supply.  I.e., the following code could fail if you
   turn on the store optimization:
   add1 ( n -- n )
   @end example
   Instead, you have to use different names, i.e.:
   add1 ( n1 -- n2 )
   @end example
   Similarly, the store optimization assumes that the stack pointer is only
   changed by Vmgen-erated code.  If your C code changes the stack pointer,
   use different names in input and output stack items to avoid a (probably
   wrong) store optimization, or turn the store optimization off for this
   VM instruction.
   To turn on the store optimization, write
   \E store-optimization on
   @end example
   at the start of the file.  You can turn this optimization on or off
   between any two VM instruction descriptions.  For turning it off again,
   you can use
   \E store-optimization off
   @end example
   @c -------------------------------------------------------------------
   @node Register Machines,  , Store Optimization, Input File Format
   @section Register Machines
   @cindex Register VM
   @cindex Superinstructions for register VMs
   @cindex tracing of register VMs
   If you want to implement a register VM rather than a stack VM with
   Vmgen, there are two ways to do it: Directly and through
   If you use the direct way, you define instructions that take the
   register numbers as immediate arguments, like this:
   add3 ( #src1 #src2 #dest -- )
   reg[dest] = reg[src1]+reg[src2];
   @end example
   A disadvantage of this method is that during tracing you only see the
   register numbers, but not the register contents.  Actually, with an
   appropriate definition of @code{printarg_src} (@pxref{VM engine}), you
   can print the values of the source registers on entry, but you cannot
   print the value of the destination register on exit.
   If you use superinstructions to define a register VM, you define simple
   instructions that use a stack, and then define superinstructions that
   have no overall stack effect, like this:
   loadreg ( #src -- n )
   n = reg[src];
   storereg ( n #dest -- )
   reg[dest] = n;
   adds ( n1 n2 -- n )
   n = n1+n2;
   add3 = loadreg loadreg adds storereg
   @end example
   An advantage of this method is that you see the values and not just the
   register numbers in tracing.  A disadvantage of this method is that
   currently you cannot generate superinstructions directly, but only
   through generating a sequence of simple instructions (we might change
   this in the future if there is demand).
   Could the register VM support be improved, apart from the issues
   mentioned above?  It is hard to see how to do it in a general way,
   because there are a number of different designs that different people
   mean when they use the term @emph{register machine} in connection with
   VM interpreters.  However, if you have ideas or requests in that
   direction, please let me know (@pxref{Contact}).
   @c ********************************************************************
   @node Error messages, Using the generated code, Input File Format, Top
   @chapter Error messages
   @cindex error messages
   These error messages are created by Vmgen:
   @table @code
   @cindex @code{# can only be on the input side} error
   @item # can only be on the input side
   You have used an instruction-stream prefix (usually @samp{#}) after the
   @samp{--} (the output side); you can only use it before (the input
   @cindex @code{prefix for this combination must be defined earlier} error
   @item the prefix for this superinstruction must be defined earlier
   You have defined a superinstruction (e.g. @code{abc = a b c}) without
   defining its direct prefix (e.g., @code{ab = a b}),
   @cindex @code{sync line syntax} error
   @item sync line syntax
   If you are using a preprocessor (e.g., @command{m4}) to generate Vmgen
   input code, you may want to create @code{#line} directives (aka sync
   lines).  This error indicates that such a line is not in th syntax
   expected by Vmgen (this should not happen; please report the offending
   line in a bug report).
   @cindex @code{syntax error, wrong char} error
   @item syntax error, wrong char
   A syntax error.  If you do not see right away where the error is, it may
   be helpful to check the following: Did you put an empty line in a VM
   instruction where the C code is not delimited by braces (then the empty
   line ends the VM instruction)?  If you used brace-delimited C code, did
   you put the delimiting braces (and only those) at the start of the line,
   without preceding white space?  Did you forget a delimiting brace?
   @cindex @code{too many stacks} error
   @item too many stacks
   Vmgen currently supports 3 stacks (plus the instruction stream); if you
   need more, let us know.
   @cindex @code{unknown prefix} error
   @item unknown prefix
   The stack item does not match any defined type prefix (after stripping
   away any stack prefix).  You should either declare the type prefix you
   want for that stack item, or use a different type prefix
   @cindex @code{unknown primitive} error
   @item unknown primitive
   You have used the name of a simple VM instruction in a superinstruction
   definition without defining the simple VM instruction first.
   @end table
   In addition, the C compiler can produce errors due to code produced by
   Vmgen; e.g., you need to define type cast functions.
   @c ********************************************************************
   @node Using the generated code, Hints, Error messages, Top
   @chapter Using the generated code
   @cindex generated code, usage
   @cindex Using vmgen-erated code
   The easiest way to create a working VM interpreter with Vmgen is
   probably to start with @file{vmgen-ex}, and modify it for your purposes.
   This chapter explains what the various wrapper and generated files do.
   It also contains reference-manual style descriptions of the macros,
   variables etc. used by the generated code, and you can skip that on
   first reading.
   * VM engine::                   Executing VM code
   * VM instruction table::        
   * VM code generation::          Creating VM code (in the front-end)
   * Peephole optimization::       Creating VM superinstructions
   * VM disassembler::             for debugging the front end
   * VM profiler::                 for finding worthwhile superinstructions
   @end menu
   @c --------------------------------------------------------------------
   @node VM engine, VM instruction table, Using the generated code, Using the generated code
   @section VM engine
   @cindex VM instruction execution
   @cindex engine
   @cindex executing VM code
   @cindex @file{engine.c}
   @cindex @file{-vm.i} output file
   The VM engine is the VM interpreter that executes the VM code.  It is
   essential for an interpretive system.
   Vmgen supports two methods of VM instruction dispatch: @emph{threaded
   code} (fast, but gcc-specific), and @emph{switch dispatch} (slow, but
   portable across C compilers); you can use conditional compilation
   (@samp{defined(__GNUC__)}) to choose between these methods, and our
   example does so.
   For both methods, the VM engine is contained in a C-level function.
   Vmgen generates most of the contents of the function for you
   (@file{@var{name}-vm.i}), but you have to define this function, and
   macros and variables used in the engine, and initialize the variables.
   In our example the engine function also includes
   @file{@var{name}-labels.i} (@pxref{VM instruction table}).
   @cindex tracing VM code
   @cindex superinstructions and tracing
   In addition to executing the code, the VM engine can optionally also
   print out a trace of the executed instructions, their arguments and
   results.  For superinstructions it prints the trace as if only component
   instructions were executed; this allows to introduce new
   superinstructions while keeping the traces comparable to old ones
   (important for regression tests).
   It costs significant performance to check in each instruction whether to
   print tracing code, so we recommend producing two copies of the engine:
   one for fast execution, and one for tracing.  See the rules for
   @file{engine.o} and @file{engine-debug.o} in @file{vmgen-ex/Makefile}
   for an example.
   The following macros and variables are used in @file{@var{name}-vm.i}:
   @table @code
   @findex LABEL
   @item LABEL(@var{inst_name})
   This is used just before each VM instruction to provide a jump or
   @code{switch} label (the @samp{:} is provided by Vmgen).  For switch
   dispatch this should expand to @samp{case @var{label}:}; for
   threaded-code dispatch this should just expand to @samp{@var{label}:}.
   In either case @var{label} is usually the @var{inst_name} with some
   prefix or suffix to avoid naming conflicts.
   @findex LABEL2
   @item LABEL2(@var{inst_name})
   This will be used for dynamic superinstructions; at the moment, this
   should expand to nothing.
   @findex NAME
   @item NAME(@var{inst_name_string})
   Called on entering a VM instruction with a string containing the name of
   the VM instruction as parameter.  In normal execution this should be
   expand to nothing, but for tracing this usually prints the name, and
   possibly other information (several VM registers in our example).
   @findex DEF_CA
   @item DEF_CA
   Usually empty.  Called just inside a new scope at the start of a VM
   instruction.  Can be used to define variables that should be visible
   during every VM instruction.  If you define this macro as non-empty, you
   have to provide the finishing @samp{;} in the macro.
   @findex NEXT_P0
   @findex NEXT_P1
   @findex NEXT_P2
   @item NEXT_P0 NEXT_P1 NEXT_P2
   The three parts of instruction dispatch.  They can be defined in
   different ways for best performance on various processors (see
   @file{engine.c} in the example or @file{engine/threaded.h} in Gforth).
   @samp{NEXT_P0} is invoked right at the start of the VM instruction (but
   after @samp{DEF_CA}), @samp{NEXT_P1} right after the user-supplied C
   code, and @samp{NEXT_P2} at the end.  The actual jump has to be
   performed by @samp{NEXT_P2} (if you would do it earlier, important parts
   of the VM instruction would not be executed).
   The simplest variant is if @samp{NEXT_P2} does everything and the other
   macros do nothing.  Then also related macros like @samp{IP},
   @samp{SET_IP}, @samp{IP}, @samp{INC_IP} and @samp{IPTOS} are very
   straightforward to define.  For switch dispatch this code consists just
   of a jump to the dispatch code (@samp{goto next_inst;} in our example);
   for direct threaded code it consists of something like
   @samp{(@{cfa=*ip++; goto *cfa;@})}.
   Pulling code (usually the @samp{cfa=*ip++;}) up into @samp{NEXT_P1}
   usually does not cause problems, but pulling things up into
   @samp{NEXT_P0} usually requires changing the other macros (and, at least
   for Gforth on Alpha, it does not buy much, because the compiler often
   manages to schedule the relevant stuff up by itself).  An even more
   extreme variant is to pull code up even further, into, e.g., NEXT_P1 of
   the previous VM instruction (prefetching, useful on PowerPCs).
   @findex INC_IP
   @item INC_IP(@var{n})
   This increments @code{IP} by @var{n}.
   @findex SET_IP
   @item SET_IP(@var{target})
   This sets @code{IP} to @var{target}.
   @cindex type cast macro
   @findex vm_@var{A}2@var{B}
   @item vm_@var{A}2@var{B}(a,b)
   Type casting macro that assigns @samp{a} (of type @var{A}) to @samp{b}
   (of type @var{B}).  This is mainly used for getting stack items into
   variables and back.  So you need to define macros for every combination
   of stack basic type (@code{Cell} in our example) and type-prefix types
   used with that stack (in both directions).  For the type-prefix type,
   you use the type-prefix (not the C type string) as type name (e.g.,
   @samp{vm_Cell2i}, not @samp{vm_Cell2Cell}).  In addition, you have to
   define a vm_@var{X}2@var{X} macro for the stack's basic type @var{X}
   (used in superinstructions).
   @cindex instruction stream, basic type
   The stack basic type for the predefined @samp{inst-stream} is
   @samp{Cell}.  If you want a stack with the same item size, making its
   basic type @samp{Cell} usually reduces the number of macros you have to
   @cindex unions in type cast macros
   @cindex casts in type cast macros
   @cindex type casting between floats and integers
   Here our examples differ a lot: @file{vmgen-ex} uses casts in these
   macros, whereas @file{vmgen-ex2} uses union-field selection (or
   assignment to union fields).  Note that casting floats into integers and
   vice versa changes the bit pattern (and you do not want that).  In this
   case your options are to use a (temporary) union, or to take the address
   of the value, cast the pointer, and dereference that (not always
   possible, and sometimes expensive).
   @findex vm_two@var{A}2@var{B}
   @findex vm_@var{B}2two@var{A}
   @item vm_two@var{A}2@var{B}(a1,a2,b)
   @item vm_@var{B}2two@var{A}(b,a1,a2)
   Type casting between two stack items (@code{a1}, @code{a2}) and a
   variable @code{b} of a type that takes two stack items.  This does not
   occur in our small examples, but you can look at Gforth for examples
   (see @code{vm_twoCell2d} in @file{engine/forth.h}).
   @cindex stack pointer definition
   @cindex instruction pointer definition
   @item @var{stackpointer}
   For each stack used, the stackpointer name given in the stack
   declaration is used.  For a regular stack this must be an l-expression;
   typically it is a variable declared as a pointer to the stack's basic
   type.  For @samp{inst-stream}, the name is @samp{IP}, and it can be a
   plain r-value; typically it is a macro that abstracts away the
   differences between the various implementations of @code{NEXT_P*}.
   @cindex IMM_ARG
   @findex IMM_ARG
   @item IMM_ARG(access,value)
   Define this to expland to ``(access)''.  This is just a placeholder for
   future extensions.
   @cindex top of stack caching
   @cindex stack caching
   @cindex TOS
   @findex IPTOS
   @item @var{stackpointer}TOS
   The top-of-stack for the stack pointed to by @var{stackpointer}.  If you
   are using top-of-stack caching for that stack, this should be defined as
   variable; if you are not using top-of-stack caching for that stack, this
   should be a macro expanding to @samp{@var{stackpointer}[0]}.  The stack
   pointer for the predefined @samp{inst-stream} is called @samp{IP}, so
   the top-of-stack is called @samp{IPTOS}.
   @findex IF_@var{stackpointer}TOS
   @item IF_@var{stackpointer}TOS(@var{expr})
   Macro for executing @var{expr}, if top-of-stack caching is used for the
   @var{stackpointer} stack.  I.e., this should do @var{expr} if there is
   top-of-stack caching for @var{stackpointer}; otherwise it should do
   @findex SUPER_END
   @item SUPER_END
   This is used by the VM profiler (@pxref{VM profiler}); it should not do
   anything in normal operation, and call @code{vm_count_block(IP)} for
   This is just a hint to Vmgen and does nothing at the C level.
   @findex MAYBE_UNUSED
   This should be defined as @code{__attribute__((unused))} for gcc-2.7 and
   higher.  It suppresses the warnings about unused variables in the code
   for superinstructions.  You need to define this only if you are using
   @findex VM_DEBUG
   @item VM_DEBUG
   If this is defined, the tracing code will be compiled in (slower
   interpretation, but better debugging).  Our example compiles two
   versions of the engine, a fast-running one that cannot trace, and one
   with potential tracing and profiling.
   @findex vm_debug
   @item vm_debug
   Needed only if @samp{VM_DEBUG} is defined.  If this variable contains
   true, the VM instructions produce trace output.  It can be turned on or
   off at any time.
   @findex vm_out
   @item vm_out
   Needed only if @samp{VM_DEBUG} is defined.  Specifies the file on which
   to print the trace output (type @samp{FILE *}).
   @findex printarg_@var{type}
   @item printarg_@var{type}(@var{value})
   Needed only if @samp{VM_DEBUG} is defined.  Macro or function for
   printing @var{value} in a way appropriate for the @var{type}.  This is
   used for printing the values of stack items during tracing.  @var{Type}
   is normally the type prefix specified in a @code{type-prefix} definition
   (e.g., @samp{printarg_i}); in superinstructions it is currently the
   basic type of the stack.
   @end table
   @c --------------------------------------------------------------------
   @node VM instruction table, VM code generation, VM engine, Using the generated code
   @section VM instruction table
   @cindex instruction table
   @cindex opcode definition
   @cindex labels for threaded code
   @cindex @code{vm_prim}, definition
   @cindex @file{-labels.i} output file
   For threaded code we also need to produce a table containing the labels
   of all VM instructions.  This is needed for VM code generation
   (@pxref{VM code generation}), and it has to be done in the engine
   function, because the labels are not visible outside.  It then has to be
   passed outside the function (and assigned to @samp{vm_prim}), to be used
   by the VM code generation functions.
   This means that the engine function has to be called first to produce
   the VM instruction table, and later, after generating VM code, it has to
   be called again to execute the generated VM code (yes, this is ugly).
   In our example program, these two modes of calling the engine function
   are differentiated by the value of the parameter ip0 (if it equals 0,
   then the table is passed out, otherwise the VM code is executed); in our
   example, we pass the table out by assigning it to @samp{vm_prim} and
   returning from @samp{engine}.
   In our example (@file{vmgen-ex/engine.c}), we also build such a table for
   switch dispatch; this is mainly done for uniformity.
   For switch dispatch, we also need to define the VM instruction opcodes
   used as case labels in an @code{enum}.
   For both purposes (VM instruction table, and enum), the file
   @file{@var{name}-labels.i} is generated by Vmgen.  You have to define
   the following macro used in this file:
   @table @code
   @findex INST_ADDR
   @item INST_ADDR(@var{inst_name})
   For switch dispatch, this is just the name of the switch label (the same
   name as used in @samp{LABEL(@var{inst_name})}), for both uses of
   @file{@var{name}-labels.i}.  For threaded-code dispatch, this is the
   address of the label defined in @samp{LABEL(@var{inst_name})}); the
   address is taken with @samp{&&} (@pxref{Labels as Values, , Labels as
   Values, gcc.info, GNU C Manual}).
   @end table
   @c --------------------------------------------------------------------
   @node VM code generation, Peephole optimization, VM instruction table, Using the generated code
   @section VM code generation
   @cindex VM code generation
   @cindex code generation, VM
   @cindex @file{-gen.i} output file
   Vmgen generates VM code generation functions in @file{@var{name}-gen.i}
   that the front end can call to generate VM code.  This is essential for
   an interpretive system.
   @findex gen_@var{inst}
   For a VM instruction @samp{x ( #a b #c -- d )}, Vmgen generates a
   function with the prototype
   void gen_x(Inst **ctp, a_type a, c_type c)
   @end example
   The @code{ctp} argument points to a pointer to the next instruction.
   @code{*ctp} is increased by the generation functions; i.e., you should
   allocate memory for the code to be generated beforehand, and start with
   *ctp set at the start of this memory area.  Before running out of
   memory, allocate a new area, and generate a VM-level jump to the new
   area (this overflow handling is not implemented in our examples).
   @cindex immediate arguments, VM code generation
   The other arguments correspond to the immediate arguments of the VM
   instruction (with their appropriate types as defined in the
   @code{type_prefix} declaration.
   The following types, variables, and functions are used in
   @table @code
   @findex Inst
   @item Inst
   The type of the VM instruction; if you use threaded code, this is
   @code{void *}; for switch dispatch this is an integer type.
   @cindex @code{vm_prim}, use
   @item vm_prim
   The VM instruction table (type: @code{Inst *}, @pxref{VM instruction table}).
   @findex gen_inst
   @item gen_inst(Inst **ctp, Inst i)
   This function compiles the instruction @code{i}.  Take a look at it in
   @file{vmgen-ex/peephole.c}.  It is trivial when you don't want to use
   superinstructions (just the last two lines of the example function), and
   slightly more complicated in the example due to its ability to use
   superinstructions (@pxref{Peephole optimization}).
   @findex genarg_@var{type_prefix}
   @item genarg_@var{type_prefix}(Inst **ctp, @var{type} @var{type_prefix})
   This compiles an immediate argument of @var{type} (as defined in a
   @code{type-prefix} definition).  These functions are trivial to define
   (see @file{vmgen-ex/support.c}).  You need one of these functions for
   every type that you use as immediate argument.
   @end table
   @findex BB_BOUNDARY
   In addition to using these functions to generate code, you should call
   @code{BB_BOUNDARY} at every basic block entry point if you ever want to
   use superinstructions (or if you want to use the profiling supported by
   Vmgen; but this support is also useful mainly for selecting
   superinstructions).  If you use @code{BB_BOUNDARY}, you should also
   define it (take a look at its definition in @file{vmgen-ex/mini.y}).
   You do not need to call @code{BB_BOUNDARY} after branches, because you
   will not define superinstructions that contain branches in the middle
   (and if you did, and it would work, there would be no reason to end the
   superinstruction at the branch), and because the branches announce
   themselves to the profiler.
   @c --------------------------------------------------------------------
   @node Peephole optimization, VM disassembler, VM code generation, Using the generated code
   @section Peephole optimization
   @cindex peephole optimization
   @cindex superinstructions, generating
   @cindex @file{peephole.c}
   @cindex @file{-peephole.i} output file
   You need peephole optimization only if you want to use
   superinstructions.  But having the code for it does not hurt much if you
   do not use superinstructions.
   A simple greedy peephole optimization algorithm is used for
   superinstruction selection: every time @code{gen_inst} compiles a VM
   instruction, it checks if it can combine it with the last VM instruction
   (which may also be a superinstruction resulting from a previous peephole
   optimization); if so, it changes the last instruction to the combined
   instruction instead of laying down @code{i} at the current @samp{*ctp}.
   The code for peephole optimization is in @file{vmgen-ex/peephole.c}.
   You can use this file almost verbatim.  Vmgen generates
   @file{@var{file}-peephole.i} which contains data for the peephole
   @findex init_peeptable
   You have to call @samp{init_peeptable()} after initializing
   @samp{vm_prim}, and before compiling any VM code to initialize data
   structures for peephole optimization.  After that, compiling with the VM
   code generation functions will automatically combine VM instructions
   into superinstructions.  Since you do not want to combine instructions
   across VM branch targets (otherwise there will not be a proper VM
   instruction to branch to), you have to call @code{BB_BOUNDARY}
   (@pxref{VM code generation}) at branch targets.
   @c --------------------------------------------------------------------
   @node VM disassembler, VM profiler, Peephole optimization, Using the generated code
   @section VM disassembler
   @cindex VM disassembler
   @cindex disassembler, VM code
   @cindex @file{disasm.c}
   @cindex @file{-disasm.i} output file
   A VM code disassembler is optional for an interpretive system, but
   highly recommended during its development and maintenance, because it is
   very useful for detecting bugs in the front end (and for distinguishing
   them from VM interpreter bugs).
   Vmgen supports VM code disassembling by generating
   @file{@var{file}-disasm.i}.  This code has to be wrapped into a
   function, as is done in @file{vmgen-ex/disasm.c}.  You can use this file
   almost verbatim.  In addition to @samp{vm_@var{A}2@var{B}(a,b)},
   @samp{vm_out}, @samp{printarg_@var{type}(@var{value})}, which are
   explained above, the following macros and variables are used in
   @file{@var{file}-disasm.i} (and you have to define them):
   @table @code
   @item ip
   This variable points to the opcode of the current VM instruction.
   @cindex @code{IP}, @code{IPTOS} in disassmbler
   @item IP IPTOS
   @samp{IPTOS} is the first argument of the current VM instruction, and
   @samp{IP} points to it; this is just as in the engine, but here
   @samp{ip} points to the opcode of the VM instruction (in contrast to the
   engine, where @samp{ip} points to the next cell, or even one further).
   @findex VM_IS_INST
   @item VM_IS_INST(Inst i, int n)
   Tests if the opcode @samp{i} is the same as the @samp{n}th entry in the
   VM instruction table.
   @end table
   @c --------------------------------------------------------------------
   @node VM profiler,  , VM disassembler, Using the generated code
   @section VM profiler
   @cindex VM profiler
   @cindex profiling for selecting superinstructions
   @cindex superinstructions and profiling
   @cindex @file{profile.c}
   @cindex @file{-profile.i} output file
   The VM profiler is designed for getting execution and occurence counts
   for VM instruction sequences, and these counts can then be used for
   selecting sequences as superinstructions.  The VM profiler is probably
   not useful as profiling tool for the interpretive system.  I.e., the VM
   profiler is useful for the developers, but not the users of the
   interpretive system.
   The output of the profiler is: for each basic block (executed at least
   once), it produces the dynamic execution count of that basic block and
   all its subsequences; e.g.,
          9227465  lit storelocal 
          9227465  storelocal branch 
          9227465  lit storelocal branch 
   @end example
   I.e., a basic block consisting of @samp{lit storelocal branch} is
   executed 9227465 times.
   @cindex @file{stat.awk}
   @cindex @file{seq2rule.awk}
   This output can be combined in various ways.  E.g.,
   @file{vmgen-ex/stat.awk} adds up the occurences of a given sequence wrt
   dynamic execution, static occurence, and per-program occurence.  E.g.,
         2      16        36910041 loadlocal lit 
   @end example
   indicates that the sequence @samp{loadlocal lit} occurs in 2 programs,
   in 16 places, and has been executed 36910041 times.  Now you can select
   superinstructions in any way you like (note that compile time and space
   typically limit the number of superinstructions to 100--1000).  After
   you have done that, @file{vmgen/seq2rule.awk} turns lines of the form
   above into rules for inclusion in a Vmgen input file.  Note that this
   script does not ensure that all prefixes are defined, so you have to do
   that in other ways.  So, an overall script for turning profiles into
   superinstructions can look like this:
   awk -f stat.awk fib.prof test.prof|
   awk '$3>=10000'|                #select sequences
   fgrep -v -f peephole-blacklist| #eliminate wrong instructions
   awk -f seq2rule.awk|            #turn into superinstructions
   sort -k 3 >mini-super.vmg       #sort sequences
   @end example
   Here the dynamic count is used for selecting sequences (preliminary
   results indicate that the static count gives better results, though);
   the third line eliminates sequences containing instructions that must not
   occur in a superinstruction, because they access a stack directly.  The
   dynamic count selection ensures that all subsequences (including
   prefixes) of longer sequences occur (because subsequences have at least
   the same count as the longer sequences); the sort in the last line
   ensures that longer superinstructions occur after their prefixes.
   But before using this, you have to have the profiler.  Vmgen supports its
   creation by generating @file{@var{file}-profile.i}; you also need the
   wrapper file @file{vmgen-ex/profile.c} that you can use almost verbatim.
   @cindex @code{SUPER_END} in profiling
   @cindex @code{BB_BOUNDARY} in profiling
   The profiler works by recording the targets of all VM control flow
   changes (through @code{SUPER_END} during execution, and through
   @code{BB_BOUNDARY} in the front end), and counting (through
   @code{SUPER_END}) how often they were targeted.  After the program run,
   the numbers are corrected such that each VM basic block has the correct
   count (entering a block without executing a branch does not increase the
   count, and the correction fixes that), then the subsequences of all
   basic blocks are printed.  To get all this, you just have to define
   @code{SUPER_END} (and @code{BB_BOUNDARY}) appropriately, and call
   @code{vm_print_profile(FILE *file)} when you want to output the profile
   on @code{file}.
   @cindex @code{VM_IS_INST} in profiling
   The @file{@var{file}-profile.i} is similar to the disassembler file, and
   it uses variables and functions defined in @file{vmgen-ex/profile.c},
   plus @code{VM_IS_INST} already defined for the VM disassembler
   (@pxref{VM disassembler}).
   @c **********************************************************
   @node Hints, The future, Using the generated code, Top
   @chapter Hints
   @cindex hints
   * Floating point::              and stacks
   @end menu
   @c --------------------------------------------------------------------
   @node Floating point,  , Hints, Hints
   @section Floating point
   How should you deal with floating point values?  Should you use the same
   stack as for integers/pointers, or a different one?  This section
   discusses this issue with a view on execution speed.
   The simpler approach is to use a separate floating-point stack.  This
   allows you to choose FP value size without considering the size of the
   integers/pointers, and you avoid a number of performance problems.  The
   main downside is that this needs an FP stack pointer (and that may not
   fit in the register file on the 386 arhitecture, costing some
   performance, but comparatively little if you take the other option into
   account).  If you use a separate FP stack (with stack pointer @code{fp}),
   using an fpTOS is helpful on most machines, but some spill the fpTOS
   register into memory, and fpTOS should not be used there.
   The other approach is to share one stack (pointed to by, say, @code{sp})
   between integer/pointer and floating-point values.  This is ok if you do
   not use @code{spTOS}.  If you do use @code{spTOS}, the compiler has to
   decide whether to put that variable into an integer or a floating point
   register, and the other type of operation becomes quite expensive on
   most machines (because moving values between integer and FP registers is
   quite expensive).  If a value of one type has to be synthesized out of
   two values of the other type (@code{double} types), things are even more
   One way around this problem would be to not use the @code{spTOS}
   supported by Vmgen, but to use explicit top-of-stack variables (one for
   integers, one for FP values), and having a kind of accumulator+stack
   architecture (e.g., Ocaml bytecode uses this approach); however, this is
   a major change, and it's ramifications are not completely clear.
   @c **********************************************************
   @node The future, Changes, Hints, Top
   @chapter The future
   @cindex future ideas
   We have a number of ideas for future versions of Vmgen.  However, there
   are so many possible things to do that we would like some feedback from
   you.  What are you doing with Vmgen, what features are you missing, and
   One idea we are thinking about is to generate just one @file{.c} file
   instead of letting you copy and adapt all the wrapper files (you would
   still have to define stuff like the type-specific macros, and stack
   pointers etc. somewhere).  The advantage would be that, if we change the
   wrapper files between versions, you would not need to integrate your
   changes and our changes to them; Vmgen would also be easier to use for
   beginners.  The main disadvantage of that is that it would reduce the
   flexibility of Vmgen a little (well, those who like flexibility could
   still patch the resulting @file{.c} file, like they are now doing for
   the wrapper files).  In any case, if you are doing things to the wrapper
   files that would cause problems in a generated-@file{.c}-file approach,
   please let us know.
   @c **********************************************************
   @node Changes, Contact, The future, Top
   @chapter Changes
   @cindex Changes from old versions
   User-visible changes between 0.5.9-20020822 and 0.5.9-20020901:
   The store optimization is now disabled by default, but can be enabled by
   the user (@pxref{Store Optimization}).  Documentation for this
   optimization is also new.
   User-visible changes between 0.5.9-20010501 and 0.5.9-20020822:
   There is now a manual (in info, HTML, Postscript, or plain text format).
   There is the vmgen-ex2 variant of the vmgen-ex example; the new
   variant uses a union type instead of lots of casting.
   Both variants of the example can now be compiled with an ANSI C compiler
   (using switch dispatch and losing quite a bit of performance); tested
   with @command{lcc}.
   Users of the gforth-0.5.9-20010501 version of Vmgen need to change
   several things in their source code to use the current version.  I
   recommend keeping the gforth-0.5.9-20010501 version until you have
   completed the change (note that you can have several versions of Gforth
   installed at the same time).  I hope to avoid such incompatible changes
   in the future.
   The required changes are:
   @table @code
   @cindex @code{TAIL;}, changes
   @item TAIL;
   has been renamed into @code{INST_TAIL;} (less chance of an accidental
   @cindex @code{vm_@var{A}2@var{B}}, changes
   @item vm_@var{A}2@var{B}
   now takes two arguments.
   @cindex @code{vm_two@var{A}2@var{B}}, changes
   @item vm_two@var{A}2@var{B}(b,a1,a2);
   changed to vm_two@var{A}2@var{B}(a1,a2,b) (note the absence of the @samp{;}).
   @end table
   Also some new macros have to be defined, e.g., @code{INST_ADDR}, and
   @code{LABEL}; some macros have to be defined in new contexts, e.g.,
   @code{VM_IS_INST} is now also needed in the disassembler.
   @c *********************************************************
   @node Contact, Copying This Manual, Changes, Top
   @chapter Contact
   To report a bug, use
   For discussion on Vmgen (e.g., how to use it), use the mailing list
   @email{bug-vmgen@@mail.freesoftware.fsf.org} (use
   @url{http://mail.gnu.org/mailman/listinfo/help-vmgen} to subscribe).
   You can find vmgen information at
   @c ***********************************************************
   @node Copying This Manual, Index, Contact, Top
   @appendix Copying This Manual
   * GNU Free Documentation License::  License for copying this manual.
   @end menu
   @node GNU Free Documentation License,  , Copying This Manual, Copying This Manual
   @appendixsec GNU Free Documentation License
   @include fdl.texi
   @node Index,  , Copying This Manual, Top
   @unnumbered Index
 @section Why interpreters?  @printindex cp

Removed from v.1.1  
changed lines
  Added in v.1.30

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