@c -*-texinfo-*-
@c This is part of the GNU Guile Reference Manual.
@c Copyright (C)  2008
@c   Free Software Foundation, Inc.
@c See the file guile.texi for copying conditions.

@node A Virtual Machine for Guile
@section A Virtual Machine for Guile

Guile has both an interpreter and a compiler. To a user, the
difference is largely transparent -- interpreted and compiled
procedures can call each other as they please.

The difference is that the compiler creates and interprets bytecode
for a custom virtual machine, instead of interpreting the
S-expressions directly. Running compiled code is faster than running
interpreted code.

The virtual machine that does the bytecode interpretation is a part of
Guile itself. This section describes the nature of Guile's virtual
machine.

@menu
* Why a VM?::                   
* VM Concepts::                 
* Stack Layout::                
* Variables and the VM::                   
* Compiled Procedures::         
* Instruction Set::
@end menu

@node Why a VM?
@subsection Why a VM?

For a long time, Guile only had an interpreter, called the evaluator.
Guile's evaluator operates directly on the S-expression representation
of Scheme source code.

But while the evaluator is highly optimized and hand-tuned, and
contains some extensive speed trickery (REFFIXME memoization), it
still performs many needless computations during the course of
evaluating an expression. For example, application of a function to
arguments needlessly conses up the arguments in a list. Evaluation of
an expression always has to figure out what the car of the expression
is -- a procedure, a memoized form, or something else. All values have
to be allocated on the heap. Et cetera.

The solution to this problem is to compile the higher-level language,
Scheme, into a lower-level language for which all of the checks and
dispatching have already been done -- the code is instead stripped to
the bare minimum needed to ``do the job''.

The question becomes then, what low-level language to choose? There
are many options. We could compile to native code directly, but that
poses portability problems for Guile, as it is a highly cross-platform
project.

So we want the performance gains that compilation provides, but we
also want to maintain the portability benefits of a single code path.
The obvious solution is to compile to a virtual machine that is
present on all Guile installations.

The easiest (and most fun) way to depend on a virtual machine is to
implement the virtual machine within Guile itself. This way the
virtual machine provides what Scheme needs (tail calls, multiple
values, call/cc) and can provide optimized inline instructions for
Guile (cons, struct-ref, etc.).

So this is what Guile does. The rest of this section describes that VM
that Guile implements, and the compiled procedures that run on it.

Note that this decision to implement a bytecode compiler does not
preclude native compilation. We can compile from bytecode to native
code at runtime, or even do ahead of time compilation. More
possibilities are discussed in REFFIXME.

@node VM Concepts
@subsection VM Concepts

A virtual machine (VM) is a Scheme object. Users may create virtual
machines using the standard procedures described later in this manual,
but that is usually unnecessary, as Guile ensures that there is one
virtual machine per thread. When a VM-compiled procedure is run, Guile
looks up the virtual machine for the current thread and executes the
procedure using that VM.

Guile's virtual machine is a stack machine -- that is, it has few
registers, and the instructions defined in the VM operate by pushing
and popping values from a stack.

Stack memory is exclusive to the virtual machine that owns it. In
addition to their stacks, virtual machines also have access to the
global memory (modules, global bindings, etc) that is shared among
other parts of Guile, including other VMs.

A VM has generic instructions, such as those to reference local
variables, and instructions designed to support Guile's languages --
mathematical instructions that support the entire numerical tower, an
inlined implementation of @code{cons}, etc.

The registers that a VM has are as follows:

@itemize
@item ip - Instruction pointer
@item sp - Stack pointer
@item fp - Frame pointer
@end itemize

In other architectures, the instruction pointer is sometimes called
the ``program counter'' (pc). This set of registers is pretty typical
for stack machines; their exact meanings in the context of Guile's VM
is described below REFFIXME.

A virtual machine executes by loading a compiled procedure, and
executing the object code associated with that procedure. Of course,
that procedure may call other procedures, tail-call others, ad
infinitum -- indeed, within a guile whose modules have all been
compiled to object code, one might never leave the virtual machine.

@c wingo: I wish the following were true, but currently we just use
@c the one engine. This kind of thing is possible tho.

@c A VM may have one of three engines: reckless, regular, or debugging.
@c Reckless engine is fastest but dangerous.  Regular engine is normally
@c fail-safe and reasonably fast.  Debugging engine is safest and
@c functional but very slow.

@node Stack Layout
@subsection Stack Layout

While not strictly necessary to understand how to work with the VM, it
is instructive and sometimes entertaining to consider the struture of
the VM stack.

Logically speaking, a VM stack is composed of ``frames''. Each frame
corresponds to the application of one compiled procedure, and contains
storage space for arguments, local variables, intermediate values, and
some bookkeeping information (such as what to do after the frame
computes its value).

While the compiler is free to do whatever it wants to, as long as the
semantics of a computation are preserved, in practice every time you
call a function, a new frame is created. (The notable exception of
course is the tail call case, @pxref{Tail Calls}.)

Within a frame, you have the data associated with the function
application itself, which is of a fixed size, and the stack space for
intermediate values. Sometimes only the former is referred to as the
``frame'', and the latter is the ``stack'', although all pending
application frames can have some intermediate computations interleaved
on the stack.

The structure of the fixed part of an application frame is as follows:

@example
             Stack
   |                  | <- fp + bp->nargs + bp->nlocs + 5
   +------------------+    = SCM_FRAME_UPPER_ADDRESS (fp)
   | Return address   |
   | MV return address|
   | Dynamic link     |
   | Heap link        |
   | External link    | <- fp + bp->nargs + bp->nlocs
   | Local variable 1 |    = SCM_FRAME_DATA_ADDRESS (fp)
   | Local variable 0 | <- fp + bp->nargs
   | Argument 1       |
   | Argument 0       | <- fp
   | Program          | <- fp - 1
   +------------------+    = SCM_FRAME_LOWER_ADDRESS (fp)
   |                  |
@end example

In the above drawing, the stack grows upward. The intermediate values
stored in the application of this frame are stored above
@code{SCM_FRAME_UPPER_ADDRESS (fp)}. @code{bp} refers to the
@code{struct scm_program*} data associated with the program at
@code{fp - 1}. @code{nargs} and @code{nlocs} are properties of the
compiled procedure, which will be discussed later.

The individual fields of the frame are as follows:

@table @asis
@item Return address
The @code{ip} that was in effect before this program was applied. When
we return from this activation frame, we will jump back to this
@code{ip}.

@item MV return address
The @code{ip} to return to if this application returns multiple
values. For continuations that only accept one value, this value will
be @code{NULL}; for others, it will be an @code{ip} that points to a
multiple-value return address in the calling code. That code will
expect the top value on the stack to be an integer -- the number of
values being returned -- and that below that integer there are the
values being returned.

@item Dynamic link
This is the @code{fp} in effect before this program was applied. In
effect, this and the return address are the registers that are always
``saved''.

@item Heap link
This field is unused and needs to be removed ASAP.

@item External link
This field is a reference to the list of heap-allocated variables
associated with this frame. A discussion of heap versus stack
allocation can be found in REFFIXME.

@item Local variable @var{n}
Lambda-local variables that are allocated on the stack are all
allocated as part of the frame. This makes access to non-captured,
non-mutated variables very cheap.

@item Argument @var{n}
The calling convention of the VM requires arguments of a function
application to be pushed on the stack, and here they are. Normally
references to arguments dispatch to these locations on the stack.
However if an argument has to be stored on the heap, it will be copied
from its initial value here onto a location in the heap, and
thereafter only referenced on the heap.

@item Program
This is the program being applied. Programs are discussed in REFFIXME!
@end table

@node Variables and the VM
@subsection Variables and the VM

Let's think about the following Scheme code as an example:

@example
  (define (foo a)
    (lambda (b) (list foo a b)))
@end example

Within the lambda expression, "foo" is a top-level variable, "a" is a
lexically captured variable, and "b" is a local variable.

That is to say: @code{b} may safely be allocated on the stack, as
there is no enclosed procedure that references it, nor is it ever
mutated.

@code{a}, on the other hand, is referenced by an enclosed procedure,
that of the lambda. Thus it must be allocated on the heap, as it may
(and will) outlive the dynamic extent of the invocation of @code{foo}.

@code{foo} is a toplevel variable, as mandated by Scheme's semantics:

@example
  (define proc (foo 'bar)) ; assuming prev. definition of @code{foo}
  (define foo 42)          ; redefinition
  (proc 'baz)
  @result{} (42 bar baz)
@end example

Note that variables that are mutated (via @code{set!}) must be
allocated on the heap, even if they are local variables. This is
because any called subprocedure might capture the continuation, which
would need to capture locations instead of values. Thus perhaps
counterintuitively, what would seem ``closer to the metal'', viz
@code{set!}, actually forces heap allocation instead of stack
allocation.

@node Compiled Procedures
@subsection Compiled Procedures

By default, when you enter in expressions at Guile's REPL, they are
first compiled to VM object code, then that VM object code is executed
to produce a value. If the expression evaluates to a procedure, the
result of this process is a compiled procedure.

A compiled procedure is a compound object, consisting of its bytecode,
a reference to any captured lexical variables, an object array, and
some metadata such as the procedure's arity, name, and documentation.
You can pick apart these pieces with the accessors in @code{(system vm
program)}. REFFIXME, for a full API reference.



@c @example
@c (use-modules (system vm program))
@c @end example

@c @deffn {Scheme Procedure} program-bytecode program
@c @deffnx {C Function} scm_program_bytecode (program)
@c Returns the object code associated with this program.
@c @end deffn

@c @deffn {Scheme Procedure} program-arity program
@c @deffnx {C Function} scm_program_arity (program)
@c Returns a representation of the ``arity'' of a program.
@c @end deffn

@c @deffn {Scheme Procedure} arity:nargs arity
@c @deffnx {Scheme Procedure} arity:nrest arity
@c @deffnx {Scheme Procedure} arity:nlocs arity
@c @deffnx {Scheme Procedure} arity:nexts arity
@c Accessors for arity objects, as returned by @code{program-arity}.

@c @code{nargs} is the number of arguments to the procedure, and
@c @code{nrest} will be non-zero if the last argument is a rest argument.

@c The other two accessors determine the number of local and external
@c (heap-allocated) variables that this procedure will need to have
@c allocated.
@c @end deffn

We can see how these concepts tie together by disassembling the
@code{foo} function to see what is going on:

@smallexample
scheme@@(guile-user)> (define (foo a) (lambda (b) (list foo a b)))
scheme@@(guile-user)> ,x foo
Disassembly of #<program foo (a)>:

Bytecode:

   0    (local-ref 0)                   ;; `a' (arg)
   2    (external-set 0)                ;; `a' (arg)
   4    (object-ref 0)                  ;; #<program #(0 28 #f) (b)>
   6    (make-closure)                                        at (unknown file):0:16
   7    (return)                        

----------------------------------------
Disassembly of #<program #(0 28 #f) (b)>:

Bytecode:

   0    (toplevel-ref 0)                ;; `list'
   2    (toplevel-ref 1)                ;; `foo'
   4    (external-ref 0)                ;; (closure variable)
   6    (local-ref 0)                   ;; `b' (arg)
   8    (goto/args 3)                                         at (unknown file):0:28
@end smallexample

At @code{ip} 0 and 2, we do the copy from argument to heap for
@code{a}. @code{Ip} 4 loads up the compiled lambda, and then at
@code{ip} 6 we make a closure -- binding code (from the compiled
lambda) with data (the heap-allocated variables). Finally we return
the closure.

The second stanza disassembles the compiled lambda. Toplevel variables
are resolved relative to the module that was current when the
procedure was created. This lookup occurs lazily, at the first time
the variable is actually referenced, and the location of the lookup is
cached so that future references are very cheap. REFFIXME xref
toplevel-ref, for more details.

Then we see a reference to an external variable, corresponding to
@code{a}. The disassembler doesn't have enough information to give a
name to that variable, so it just marks it as being a ``closure
variable''. Finally we see the reference to @code{b}, then a tail call
(@code{goto/args}) with three arguments.

@node Instruction Set
@subsection Instruction Set

CISC, etc. Link to when to add instructions.

@menu
* Environment Control Instructions::  
* Branch Instructions::         
* Subprogram Control Instructions::  
* Data Control Instructions::   
* Miscellaneous Instructions::  
* Inlined Scheme Instructions::  
* Inlined Mathematical Instructions::  
@end menu

@node Environment Control Instructions
@subsubsection Environment Control Instructions

@deffn Instruction link-now
Pop a symbol from the stack, and look it and push the corresponding
variable object onto the stack. If the symbol is not bound yet, an
error will be signalled.
@end deffn

@deffn Instruction variable-ref
Dereference the variable object which is on top of the stack and
replace it by the value of the variable it represents.
@end deffn

@deffn Instruction variable-set
Set the value of the variable on top of the stack (at @code{sp[0]}) to
the object located immediately before (at @code{sp[-1]}).
@end deffn

@deffn Instruction local-ref offset
Push onto the stack the value of the local variable located at
@var{offset} within the current stack frame.
@end deffn

@deffn Instruction local-set offset
Pop the Scheme object located on top of the stack and make it the new
value of the local variable located at @var{offset} within the current
stack frame.
@end deffn

@deffn Instruction external-ref offset
Push the value of the closure variable located at position
@var{offset} within the program's list of external variables.
@end deffn

@deffn Instruction external-set offset
Pop the Scheme object located on top of the stack and make it the new
value of the closure variable located at @var{offset} within the
program's list of external variables.
@end deffn

@deffn Instruction externals
something here...
@end deffn

@deffn Instruction toplevel-ref offset
Foo...
@end deffn

@deffn Instruction toplevel-ref offset
Bar...
@end deffn

@deffn Instruction make-closure
Pop the program object from the stack and assign it the current
closure variable list as its closure.  Push the result program
object.
@end deffn

@node Branch Instructions
@subsubsection Branch Instructions

All the conditional branch instructions described below work in the
same way:

@itemize
@item They take the Scheme object located on the stack and use it as
the branch condition;
@item If the condition is false, then program execution continues with
the next instruction;
@item If the condition is true, then the instruction pointer is
increased by the offset passed as an argument to the branch
instruction;
@item Finally, when the instruction finished, the condition object is
removed from the stack.
@end itemize

Note that the offset passed to the instruction is encoded on two 8-bit
integers which are then combined by the VM as one 16-bit integer.

@deffn Instruction br offset
Jump to @var{offset}.
@end deffn

@deffn Instruction br-if offset
Jump to @var{offset} if the condition on the stack is not false.
@end deffn

@deffn Instruction br-if-not offset
Jump to @var{offset} if the condition on the stack is false.
@end deffn

@deffn Instruction br-if-eq offset
Jump to @var{offset} if the two objects located on the stack are
equal in the sense of @var{eq?}.  Note that, for this instruction, the
stack pointer is decremented by two Scheme objects instead of only
one.
@end deffn

@deffn Instruction br-if-not-eq offset
Same as @var{br-if-eq} for non-@code{eq?} objects.
@end deffn

@deffn Instruction br-if-null offset
Jump to @var{offset} if the object on the stack is @code{'()}.
@end deffn

@deffn Instruction br-if-not-null offset
Jump to @var{offset} if the object on the stack is not @code{'()}.
@end deffn


@node Subprogram Control Instructions
@subsubsection Subprogram Control Instructions

Programs (read: ``compiled procedure'') may refer to external
bindings, like variables or functions defined outside the program
itself, in the environment in which it will evaluate at run-time.  In
a sense, a program's environment and its bindings are an implicit
parameter of every program.

@cindex object table
In order to handle such bindings, each program has an @dfn{object
table} associated to it.  This table (actually a Scheme vector)
contains all constant objects referenced by the program.  The object
table of a program is initialized right before a program is loaded
with @code{load-program}.

Variable objects are one such type of constant object: when a global
binding is defined, a variable object is associated to it and that
object will remain constant over time, even if the value bound to it
changes. Therefore, toplevel bindings only need to be looked up once.
ThereafterReferences to the corresponding toplevel variables from within the
program are then performed via the @code{object-ref} instruction and
are almost as fast as local variable references.

Let us consider the following program (procedure) which references
external bindings @code{frob} and @code{%magic}:

@example
(lambda (x)
  (frob x %magic))
@end example

This yields the following assembly code:

@example
(make-int8 64)   ;; number of args, vars, etc. (see below)
(link "frob")
(link "%magic")
(vector 2)       ;; object table (external bindings)
...
(load-program #u8(20 0 23 21 0 20 1 23 36 2))
(return)
@end example

All the instructions occurring before @var{load-program} (some were
omitted for simplicity) form a @dfn{prologue} which, among other
things, pushed an object table (a vector) that contains the variable
objects for the variables bound to @var{frob} and @var{%magic}.  This
vector and other data pushed onto the stack are then popped by the
@var{load-program} instruction.

Besides, the @var{load-program} instruction takes one explicit
argument which is the bytecode of the program itself.  Disassembled,
this bytecode looks like:

@example
(object-ref 0)  ;; push the variable object of `frob'
(variable-ref)  ;; dereference it
(local-ref 0)   ;; push the value of `x'
(object-ref 1)  ;; push the variable object of `%magic'
(variable-ref)  ;; dereference it
(tail-call 2)   ;; call `frob' with two parameters
@end example

This clearly shows that there is little difference between references
to local variables and references to externally bound variables since
lookup of externally bound variables if performed only once before the
program is run.

@deffn Instruction load-integer length
embeds 32-bit int in instruction stream
@end deffn
@deffn Instruction load-number length
embeds arbitrary number in instruction stream (as string)
@end deffn
@deffn Instruction load-string length
embeds string in instruction stream
@end deffn
@deffn Instruction load-symbol length
embeds symbol in instruction stream
@end deffn
@deffn Instruction load-keyword length
embeds keyword in instruction stream
@end deffn

@deffn Instruction link-now
FIXME: should not be in the loaders
Pops a symbol, pushes a variable.
@end deffn

@deffn Instruction define
Pulls a symbol from the instruction stream, pushes the variable.
@end deffn

FIXME: remove late-bind instruction

@deffn Instruction load-program bytecode
Load the program whose bytecode is @var{bytecode} (a u8vector), pop
its meta-information from the stack, and push a corresponding program
object onto the stack.  The program's meta-information may consist of
(in the order in which it should be pushed onto the stack):

@itemize
@item optionally, a pair representing meta-data (see the
@var{program-meta} procedure); [FIXME: explain their meaning]
@item optionally, a vector which is the program's object table (a
program that does not reference external bindings does not need an
object table);
@item either one immediate integer or four immediate integers
representing respectively the number of arguments taken by the
function (@var{nargs}), the number of @dfn{rest arguments}
(@var{nrest}, 0 or 1), the number of local variables (@var{nlocs}) and
the number of external variables (@var{nexts}) (@pxref{Environment
Control Instructions}).
@end itemize

@end deffn

@deffn Instruction object-ref n
Push @var{n}th value from the current program's object vector.
@end deffn

@deffn Instruction return
Free the program's frame.
@end deffn

@deffn Instruction call nargs
Call the procedure, continuation or program located at
@code{sp[-nargs]} with the @var{nargs} arguments located from
@code{sp[0]} to @code{sp[-nargs + 1]}.  The
procedure/continuation/program and its arguments are dropped from the
stack and the result is pushed.  When calling a program, the
@code{call} instruction reserves room for its local variables on the
stack, and initializes its list of closure variables and its vector of
externally bound variables.
@end deffn

@deffn Instruction goto/args nargs
Same as @code{call} except that, for tail-recursive calls to a
program, the current stack frame is re-used, as required by RnRS.
This instruction is otherwise similar to @code{call}.
@end deffn

@deffn Instruction call/nargs
@end deffn
@deffn Instruction goto/nargs
@end deffn
@deffn Instruction apply
@end deffn
@deffn Instruction goto/apply
@end deffn

@deffn Instruction call/cc
@end deffn
@deffn Instruction goto/cc
@end deffn

@deffn Instruction mv-call
@end deffn
@deffn Instruction return/values
@end deffn
@deffn Instruction return/values*
@end deffn
@deffn Instruction return/values*
@end deffn
@deffn Instruction truncate-values
@end deffn


@node Data Control Instructions
@subsubsection Data Control Instructions

@deffn Instruction make-int8 value
Push @var{value}, an 8-bit integer, onto the stack.
@end deffn

@deffn Instruction make-int8:0
Push the immediate value @code{0} onto the stack.
@end deffn

@deffn Instruction make-int8:1
Push the immediate value @code{1} onto the stack.
@end deffn

@deffn Instruction make-int16 value
Push @var{value}, a 16-bit integer, onto the stack.
@end deffn

@deffn Instruction make-false
Push @code{#f} onto the stack.
@end deffn

@deffn Instruction make-true
Push @code{#t} onto the stack.
@end deffn

@deffn Instruction make-eol
Push @code{'()} onto the stack.
@end deffn

@deffn Instruction make-char8 value
Push @var{value}, an 8-bit character, onto the stack.
@end deffn

@deffn Instruction list n
Pops off the top @var{n} values off of the stack, consing them up into
a list, then pushes that list on the stack. What was the topmost value
will be the last element in the list.
@end deffn

@deffn Instruction vector n
Create and fill a vector with the top @var{n} values from the stack,
popping off those values and pushing on the resulting vector.
@end deffn

@deffn Instruction mark
Pushes a special value onto the stack that other stack instructions
like @code{list-mark} can use.
@end deffn

@deffn Instruction list-mark
Create a list from values from the stack, as in @code{list}, but
instead of knowing beforehand how many there will be, keep going until
we see a @code{mark} value.
@end deffn

@deffn Instruction cons-mark
As @code{cons*} is to @code{list}, so @code{cons-mark} is to
@code{list-mark}.
@end deffn

@deffn Instruction vector-mark
Like @code{list-mark}, but makes a vector instead of a list.
@end deffn

@deffn Instruction list-break
The opposite of @code{list}: pops a value, which should be a list, and
pushes its elements on the stack.
@end deffn

@node Miscellaneous Instructions
@subsubsection Miscellaneous Instructions

@deffn Instruction nop
Does nothing!
@end deffn

@deffn Instruction halt
Exits the VM, returning a SCM value. Say more about this.
@end deffn

@deffn Instruction break
Does nothing, but invokes the break hook.
@end deffn

@deffn Instruction drop
Pops off the top value from the stack, throwing it away.
@end deffn

@deffn Instruction dup
Re-pushes the top value onto the stack.
@end deffn

@deffn Instruction void
Pushes ``the unspecified value'' onto the stack.
@end deffn

@node Inlined Scheme Instructions
@subsubsection Inlined Scheme Instructions

@deffn Instruction not x
@end deffn
@deffn Instruction not-not x
@end deffn
@deffn Instruction eq? x y
@end deffn
@deffn Instruction not-eq? x y
@end deffn
@deffn Instruction null?
@end deffn
@deffn Instruction not-null?
@end deffn
@deffn Instruction eqv? x y
@end deffn
@deffn Instruction equal? x y
@end deffn
@deffn Instruction pair? x y
@end deffn
@deffn Instruction list? x y
@end deffn
@deffn Instruction set-car! pair x
@end deffn
@deffn Instruction set-cdr! pair x
@end deffn
@deffn Instruction slot-ref struct n
@end deffn
@deffn Instruction slot-set struct n x
@end deffn
@deffn Instruction cons
@end deffn
@deffn Instruction car
@end deffn
@deffn Instruction cdr
@end deffn

@node Inlined Mathematical Instructions
@subsubsection Inlined Mathematical Instructions

@deffn Instruction add
@end deffn
@deffn Instruction sub
@end deffn
@deffn Instruction mul
@end deffn
@deffn Instruction div
@end deffn
@deffn Instruction quo
@end deffn
@deffn Instruction rem
@end deffn
@deffn Instruction mod
@end deffn
@deffn Instruction ee?
@end deffn
@deffn Instruction lt?
@end deffn
@deffn Instruction gt?
@end deffn
@deffn Instruction le?
@end deffn
@deffn Instruction ge?
@end deffn