develooper Front page | perl.perl6.internals | Postings from July 2001

PDD5: opcode function specs

From:
Dan Sugalski
Date:
July 16, 2001 10:50
Subject:
PDD5: opcode function specs
Message ID:
5.1.0.14.0.20010716134834.01f7d8e8@24.8.96.48
Here's a prelim version of the opcode function specification. It's pretty 
short, but there's not all that much to it.

I hope to have a working version of the interpreter soonish, by the end of 
TPC 5 at least. (It won't *do* much, mind, but it will whiz through an 
opcode stream)

----Cut here---

=head1 TITLE

Opcode Function specs

=head1 VERSION

1.0

=head2 CURRENT

    Maintainer: Dan Sugalski <dan@sidhe.org>
    Class: Internals
    PDD Number: 5
    Version: 1.0
    Status: Developing
    Last Modified: 16 Jul 2001
    PDD Format: 1
    Language: English

=head2 HISTORY

None. First version

=head1 CHANGES

None. First version

=head1 ABSTRACT

This PDD specifies how the opcode functions should behave and how they
are called by the Parrot interpreter.

=head1 DESCRIPTION

The opcode functions are the workhorse of the Parrot engine. They
control program flow and do most of the work in a program. (The rest
being done by the variable vtable functions)

Opcode functions have very few limitations or restrictions on them. In
particular, opcode functions:

=over 4

=item *

Can assume a working interpreter

=item *

Must leave all interpreter registers the way they were found, unless the 
opcode signature indicates otherwise

=back

Each opcode has two separate functions. The first function takes two
parameters, the current interpreter pointer and current interpreter
PC, and returns the address of the next opcode to execute. The second
function takes zero or more paramters as addresses, register numbers,
integers, or floating point numbers and optionally returns either the
address of the next opcode or the register number holding the address
of the next opcode. These are referred to as the I<wrapping function>
and the I<inner function>, respectively.

The I<wrapping function> is required, as this is the code that the
interpreter will call. Normally this is automatically generated.

The I<inner function> is the code that gets directly executed when
perl gets TIL-ified. If there is no I<inner function> for some reason,
then your opcode will likely run slower (as the interpreter would need
to set up the registers and other stuff that would normally get
stripped away for speed)

=head2 The wrapping function

This is the function that the interpreter actually executes. It has
all the intimate knowledge of its parameters embedded in it, and is
responsible for figuring out what register data it needs and from
where.

This function is generally created automatically by
C<opcode_process.pl>, so the programmer doesn't have to create it. If,
for some reason, you do need or want to write it (for example if you
have no inner function) that's fine.

=head2 The inner function

The inner function is the code that actually does the work. This is
generally a chunk of C code, though the interpreter will be able to
call perl code soon.

=head1 IMPLEMENTATION

=head2 Prototype declaration of inner function

   RETURN function(INPUT[, INPUT[, INPUT...]])

The C<RETURN> type may be one of:

=over 4

=item void

Indicates the function returns nothing. The I<wrapping function> will
automagically figure out what address to return based on the size of
the current opcode.

=item void *

Indicates the function returns the address of the next opcode to execute.

=item I

Indicates the function returns the number of the C<PMC> register that
holds address of the next opcode to be execute.

=back

The C<ITEM> may be one of:

=over 4

=item IV

Indicates the item is an integer

=item IV *

Indicates the item is a pointer to an integer

=item NV

Indicates the item is a float

=item NV *

Indicates the item is a pointer to a float

=item STRING

Indicates the item is a perl string pointer

=item PMC

Indicates the item is a pointer to a PMC

=item INT

Indicates the item is a pointer to an bigint structure

=item NUM

Indicates the item is a pointer to a bignum structure

=item Ix

Indicates the item is an integer register number.

=item Nx

Indicates the item is a float register number.

=item Sx

Indicates the item is a string register number.

=item Px

Indicates the item is a PMC register number.

=back

The function starts with the first open brace, which should generally
be on the first non-empty line.

For example:

     void addI(Ix out, Ix in1, Ix in2)
     {
       INTREG(out) = INTREG(in1) + INTREG(in2);
     }

is a simple opcode function that corresponds to the C<addI>
opcode.

=head1 TODO

=over 4

=item write opcode_process.pl

=back

=head1 REFERENCES

Oploop PDD, PDD 4 (Internal types)

=head1 FOOTNOTES


					Dan

--------------------------------------"it's like this"-------------------
Dan Sugalski                          even samurai
dan@sidhe.org                         have teddy bears and even
                                      teddy bears get drunk




nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About