Make compiler output actually compatible with assembler

[app-scheme79asm.git] / lib / App / Scheme79asm.pm

package App::Scheme79asm;

use 5.014000;
use strict;
use warnings;

use Data::Dumper qw/Dumper/;
use Data::SExpression qw/consp scalarp/;
use Scalar::Util qw/looks_like_number/;

our $VERSION = '0.004';

our %TYPES = (
	LIST => 0,
	SYMBOL => 1,
	NUMBER => 1,
	VAR => 2,
	VARIABLE => 2,
	CLOSURE => 3,
	PROC => 4,
	PROCEDURE => 4,
	IF => 5,
	COND => 5,
	CONDITIONAL => 5,
	CALL => 6,
	QUOTE => 7,
	QUOTED => 7,

	MORE => 0,
	CAR => 1,
	CDR => 2,
	CONS => 3,
	ATOM => 4,
	PROGN => 5,
	'REVERSE-LIST' => 6,
	FUNCALL => 7,
);

*consp = *Data::SExpression::consp;
*scalarp = *Data::SExpression::scalarp;

sub process {
	my ($self, $sexp, $location) = @_;
	die 'Toplevel is not a list: ', Dumper($sexp), "\n" unless ref $sexp eq 'ARRAY';
	my ($type, @addrs) = @$sexp;
	my $addr;

	die 'Type of toplevel is not atom: '. Dumper($type), "\n" unless scalarp($type);

	if (@addrs > 1) {
		$addr = $self->{freeptr} + 1;
		$self->{freeptr} += @addrs;
		$self->process($addrs[$_], $addr + $_) for 0 .. $#addrs;
	} else {
		$addr = $addrs[0];
	}

	$addr = $self->process($addr) if ref $addr eq 'ARRAY';
	die 'Addr of toplevel is not atom: ', Dumper($addr), "\n" unless scalarp($addr);

	my ($comment_type, $comment_addr) = ($type, $addr);

	unless (looks_like_number $addr) { # is symbol
		unless (exists $self->{symbols}{$addr}) {
			$self->{symbols}{$addr} = $self->{nsymbols};
			$self->{nsymbols}++;
		}
		$addr = $self->{symbols}{$addr}
	}

	die 'Computed addr is not a number: ', Dumper($addr), "\n" unless looks_like_number $addr;

	if (!looks_like_number $type) {
		die "No such type: $type\n" unless exists $TYPES{$type};
		$type = $TYPES{$type};
	}

	$addr += (1 << $self->{addr_bits}) if $addr < 0;
	die "Type too large: $type\n" unless $type < (1 << $self->{type_bits});
	die "Addr too large: $addr\n" unless $addr < (1 << $self->{addr_bits});
	my $result = ($type << $self->{addr_bits}) + $addr;
	unless ($location) {
		$self->{freeptr}++;
		$location = $self->{freeptr}
	}
	$self->{memory}[$location] = $result;
	$self->{comment}[$location] = "$comment_type $comment_addr";
	$location
}

sub parse {
	my ($self, $string) = @_;
	my $ds = Data::SExpression->new({symbol_case => 'up', use_symbol_class => 1, fold_lists => 1});

	my $sexp;
	while () {
		last if $string =~ /^\s*$/;
		($sexp, $string) = $ds->read($string);
		$self->process($sexp)
	}
}

sub finish {
	my ($self) = @_;
	$self->{memory}[5] = $self->{memory}[$self->{freeptr}];
	$self->{comment}[5] = $self->{comment}[$self->{freeptr}];
	$self->{memory}[4] = $self->{freeptr};
	delete $self->{memory}[$self->{freeptr}]
}

sub new {
	my ($class, %args) = @_;
	$args{type_bits} //= 3;
	$args{addr_bits} //= 8;
	$args{freeptr} //= 6;
	$args{memory} //= [0, 0, (1<<$args{addr_bits}), (1<<$args{addr_bits}), 0, 0, 0];
	$args{symbols}{T} = 2;
	$args{nsymbols} = 3;
	$args{comment} = ['(cdr part of NIL)', '(car part of NIL)', '(cdr part of T)', '(car part of T)', '(free storage pointer)', '', '(result of computation)'];
	bless \%args, $class
}

sub print_binary16 {
	my ($self, $fh) = @_;
	$fh //= \*STDOUT;

	die "addr_bits + type_bits >= 16\n"if $self->{addr_bits} + $self->{type_bits} > 16;

	my $length = @{$self->{memory}};
	print $fh pack('n', $length);
	for (@{$self->{memory}}) {
		print $fh pack('n', $_)
	}
}

sub print_verilog {
	my ($self, $fh) = @_;
	$fh //= \*STDOUT;

	my $bits = $self->{type_bits} + $self->{addr_bits};
	my $index_length = length $#{$self->{memory}};
	my $index_format = '%' . $index_length . 'd';
	for my $index (0 .. $#{$self->{memory}}) {
		my $val = $self->{memory}[$index];
		my $comment = $self->{comment}[$index];
		if ($index == 4) {
			$val = "${bits}'d$val"
		} else {
			$val = $val ? sprintf "%d'b%0${bits}b", $bits, $val : '0';
		}
		my $spaces = ' ' x ($bits + 5 - (length $val));
		$index = sprintf $index_format, $index;
		say $fh "mem[$index] <= $val;$spaces // $comment"
	}

}
sub parse_and_print_binary16 {
	my ($self, $string, $fh) = @_;
	$self->parse($string);
	$self->finish;
	$self->print_binary16($fh);
}

sub parse_and_print_verilog {
	my ($self, $string, $fh) = @_;
	$self->parse($string);
	$self->finish;
	$self->print_verilog($fh);
}

1;
__END__

=encoding utf-8

=head1 NAME

App::Scheme79asm - assemble sexp to Verilog ROM for SIMPLE processor

=head1 SYNOPSIS

  use App::Scheme79asm;
  my $asm = App::Scheme79asm->new(type_bits => 3, addr_bits => 5);
  $asm->parse_and_print_verilog('(number 70)');

=head1 DESCRIPTION

SIMPLE is a LISP processor defined in the 1979
B<Design of LISP-Based Processors> paper by Steele and Sussman.

The SIMPLE processor expects input in a particular tagged-pointer
format. This module takes a string containing a sequence of
S-expressions. Each S-expression is a list of one of three types:

C<(tag value)>, for example C<(symbol nil)>, represents a value to be
put in memory (for example a number, or a symbol, or a variable
reference).

C<(tag list)>, where C<list> is of one of these three types,
represents a tagged pointer. In this case, C<list> is (recursively)
laid out in memory as per these rules, and a pointer to that location
(and tagged C<tag>) is put somewhere in memory.

C<(tag list1 list2)>, where C<list1> and C<list2> are of one of these
three types (not necessarily the same type). In this case, C<list1>
and C<list2> are (recursively) laid out in memory such that C<list1>
is at position X and C<list2> is at position X+1, and a pointer of
type tag and value X is put somewhere in memory.

After this process the very last pointer placed in memory is moved to
the special location 5 (which is where SIMPLE expects to find the
expression to be evaluated).

In normal use a single S-expression will be supplied, representing an
entire program.

The C<tag> is either a number, a type, or a primitive.
The available types are:

=over

=item LIST

=item SYMBOL (syn. NUMBER)

=item VAR (syn. VARIABLE)

=item CLOSURE

=item PROC (syn. PROCEDURE)

=item IF (syn. COND, CONDITIONAL)

=item CALL

=item QUOTE (syn. QUOTED)

=back

The available primitives are:

=over

=item MORE

=item CAR

=item CDR

=item CONS

=item ATOM

=item PROGN

=item REVERSE-LIST

=item FUNCALL

=back

The following methods are available:

=over

=item App::Scheme79asm->B<new>([key => value, key => value, ...])

Create a new assembler object. Takes a list of keys and values, here
are the possible keys:

=over

=item type_bits

=item address_bits

A word is made of a type and an address, with the type occupying the
most significant C<type_bits> (default 3) bits, and the address
occupying the least significant C<address_bits> (default 8) bits.
Therefore the word size is C<type_bits + address_bits> (default 13).

=item freeptr

A pointer to the last used byte in memory (default 6). The program
will be laid out starting with location C<freeptr + 1>.

=item memory

The initial contents of the memory. Note that locations 4, 5, 6 will
be overwritten, as will every location larger than the value of
C<freeptr>.

=item comment

The initial comments for memory entries. C<< $comment->[$i] >> is the
comment for C<< $memory->[$i] >>.

=item symbols

The initial symbol map, as a hashref from symbol name to the index of
that symbol. Defaults to C<< {T => 2} >>.

=item nsymbols

The number to give to the "next" symbol (default 3, because T is
defined to be 2).

=back

=item $asm->B<parse>(I<$string>)

Parse a sequence of S-expressions and lay it out in memory.
Can be called multiple times to lay out multiple sequences of
S-expressions one after another.

=item $asm->B<process>(I<$sexp>)

Given an already-parsed sexp (meaning a
L<Data::SExpression> object), lay it out in memory.
Can be called multiple times to lay out multiple sequences of
S-expressions one after another.

=item $asm->B<finish>

Move the last pointer to position 5, and put the free pointer at
position 4. After all sequences of S-expressions have been given to
B<parse>, this method should be called.

=item $asm->B<print_binary16>([I<$fh>])

Print the length of the memory (as a big-endian 16-bit value),
followed by the memory contents as a sequence of big-endian 16-bit
values to the given filehandle (default STDOUT). Dies if
C<addr_bits + type_bits> is more than 16.

Big-endian 16-bit values can be decoded with C<unpack 'n', $value>.

=item $asm->B<print_verilog>([I<$fh>])

Print a block of Verilog code assigning the memory contents to an
array named C<mem> to the given filehandle (default STDOUT).

=item $asm->B<parse_and_print_binary16>(I<$string>[, I<$fh>])

Convenience method that calls B<parse>($string), B<finish>, and then
B<print_binary16>($fh).

=item $asm->B<parse_and_print_verilog>(I<$string>[, I<$fh>])

Convenience method that calls B<parse>($string), B<finish>, and then
B<print_verilog>($fh).

=back

=head1 SEE ALSO

L<http://repository.readscheme.org/ftp/papers/ai-lab-pubs/AIM-514.pdf>

=head1 AUTHOR

Marius Gavrilescu, E<lt>marius@ieval.roE<gt>

=head1 COPYRIGHT AND LICENSE

Copyright (C) 2018 by Marius Gavrilescu

This library is free software; you can redistribute it and/or modify
it under the same terms as Perl itself, either Perl version 5.24.3 or,
at your option, any later version of Perl 5 you may have available.


=cut