The ICWS'94 draft (extended)
Opcodes:
DAT terminate process
MOV move from A to B
ADD add A to B, store result in B
SUB subtract A from B, store result in B
MUL multiply A by B, store result in B
DIV divide B by A, store result in B if A <> 0, else terminate
MOD divide B by A, store remainder in B if A <> 0, else terminate
JMP transfer execution to A
JMZ transfer execution to A if B is zero
JMN transfer execution to A if B is non-zero
DJN decrement B, if B is non-zero, transfer execution to A
SPL split off process to A
SLT skip next instruction if A is less than B
CMP same as SEQ
SEQ Skip next instruction if A is equal to B
SNE Skip next instruction if A is not equal to B
NOP No operation
LDP Load P-space cell A into core address B
STP Store A-number into P-space cell B
Pseudo opcodes:
[labels] EQU text replaces [labels] by text
[EQU text] (*) multi-line EQU: text continues
ORG start specifies execution start
END [start] end of assembly (optional execution start)
[count] FOR expression (*) repeat enclosed instructions "expression"
ROF times, counter is incremented starting with 01
PIN number (+) P-space identification number, warriors
with same number share P-space
Modifiers:
.A Instructions read and write A-fields.
.B Instructions read and write B-fields.
.AB Instructions read the A-field of the A-instruction and
the B-field of the B-instruction and write to B-fields.
.BA Instructions read the B-field of the A-instruction and
the A-field of the B-instruction and write to A-fields.
.F Instructions read both A- and B-fields of the the A-
and B-instruction and write to both A- and B-fields (A
to A and B to B).
.X Instructions read both A- and B-fields of the the A-
and B-instruction and write to both A- and B-fields
exchanging fields (A to B and B to A).
.I Instructions read and write entire instructions.
Addressing Modes:
# immediate
$ direct
@ indirect using B-field
< predecrement indirect using B-field
> postincrement indirect using B-field
* indirect using A-field
{ predecrement indirect using A-field
} postincrement indirect using A-field
Directives:
;redcode code follows, preceding text is ignored
;name name of warrior follows
;author name of author follows
;assert (*) expression that must evaluate to true
;trace [off] (*) toggle trace bit for following instructions
;break (*) set trace bit for next instruction
;debug [static|off] (*) enable/disable setting [static] trace bits
Predefined variables:
CORESIZE value of -s parameter (default: 8000)
MAXPROCESSES value of -p parameter (default: 8000)
MAXCYCLES value of -c parameter (default: 80000)
MAXLENGTH value of -l parameter (default: 100)
MINDISTANCE value of -d parameter (default: 100)
ROUNDS value of -r parameter (default: 1)
PSPACESIZE value of -S parameter (default: 1/16th CORESIZE)
CURLINE current line in generated assembly (starts with 0)
VERSION pMARS version ("60" is v0.6.0)
WARRIORS number of warriors specified on command line
Expression Operators:
Arithmetic:
+ addition or unary plus
- subtraction or unary minus
/ division
% modulo (remainder of division)
Comparison (*):
== equality
!= inequality
< less than
> greater than
<= less than or equal
>= greater than or equal
Logical (*):
&& and
|| or
! unary negation
Assignment (*):
= (to register variables a..z)
Comparison and logical operators return 1 for true and 0 for false.
Parentheses can be used to override this precedence order:
1) ! - + (unary)
2) * / %
3) - + (binary)
4) == != < > <= >=
5) &&
6) ||
7) =
Redcode Grammar:
statement_list :: statement statement_list | e ;
statement :: normal_stmt<1> |
equ_stmt |
forrof_stmt |
comment_stmt |
substitution_stmt<2> ;
num :: [0-9] ;
number :: num number | num ;
alpha :: [a-zA-Z] | "_" ;
alphanum :: alpha | num ;
alphanums :: alphanum alphanums | e ;
label :: alpha alphanums ;
label1 :: label label1 | label ;
labels :: label1 "\n" labels | label1 ;
stringization :: stringization"&"label | label ;
this_string :: (^\n)* ;
comment :: ";" this_string "\n" | e ;
equ_stmt :: labels equ_strings ;
equ_string :: "equ" this_string comment "\n" ;
equ_strings :: equ_string "\n" equ_strings | equ_string ;
forrof_stmt :: labels index "for" expression<3> comment "\n"
statement_list
"rof" this_string "\n" ;
index :: label
comment_stmt :: info_comment | debug_comment | ignore ;
ignore :: ";" this_string "\n" ;
info_comment :: ";redcode" this_string "\n" |
";" "name" this_string "\n" |
";" "author" this_string "\n" |
";" "date" this_string "\n" |
";" "version" this_string "\n" |
";" "assert" expression<4> "\n" ;
debug_comment :: ";" debug "\n" | ";" trace "\n" | ";" break "\n" ;
debug :: "debug" | "debug" "off" | "debug" "static" ;
trace :: "trace" | "trace" "off" ;
break :: "break" ;
Note:
1. Normal statements are statements in the following form:
opcode [address mode] operand [, [address mode] operand] [comment]
More details about the grammar are given in the '88 or '94 proposal.
2. Substitution statements are labels that have been declared by EQU.
If a label is declared this way: "IMP_instr equ imp mov imp, imp +
1", the label name IMP_instr can be thought of as a statement.
Therefore, whenf IMP_instr is used after its declaration, it will
be replaced by "imp mov imp, imp + 1". It has effect of declaring
a label 'imp' and inserting the statement 'mov imp, imp + 1'.
3. Valid expression for "FOR" statement is very close to C expression in
which operator '()' has the highest precedence, followed by 'unary +,
unary -', '*, / and %', 'binary + and binary -', '<, <=, >, >=',
'==, !=', '&&', '||', and the lowest '='.
Beside numbers, it can also has labels as its terms. All of its labels
have to be declared before "FOR" statement is invoked.
4. Valid expression for "ASSERT" statement is the same as that for "FOR".
Case sensitivity
Opcode and pseudo-opcode names are case insensitive; labels are case
sensitive.
Label declaration
Labels are declared in three ways:
o Using EQU. When a label name that first appears (has not been declared)
is declared with EQU, all subsequent occurences of that label name
will be replaced by the strings following the EQU.
The following equates label THIS with "num + 1". THIS equ num + 1.
If the string substituting the label contains other labels, those labels
are also replaced by their substituting string. Recursive reference of
labels are flagged as error.
More than one statement can be declared as a label name. To achieve this,
declare a blank label with EQU following the statement that declares
as part of equation of a named EQU label. Thus, the declaration of the
following:
core_clear equ spl 0
equ mov 2, <-1
equ jmp -1
causes three statements are declared as strings of label core_clear.
o As an offset relative to the current normal statement. If some labels
appear just before any '88 or '94 opcodes, they are automatically
declared as constants that are relative to the current statement.
For example:
first spl first 0000 spl 0
imp mov imp, imp + 1 -----> 0001 mov 0, 1
o As an index belonging to FOR statement. Unlike other statements, labels
declared with FOR consist of two ingredients: the last label serving
as FOR index and the remaining labels serving as the same offset pointing
at the first of the FOR statement. This allows the implementation of
base[index] kind.
This declaration:
base
index for 3
mov base, base + index - 1
rof
translates into:
0000 mov 0, 0
0001 mov 0, 1
0002 mov 0, 2
Stringization and declaration inside FOR/ROF:
FOR 5
imp mov imp, imp + 5
ROF
Declaration of the above example causes the assembler to complain for
duplicating declarations. pMARS however offers the stringization feature
to accomplish the same goal. Its syntax is: label"&"label"&"...
The first label can be any valid alphanums and it goes untranslated.
The rest of the labels have to be a FOR index and it is to be substituted
accordingly. Thus:
N FOR 5
imp&N mov imp&N, imp&N + 1
ROF
are expanded into:
imp01 mov imp01, imp01 + 5
imp02 mov imp02, imp02 + 5
imp03 mov imp03, imp03 + 5
imp04 mov imp04, imp04 + 5
imp05 mov imp05, imp05 + 5
It is then correctly compiled.
The following form is also valid:
prime01 equ 2
prime02 equ 3
prime03 equ 5
prime04 equ 7
prime05 equ 11
N FOR 5
dat prime&N
ROF
The conjunction of FOR statements and EQU statements:
If labels that are not stringized are declared inside FOR/ROF statements,
the result is duplicating declaration. Although in the future it might be
allowed, it is well-advised that such labels are to be declared outside of
the FOR/ROF block.
If FOR statements are declared as strings of a label name using EQU such as:
THIS EQU N FOR 3
EQU mov 0, 3
EQU ROF
The expansion of such is feasible providing that both the FOR and ROF are
present in the same label name.
Therefore, the following will not work:
THIS EQU N FOR 3
THAT EQU ROF
THIS
mov 0, 3
THAT
or
TEST EQU N FOR 3
mov 0, 3
ROF
PMARS parsing background
Parsing is done in three steps:
1. Reading from input
2. First pass assembly
3. Second pass assembly
- During reading: info_comment is parsed, comments to be ignored are removed,
the remaining are copied into memory. If the first ';redcode' appears,
all the current contents are erased from memory. Reading is continued
until it encounters the next ';redcode'. Reading always stops when it
encounters a line containing a word END or end-of-file.
- During the first pass, all labels are collected. All expansions and
removals are done in this pass. All labels before opcode are substituted
if they have been declared or declared otherwise. All labels after opcode
are preserved for the second pass. This allows forward declaration
mechanism for labels after opcode.
- During the second pass, no labels are declared and collected. There is
no further statement expansions and removals. All labels that have not
been substituted are substituted. Syntax checking to meet with '88 or '94
requirement is done in this stage.
|
