Used to define opcodes.
Format
$OPDEF MNEMONIC, FORMAT, ENCODING [,DEFAULTS]
1 – Parameters
MNEMONIC
MNEMONIC is the mnemonic name by which the instruction is called.
You may optionally specify a qualifier list separated from the
name by a slash (/). A qualifier list is a sequence of one or
more letters or digits with no intervening spaces or separators.
FORMAT
FORMAT is one of the following arguments:
Format Description
MEMORY Standard memory format instructions.
MEMORY_FUNCTION Memory format instructions with a
function code.
JUMP Memory format instructions formatted
like jump instructions.
BRANCH Standard branch format instructions.
OPERATE Standard operate instructions.
FLOATING_OPERATE Standard floating-point operate
instructions.
PAL Standard PALcode instructions.
<CUSTOM=operand_type_list> Custom format.
With the CUSTOM format, you may optionally specify a list of the
types of operands the instruction is to accept. If you specify
a list of operand types, you must enclose the entire FORMAT
argument within angle brackets, and you must specify the operand
types in the order they are to be used with the instruction.
$OPDEF supports the following operand types:
IREGISTER Integer register, such as R5 or SP.
FREGISTER Floating-point register, such as F7.
LITERAL Integer literal, such as #123 or 32767.
LIT_IREG Integer literal, such as #123 or 32767, or integer
register, such as R5 or SP.
INDIRECT Indirect integer register notation such as (R7).
DISPLACEMENT Indirect integer register notation with an integer
literal displacement, such as FOO(R12).
BRANCH_OFFSET Label or address expression, such as L1.
For example:
FORMAT=<CUSTOM=IREGISTER,DISPLACEMENT>
The following example shows the definition of the ADDQ
instruction, which takes either an integer register or literal
as its second argument:
$OPDEF ADDQ, -
FORMAT=<CUSTOM=IREGISTER,LIT_IREG,IREGISTER>, -
ENCODING=<26:31=^x10, -
21:25=%OP1, -
12:20=%OP2, -
5:11=^x20, -
0:4=%OP3>
For a detailed description of instruction formats, see the Alpha
Architecture Reference Manual.
ENCODING
ENCODING is the numeric encoding for the instruction opcode.
The default radix is decimal, as is the case for all assembler
constants. Prefix the value with ^X for hexadecimal. Certain
instruction formats allow multipart encoding:
Format Encoding Description
MEMORY_FUNCTION Specify the base opcode, followed by a dot,
followed by the function code. For example:
ENCODING=^X10.F000
JUMP Specify the base opcode, optionally followed
by a dot, and the hardware-hint bits
optionally followed by a dot and the software-
hint bits. For example:
ENCODING=^X1A.2.1
OPERATE Specify the base opcode, followed by a dot,
followed by the function code. For example:
ENCODING=^X12.3C
FLOATING_OPERATE Specify the base opcode, followed by a dot and
the function code. For example:
ENCODING=^X17.02B
PAL Specify the base opcode, optionally followed
by a dot and the function code. Omit the
function code for a generic PAL instruction
that accepts an integer-expression argument.
For example:
ENCODING=^X0.0080
ENCODING=^X19
CUSTOM Specify a comma-separated list of bit ranges
and values to place in those bit ranges. For
example:
ENCODING = < 26:31=^X14, 21:25=%OP1, -
16:20=%OP2.REG, 0:15=%OP2.DISP >
For CUSTOM format instructions, specify the ENCODING argument as
a comma-separated list of bit ranges and values to place in those
bit ranges. Enclose the list within angle brackets.
Specify a bit range as start:end where start and end are integer
constant expressions. For a 1-bit bit range, start and end are
equal. Bit positions range from 0 to 31. Place an equal sign (=)
after the bit-range specifier followed by the value you wish to
put in the bit range. You can place either a constant integer
expression or an operand into the bit range. Start and end
expressions and integer constant expressions must not reference
any external symbols or symbols not yet defined in the assembly.
$OPDEF evaluates these expressions at the time that it defines
the instruction as opposed to when the defined instruction is
referenced.
Operand names are of the form %OPn, where n is the ordinal number
of the operand as specified from left to right with the FORMAT
argument.
For the IREGISTER, FREGISTER, and INDIRECT operands, $OPDEF
places the 5-bit register number into the bit positions you
specify.
For LITERAL operands, $OPDEF places the literal value of the
operand into the bit positions you specify. Operand literal
values can be up to 32 bits long. The most significant bits
are truncated if the size of the operand literal value exceeds
the bit range you specify. Forward and external references are
allowed.
For LIT_IREG operands, $OPDEF places either a literal value or
a 5-bit register number into the bit positions you specify. If a
literal, the low bit is 1, and the literal value is placed in the
upper bits. If an integer register, the low four bits are 0, and
the high five bits contain the register number.
For DISPLACEMENT operands, $OPDEF defines two parts: a 5-bit
register number and a displacement value that can be up to 32
bits long. The most significant bits are truncated from the
displacement value if the size of the displacement value exceeds
the bit range you specify. You can reference the register number
by placing .REG after the operand name. For example: %OP2.REG.
Similarly, you can reference the displacement value by placing
.DISP after the operand name. For example: %OP2.DISP. Forward
references are allowed. Relocatable expressions are not allowed.
For BRANCH_OFFSET operands, $OPDEF stores the signed longword
offset between the next instruction and the specified address in
the bit positions you specify. The address expression specified
for a BRANCH_OFFSET operand can be a backward or forward
reference to a location within the current psect. It cannot
be an external address or an address in a different psect. The
resulting offset can be up to 32 bits in size. If the size of the
offset exceeds the bit range you specify, the most significant
bits are truncated.
$OPDEF fills any bit positions you leave unspecified with zeros.
DEFAULTS
DEFAULTS is an optional list of operand defaults of the form
<%OPn=value, ...>, where n is the number of the operand to which
the value is to apply as a default. Operand defaults may be
specified in any order. If you specify a default for one or more
operands, you need not specify a default for all operands.
The following example specifies a default of R31 for the first
instruction argument:
$OPDEF RET, FORMAT=<CUSTOM=IREGISTER,INDIRECT>, -
ENCODING=<26:31=^x1A, -
21:25=%OP1, -
16:20:%OP2, -
14;14=^x2,0:13=0>, -
DEFAULTS=<%OP1=R31>
2 – Description
You can use the $OPDEF macro to define your own opcodes.
$OPDEF defines a macro using an unqualified version of the
mnemonic name you specify. When this macro is invoked with the
instruction qualifiers you specify when you define it with $OPDEF
(if any), it expands to the instruction representation you have
defined with $OPDEF. You can specify the qualifiers in any order
as long as the combination of qualifiers you use is the same.
Other uses of the mnemonic name remain valid provided you do not
use the mnemonic name as a macro name in a .MACRO directive. For
instance, you can use the same mnemonic name in the opcode field
with different or no qualifiers. If the qualifiers (or absence
thereof) do not match those specified in your $OPDEF instruction
definition, the macro defined by $OPDEF processes as though you
had not defined an instruction by that mnemonic name. To do
so, it expands to a single statement. This expansion statement
is identical to the mnemonic-name macro invocation statement,
except it is processed in a context that prevents the mnemonic-
name macro from expanding recursively. Instead, the statement
is processed as a normal, MACRO-64 instruction statement. In
this case, you may notice references to the mnemonic-name macro
expansion in a MACAUXMSG diagnostic message if the instruction
statement contains errors.
For instance, if you define a STQ/P instruction using $OPDEF,
you can still use the STQ instruction without the /P qualifier.
If you do, and your STQ instruction statement contains an error,
the assembler generates a MACAUXMSG message indicating that the
error occurred during the expansion of macro STQ. Aside from
the fact that the STQ instruction is processed in the context of
the expansion of the STQ macro, $OPDEF's definition of the STQ/P
instruction has no effect on your use of the STQ instruction.
3 – Example
$OPDEF MNEMONIC=BANG, FORMAT=PAL, -
ENCODING=^X0.0099