awesome-cpus/PDP-11/The_PDP-11_FAQ.htm

370 lines
18 KiB
HTML

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<!-- saved from url=(0051)http://www.village.org/pdp11/faq.pages/PDPinst.html -->
<HTML><HEAD><TITLE>The PDP-11 FAQ</TITLE>
<META http-equiv=Content-Type content="text/html; charset=windows-1252">
<META content="MSHTML 6.00.2900.3059" name=GENERATOR></HEAD>
<BODY>
<H1><A name=PDPinst>What is the PDP-11 instruction set?</A></H1>
<P>The instruction set of the PDP-11 was designed towards a clean, general,
symmetric instruction set. It can be used as a register-based, stack-based, or
memory-based machine, depending on the programmer's preferences. Interrupt
responsiveness is also important, supported with multiple interrupt levels for
real-time computing as well as allowing for a separate interrupt handler for
each device that generates interrupts. </P>
<P>Word length is 16 bits with the leftmost, most significant bit (MSB) being
bit 15. There are eight general registers of 16 bits each. Register 7 is the
program counter (PC) and, by convention, Register 6 is the stack pointer (SP).
There is also a Processor Status Register/Word (PSW) which indicates the 4
condition code bits (N, Z, V, C), the Trace Trap bit, processor interrupt
priority, and 4 bits for current and previous operating modes. Addressing on the
-11 is linear from memory address 0 through 177777. Memory management allows
access to physical memory with addresses of up to 22 bits (17777777). All I/O
devices, registers etc are addressed as if they were part of memory. These live
in the 4KW of reserved memory space at the top of the addressing range.
Additionally, on most implementations of the PDP-11 architecture, the
processor's registers are memory-mapped to the range 17777700-17777717 (there
are many control registers beyond just the general registers, the specifics vary
between implementations). Thus Register 2 (R2) has an address of 17777702. All
word memory addresses are even, except for registers. In byte operations, an
even address specifies the least-significant byte and an odd address specifies
the most-significant byte. Specifying an odd byte in a word operation will
return an odd address trap. Memory addresses from 0 to 400 octal are reserved
for various exception traps such as timeouts, reserved instructions, parity,
etc., and device interrupts. </P>
<P>Addressing for the Single Operand, Double Operand and Jump instructions is
achieved via six bits: </P><PRE> _ _ _ _ _ _
|x|x|x|_|_|_|
|Mode |Reg |
</PRE>
<P>where the modes are as follows: (Reg = Register, Def = Deferred) </P><PRE> Mode 0 Reg Direct addressing of the register
Mode 1 Reg Def Contents of Reg is the address
Mode 2 AutoIncr Contents of Reg is the address, then Reg incremented
Mode 3 AutoIncrDef Content of Reg is addr of addr, then Reg Incremented
Mode 4 AutoDecr Reg is decremented then contents is address
Mode 5 AutoDecrDef Reg is decremented then contents is addr of addr
Mode 6 Index Contents of Reg + Following word is address
Mode 7 IndexDef Contents of Reg + Following word is addr of addr
</PRE>
<P>Note that the right-most bit of the mode is an indirection bit.
<P>Although not special cases, when dealing with R7 (aka the PC), some of these
operations are called different things: </P><PRE> _ _ _ _ _ _
|x|x|x|1|1|1|
|Mode | R7 |
Mode 2 Immediate Operand follows the instruction
Mode 3 Absolute Address of Operand follows the instruction
Mode 6 Relative Instr address+4+Next word is Address
Mode 7 RelativeDef Instr address+4+Next word is Address of address
</PRE>
<P>Mainstream instructions are broken into Single operand and Double operand
instructions, which in turn can be word or byte instructions. </P>
<H2>Double Operand Instructions</H2><PRE> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|b|i|i|i|s|s|s|s|s|s|d|d|d|d|d|d|
| | | : | : |
| | Op | Source | Dest |
</PRE>
<P>Bit 15, b, generally selects between word-sized (b=0) and byte-sized (b=1)
operands. In the table below, the mnemonics and names are given in the order
b=0/b=1. </P>
<P>The double operand instructions are: </P>
<DL>
<DT>b 000 ssssss dddddd
<DD>Non-double-operand instructions.
<P></P>
<DT>b 001 ssssss dddddd -- MOV/MOVB <I>Move Word/Byte</I>
<DD>Moves a value from source to destination.
<P></P>
<DT>b 010 ssssss dddddd -- CMP/CMPB <I>Compare Word/Byte</I>
<DD>Compares values by subtracting the destination from the source, setting
the condition codes, and then discarding the result of the subtraction.
<P></P>
<DT>b 011 ssssss dddddd -- BIT/BITB <I>Bit Test Word/Byte</I>
<DD>Performs a bit-wise AND of the source and the destination, sets the
condition codes, and then discards the result of the AND.
<P></P>
<DT>b 100 ssssss dddddd -- BIC/BICB <I>Bit Clear Word/Byte</I>
<DD>For each bit set in the source, that bit is cleared in the destination.
This is accomplished by taking the ones-complement of the source and ANDing it
with the destination. The result of the AND is stored in the destination.
<P></P>
<DT>b 101 ssssss dddddd -- BIS/BISB <I>Bit Set Word/Byte</I>
<DD>For each bit set in the source, that bit is set in the destination. This
is accomplished by ORing the source and destination, and storing the result in
the destination.
<P></P>
<DT>b 110 ssssss dddddd -- ADD/SUB <I>Add/Subtract Word</I>
<DD>Adds the source and destination, storing the results in the destination.
<P>Subtracts the source from the destination, storing the results in the
destination.
<P>Note that this is a special case for b=1, in that it does not indicate that
byte-wide operands are used.
<P></P>
<DT>b 111 xxxxxx xxxxxx
<DD>Arithmetic functions not supported by all implementations of the PDP-11
architecture. </DD></DL>
<H2>Single Operand Instructions</H2><PRE> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|b|0|0|0|i|i|i|i|i|i|d|d|d|d|d|d|
| | | : | : |
| | |Instruction| Dest |
</PRE>
<P>Bit 15, b, generally selects between word-sized (b=0) and byte-sized (b=1)
operands. In the table below, the mnemonics and names are given in the order
b=0/b=1. Unless otherwise stated, the operand is read for the data to operate
on, and the result is then written over that data. </P>
<P>The single operand instructions are: </P>
<DL>
<DT>b 000 000 011 dddddd -- SWAB/BPL <I>Swap Bytes/Branch Plus</I>
<DD>Swap bytes exchanges the two bytes found in the destination, writing the
result back to it.
<P>The branch (b=1) is described in the section on branches, below.
<P>Note that SWAB is actually a bit pattern from the range reserved for
branches. This particular pattern is otherwise unused, as it would be a
modification of BR, Branch Always, which has no obvious semantics.
<P></P>
<DT>b 000 101 000 dddddd -- CLR/CLRB <I>Clear Word/Byte</I>
<DD>Sets all the bits in destination to zero.
<P></P>
<DT>b 000 101 001 dddddd -- COM/COMB <I>Complement Word/Byte</I>
<DD>Calculates the ones-complement of the operand, and stores it. The
ones-complement is formed by inverting each bit (0-&gt;1, 1-&gt;0)
independently.
<P></P>
<DT>b 000 010 010 dddddd -- INC/INCB <I>Increment Word/Byte</I>
<DD>Adds one to the destination.
<P></P>
<DT>b 000 101 011 dddddd -- DEC/DECB <I>Decrement Word/Byte</I>
<DD>Subtracts one from the destination.
<P></P>
<DT>b 000 101 100 dddddd -- NEG/NEGB <I>Negate Word/Byte</I>
<DD>Calculates the twos-complement of the operand, and stores it. The
twos-complement is formed by adding one to the ones-complement. The effect is
the same as subtracting the operand from zero.
<P></P>
<DT>b 000 101 101 dddddd -- ADC/ADCB <I>Add Carry Word/Byte</I>
<DD>Adds the current value of the carry flag to the destination. This is
useful for implementing arithmetic subroutines with more than word-sized
operands.
<P></P>
<DT>b 000 101 110 dddddd -- SBC/SBCB <I>Subtract Carry Word/Byte</I>
<DD>Subtracts the current value of the carry flag from the destination. This
is useful for implementing arithmetic subroutines with more than word-sized
operands.
<P></P>
<DT>b 000 101 111 dddddd -- TST/TSTB <I>Test Word/Byte</I>
<DD>Sets the N (negative) and Z (zero) condition codes based on the value of
the operand.
<P></P>
<DT>b 000 110 000 dddddd -- ROR/RORB <I>Rotate Right Word/Byte</I>
<DD>Rotates the bits of the operand one position to the right. The right-most
bit is placed in the carry flag, and the carry flag is copied to the left-most
bit (bit 15) of the operand.
<P></P>
<DT>b 000 110 001 dddddd -- ROL/ROLB <I>Rotate Left Word/Byte</I>
<DD>Rotates the bits of the operand one position to the left. The left-most
bit is placed in the carry flag, and the carry flag is copied to the
right-most bit (bit 0) of the operand.
<P></P>
<DT>b 000 110 010 dddddd -- ASR/ASRB <I>Arithmetic Shift Right Word/Byte</I>
<DD>Shifts the bits of the operand one position to the right. The left-most
bit is duplicated. The effect is to perform a signed division by 2.
<P></P>
<DT>b 000 110 011 dddddd -- ASL/ASLB <I>Arithmetic Shift Left Word/Byte</I>
<DD>Shifts the bits of the operand one position to the left. The right-most
bit is set to zero. The effect is to perform a signed multiplication by 2.
<P></P>
<DT>b 000 110 100 dddddd -- MARK/MTPS <I>Mark/Move To Processor Status</I>
<DD>Mark is used as part of one of the subroutine call/ return sequences. The
operand is the number of parameters.
<P>MTPS is only on LSI-11s, and is used to move a byte to the processor status
word. This is needed because the LSI-11 does not support accessing registers
via memory addresses.
<P></P>
<DT>b 000 110 101 dddddd -- MFPI/MFPD <I>Move From Prev. Instruction/Data</I>
<DD>Pushes a word onto the current R6 stack from the operand address in the
previous address space, as indicated in the PSW. On PDP-11s that do not
support separate instruction and data spaces, MFPD is treated the same as
MFPI.
<P></P>
<DT>b 000 110 110 dddddd -- MTPI/MTPD <I>Move To Previous Instruction/Data</I>
<DD>Pops a word from the current stack as indicated in the PSW to the operand
address in the previous address space, as indicated in the PSW. On PDP-11s
that do not support separate instruction and data spaces, MTPD is treated the
same as MTPI.
<P></P>
<DT>b 000 110 111 dddddd -- SXT/MFPS <I>Sign Extend/Move From Processor
Status</I>
<DD>SXT sets the destination to zero if the N (negative) flag is clear, or to
all ones if N is set. This is useful for implementing arithmetic subroutines
with more than word-sized operands.
<P>MFPS copies the processor status byte to the indicated register. This only
exists on LSI-11s, and is needed there because those systems don't support
accessing registers via memory addresses. </P></DD></DL>
<H2>Branches</H2><PRE> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|b|0|0|0|b|b|b|b|d|d|d|d|d|d|d|d|
| Branch Code | Destination |
</PRE>
<P>The destination of a branch is +127 to -128 words from the word following the
branch instruction itself. This seems slightly odd, until you realize the
sequence of events: the branch instruction is read from memory and the PC
incremented. If the branch is to be taken, the offset is then added to the
current value of the PC. Since the PC has already been incremented, the offset
is thus relative to the following word. Note that all branch instructions are
one word long. </P>
<P>The various branches test the values of specific condition codes, and if the
tests succeed, the branch is taken. The condition codes are N (negative), Z
(zero), C (carry), and V (overflow). In the table below, the branch tests are
shown as boolean expressions. `x' stands for exclusive-OR. `v' stands for
inclusive-OR. </P>
<DL compact>
<DT>0 000 000 1dd dddddd
<DD>BR: Branch Always
<DT>0 000 001 0dd dddddd
<DD>BNE: Branch if Not Equal (Z==0)
<DT>0 000 001 1dd dddddd
<DD>BEQ: Branch if EQual (Z==1)
<DT>0 000 010 0dd dddddd
<DD>BGE: Branch if Greater or Equal (NxV == 0)
<DT>0 000 010 1dd dddddd
<DD>BLT: Branch if Less Than (NxV == 1)
<DT>0 000 011 0dd dddddd
<DD>BGT: Branch if Greater Than (Zv(NxV) == 0)
<DT>0 000 011 1dd dddddd
<DD>BLE: Branch if Less or Equal (Zv(NxV) == 1)
<DT>1 000 000 0dd dddddd
<DD>BPL: Branch if PLus (N == 0)
<DT>1 000 000 1dd dddddd
<DD>BMI: Branch if MInus (N == 1)
<DT>1 000 001 0dd dddddd
<DD>BHI: Branch if HIgher (C==0 and Z==0)
<DT>1 000 001 1dd dddddd
<DD>BLOS: Branch if Lower Or Same (CvZ == 1)
<DT>1 000 010 0dd dddddd
<DD>BVC: Branch if oVerflow Clear (V == 0)
<DT>1 000 010 1dd dddddd
<DD>BVS: Branch if oVerflow set (V == 1)
<DT>1 000 011 0dd dddddd
<DD>BCC: Branch if Carry Clear (C == 0) <BR><I>also known as</I><BR>BHIS:
Branch if Higher Or Same
<DT>1 000 011 1dd dddddd
<DD>BCS: Branch if Carry Set (C == 1) <BR><I>also known as</I><BR>BLO: Branch
if Lower </DD></DL>
<H2>Condition Code Operations</H2><PRE> _ _ _ _ _ _ _ _ _ _:_ _ _:_ _ _
|0|0|0|0|0|0|0|0|1|0|1|s|N|Z|V|C|
| O p c o d e | | Mask |
</PRE>
<P>General opcode 000240x. Set/Clear corresponding bits depending on sense of
bit 04 (set=1, clear=0). Codes 240 and 260 set/clear no bits and are, thus, used
as NOP. Although specific mnemonic are provided for each flag and all flags, any
combination may actually be set or cleared at a time. </P>
<P>General mnemonics are: </P>
<DL compact>
<DT>CLx
<DD>Clear x, where x is N, Z, V, or C
<DT>SEx
<DD>Set x, where x is N, Z, V, or C
<DT>CCC
<DD>Clear all condition codes
<DT>SCC
<DD>Set all condition codes </DD></DL><BR>
<DL compact>
<DT>0 000 000 010 1s0 000
<DD>NOP/NOP: No Operation
<DT>0 000 000 010 1s0 001
<DD>SEC/CLC: Set/Clear Carry
<DT>0 000 000 010 1s0 010
<DD>SEV/CLV: Set/Clear Overflow
<DT>0 000 000 010 1s0 100
<DD>SEZ/CLZ: Set/Clear Zero
<DT>0 000 000 010 1s1 000
<DD>SEN/CLN: Set/Clear Negative
<DT>0 000 000 010 1s1 111
<DD>SCC/CCC: Set/Clear All Condition Codes </DD></DL>
<H2>Other, Miscellaneous</H2><PRE> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|0|0|0|0|1|0|0|s|s|s|d|d|d|d|d|d|
| Opcode |Stack|Destination|
</PRE>
<P>
<DL compact>
<DT>0 000 100 sss dddddd -- JSR <I>Jump to Subroutine</I>
<DD>The actual sequence of steps taken is: <PRE> MOV &lt;source&gt;,-(R6)
MOV PC,&lt;source&gt;
JMP &lt;destination&gt;
</PRE><BR>
<P>Thus, it loads the calling address into the specified source register
(after saving the original contents). It then jumps to the destination. The
fun part is (as usual with the PDP-11) that the PC is a general register, and
the description above is the result when the PC is used as the source.
</P></DD></DL><PRE> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|0|0|0|0|0|0|0|0|1|0|0|0|0|s|s|s|
| Opcode |Stack|
</PRE>
<P>
<DL compact>
<DT>0 000 000 010 000 sss -- RTS <I>ReTurn from Subroutine</I>
<DD>Undoes the effects of a JSR. For predictable results, it is suggested that
the same register should be used as was named in the corresponding JSR
instruction.
<P>The actual operations involved are: <PRE> MOV &lt;source&gt;,PC
MOV (R6)+,&lt;source&gt;
</PRE><BR>
<P>This is the reverse of JSR. Obviously, the finesse here too is that you can
use the PC, to get what people normally consider a CALL/RETURN function.
<P>Why is it done like this then? Well, consider this example: <PRE> ...
JSR R0,FOO
.WORD A
.WORD B
MOV R1,C
...
FOO: MOV @(R0)+,R1
ADD @(R0)+,R1
RTS R0
</PRE><BR>
<P>This type of parameter passing is used extensively in the PDP-8 and
PDP-10), for example. Also, the FORTRAN runtime system on the PDP-11 do it
this way. (It is fairly easy to write a compiler who generates such a calling
sequence, and then have a library of functions which expect this calling
convention.) </P></DD></DL><PRE> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|0|0|0|0|0|0|0|0|0|1|d|d|d|d|d|d|
| Opcode |Destination|
</PRE>
<P>
<DL compact>
<DT>0 000 000 001 ddd ddd -- JMP <I>JuMP</I>
<DD>Loads the destination address into the PC, thus effecting an unconditional
jump. Note that a trap will occur on some systems if an odd address is
specified. On others, the destination is silently rounded down to the
next-lower even address (i.e., the right-most bit is ignored). </DD></DL><PRE> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|0|0|0|0|0|0|0|0|0|0|0|0|0|i|i|i|
| | | | | | Op |
</PRE>
<P>
<DL compact>
<DD>0 000 000 000 000 000 -- HALT <I>Halts the machine</I>
<DD>Ceases I/O, and gives control to the console. Operator intervention is
required to continue or restart the system.
<P></P>
<DD>0 000 000 000 000 001 -- WAIT <I>WAIT for interrupt</I>
<DD>0 000 000 000 000 010 -- RTI <I>ReTurn from Interrupt</I>
<DD>0 000 000 000 000 100 -- BPT <I>BreakPoint Trap</I>
<DD>0 000 000 000 000 101 -- RESET <I>Initializes the system</I> </DD></DL>
<P>The following opcode ranges are all unused (using three bits per digit):
<P>
<DL compact>
<DD>00 00 07 .. 00 00 77
<DD>00 02 10 .. 00 02 27
<DD>00 70 00 .. 00 77 77
<DD>07 50 40 .. 07 67 77
<DD>10 64 00 .. 10 64 77
<DD>10 67 00 .. 10 77 77 </DD></DL>
<P>Other arithmetic and floating point instructions were added to the basic set
over the years, but those listed above form the core PDP-11 instruction set.
<P>There is a comparison of PDP-11 and 80x86 floating point formats available
at: <A
href="ftp://ftp.dbit.com/pub/pdp11/info/fpp.txt">ftp://ftp.dbit.com/pub/pdp11/info/fpp.txt</A>
<HR>
<A href="http://www.village.org/pdp11/faq.html">Table of Contents</A>
</BODY></HTML>