Oakland CPM Archive/oakcpm.iso/cpm/asmutl/macros.tqp/MACROS.TIP

Text File | 1986-05-06 | 6.6 KB | 203 lines

     Here  is a    tip for    users of 8080 macro assemblers which
I've never seen    in print, but which I've found to be useful.
It  pertains  to the use of the    standard condition codes (Z,
NZ,  C,     NC,  M,  P,  PO,  PE)    as  parameters    of  a  macro
instruction.
     Let's say you've written a    useful macro to    do something
or  other.  As an example most CP/M programmers    are familiar
with,    let's  consider     the  CPM  macro.  It  usually    goes
something like this:

CPM    MACRO    FUNC,VALUE
     IF     NOT NUL VALUE
     LXI     D,VALUE     ;;Load    D&E with value.
     ENDIF
     IF     NOT NUL FUNC
     MVI     C,FUNC         ;;Load    C reg with Function.
     ENDIF
     CALL     BDOS
     ENDM

     The  examples  are    written    for the    Digital    Research MAC
assembler,  but    would probably be similar for others. To use
the macro, one might say:

     CPM     CON,CR     ;Send carriage    return to console

Assuming  that    CON had    been EQUated to    2, the CP/M function
for  console  output,  and  CR    had been EQUated to 0DH, the
macro would expand to:

     LXI     D,0DH
     MVI     C,02H
     CALL     0005H

which    would    send  a     carriage  return  to  the  console.
Similarly,  if    KEY had    been EQUated to    1, the console input
function, the macro call

     CPM     KEY

would expand to:

     MVI     C,01H
     CALL     0005H

which  would  get a keyboard character into register A.    This
macro  is  in fact used    quite frequently by many programmers
who work with CP/M. Note that the IF's prevent unneeded    code
from being assembled where the parameter is not    supplied.
     Now  consider  a  case  where we want to read a console
character only when the    carry flag is set. Perhaps the carry
flag  indicates     an error condition, and we want the program
to pause. The usual way    of doing this is:

     ...         ;Code to set carry on error
     JNC     OK     ;Jump around input routine
     CPM     KEY     ;Get a    character from console.
OK:
     ...         ;Continue with    operation.

     To    save programmer    time, we can modify the    CPM macro to
allow a    condition code to be specified as a third parameter.
Then we    could write lines like:

     CPM     KEY,,C         ;Get char if Carry set
     ...or...
     PUSH     PSW         ;Save output char.
     MOV     E,A         ;Move to E for    output
     CPM     CON         ;Output it
     POP     PSW         ;Get back character
     CPI     CR         ;Was it a carriage return?
     CPM     CON,LF,Z     ;If so, follow    with Line Feed.
     ...

without     having     to  code  the    jump  instructions which are
necessary  to  avoid executing the macro code. We do this by
coding    the  jump instruction into the macro itself. This is
made  easier  by  the  fact  that most macro assemblers    will
allow  us  to  use  the     value    of  an    opcode    as data. For
compatability  with  Intel  standards,    it  should  be coded
within    parentheses  (required    by  some  assemblers) and if
we're  using  MAC,  should  have spaces    around the opcode to
avoid a    little-known glitch in MAC.
     Let's  work  with an even simpler macro to    see how    this
might  operate.     We'll    invent    the JUMPIF macro, which    does
nothing    but cause a conditional    jump.

JUMPIF    MACRO    COND,ADDR
     DB     ( J&COND )
     DW     ADDR
     ENDM

     In    this simple example, coding the    line:
     JUMPIF     NC,EOJ
would expand to:
     DB     ( JNC )
     DW     EOJ
     "Big deal!", you say. Why not just    code:
     JNC     EOJ
     Well,  we want to work the    condition code into a larger
macro,    and  generate  a  jump around the inline code if the
condition  is  false.  Ah, but there's the rub.    In order for
the macro to work properly, the    jump instruction has to    jump
if  the    condition given    is NOT true, and fall through to the
inline    code  if it IS true. To    illustrate, let's invent the
opposite  of  the  JUMPIF  macro, the JMPUNLES (Jump Unless)
macro. We could    do something like this:

JMPUNLES MACRO    COND,ADDR
     J&COND     LABL1     ;Jump around the next instr.
     JMP     ADDR     ;Do the real jump.
LABL1:
     ENDM

But  this requires the use of two jump instructions where we
know  only  one     is  required. In practice, of course, LABL1
would be declared LOCAL    or we could only use this macro    once
per program without getting duplicate label errors. There is
a  way    out.  In  all  8080 instructions involving condition
codes,    there  is  a  one bit difference between a condition
code  and  it's    opposite, and it is always the same bit. The
bit  used  is  the  bit     with  a  value    of 8. Compare the JZ
instruction with the JNZ instruction:

     1     1   0     0   1     0   1     0     Jump If Zero
     1     1   0     0   0     0   1     0     Jump If Not Zero
   --- --- --- --- --- --- --- ---
   128    64  32    16  *8*     4   2     1     Bit Values

Changing  this    one bit    changes    a JC to    a JNC, a JP to a JM,
and  a    JPO  to     a JPE.    The bit    has the    same function in the
conditional call and return instructions as well.
     Combining    all  these facts, we can write our CPM macro
with the condition code    parameter. It comes out    looking    like
this:

CPM    MACRO    FUNC,VALUE,COND
     LOCAL     LABLX         ;Generate one-time label
     IF     NOT NUL COND     ;Generate jump    only when needed
     DB     ( J&COND ) XOR    8 ;Change condition code to its
                 ;    opposite.
     DW     LABLX         ;Address field    of jump    instr.
     ENDIF
     IF     NOT NUL VALUE
     LXI     D,VALUE
     ENDIF
     IF     NOT NUL FUNC
     MVI     C,FUNC
     ENDIF
     CALL     BDOS
LABLX:    ;Jump here if COND wasn't true.     Continue...
     ENDM

     Coding  CPM     CON,LF,NZ    would cause the    opcode to be
expanded as:
     DB     ( JNZ ) XOR 8
which is the same as:
     DB     ( JZ )
Note  that  the     COND  parameter  could    have been any of the
valid 8080 condition codes: Z, NZ, C, NC, P, M,    PO, PE.
     Also,  we    could  use  conditional     calls ( C&COND    ) or
conditional  returns  (    R&COND)    and changed the    sense of the
condition  by  XORing  the  opcode  with  8  to     change     the
necessary  bit.    This technique can be quite useful in macros
of all types to    allow the use of condition codes on the    same
line   as  the    macro  call,  saving  programmer  time,     and
lessening  the    chances    of error. For your information,    here
is  a  list  of     8080  opcodes    which are changed into their
opposites by XORing with 8. Although the ones with condition
codes  will  probably be the most useful, some of the others
may  come in handy some    day. The instruction on    the left has
the "8's" bit as a 0, on the right as a    1.
     STAX     LDAX
     RLC     RRC
     RAL     RAR
     INX     DCX
     SHLD     LHLD
     JNC     JC
     JNZ     JZ
     JP     JM
     JPO     JPE
     CNC     CC
     CNZ     CZ
     CP     CM
     CPO     CPE
     RNC     RC
     RNZ     RZ
     RP     RM
     RPO     RPE
     DI     EI
     OUT     IN

Do yourself a favor and    use this trick only where it will
genuinely improve efficiency, not just to make simple code hard
to read.  Happy    Hacking!

            Gary P. Novosielski
           [phone] Eve's.
iciency, not just to make simple code hard
to read.  Happy    Hacking!