CHAPTER 3 Macro - Assembler

From Atari Wiki
Revision as of 16:19, 28 January 2012 by Admin (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Here we are yet again.  Chapter 3 of the huge Devpac Docs typed by 
the Animal house and edited by Sewer Rat for Doc Disk Number 8.


                            CHAPTER 3

                         MACRO ASSEMBLER


INTRODUCTION

     GenST  is a powerful,  fast,  full  specification  assembler, 
available  instantly  from within the editor or as a  stand  alone 
program.  It  converts the text typed or loaded into  the  editor, 
optionally  together  with the files read from the  disk,  into  a 
binary file suitable for immediate execution or linking, or into a 
memory image for immediate execution from the editor.


INVOKING THE ASSEMBLER

FROM THE EDITOR

     The  assembler  is  invoked from the editor  by  clicking  on 
Assemble from the Program menu,  or by pressing Alt-A.  A dialogue 
box will appear which looks like this (almost),


-------------------------------------------------
                    Assembly Options

Program type   Exec        GST        DRI 
Symbols case   Dependant     Independant
Debug info     None    Normal   Extended
List           None   Screen  Printer  Disk

               Output to 

None     Memory  max 10_k

Disk |__________________________________

Cancel                        Assemble

__________________________________________________

     PROGRAM  TYPE This lets you select between executable GST  or 
DRI  format  output.  The differences between these  are  detailed 
later.

     SYMBOLS  CASE  This lets you select whether labels  are  case 
dependant or not. If case Dependant is selected then Test and test  
would  be different labels,  if the case Independant  is  selected 
then they would be the same.

     DEBUG  INFO  If  you wish to debug your  program  using  your 
original  symbols you can select Normal or Extended  debug  modes. 
The  advantage  of extended debug is that up to 22  characters  of 
each symbol are included in the debug information,  whereas normal 
mode restricts symbols to eight characters.

     LIST  selecting  Printer  will  divert  the  listing  to  the 
current printer port, or selecting Disk will send the listing to a 
file based on the source filename, but with the extension .LST.

     OUTPUT TO This lets you select where the output file is to be 
created. None means it is 'thrown away', ideal for syntax checking 
a program;  Memory means it is assemble into a buffer allowing  it 
to be run or debugged instantly form the editor without having  to 
create  a  disk  file:  Disk means a file  will  be  created.  The 
selection  of the name of this file can be left to the  assembler, 
using rules described shortly.
     The  first  two options may be specified in the  source  file 
using the OPT directive.
     Having selected your required options you should click on the 
Assemble  button (or press Return) to start the assembly.  At  the 
end of assembly you should press any key to return to the  editor. 
If any errors occurred the cursor will be positioned on the  first 
offending line.


STAND ALONE ASSEMBLER

     If  the  .TTP version of the assembler is  invoked,  the  one 
without  a  command line,  the programmer will be asked  for  one, 
confirming to the rules below,  or press Return to abort.  At  the 
end of assembly there will be a pause,  pressing any key will exit 
the  program.  If a command line has been supplied  the  assembler 
will not wait for a key as it assumes it has been run from the CLI 
or batch file.


COMMAND LINE FORMAT

     The command line should be the form

          mainfile    <-options>  [-options]

     The  mainfile  should  be  the name  of  the  file  requiring 
assembly and if no extension is specified defaults to ,S.  Options 
should  follow  this  donated by a -  sign  then  an  alphabetical 
character.   Allowed   options  are  shown  below  together   with 
equivalent OPT directives :

     B    no binary file should be created
     C    case insensitive labels (OPT C-)
     D    debug (OPT D+)
     L    GST linkable code (OPT L+)
     L2   DRI linkable code (OPT L2)
     O    specify output filename (follows immediately after O)
     P    specify  listing  filename  (should  follow  immediately 
          after P) defaults to source filename with extension .LST
     Q    pause for key press after assembly
     X    extended debugging (OPT X+)

     The  default is to create an executable binary file with  the 
name  based on the source file and output file type,  no  listing, 
with no case sensitive labels. For example,

          test  -b

assembles test.s with no binary output file

          test  -om:test.prg -p

assembles test.s into a binary file m:test.prg and a listing  file 
to test.lst

          test  -L2dpprn:

assembles  test.s into DRI linkable code with debug and a  listing 
to  the  parallel  port.  (A listing to the  serial  port  can  be 
obtained by specifying AUX: as the listing name.


OUTPUT FILENAME

     GenST  has  certain rules regarding the  calculation  of  the 
output filename, using a combination of that specified at assembly 
time  (either in the Disk:  filename field in the dialogue box  or 
using the -O option on the command line) and the Output directive:
     If  an output filename is explicitly given at  assembly  time 
then 
     name = explicit filename 
else
     if the Output directive has not been used then
          name = source filename + .PRG, .BIN or .O
     else if the Output directive specifies an extension then 
          name = source filename + extension in Output
     else
          name = name in Output


ASSEMBLY PROCESS

     GenST is a two-pass assembler; during the first pass it scans 
all  the text in memory and from disk if required,  building up  a 
symbol  table.  If  syntax  errors are found  on  the  first  pass 
assembly these will be reported and assembly will stop at the  end 
of  the  first  pass,   otherwise,  during  the  second  pass  the 
instructions are converted into bytes,  a listing may be  produced 
if required and a binary file can be created on the  disk.  During 
the  second  pass any further errors and warnings will  be  shown, 
together with a full listing and symbol table if required.
     During assembly, any screen can be paused by pressing Ctrl-S, 
pressing  Ctrl-Q  will  resume it.  Assembly  may  be  aborted  by 
pressing Ctrl-C, although doing so will make any binary file being 
created  invalid  as  it  will be incomplete  and  should  not  be 
executed.


ASSEMBLY TO MEMORY
     
     To  reduce  development time GenST can assemble  programs  to 
memory, allowing immediate execution or debugging from the editor. 
To  do  this  a program buffer is used.,  the  size  of  which  is 
specified in the Assembly Options dialogue box. If no debug option 
is  specified the size given can be just a little larger than  the 
output  program,  but if either form of debug is required  a  much 
larger buffer may be needed.

   A  program running from memory is just like any other  normal 
GEMDOS  program and should terminate using either pterm or  pterm0 
GEMDOS calls, foe example
          
               clr.w -(a7)
               trap  #1

     Programs may self-modify if required as a re-executed program 
will be in its original state.
     The  program buffer size and current assembly options can  be 
made  the default on re-loading the editor if Save Preferences  is 
used.


BINARY FILE TYPES

     There are six types of binary files which may be produced  by 
GenST, for different types of applications. They are distinguished 
by the extension on the filename:

     .PRG   GEM-type application i.e. one that uses windows
     .TOS   TOS-type application i.e. one that doesn't use windows
     .TTP   TOS-type application that requires a command line 
     .ACC   desk accessory program file 
     .BIN   non-executable  file  suitable for linking  with  GST-
            format files and libraries
     .O     non-executable  file  suitable for linking  with  DRI-
            format files and libraries  

     It can also assemble executable code directly to memory  when 
using  the  integrated version allowing very  fast  edit-assemble-
debug-run times.
     The  first three are double-clickable,  can be run  from  the 
Desktop  and  are  known  as  executable.   They  differ  in   the 
initialisation performed before the execution. With .PRG files the 
screen is cleared to the Desktop's pattern,  while with the  other 
two  the screen clears to white,  the flashing cursor appears  and 
the  mouse  is disabled.  When you double-click a  .TTP  file  the 
Desktop will prompt you for a command line to pass to it.
  
     .ACC  files  are executable but cannot be  double-clicked  on 
from the desktop. They will only run successfully when executed by 
the AES during the boot sequence of the machine   
     .BIN and .O files cannot be run immediately,  but have to  be 
read into a linker,  usually with other sections, and are known as 
linkable object modules. There are two different linker formats on 
the ST,  .BIN files are GST format,  .O files are DRI format.  The 
difference between these is discussed later in this chapter.
     The above extensions are not absolute rules;  for example, if 
you  have a TOS type program you may give it a .PRG extension  and 
use  the Install Application function from the Desktop,  but  it's 
usually much easier to use the normal extensions. One exception is 
for  programs  that  are to be placed in an AUTO  folder  so  they 
execute  during  the  boot sequence.  They have  to  be  TOS  type 
programs,  but  need the extension .PRG for the boot  sequence  to 
find them.

**   Certain versions of the French ST ROM's do not recognise .TTP 
files  from  the  Desktop so they have to  be  renamed  .TOS  then 
installed as TOS Takes Parameters.


TYPES OF CODE

     Unlike  most 8-bit operating systems,  but like  most  16-bit 
systems,  an executable program under GEMDOS will not be loaded at 
a  particular  address  but,  instead,  be loaded  at  an  address 
depending on the exact free memory configuration at that time.
     To get around the problem of absolute addressing the ST  file 
format includes relocation information allowing GEMDOS to relocate 
the  program  after it has loaded it but before  running  it.  For 
example the following program segment 

               move.I #string, a0
               .
               .
               .
     string    dc.b   'Press any key',0

places the absolute address of string into a register, even though 
at  assembly  time the real address of string cannot  possibly  be 
known.  Generally  the programmer may treat addresses as  absolute 
even though the real addresses will not be known to him, while the 
assembler  (or  linker) will look after the  necessary  relocation 
information.

**   For  certain  programs,   normally  games  or  cross  machine 
development an absolute address may be required,  for this  reason 
the ORG directive is supported.

The syntax of the assembler will now be described.


ASSEMBLER STATEMENT FORMAT

     Each  line  that is to be processed by the  assembler  should 
have the following format;
     
     Label     Mnemonic       Operand(s)          Comment
     Start     move.l         d0,(a0)+            store the result

     Exceptions  to  this  are  comment  lines,  which  are  lines 
starting with an asterisk or semi-colon,  and blank  lines,  which 
are  ignored.  Each field has to be separated from the  others  by 
white space - any number or mixture of space and tab characters.


LABEL FIELD

     The label should normally start at line 1,  but if a label is 
required  to start at another position then it should be  followed 
immediately   by  a  colon  (:).   Labels  are  allowed   on   all 
instructions, but are prohibited on some assembler directives, and 
absolutely  required  on  others.  A  label  may  start  with  the 
characters A-Z,  a-z,  or underline (_),  and may continue with  a 
similar  set together with the addition of the digits 0-9 and  the 
period (.).
     Labels  starting with a period are  local  labels,  described 
later.  Macro  names  and  register equate symbols  may  not  have 
periods in them,  though macro names may start with a  period.  By 
default the first 127 characters of labels are significant, though 
this can be reduced if required.  Labels should not be the same as 
register names, or the reserved words SR, CCR or USP.
     
     By  default  labels  are case-sensitive though  this  may  be 
changed.

     Some example legal labels are ;

test, Test, TEST, _test, _test.end, test5, _test5

     Some illegal labels are;
     
     5test, _&e, test>,

     There  are  certain reserved symbols  in  GenST,  denoted  by 
starting with two underline characters.  These are __LK,  __RS and 
__G2.


MNEMONIC FIELDS

     The  mnemonic  field  comes after the  label  field  and  can 
consist  of 68000 assembler instructions,  assembly directives  or 
macro  calls.  Some  instructions  and  directives  allow  a  size 
specifier,  separated from the mnemonic by a period. Allowed sizes 
are .B for byte,  .W for word, .L for long and .S for short. Which 
size specifiers are allowed in each case depends on the particular 
instruction or directive.  GenST is case-sensitive to mnemonic and 
directive names, so Move is the same as move and the same as mOvE, 
for example.


OPERAND FIELD

     For those instructions or directives which require  operands, 
this field contains one or more parameters,  separated by  commas. 
GenST is case-sensitive regarding register names so they may be in 
either or mixed case.


COMMENT FIELD

     Any  white space not within quotes found after  the  expected 
operand(s)  is  treated  as a delimiter before the  start  of  the 
comment, which will be ignored by the assembler.

EXAMPLES OF VALID LINES

               move.l  d0,(a0)+         comment is here
     loop      TST.W       d0
     lonely.label
        rts
* this is a complete line of comment
; and so is this
     indented: link A6,#-10 make room 
a_string: dc.b 'spaces allowed in quotes' a string


EXPRESSIONS

     GenST  allows complex expressions and supports full  operator 
precedence, parenthesis and logical operators.
     Expressions are in two types -absolute and relative - and the 
distinction is important. Absolute expressions are constant values 
which are known at assembly time. Relative expressions are program 
addresses  which  are  not known at assembly-time  as  the  GEMDOS 
loader  can  put  the  program where  it  likes  in  memory.  Some 
instructions and directives place restrictions on which types  are 
allowed  and  some  operators cannot be used  with  certain  type-
combinations.

OPERATORS

     The operators available,  in decreasing order of  precedence, 
are:
     monadic minus (-) and plus (+)
     bitwise not (~)
     shift left (<<) and shift right (>>)
     bitwise And (&), Or (!), and Xor (^)
     multiply (*) and divide (/)
     addition (+) and subtraction (-)
     equality (=), less than (<), greater than (>)

     The comparison operators are signed and return 0 if false  or 
-1  ($FFFFFFFF) if true.  The shift operators take the  left  hand 
operand and shift it the number of bits in the right hand  operand 
and vacated bits are filled with zeros.

     This precedence can be overridden by the parentheses ( and ). 
With operators of equal precedence, expressions are evaluated from 
left-to-right.  Spaces  in  expressions (other than  those  within 
quotes  as ASCII constants) are not allowed as they are  taken  as 
the separator to the comment.

     All expression evaluation is done using 32-bit signed-integer 
arithmetic, with no checking of overflow.

NUMBERS

     Absolute numbers may be in various forms;

          decimal constants, e.g. 1029
          hexadecimal constants, e.g. $12f
          octal constants, e.g. @730
          binary constants, e.g. %1100010
          character constants, e.g. 'X'

     $ is used to denote hex numbers,  % for binary,  @ for  octal 
and single ' or double quotes " for character constants.

CHARACTER CONSTANTS

     Whichever  quote is used to mark the start of a  string  must 
also  be used to denote its end and quotes themselves may be  used 
in  strings delimited with the same quote character by  having  it 
occur  twice.  Character  constants can be upto  4  characters  in 
length and evaluate to right-justified longs with null-padding  if 
required.  For example, her are some character constants and their 
ASCII and hex values:

     "Q"       Q              $00000051
     "hi"      hi             $00006869
     "Test"    test           $54657374
     "it's"    it's           $6974277C
     'it''s'   it's           $6974277C

     Strings  used  in DC.B statements follow  slightly  different 
justification rules, detailed with the directive later.

     Symbols  used  in  expressions will  be  either  relative  or 
absolute,  depending on how they were defined.  Labels within  the 
source  will  be  relative,  while those  defined  using  the  EQU 
directive  will be the same type as the expression to  which  they 
are equated.
     The  use of an asterisk (*) denotes the value of the  program 
counter at the start of the instruction or directive and is always 
a relative quantity.

ALLOWED TYPE COMBINATIONS

     The  table below summarises for each operator the results  of 
the various type combinations of parameter an d which combinations 
are  not allowed.  An R denotes a Relative result,  an  A  denotes 
absolute  and a * denotes that the combination is not allowed  and 
will produce an error message if attempted.

                    A op A   A op R   R op A   R op R
Shift operators       A        *        *         *
Bitwise operators     A        *        *         *
Multiply              A        *        *         *
Divide                A        *        *         *
Add                   A        R        R         *
Subtract              A        *        R         A
Comparisons           A        *        *         A

                    ALLOWED TYPE COMBINATIONS


ADDRESSING MODES

     The available addressing modes are shown in the table  below. 
Please   note  that  GenST  is  case-insensitive   when   scanning 
addressing modes, so D0 and a3 are both valid registers.

FORM      MEANING                                 EXAMPLE
Dn        data register direct                    D3
An        address register direct                 A5
(An)      address register indirect               (A1)
(An)+     address reg indirect + post-increment.  (A5)+
-(An)     address reg indirect + pre-increment.   -(A0)
d(An)     address reg indirect with displacement  20(A7)
d(An,Rn.s)address register indirect with index    4(A6,D4.L)
d.W       absolute short address                  $0410.W
d.L       absolute long address                   $12000.L
d(PC)     program counter relative with offset    NEXT(PC)
d(PC,Rn.s)program counter relative with index     NEXT(PC,A2.W)
#d        immediate data                          #26

n         denotes register number from 0 to 7
d         denotes a number 
R         denotes index register, either A or D
s         denotes size, either W or L, when omitted defaults to W

When  using address register indirect with index the  displacement 
may be omitted, for example

move.l    (a3,d2.l),d0

will assemble to the same as 

move.l    0(a3,d2.l),d0


SPECIAL ADDRESSING MODES

CCR       condition code register
SR        status register
USP       user stack pointer

In  addition to the above,  SP can be used in place of A7  in  any 
addressing mode, e.g. 4(SP,D3.W)
     The data and address registers can also be denoted by use  of 
the reserved symbols R0 through R15. R0 to R7 are equivalent to D0 
to D7,  R8 to R15 are equivalent to A0 to A7. This is included for 
compatibility with other assemblers.

LOCAL LABELS

     GenST supports local labels,  that is labels which are  local 
to  a  particular area of the source code.  These are  denoted  by 
starting  with  a period and are attached to  the  last  non-local 
label, for example;

lenl      move.l    4(sp),a0
.loop     tst.b     (a0)+
          bne.s     .loop
          rts
len2      move.l    4(sp),a0
.loop     tst.b     -(a0)
          bne.s     .loop
          rts

There  are  two labels called .loop in this code segment  but  the 
first is attached to lenl, the second to len2.
The local labels .W and .L are not allowed to avoid confusion with 
the absolute addressing syntax.

SYMBOLS AND PERIODS

     Symbols which include the period character can cause problems 
with GenST due to absolute short addressing.
The  Motorola  standard way of denoting absolute  short  addresses 
causes  problems as periods are considered to be part of a  label, 
best illustrated by an example:

          move.l    vector.w,d0

where vector is an absolute value, such as a system variable. This 
would  generate an undefined label error,  as the label  would  be 
scanned as vector.w.  To get around this,  the expression, in this 
case a symbol, may be enclosed in brackets, e.g.

          move.l    (vector).w,d0

though  the  period mat still be used after  numeric  expressions, 
e.g.

          move.l    $402.w,d0

**   GenST  version  1 also supported the use of \  instead  of  a 
period to denote short word addressing and this is still supported 
in this version,  but this is not recommended due to the potential 
for \W and \L to be mistaken for macro parameters.


                         INSTRUCTION SET

WORD ALIGNMENT

     All  instructions  with the exception of DC.B  and  DS.B  are 
always  assembled on a word boundary.  Should you require  a  DC.B 
explicitly on a word boundary,  the EVEN directive should be  used 
before  it.  Although all instructions that require it  are  word-
aligned,  labels with nothing following them are not word  aligned 
and can have odd values. This is best illustrated by an example;

          nop                 this is always word aligned
          dc.b      'odd'
start
          tst.l     (a0)+
          bne.s     start

The  above  code would not produce the required  result  as  start 
would have an odd value.  To help in finding such an  instructions 
the assembler will produce an error if it finds an odd destination 
in a BSR or BRA operand. Note that such checks are not made on any 
other  instructions,  so it is recommended that you  precede  such 
labels  with  EVEN  directives if you require  them  to  be  word-
aligned.  A  common error is deliberately not to do this,  as  you 
know  the preceding string is an even number of  bytes  long.  All 
will be well until the day you change the string...

INSTRUCTION SET EXTENSIONS

     The  complete 68000 instruction set is supported and  certain 
shorthands are automatically accepted,  detailed below. A complete 
description of the instruction set including syntax and addressing 
modes can be found in any 68000 reference guide or in the supplied 
pocket guide.

CONDITION CODE

The altenate condition codes HS and LO are supported in Bcc,  DBcc 
and Scc instructions, equivalent to CC and CS, respectively.

BRANCH INSTRUCTIONS

     To force a short branch use Bcc.B or Bcc.S,  to force a  word 
branch  use  Bcc.W or to leave the optimiser  use  Bcc.  Bcc.L  is 
supported for compatibility with GenST 1 with a warning as it  is, 
strictly  speaking,   a  68020  ibnstruction.   A  BRA.S  to   the 
immediately following instruction is not allowed and is converted, 
with  a warning,  to a NOP.  A BSR.S to the immediately  following 
instruction is not allowed and will produce an error.

BTST INSTRUCTION

     BTST is unique among bit-test instructions in supporting  PC-
relative addressing modes.

CLR INSTRUCTION

     CLR An is not allowed,  use SUB.L An,An instead (though  note 
that the flags are not affected).

CMP INSTRUCTION

     If  the source is immediate then CMPI is used,  else  if  the 
destination is an address register then CMPA is used, else if both 
addressing modes are post-incremented then CMPM is used.

DBcc INSTRUCTION

     DBRA is accepted as an equivalent to DBF.

ILLEGAL INSTRUCTION

     This generates the op-cope word $4AFC.

LINK INSTRUCTION

If the displacement is positive or not even a warning is given.

MOVE FROM CCR INSTRUCTION

     This  is a 68010 and upwards instruction,  converted  with  a 
warning to MOVE from SR.

MOVEQ INSTRUCTION

     If the data is in the range 128-255 inclusive a warning  will 
be  given.  It  may be disabled by specifying a long size  on  the 
instruction.


ASSEMBLER DIRECTIVES

     Certain  pseudo-mnemonics  are  recognised  by  GenST.  These 
assembler  directives,  as  they are called,  are  not  (normally) 
decoded into opcodes, but instead direct the assembler time. These 
actions have the effect of changing the object code  produced,  or 
the  format of the listing.  Directives are scanned  exactly  like 
executable  instructions and some may be preceded by a label  (for 
some  it is obligatory) and may be followed by a comment.  If  you 
put  a  label on a directive for which it  is  not  relevant,  the 
result  will  be undefined but will usually result  in  the  label 
being ignored.
     Each  directive will now be described in  turn.  Please  note 
that  the case of a directive name is not important,  though  they 
generally are shown in upper case.  The use of angled brackets ( < 
>  )  in descriptions denote optional  items,  ellipses  (  ...  )  
denote repeated items.

ASSEMBLY CONTROL

          END

     This directive signals that no more text is to be examined on 
the current pass of the assembler. It is not obligatory. 

          INCLUDE filename 

     This directive will cause source code to be taken from a file 
on  disk  and assembled exactly as though it were present  in  the 
text.  The  directive  must be followed by a  filename  in  normal 
GEMDOS format.
     If the filename has a space in it the name should be enclosed 
in  single  or double quotes.  A drive  specifier,  directory  and 
extension may be included as required, e.g. 

     include b:constants/header.s

     Include  directives may be nested as deeply as memory  allows 
and  if any error occurs when trying to open the file or read  it, 
assembly will be aborted with a fatal error. 
     If  no  drive or path name is specified,  that  of  the  main 
source file will be used when trying to open the file. 

**   The more memory the better,  GenST will read the whole of the 
file in one go if it can and not bother to re-read the file during 
pass 2.

     INCBIN    filename 

     This  takes a given binary file and  includes  it,  verbatim, 
into the output file.  Suggested uses include screen data,  sprite 
date and ASCII files.
          
     OPT       option <,option>...

     This  allows  control over various options within  GenST  and 
each one is denoted by a single character normally followed by a + 
or - sign. Multiple options may be specified, separated by commas. 
The allowed options are:

OPTION C - CASE SENSITIVITY AND SIGNIFICANCE.

     By default,  GenST is sensitive to label case and labels  are 
significant to 127 characters.  This can be overridden , using  C- 
for case-sensitivity or C+ for case-insensitivity. The significan�
ce  may be specified by specifying a decimal number between the  C 
and  the sign,  for example C16+ denotes case  insensitive  labels 
with  16 character significance.  This option may be used  at  any 
time  in  a  program but normally only makes  sense  at  the  very 
beginning of a source file.

OPTION D - DEBUGGING INFORMATION

     The  GEMDOS  binary file format supports the inclusion  of  a 
symbol  table at the end,  which may be read by debuggers such  as 
MONST  and  can be extremely useful when  debugging  programs.  By 
default  this is switched off but it may be activated with D+  and 
de-activated with D-.  The first 8 characters only of all relative 
labels are written to the file and will be upper-cased if GenST is 
in  case-sensitive mode.  The 8 character limit is due to the  DIR 
standard file format and may be improved on by using the  extended 
debug option, described below.

OPTION L - LINKER MODE.

     The  default for GenST is to produce executable code  but  L+ 
will  make it produce GST linkable code,  L2 will make it  produce 
DIR linkable code,  or L- will make it revert to executable.  This 
directive must be the very first line, in the first text file.

OPTION M - MACRO EXPANSIONS.

     When an assembly listing is produced,  macro calls are  shown 
in  the same form as in the source.  If you wish the  instructions 
within  macros to be listed,  use M+,  while M- will  disable  the 
option. You can use this directive as often as required.

OPTION O - OPTIMISING.

     GenST  is capable of optimising certain statements to  faster 
and  smaller versions.  By default all optimising is off but  each 
type  can  be  abled and disabled as  required.  This  option  has 
several forms:

OPT  O1+       will optimise backward branches to short if  within 
               range, can be disabled with O1-.
OPT  O2+       will   optimise  address  register  indirect   with 
               displacement addressing modes to addresss  register 
               indirect,  if the displacemant evalues to zero, and 
               can be disabled with O2-. For example,

          move.l   next(a0),d3

     will be optimised to 

          move.l   (a0),d3

     If the value of next is zero.

OPT O+         will turn all optimising on.
OPT O-         will turn all optimising off.
OPT O1-, OPT O2-    will disable the relevant optimisation.
OPT OW-        will disable the warning messages generated by each 
               optimisation, OW+ will enable them.

     If any optimising has been done during an assembly the number 
of optimisations made and bytes saved will be shown at the end  of 
assembly.

OPTION P - POSITION INDEPENDANT CHECKS.

     With  this  enabled with P+ GenST will check  that  all  code 
generated is position-independant,  generating errors on any lines 
which require relocation.  It can be disabled with P- and defaults 
to off.

OPTION S - SYMBOL TABLE

     When  a listing is turned on a symbol table will be  produced 
at the end.  If you wish to change this S- will disable it,  while 
S+ will re-enable it.  If you use this directive ore than once  the 
last one will be taken into account.

OPTION T - TYPE CHECKING.

     GenST  can  often spot programming errors as  it  checks  the 
types of certain expressions.  For some applications or styles  of 
programming  this can be more of a hindrance than a  help.  So  T- 
will  turn checks off.  T+ turning them back on.  For example  the 
program segment
          
     main      bsr            initialise
               lea            main(a6),a0
               move.l         (main).w,a0

     will  normally  produce  an  error  as  main  is  a  relative 
expression  where as the assembler expects an absolute  expression 
in both cases.  However if this code is designed to run on another 
68000  machine this may be perfectly valid,  so the type  checking 
should be disabled.

OPTION W - WARNINGS

     If  you wish to disable the warnings that GenST can  produce, 
you can do so with W-.  To re-enable them use W+.  This  directive 
can be used as often as required.

OPTION X - EXTENDED DEBUG

     This is a special version of option D which uses the HiSoft 
Extended  Debug format  to  generate  debugging  information  with 
symbols of up to 22 character significance. 

OPTION SUMMARY

     The  defaults  are  shown  in  brackets  after  each  options 
description;

C    case-sensitivity and significance (C127+)
D    include debugging info, (D-)
L-   produce executable code (default)
L+   produce GST linkable code
L2   produce DIR linkable code
M    expand macros enlisting (M+)
O    optimising control (O-)
P    position independant code checks (P-)
S    symbol table listing 
T    type checking (T+)
W    warning control (W+)
X    extended debug (X-)

For example -   opt  m+,s+,w-

will  turn macro expansion on,  enable the symbol table  list  and 
turn warnings off.

<label> EVEN

     This  directive  will force the program counter to  be  even, 
i.e.   word  aligned.   As  GenST  automatically  word-aligns  all 
instructions (except DC.Bs and DS.Bs).  It should not be  required 
very often but can be useful for ensuring buffers and strings  are 
word-aligned when required

     CNOP      offset,alignment

This  directive  will align the program counter  using  the  given 
offset  and alignment.  An alignment of 2 means  word-aligned,  an 
alignment of 4 means long-word-aligned and so on. The alignment is 
relative to the start of the current section. For example,

          cnop   1,4

aligns  the  program  counter  a  byte  past  the  next  long-word 
boundary.

<label>   DC.B      expression<,expression> ...
<label>   DC.W      expression<,expression> ...
<label>   DC.L      expression<,expression> ...

     These  directives define constants in memory.  They may  have 
one or more operands,  separated by commas.  The constants will be 
aligned  on  word boundaries for DC.W and DC.L no  more  than  128 
bytes can be generated by a single DC directive.
     DC.B  treats  strings slightly different to those  in  normal 
expressions.  While the rules described previously about quotation 
marks still applies.  No padding of the bytes will occur,  and the 
length of any string can be upto 128 bytes.
     Be very careful of spaces in any DC directives, as a space is 
the delimiter before a comment. For example, the line 

          dc.b      1,2,3 ,4

will only generate 3 bytes - the ,4 will be taken as a comment.

<label>   DS.B      expression
<label>   DS.W      expression
<label)   DS.L      expression

     These  directives  will  reserve  memory  locations  and  the 
contents will be initialised to zeros. If there is a label then it 
will be set to the start of the area defined,  which will be on  a 
word  boundary  for  DS.W  and  DS.L  directives.   There  is   no 
restriction  on the size though the larger the area the longer  it 
will take to save to disk.  For example,  all of these lines  will 
reserve 8 bytes of space, in different ways :

          ds.b 8
          ds.w 4
          ds.l 2

<label>   DCB.B     number,value
<label>   DCB.W     number,value
<label>   DCB.L     number,value

     This directive allows constant blocks of data to be generated 
of the size specified,  number specifies how many times the  value 
should be repeated.

          FAIL

     This directive will produce the error user error.  It can  be 
used  for  such things as warning the programmer if  an  incorrect 
number of parameters have been passed to a macro.

          OUTPUT    filename

     This directive sets the normal output filename though can  be 
overridden by specifying a filename at the start of  assembly.  If 
filename starts with a period then it is used as an extension  and 
the output name is built up as described previously.

          __G2      (reserved symbol)

     This is a reserved symbol that can be used to detect  whether 
GenST is being used to assemble a file using the IFD  conditional. 
The  value of this symbol depends on the version of the  assembler 
and is always absolute.
 

REPEAT LOOPS

     It  is  often  useful  to  be able  to  repeat  one  or  more 
instructions  a  particular number of times and  the  repeat  loop 
construct allows this.

<label>   REPT      expression
          ENDR

     Lines to be repeated should be enclosed within REPT and  ENDR 
directives  and will be repeated the number of times specified  in 
the expression. If the expression is zero or negative then no code 
will be generated.  It is not possible to nest repeat  loops.  For 
example

          REPT      512/4             copy a sector quickly
          move.l    (a0)+,(a1)+
          ENDR

**   Program  labels should not be define within repeat  loops  to 
prevent label defined twice errors.

LISTING CONTROL

          LIST

     This  will  turn the assembly listing on during  pass  2,  to 
whatever  device was selected at the start of the assembly (or  to 
the  screen of None was initially chosen).  All  subsequent  lines 
will be listed until an END directive is reached,  the end of  the 
text is reached, or a Nolist directive is encountered.
     Greater  control  over  listing sections of  program  can  be 
achieved using LIST+ or LIST- directives. A counter is maintained, 
the state of which dictates whether listing is on or off.  A LIST+ 
directive  adds one to the counter and a LIST- subtracts  one.  If 
the  counter  is zero or positive then listing is  on,  if  it  is 
negative  then listing is off.  The default starting value  is  -1 
(i.e.  listing  off)  unless  a  listing  is  specified  when  the 
assembler was invoked,  when it is set to zero. This system allows 
a  considerable  degree of control over listing  particularly  for 
include  files.  The normal LIST directive sets the counter to  0, 
NOLIST sets it to -1.

          NOLIST

     This will turn off any listing during pass 2.

     When  a listing is requested onto a printer or to  disk,  the 
output is formatted into pages,  with a header at the top of every 
page.  The header itself consists of a line containing the program 
title, date, time and page number, then a line showing the program 
title,  then a line showing the sub-title,  then a blank line. The 
date  format  will be printed in the  form  DD/MM/YY,  unless  the 
assembler  is  running  on a US Atari ST,  then  it  is  MM/DD/YY. 
Between  pages  a  form-feed character (ASCII  FF,  value  12)  is 
issued.

          PLEN      expression

     This  will  set the page length of the assembly  listing  and 
defaults to 60. The expression must be between 12 and 255.

          LLEN      expression

     This  will  set the line width of the  assembly  listing  and 
defaults  to 132.  The value of the expression must be between  38 
and 255.

          TTL       string

     This  will set the title printed at the top of each  page  to 
the  given string,  which may be enclosed in  single  quotes.  The 
first TTL directive will set the title of the first printed  page. 
If  no  title is specified the current include file name  will  be 
used.

          SUBTTL    string

     Sets  the  subtitle printed at the top of each  page  to  the 
given string,  which may be enclosed in single quotes.  The  first 
such directive will set the sub-title of the first printed page. 

          SPC       expression

     This  will  output  the number of blank lines  given  in  the 
expression in the assembly listing, if active.

          PAGE

     Causes a new page in the listing to be started.

          LISTCHAR  expression<,expression> ...

     This will send the characters specified to the listing device 
(except  the  screen)  and is intended for doing  things  such  as 
setting  condensed mode on printers.  For example,  on Epsons  and 
compatibles the line 

          listchar  15
    
will set the printer to 132-column mode.

          FORMAT    parameter<,parameter> ...

     This allows exact control over the listed format of a line of 
source  code.  Each parameter controls a field in the listing  and 
must consist of a digit from 0 to 2 inclusive followed by a +  (to 
enable the field) or a - (to disable it):

0    line number, in decimal
1    section name/number and program counter
2    hex data words, up to 10 words unless printer is less than 80 
     characters wide, when up to three words are listed.

LABEL DIRECTIVES

label     EQU       expression

     This directive will set the value and type of the given label 
to  the  results of the expression.  It may  not  include  forward 
references,  or  external  labels.  If there is any error  in  the 
expression,  the  assignment  will  not  be  made.  The  label  is 
compulsory and must not be a local label.

label     =         expression

Alternate form of EQU statement

label     EQUR      register

     This  directive  allows  a data or  address  register  to  be 
referred  to  by  a  user-name,  supplied as  the  label  to  this 
directive.  This is known as a register equate.  A register equate 
must be defined before it is used.

label     SET       expression

     This is similar to EQU,  but the assignment is only temporary 
and  can  be  changed with a  subsequent  SET  directive.  Forward 
references  cannot  be used in the expression.  It  is  especially 
useful for counters within macros, for example, using a line

zcount    set       zcount+1

(assuming zcount is set to 0 at the start of the source).  At  the 
start of pass 2 all SET labels are made undefined, so their values 
will always be the same on both passes

label     REG       register-list

     This  allows  a symbol to be used to denote a  register  list 
within MOVEM instructions,  reducing the likelihood of having  the 
list at the start of a routine different from the list at the  end 
of the routine. A label defined with REG can only be used in MOVEM 
instructions.

<label>   RS.B      expression
<label>   RS.W      expression
<label>   RS.L      expression

     These  directives  let you set up lists of  constant  labels, 
which is very useful for data structures and global variables  and 
is best illustrated by a couple of examples
     Let's  assume you have a data structure which consists  of  a 
long word,  a byte and another long word,  in that order.  To make 
your code more readable and easier to update should the  structure 
change, you could use lines such as 

          rsreset
d_next    rs.l      1
d_flag    rs.b      1
d_where   rs.l      1

and then you could access them with lines like 

          move.l    d_next(a0),a1
          move.l    d_where(a0),a2
          tst.b     d_flag(a0)

As  another  example  let's assume you are  referencing  all  your 
variables  off register A6 (as done in GenST and MonST) you  could 
define them with lines such as

onstate   rs.b      1
start     rs.l      1
end       rs.l      1

you then could reference them with lines such as 

          move.b    onstate(a6),d1
          move.l    start(a6),d0
          cmp.l     end(a6),d0

Each such directive uses its own internal counter,  which is reset 
to 0 at the beginning of each pass. Every time the assembler comes 
across  the directive it sets the label according to  the  current 
value  (with word alignment if it is .W or .L) then increments  it 
according to the size and magnitude of the directive. If the above 
definitions  were  the first RS directives,  onstate would  be  0, 
start would be 2 and end would b e6.

          RESET

This  directive will reset the internal counter as used by the  RS 
directive.

          RSSET     expression

This allows the RS counter to be set to a particular value.

          __RS      (reserved symbol)

This  is  a  reserved symbol having the current value  of  the  RS 
counter.


CONDITIONAL ASSEMBLY

     Conditional  assembly  allows  the  programmer  to  write   a 
comprehensive  source  program  that can  cover  many  conditions. 
Assembly  conditionals  may  be  specified  through  the  use   of 
arguments,  in the case of macros,  and through the definition  of 
symbols  in EQU or SET directives.  Variations in these  can  then 
cause  assembly  of only those parts necessary for  the  specified 
conditions.

     There   are  a  wide  range  of  directives  concerned   with 
conditional assembly. At the start of the conditional blocks there 
must  be  one of many IF directives and at the end of  each  block 
there must be an ENDC directive.  Conditional blocks may be nested 
up to 65535 levels. 
     Labels  should not be placed on IF or ENDC directives as  the 
directives will be ignored by the assembler.

           
          IFEQ      expression
          IFNE      expression
          IFGT      expression
          IFGE      expression
          IFLT      expression
          IFLE      expression

These  directives will evaluate the expression,  compare  it  with 
zero and then turn the conditional assembly on or off depending on 
the  result.  The  conditions  correspond  exactly  to  the  68000 
condition codes.  For example, if the label DEBUG had the value 1, 
then with the following code,

          IFEQ DEBUG
logon     dc.b      'enter a command:'.0
          ENDC
          IFNE DEBUG
          opt       d+        labels please
logon     dc.b      'Yeah, gimme man:',0
          ENDC

the first conditional would turn assembly off as 1 is not EQ to 0, 
while the second conditional would turn it on as 1 is NE to 0.

**   IFNE   corresponds  to  IF  in  assemblers  with   only   one 
conditional directive.
     The  expressions  used in these conditional  statements  must 
evaluate correctly.

          IFD       label
          IFND      label

     These  directives  allow  conditional  control  depending  on 
whether a label is defined or not.  With IFD, assembly is switched 
on if the label is defined, whereas with IFND assembly is switched 
on  if the label is not defined.  These directives should be  used 
with  care otherwise different object codes could be generated  on 
pass  1 and pass 2 which will produce incorrect code and  generate 
phasing errors. Both directives also work on reserved symbols.

          IFC       'string1','string2'

This  directive will compare two strings,  each of which  must  be 
surrounded  by single quotes.  If they are identical  assembly  is 
switched  on,  else  it i switched off.  The  comparison  is  case 
sensitive.

          IFNC      'string1','string2'

This directive is similar to the above, but only switches assembly 
on  if  the strings are not identical.  This may at  first  appear 
somewhat useless, but when one or both of the parameters are macro 
parameters it can b every useful, as shown in the next section.

          ELSEIF

This  directive toggles conditional assembly from on  to  off,  or 
vice versa.

          ENDC

This  directive  will terminate the current level  of  conditional 
assembly.  If  there  are more IF's then ENDC's an error  will  be 
reported at the end of assembly.

          IIF       expression   instruction

This  is  a  short form of the IFNE directive  allowing  a  single 
instruction  or directive to be assembled conditionally.  No  ENDC 
should be used with IIF directives.


                         MACRO OPERATIONS

GenST  fully supports Motorola-style macros,  which together  with 
conditional  assembly  allows you greatly  to  simplify  assembly-
language programming and the readability of your code.
     A macro is a way for a programmer to specify a whole sequence 
of  instructions  or  directives  that  are  used  together   very 
frequently. A macro is first defined, then its name can be used in 
a macro call like a directive with up to 36 parameters.

label     MACRO

This  starts  a  macro definition and causes  GenST  to  copy  all 
following  lines  to  a macro buffer until an  ENDM  directive  is 
encountered. Macro definitions may not be nested.

          ENDM

This terminates the storing of a macro definition,  after a  MACRO 
directive.

          MEXIT

This  stops  prematurely the current macro expansion and  is  best 
illustrated by the INC example given later.

          NARG      (reserved label)

This  is not a directive but a reserved symbol,  Its value is  the 
number  of parameters passed to the current macro,  or 0  if  used 
when not within any macro.  If GenST is in the case sensitive mode 
then the name should be all upper-case.

     Once a macro has been defined with the Macro directive it can 
be invoked by using its name as a directive,  followed by upto  36 
parameters.  In the macro itself the parameters may be referred to 
by  using a backslash character (\) followed by  an  alpha-numeric 
(1-9,  A-Z  or  a-z)  which will be  replaced  with  the  relevant 
parameter when expanded or with nothing if no parameter was given. 
There  is  also the special macro parameter \0 which is  the  size 
appended to the macro call and defaults to W if none is given.  If 
a  macro  parameter  is  to include  spaces  or  commas  then  the 
parameter should be enclosed in between < and > symbols;  in  this 
case  a  >  symbol  may  be  included  within  the  parameters  by 
specifying >>.

     A special form of macro expansion allows the conversion of  a 
symbol to a decimal or hexadecimal sequence of digits,  using  the 
syntax \<symbol> or \$<symbol>, the latter denoting hex expansion. 
The symbol mast be defined and absolute at the time of expansion.

     The  parameter \@ can be useful for generating unique  labels 
with each macro call and is replaced when the macro is expanded by 
the  sequence  _nnn where nnn is a number which increases  by  one 
with  every  macro call.  It may be expanded up to  5  digits  for 
very large assemblies.

A true \ may be included in a macro definition by specifying \\.     

     A  macro  call  may  be  spread  over  more  than  one  line, 
particularly  useful for macros with large numbers of  parameters. 
This  can  be  done  by ending a macro  call  with  a  comma  then 
starting  the next line with an & followed by tabs or spaces  then 
the continuation of the parameters. 

In the assembly listing the default is to show just the macro call 
and not the code produced by it. However, macro expansion listings 
can  be  switched on and off using the OPT M  directive  described 
previously.

Macro  calls may be nested as deeply as memory  permits,  allowing 
recursion if required.

Macro  names  are  stored in a separate  symbol  table  to  normal 
symbols so will not clash with similarly named routines,  and  may 
start with a period.

MACRO EXAMPLES

EXAMPLE 1 - CALLING THE BDOS 

As the first example,  the general GEMDOS calling-sequence for the 
BDOS is;

     put the word parameters on the stack
     invoke a TRAP #1
     correct the stack afterwards

A macro to follow these specifications could be, 

call_gemdos         MACRO
          move.w    #\1,-(a7)           function
          trap      #1
          lea       \2(a7),a7           correct stack
          ENDM

The directives are in capitals only to make them stand  out:  they 
don't have to be.  If you wanted to call this macro to use  GEMDOS 
function 2 (print a character) the code would be 

          move.w    #'X',-(a7)
          call_gemdos   2,4

When this macro call is expanded,  \1 is replaced with 2 and \2 is 
replaced with 4. \0, if it occurred in the macro, would be W as no 
size  is given on the call.  So the above call would be  assembled 
as:

          move.w    #2,-(a7)
          trap      #1
          lea       4(a7),a7


EXAMPLE 2 - AN INC INSTRUCTION]

The 68000 does not have the INC instruction like other  processors 
but the same effect can be achieved using an ADDQ #1  instruction. 
A macro may be used to do this, like so:

inc       MACRO     
          IFC       '','\1'
          fail      missing parameter!
          MEXIT
          ENDC
          addq.\0   #1,\1
          ENDM

An example call would be

          inc.1     a0

which would expand to 

          addq.1    #1,a0

The  macro starts by comparing the first parameter with  an  empty 
string and causing an error message to be issued using FAIL if  it 
is equal.  The MEXIT directive is used to leave the macro  without 
expanding the rest of it.  Assuming there is a non-null parameter, 
the next line does the ADDQ instruction, using the \0 parameter to 
get the correct size.

EXAMPLE 3 - A FACTORIAL MACRO

Although  unlikely actually to be used as it  stands,  this  macro 
defines  a  label to be the factorial of a number.  It  shows  how 
recursion  can work in macros.  Before showing the  macro,  it  is 
useful to examine how the same thing would be done in a high level 
language such as Pascal.

function  factor(n:integer):integer;
begin
          if n>0 then
                    factor:=n*factor(n-1)
          else
                    factor:=1
end;

The  macro  definition for this uses the SET directive to  do  the 
multiplication n*(n-1)*(n-2) etc. in this way

*    parameter 1= label,   parameter 2= 'n'
factor    MACRO
          IFND      \1
\1        set       1         set if not yet defined
          ENDC
          IFGT      \2
          factor    \1,\2-1   work out next level down
\1        set       \1*(\2)   n = n * factor(n-1)
          ENDC
          ENDM
*    a sample call
          factor test,3

The  net  result  of the previous code is to set  test  to  3!  (3 
factorial). The reason the second set  has (\2) instead of just \2 
is  that  the  parameter  will  not  normally  be  just  a  simple 
expression,  but a list of numbers separated by minus signs, so it 
could assemble to 

test      set       test*5-1-1-1

(i.e. test*5-3) instead of the correct

test      set       test*(5-1-1-1)

(i.e. test *2)

EXAMPLE 4 - CONDITIONAL RETURN INSTRUCTION

The 68000 lacks the conditional return instruction found on  other 
processors,  but macros can be defined to implement them using the 
\@ parameter. For example, a return if EQ macro could look like:

rtseq     MACRO
          bne.s     \@
          rts
\@
          ENDM

The  \@ parameter has been used to generate a unique  label  every 
time  the macro is called,  so will generate in this  case  labels 
such as _002 and _017.

EXAMPLE 5 - NUMERIC SUBSTITUTION

Suppose you have a constant containing the version number of  your 
program and wish it to appear as ASCII in a message:

showname  MACRO
          dc.b      \1,'\<version>',0
          ENDM
          .    
          .
version   equ       42
          showname  <'Real Ale Search Program v'>

will expand to the line

          dc.b      Real Ale Search Program v','42',0

Note  the  way  the string parameter is enclosed in  <  >s  as  it 
contains spaces.

EXAMPLE 6 - COMPLEX MACRO CALL

Suppose your program needs a complicated table structure which can 
have  a varying number of fields.  A macro can only be written  to 
using those parameters that are specified,  for example:

table_entry         macro
          dc.b      .end\0-*            length byte
          dc.b      \1                  always 
          IFNC      '\2',''
          dc,w      \2,\3               2nd and 3rd together
          ENDC
          dc.l      \4,\5,\6,\7
          IFNC      '\8',''
          dc.b      '\8'                text
          ENDC
          dc.b      \9
.end\@    dc.b      0
          ENDM

* a sample call 
          table_entry                   &42,,,t1,t2,t3,t4,
&                                       <Enter Name:>,%0110

This   is  a  non-trivial  example  of  how  macros  can  make   a 
programmer's  life so much easier when dealing with  complex  data 
structures.  In  this  case the table consists of a  length  byte, 
calculated in the macro using \@,  two optional words, four longs, 
an optional string, a byte, then a zero byte. The code produced in 
this example would be :

          dc.b      .end_001
          dc.b      $42
          dc.l      t1,t2,t3,t4
          dc,b      'Enter Name:'
          dc,b      %0110
.end_001  dc.b      0


OUTPUT FILE FORMATS

     GenST is very flexible in terms of output file formats. These 
are detailed in this section together with notes on the advantages 
and  disadvantages  of each.  Certain  directives  take  different 
actions, depending on what output file format is specified.

     The exact details of using each format will now be described.

EXECUTABLE FILES

     These  are  directly  executable,   for  example,  by  double 
clicking  from  the  desktop.  The  file  may  include  relocation 
information  and/or symbolic information.  Normal file  extensions 
for this of type file .PRG, .TOS, .TTP and .ACC.

ADVANTAGES     true BSS sections, reduced development time.

DISADVANTAGES  messy if more than one programmer.

GST LINKABLE FILES

     When  writing  larger programs,  or  when  writing  assembly 
language  modules  for  use  from  the  high  level  language,   a 
programmer needs to generate a linkable file.  The GST link format 
is  supported  by  most of the high level  languages  produced  in 
Britain as well as others,  for example HiSoft Basic,  Lattice  C, 
Prospero  Fortran and Prospero Pascal.  GST format files  normally 
have the extension .BIN.

Advantages     great  degree of freedom - imported labels  can  be 
used  practically anywhere including within arbitrary  expressions, 
libraries can be created from the assembler,  import methods means 
the assembler can detect type conflicts.

Disadvantages  library format means selective library linking  can 
be  slow,  true GEMDOS sections not supported as standard  (though 
LinkST can create true BSS sections).

DRI LINKABLE FILES

     This  is the original linker format for the Atari ST  created 
by  Digital  Research originally for CP/M 68k.  It  is  supported, 
often via a conversion utility,  by the majority of US high  level 
languages. DRI format files normally have the extension .O.

Advantages     selective  libraries  are faster to link  than  GST 
format, GEMDOS sections are fully supported.

Disadvantages  very restrictive on use of imported labels;  object 
files  twice  as big as executable files,  8  character  limit  on 
symbols.

CHOOSING THE RIGHT FILE FORMAT

     If  you wish to link with a high level language  there  isn't 
usually  much  choice  -  you have  to  use  whichever  format  is 
supported by the language.
     If  you  are writing entirely in assembly language  then  the 
normal  choice has to be executable - it is fast to  assemble,  no 
linking  required,  and  allows assembly to memory  for  decreased 
development time.
     If  you  are writing a larger program,  say bigger  than  32k 
object,  or writing a program as a team,  then linkable code often 
makes  most sense.  We recommend GST-linkable over DRI because  of 
the much greater flexibility of the format.

OUTPUT FILE DIRECTIVES

     This section details those directives whose actions depend on 
the  output  file format chosen.  The file format  itself  can  be 
chosen by one of the following methods; command line options using 
GENST2.TTP;  clicking on the radio buttons in the Assembly Options 
dialogue box from the editor,  or with the OPT L directive at  the 
beginning of the source file.

Icons are used to denote those sections specific to a file format, 
viz

*EXEC*    Executable-code, also assembled to memory code.

*GST*     GST-linkable code

*DRI*     DRI-linkable code.


MODULES AND SECTIONS

          MODULE    modulename

     This  defines  the start of a new  module.  The  module  name 
should be contained within quotes if it contains spaces.  There is 
a  default  module  called ANON_MODULE so  the  directive  is  not 
obligatory.

*EXEC*    This directive is ignored

*DRI*     This directive is ignored

*GST*     This directive allows assembly-language library files to 
be  created using multiple modules.  Each module is like  a  self-
contained  program  segment,  with its own  imports  and  exports. 
Relative labels are local to there own modules, so you can use two 
labels with the same name in different modules with no danger of a 
clash.  Absolute  labels  are global to  all  modules,  ideal  for 
constants and the like.                  

          SECTION   sectionname

     This defines a switch to the named section.  The program  may 
consist  of several sections which will be  concatenated  together 
with other sections of the same name in the final executable file. 
By default assembly starts in the TEXT section.  You may switch to 
any section during the assembly.

*EXEC*    Allowed section names are TEXT, the normal program area, 
DATA,  for  initialised  data,  and BSS a special area  of  memory 
reserved by the GEMDOS program loaded.  It is initialised to  zero 
and takes up no space within the disk file.  When in a BSS section 
no  code  generating  instructions  are  allowed  except  the   DS 
directive.  Using  a  BSS section for global  variables  can  save 
valuable disk space.

*DRI*     The rules described above for executable files apply.

*GST*     There are no rigid rules about section  names.  Sections 
with  the same names from different files will be concatenated  by 
the linker. The default ordering of sections is the order they are 
first used in.

IMPORTS AND EXPORTS

     With both linkable types of program it is crucial to be  able 
to import and export symbols,  both relative symbols (i.e. program 
references) and absolute symbols (i.e.  constants). The GST format 
distinguishes between these types whereas the DRI format does not. 
The GST format allows the assembler to type check,  often  finding 
programming errors that would otherwise be missed.

          XDEF      export<,export>...

     This  defines  labels  for  export  to  other  programs   (or 
modules).  If any of the labels specified are not defined an error 
will occur. It is not possible to export local labels.

*EXEC*    This directive is ignored

*DRI*     Note  that  all  symbols  will  be  truncated   (without 
warning) before exporting. OPT C8 is therefore recommended.

          XREF      import<,import>...
          XREF.L    import<,import>...

     This  defines  labels to be imported from other  programs  or 
modules.  If any of the labels specified are defined an error will 
occur.  The  normal  XREF  statement should be used  to  import  a 
relative label (i.e.  program reference),  while XREF.L should  be 
used to import absolute labels (i.e. constants). Importing a label 
more than once will not produce an error.

*EXEC*    This directive is ignored

*DRI*     The  DRI format does not actually need to know the  type 
of imports but it is recommended that both forms of XREF are  used 
to  allow  the assembler to type check.  If you do not  type  your 
imports you should turn type checking off using OPT T-. DRI labels 
are only significant to the first 8 characters.

*GST*     Care  should  be taken to import labels of  the  correct 
type otherwise the relocation information will not be correct.


USING IMPORTS IN EXPRESSIONS

*EXEC*    There are no imports!

*DRI*     Imports  may be used in expressions but only one  import 
per  expression is allowed.  The results of an expression with  an 
import  in,  must be of the form import+number  or  import-number.
Imports can be combined with arbitrarily complex  expressions,  so 
long as the complex expression lexically precedes it, for example

          move.l    3+(1<<count+5)+import

*GST*     Imports may be used in expressions,  with up to ten  per 
expression.  They may only be added or subtracted from each  other 
though  can be combined with arbitrarily complex  expressions,  so 
long as the complex expression lexically precedes it, for example 

          move.l    3+(1<<count+5)+import1-import2

     Where  exactly an expression involving an import can be  used 
depends  on the file format.  The following file shows  which  are 
allowed.

Expression     GST            DRI       Example
PC-byte        Y              N         move.w import(pc,d3.w)
                                        bsr.s  import
PC-word        Y              Yž        move.w import(pc),a0
                                        bsr    import
byte           Y              N         move.b #import,do
word           Y              Y         move.w import(a3),d0
long           Y              Y         move.l import,d0

ž so  long as it is not a reference to a different section in  the 
same program, which is not allowed.

*DRI*     *GST*     Note  that  a  reference  to  a  symbol  in  a 
different  section  is regarded as an import and  subject  to  the 
above rules.

          COMMENT   comment string

*EXEC*    This directive is ignored

*DRI*     This directive is ignored

*GST*     The  directive passes the following string,  exactly  as 
entered, into the .BIN file and will be shown by the linker.

          ORG       expression

     This will make the assembler generate position dependant code 
and  set  the program counter to the given  value.  Normal  GEMDOS 
program,s do not need an ORG statement even if position dependant. 
It  is included to allow code to be generated for the ROM port  or 
for other 68000 machines.  More than one ORG statement is  allowed 
in a source file but no padding of the file is done.

*EXEC*    It  should  be used with great care as the  binary  file 
generated will probably not execute correctly when double clicked. 
as  no  relocation information is written  out.  The  binary  file 
produced  has  the standard GEMDOS header at  the  front,  but  no 
relocation information.

*DRI*     This directive is not allowed,  absolute code generation 
is an option in the linker.
*GST*     This  sends the ORG directive to the linker  which  will 
pad the file with zeros to the given address.

*NOTE*    This  directive  is  very unlikely to  make  sense  when 
assembling to memory.

          OFFSET    <expression>

     This  switches  code  generation  to  a  special  section  to 
generate absolute labels. The optional expression sets the program 
counter for the start of this section. No bytes are written to the 
disk and the only code generating directive allowed is DS.  Labels 
defined in this section will be absolute.  For example,  to define 
some of the system variables of the ST;

          offset    $400
etv_timer           ds.l      1         will be $400
etv_critic          ds.l      1         404
etv_term            ds.l      1         408
ext_extra           ds.l      5         40C
memvalid            ds.l      1         420
memcntlr            ds.w      1         424

          __LK      (reserved symbol)

     This  is a reserved symbol that can be used to  detect  which 
output  mode  is specified.  The value of this symbol  is  alwayus 
absolute and one of the following,

          0    executable
          1    GST linkable
          2    DRI linkable

other values are reserved for future expansion.

DRI DEBUG OPTION

     Normally  only explicitly XDEF'ed labels are included in  the 
symbol  table within the output file.  However the  format  allows 
what  it calls local labels (not to be confused with  GenST  local 
labels)  which are not true exports and cannot be referred  to  in 
any other modules but will be included in the symbol table in  the 
file  output file for debugging purposes.  OPT D+ will  cause  all 
relative labels to be output as DRI local labels.
�9[....�....�....�....�.........�.........�.........�.........�...�]0110
WRITING GST LIBRARIES

     When using multiple MODULEs to generate a GST format  library 
file  care  must  be taken with backward  references  to  imports. 
within  a  library file,  Higher level routines should  be  first, 
lower  level routines last.  For example the source file  skeleton 
overleaf will not link when used as a selective library.

          MODULE    low_level
          XDEF      low_output
low_output
          etc
          MODULE    high_level
          XDEF      high_output
          XREF      low_output
high_output
          etc

This is because the second module references a label defined in an 
earlier module, which is not allowed. The corrected version is:

          MODULE    high_level
          XDEF      high_output
          XREF      low_output
high_output
          etc
          MODULE    low_level
          XDEF      low_output
low_output     
          etc

SIMPLE FILE FORMAT EXAMPLES

     This section shows a (non-functional and incomplete)  example 
of the use of each file format.

Executable

          SECTION   TEXT
start     lea       string(pc),a0
          move.l    a0,save_str
          bsr       printstring
          bra       quit
          SECTION   DATA
string    dc.b      'Enter Your Name,0
          SECTION   TEXT
printstring
          move.l    a0,-(sp)
          move.w    #9,-(sp)
          trap      #1
          addq.l    #6,sp
          rts
          SECTION   BSS
save_str  ds.l      1
          END

DRI Linkable

          XREF.L    quit
          SECTION   TEXT
start     move.l    #string,a0
          move.l    a0,save_str
          bsr       printstring
          bra       quit
          SECTION   DATA
string    dc.b      'Enter Your Name,0
          SECTION   TEXT
printstring
          move.l    a0,-(sp)
          move.w    #9,-(sp)
          trap      #1
          addq.l    #6,sp
          rts
          SECTION   BSS
save_str  ds.l      1
          END

     note  the  way  the instruction has been  changed  as  a  PC-
relative reference is not allowed between sections.

GST Linkable

          MODULE    TESTPROG
          COMMENT   needs work
          XREF.L    quit
          SECTION   TEXT
start     lea       string(pc),a0
          move.l    a0,save_str
          bsr       prinrstring
          bra       quit
          SECTION   DATA
string    dc.b      'Enter Your Name,0
          SECTION   TEXT
printstring
          move.l    a0,-(sp)
          move.w    #9,-(sp)
          trap      #1
          addq.l    #6,sp
          rts
          SECTION   BSS
save_str  ds.l      1
          END


                        DIRECTIVE SUMMARY

Assembly Control

END       terminate source code
INCLUDE   read source file from disk
INCBIN    read binary file from disk
OPT       option control
EVEN      ensure PC even
CNOP      align PC arbitrarily
DC        define constant
DS        define space
DCB       define constant block
FAIL      force assembly error

Repeat Loop

REPT      start repeat block
ENDR      end repeat block

Listing Control

LIST      enable list
NOLIST    disable list
PLEN      set page length
LLEN      set line length
TTL       set title
SUBTTL    set sub-title
PAGE      start new page
LISTCHAR  send control character
FORMAT    define listing format

Label Directives

EQU       define label value
EQUR      define register equate
SET       define label value temporarily
REG       define register list
RS        reserve space
RSRESET   reset RS counter
RSSET     set RS counter

Conditional Assembly

IFEQ      assemble if zero
IFNE      assemble in non-zero
IFGT      assemble if greater than
IFGE      assemble if greater than or equal to
IFLT      assemble if less than
IFLE      assemble if less than or equal to
IFD       assemble if label defined
IFND      assemble if label not defined
IFC       assemble if strings same
IFNC      assemble if strings different
ELSEIF    switch assembly state
IIFC      immediate IF

Macros

MACRO     define macro
ENDM      end macro definition

Output File Directives

MODULE    start new module
SECTION   switch section
XDEF      define label for export
XREF      define label for import
COMMENT   send linker comment
ORG       set absolute code generation
OFFSET    define offset table

Reserved Symbols

NARG      number of macro parameters
__G2      internal version number
__RS      RS counter
__LK      output file type


Back to HiSoft DevpacST Version 2 - Manual