Implemented new .PUSHCPU and .POPCPU commands.

[cc65] / doc / ca65.sgml

<!doctype linuxdoc system>

<article>
<title>ca65 Users Guide
<author>Ullrich von Bassewitz, <htmlurl url="mailto:uz@cc65.org" name="uz@cc65.org">
<date>2000-07-19, 2000-11-29, 2001-10-02, 2005-09-08

<abstract>
ca65 is a powerful macro assembler for the 6502, 65C02 and 65816 CPUs. It is
used as a companion assembler for the cc65 crosscompiler, but it may also be
used as a standalone product.
</abstract>

<!-- Table of contents -->
<toc>

<!-- Begin the document -->

<sect>Overview<p>

ca65 is a replacement for the ra65 assembler that was part of the cc65 C
compiler, originally developed by John R. Dunning. I had some problems with
ra65 and the copyright does not permit some things which I wanted to be
possible, so I decided to write a completely new assembler/linker/archiver
suite for the cc65 compiler. ca65 is part of this suite.

Some parts of the assembler (code generation and some routines for symbol
table handling) are taken from an older crossassembler named a816 written
by me a long time ago.


<sect1>Design criteria<p>

Here's a list of the design criteria, that I considered important for the
development:

<itemize>

<item>  The assembler must support macros. Macros are not essential, but they
        make some things easier, especially when you use the assembler in the
        backend of a compiler.
<item>  The assembler must support the newer 65C02 and 65816 CPUs. I have been
        thinking about a 65816 backend for the C compiler, and even my old
        a816 assembler had support for these CPUs, so this wasn't really a
        problem.
<item>  The assembler must produce relocatable code. This is necessary for the
        compiler support, and it is more convenient.
<item>  Conditional assembly must be supported. This is a must for bigger
        projects written in assembler (like Elite128).
<item>  The assembler must support segments, and it must support more than
        three segments (this is the count, most other assemblers support).
        Having more than one code segments helps developing code for systems
        with a divided ROM area (like the C64).
<item>  The linker must be able to resolve arbitrary expressions. It should
        be able to get things like
<tscreen><verb>
        .import S1, S2
        .export Special
        Special = 2*S1 + S2/7
</verb></tscreen>
        right.
<item>  True lexical nesting for symbols. This is very convenient for larger
        assembly projects.
<item>  "Cheap" local symbols without lexical nesting for those quick, late
        night hacks.
<item>  I liked the idea of "options" as Anre Fachats .o65 format has it, so I
        introduced the concept into the object file format use by the new cc65
        binutils.
<item>  The assembler will be a one pass assembler. There was no real need for
        this decision, but I've written several multipass assemblers, and it
        started to get boring. A one pass assembler needs much more elaborated
        data structures, and because of that it's much more fun:-)
<item>  Non-GPLed code that may be used in any project without restrictions or
        fear of "GPL infecting" other code.
</itemize>
<p>


<sect>Usage<p>


<sect1>Command line option overview<p>

The assembler accepts the following options:

<tscreen><verb>
---------------------------------------------------------------------------
Usage: ca65 [options] file
Short options:
  -D name[=value]       Define a symbol
  -I dir                Set an include directory search path
  -U                    Mark unresolved symbols as import
  -V                    Print the assembler version
  -W n                  Set warning level n
  -g                    Add debug info to object file
  -h                    Help (this text)
  -i                    Ignore case of symbols
  -l                    Create a listing if assembly was ok
  -mm model             Set the memory model
  -o name               Name the output file
  -s                    Enable smart mode
  -t sys                Set the target system
  -v                    Increase verbosity

Long options:
  --auto-import         Mark unresolved symbols as import
  --bin-include-dir dir Set a search path for binary includes
  --cpu type            Set cpu type
  --debug-info          Add debug info to object file
  --feature name        Set an emulation feature
  --forget-inc-paths    Forget include search paths
  --help                Help (this text)
  --ignore-case         Ignore case of symbols
  --include-dir dir     Set an include directory search path
  --listing             Create a listing if assembly was ok
  --list-bytes n        Maximum number of bytes per listing line
  --macpack-dir dir     Set a macro package directory
  --memory-model model  Set the memory model
  --pagelength n        Set the page length for the listing
  --smart               Enable smart mode
  --target sys          Set the target system
  --verbose             Increase verbosity
  --version             Print the assembler version
---------------------------------------------------------------------------
</verb></tscreen>


<sect1>Command line options in detail<p>

Here is a description of all the command line options:

<descrip>

  <label id="option--bin-include-dir">
  <tag><tt>--bin-include-dir dir</tt></tag>

  Name a directory which is searched for binary include files. The option
  may be used more than once to specify more than one directory to search. The
  current directory is always searched first before considering any
  additional directories. See also the section about <ref id="search-paths"
  name="search paths">.


  <label id="option--cpu">
  <tag><tt>--cpu type</tt></tag>

  Set the default for the CPU type. The option takes a parameter, which
  may be one of

        6502, 65SC02, 65C02, 65816, sunplus, sweet16, HuC6280

  The sunplus cpu is not available in the freeware version, because the
  instruction set is "proprietary and confidential".


  <label id="option--feature">
  <tag><tt>--feature name</tt></tag>

  Enable an emulation feature. This is identical as using <tt/.FEATURE/
  in the source with two exceptions: Feature names must be lower case, and
  each feature must be specified by using an extra <tt/--feature/ option,
  comma separated lists are not allowed.

  See the discussion of the <tt><ref id=".FEATURE" name=".FEATURE"></tt>
  command for a list of emulation features.


  <label id="option--forget-inc-paths">
  <tag><tt>--forget-inc-paths</tt></tag>

  Forget the builtin include paths. This is most useful when building
  customized assembler modules, in which case the standard header files should
  be ignored.


  <label id="option-g">
  <tag><tt>-g, --debug-info</tt></tag>

  When this option (or the equivalent control command <tt/.DEBUGINFO/) is
  used, the assembler will add a section to the object file that contains
  all symbols (including local ones) together with the symbol values and
  source file positions. The linker will put these additional symbols into
  the VICE label file, so even local symbols can be seen in the VICE
  monitor.


  <label id="option-h">
  <tag><tt>-h, --help</tt></tag>

  Print the short option summary shown above.


  <label id="option-i">
  <tag><tt>-i, --ignore-case</tt></tag>

  This option makes the assembler case insensitive on identifiers and labels.
  This option will override the default, but may itself be overridden by the
  <tt><ref id=".CASE" name=".CASE"></tt> control command.


  <label id="option-l">
  <tag><tt>-l, --listing</tt></tag>

  Generate an assembler listing. The listing file will always have the
  name of the main input file with the extension replaced by ".lst". This
  may change in future versions.


  <label id="option--list-bytes">
  <tag><tt>--list-bytes n</tt></tag>

  Set the maximum number of bytes printed in the listing for one line of
  input. See the <tt><ref id=".LISTBYTES" name=".LISTBYTES"></tt> directive
  for more information. The value zero can be used to encode an unlimited
  number of printed bytes.


  <label id="option--macpack-dir">
  <tag><tt>--macpack-dir dir</tt></tag>

  This options allows to specify a directory containing macro files that are
  used instead of the builtin images when a <tt><ref id=".MACPACK"
  name=".MACPACK"></tt> directive is encountered. If <tt>--macpack-dir</tt>
  was specified, a <tt>.mac</tt> extension is added to the package name and
  the resulting file is loaded from the given directory. This is most useful
  when debugging the builtin macro packages.


  <label id="option-mm">
  <tag><tt>-mm model, --memory-model model</tt></tag>

  Define the default memory model. Possible model specifiers are near, far and
  huge.


  <label id="option-o">
  <tag><tt>-o name</tt></tag>

  The default output name is the name of the input file with the extension
  replaced by ".o". If you don't like that, you may give another name with
  the -o option. The output file will be placed in the same directory as
  the source file, or, if -o is given, the full path in this name is used.


  <label id="option--pagelength">
  <tag><tt>--pagelength n</tt></tag>

  sets the length of a listing page in lines. See the <tt><ref
  id=".PAGELENGTH" name=".PAGELENGTH"></tt> directive for more information.


  <label id="option-s">
  <tag><tt>-s, --smart-mode</tt></tag>

  In smart mode (enabled by -s or the <tt><ref id=".SMART" name=".SMART"></tt>
  pseudo instruction) the assembler will track usage of the <tt/REP/ and
  <tt/SEP/ instructions in 65816 mode and update the operand sizes
  accordingly. If the operand of such an instruction cannot be evaluated by
  the assembler (for example, because the operand is an imported symbol), a
  warning is issued.

  Beware: Since the assembler cannot trace the execution flow this may
  lead to false results in some cases. If in doubt, use the .ixx and .axx
  instructions to tell the assembler about the current settings. Smart
  mode is off by default.


  <label id="option-t">
  <tag><tt>-t sys, --target sys</tt></tag>

  Set the target system. This will enable translation of character strings
  and character constants into the character set of the target platform.
  The default for the target system is "none", which means that no translation
  will take place. The assembler supports the same target systems as the
  compiler, see there for a list.


  <label id="option-v">
  <tag><tt>-v, --verbose</tt></tag>

  Increase the assembler verbosity. Usually only needed for debugging
  purposes. You may use this option more than one time for even more
  verbose output.


  <label id="option-D">
  <tag><tt>-D</tt></tag>

  This option allows you to define symbols on the command line. Without a
  value, the symbol is defined with the value zero. When giving a value,
  you may use the '&dollar;' prefix for hexadecimal symbols. Please note
  that for some operating systems, '&dollar;' has a special meaning, so
  you may have to quote the expression.


  <label id="option-I">
  <tag><tt>-I dir, --include-dir dir</tt></tag>

  Name a directory which is searched for include files. The option may be
  used more than once to specify more than one directory to search. The
  current directory is always searched first before considering any
  additional directories. See also the section about <ref id="search-paths"
  name="search paths">.


  <label id="option-U">
  <tag><tt>-U, --auto-import</tt></tag>

  Mark symbols that are not defined in the sources as imported symbols. This
  should be used with care since it delays error messages about typos and such
  until the linker is run. The compiler uses the equivalent of this switch
  (<tt><ref id=".AUTOIMPORT" name=".AUTOIMPORT"></tt>) to enable auto imported
  symbols for the runtime library. However, the compiler is supposed to
  generate code that runs through the assembler without problems, something
  which is not always true for assembler programmers.


  <label id="option-V">
  <tag><tt>-V, --version</tt></tag>

  Print the version number of the assembler. If you send any suggestions
  or bugfixes, please include the version number.


  <label id="option-W">
  <tag><tt>-Wn</tt></tag>

  Set the warning level for the assembler. Using -W2 the assembler will
  even warn about such things like unused imported symbols. The default
  warning level is 1, and it would probably be silly to set it to
  something lower.

</descrip>
<p>



<sect>Search paths<label id="search-paths"><p>

Normal include files are searched in the following places:

<enum>
<item>The current directory.
<item>A compiled-in directory, which is often <tt>/usr/lib/cc65/asminc</tt>
      on Linux systems.
<item>The value of the environment variable <tt/CA65_INC/ if it is defined.
<item>A subdirectory named <tt/asminc/ of the directory defined in the
      environment variable <tt/CC65_HOME/, if it is defined.
<item>Any directory added with the <tt/<ref id="option-I" name="-I">/ option
on the command line.
</enum>

Binary include files are searched in the following places:

<enum>
<item>The current directory.
<item>Any directory added with the <tt/<ref id="option--bin-include-dir"
name="--bin-include-dir">/ option on the command line.
</enum>



<sect>Input format<p>

<sect1>Assembler syntax<p>

The assembler accepts the standard 6502/65816 assembler syntax. One line may
contain a label (which is identified by a colon), and, in addition to the
label, an assembler mnemonic, a macro, or a control command (see section <ref
id="control-commands" name="Control Commands"> for supported control
commands). Alternatively, the line may contain a symbol definition using
the '=' token. Everything after a semicolon is handled as a comment (that is,
it is ignored).

Here are some examples for valid input lines:

<tscreen><verb>
        Label:                          ; A label and a comment
                lda     #$20            ; A 6502 instruction plus comment
        L1:     ldx     #$20            ; Same with label
        L2:     .byte   "Hello world"   ; Label plus control command
                mymac   $20             ; Macro expansion
                MySym = 3*L1            ; Symbol definition
        MaSym   = Label                 ; Another symbol
</verb></tscreen>

The assembler accepts

<itemize>
<item>all valid 6502 mnemonics when in 6502 mode (the default or after the
      <tt><ref id=".P02" name=".P02"></tt> command was given).
<item>all valid 6502 mnemonics plus a set of illegal instructions when in
      <ref id="6502X-mode" name="6502X mode">.
<item>all valid 65SC02 mnemonics when in 65SC02 mode (after the
      <tt><ref id=".PSC02" name=".PSC02"></tt> command was given).
<item>all valid 65C02 mnemonics when in 65C02 mode (after the
      <tt><ref id=".PC02" name=".PC02"></tt> command was given).
<item>all valid 65618 mnemonics when in 65816 mode (after the
      <tt><ref id=".P816" name=".P816"></tt> command was given).
<item>all valid SunPlus mnemonics when in SunPlus mode (after the
      <tt><ref id=".SUNPLUS" name=".SUNPLUS"></tt> command was given).
</itemize>


<sect1>65816 mode<p>

In 65816 mode several aliases are accepted in addition to the official
mnemonics:

<tscreen><verb>
        BGE is an alias for BCS
        BLT is an alias for BCC
        CPA is an alias for CMP
        DEA is an alias for DEC A
        INA is an alias for INC A
        SWA is an alias for XBA
        TAD is an alias for TCD
        TAS is an alias for TCS
        TDA is an alias for TDC
        TSA is an alias for TSC
</verb></tscreen>



<sect1>6502X mode<label id="6502X-mode"><p>

6502X mode is an extension to the normal 6502 mode. In this mode, several
mnemonics for illegal instructions of the NMOS 6502 CPUs are accepted. Since
these instructions are illegal, there are no official mnemonics for them. The
unofficial ones are taken from <htmlurl
url="http://oxyron.net/graham/opcodes02.html"
name="http://oxyron.net/graham/opcodes02.html">. Please note that only the
ones marked as "stable" are supported. The following table uses information
from the mentioned web page, for more information, see there.

<itemize>
<item><tt>ALR: A:=(A and #{imm})*2;</tt>
<item><tt>ANC: A:=A and #{imm};</tt> Generates opcode &dollar;0B.
<item><tt>ARR: A:=(A and #{imm})/2;</tt>
<item><tt>AXS: X:=A and X-#{imm};</tt>
<item><tt>DCP: {adr}:={adr}-1; A-{adr};</tt>
<item><tt>ISC: {adr}:={adr}+1; A:=A-{adr};</tt>
<item><tt>LAS: A,X,S:={adr} and S;</tt>
<item><tt>LAX: A,X:={adr};</tt>
<item><tt>RLA: {adr}:={adr}rol; A:=A and {adr};</tt>
<item><tt>RRA: {adr}:={adr}ror; A:=A adc {adr};</tt>
<item><tt>SAX: {adr}:=A and X;</tt>
<item><tt>SLO: {adr}:={adr}*2; A:=A or {adr};</tt>
<item><tt>SRE: {adr}:={adr}/2; A:=A xor {adr};</tt>
</itemize>



<sect1>sweet16 mode<label id="sweet16-mode"><p>

SWEET 16 is an interpreter for a pseudo 16 bit CPU written by Steve Wozniak
for the Apple ][ machines. It is available in the Apple ][ ROM. ca65 can
generate code for this pseudo CPU when switched into sweet16 mode. The
following is special in sweet16 mode:

<itemize>

<item>The '@' character denotes indirect addressing and is no longer available
for cheap local labels. If you need cheap local labels, you will have to
switch to another lead character using the <tt/<ref id=".LOCALCHAR"
name=".LOCALCHAR">/ command.

<item>Registers are specified using <tt/R0/ .. <tt/R15/. In sweet16 mode,
these identifiers are reserved words.

</itemize>

Please note that the assembler does neither supply the interpreter needed for
SWEET 16 code, nor the zero page locations needed for the SWEET 16 registers,
nor does it call the interpreter. All this must be done by your program. Apple
][ programmers do probably know how to use sweet16 mode.

For more information about SWEET 16, see
<htmlurl url="http://www.6502.org/source/interpreters/sweet16.htm"
name="http://www.6502.org/source/interpreters/sweet16.htm">.


<sect1>Number format<p>

For literal values, the assembler accepts the widely used number formats: A
preceding '&dollar;' or a trailing 'h' denotes a hex value, a preceding '%'
denotes a binary value, and a bare number is interpreted as a decimal. There
are currently no octal values and no floats.


<sect1>Conditional assembly<p>

Please note that when using the conditional directives (<tt/.IF/ and friends),
the input must consist of valid assembler tokens, even in <tt/.IF/ branches
that are not assembled. The reason for this behaviour is that the assembler
must still be able to detect the ending tokens (like <tt/.ENDIF/), so
conversion of the input stream into tokens still takes place. As a consequence
conditional assembly directives may <bf/not/ be used to prevent normal text
(used as a comment or similar) from being assembled. <p>


<sect>Expressions<p>


<sect1>Expression evaluation<p>

All expressions are evaluated with (at least) 32 bit precision. An
expression may contain constant values and any combination of internal and
external symbols. Expressions that cannot be evaluated at assembly time
are stored inside the object file for evaluation by the linker.
Expressions referencing imported symbols must always be evaluated by the
linker.


<sect1>Size of an expression result<p>

Sometimes, the assembler must know about the size of the value that is the
result of an expression. This is usually the case, if a decision has to be
made, to generate a zero page or an absolute memory references. In this
case, the assembler has to make some assumptions about the result of an
expression:

<itemize>
<item>  If the result of an expression is constant, the actual value is
        checked to see if it's a byte sized expression or not.
<item>  If the expression is explicitly casted to a byte sized expression by
        one of the '&gt;', '&lt;' or '^' operators, it is a byte expression.
<item>  If this is not the case, and the expression contains a symbol,
        explicitly declared as zero page symbol (by one of the .importzp or
        .exportzp instructions), then the whole expression is assumed to be
        byte sized.
<item>  If the expression contains symbols that are not defined, and these
        symbols are local symbols, the enclosing scopes are searched for a
        symbol with the same name. If one exists and this symbol is defined,
        it's attributes are used to determine the result size.
<item>  In all other cases the expression is assumed to be word sized.
</itemize>

Note: If the assembler is not able to evaluate the expression at assembly
time, the linker will evaluate it and check for range errors as soon as
the result is known.


<sect1>Boolean expressions<p>

In the context of a boolean expression, any non zero value is evaluated as
true, any other value to false. The result of a boolean expression is 1 if
it's true, and zero if it's false. There are boolean operators with extreme
low precedence with version 2.x (where x &gt; 0). The <tt/.AND/ and <tt/.OR/
operators are shortcut operators. That is, if the result of the expression is
already known, after evaluating the left hand side, the right hand side is
not evaluated.


<sect1>Constant expressions<p>

Sometimes an expression must evaluate to a constant without looking at any
further input. One such example is the <tt/<ref id=".IF" name=".IF">/ command
that decides if parts of the code are assembled or not. An expression used in
the <tt/.IF/ command cannot reference a symbol defined later, because the
decision about the <tt/.IF/ must be made at the point when it is read. If the
expression used in such a context contains only constant numerical values,
there is no problem. When unresolvable symbols are involved it may get harder
for the assembler to determine if the expression is actually constant, and it
is even possible to create expressions that aren't recognized as constant.
Simplifying the expressions will often help.

In cases where the result of the expression is not needed immediately, the
assembler will delay evaluation until all input is read, at which point all
symbols are known. So using arbitrary complex constant expressions is no
problem in most cases.



<sect1>Available operators<label id="operators"><p>

<table>
<tabular ca="clc">
<bf/Operator/| <bf/Description/| <bf/Precedence/@<hline>
| Built-in string functions| 0@
||~@
| Built-in pseudo-variables| 1@
| Built-in pseudo-functions| 1@
+| Unary positive| 1@
-| Unary negative| 1@
&tilde;<newline>
.BITNOT| Unary bitwise not| 1@
&lt;<newline>
.LOBYTE| Unary low-byte operator| 1@
&gt;<newline>
.HIBYTE| Unary high-byte operator| 1@
^<newline>
.BANKBYTE| Unary bank-byte operator| 1@
||~@
*| Multiplication| 2@
/| Division| 2@
.MOD| Modulo operator| 2@
&amp;<newline>
.BITAND| Bitwise and| 2@
^<newline>
.BITXOR| Binary bitwise xor| 2@
&lt;&lt;<newline>
.SHL| Shift-left operator| 2@
&gt;&gt;<newline>
.SHR| Shift-right operator| 2@
||~@
+| Binary addition| 3@
-| Binary subtraction| 3@
&verbar;<newline>
.BITOR| Bitwise or| 3@
||~@
= | Compare operator (equal)| 4@
&lt;&gt;| Compare operator (not equal)| 4@
&lt;| Compare operator (less)| 4@
&gt;| Compare operator (greater)| 4@
&lt;=| Compare operator (less or equal)| 4@
&gt;=| Compare operator (greater or equal)| 4@
||~@
&amp;&amp;<newline>
.AND| Boolean and| 5@
.XOR| Boolean xor| 5@
||~@
&verbar;&verbar;<newline>
.OR| Boolean or| 6@
||~@
!<newline>
.NOT| Boolean not| 7@<hline>
</tabular>
<caption>Available operators, sorted by precedence
</table>

To force a specific order of evaluation, parentheses may be used, as usual.



<sect>Symbols and labels<p>

A symbol or label is an identifier that starts with a letter and is followed
by letters and digits. Depending on some features enabled (see
<tt><ref id="at_in_identifiers" name="at_in_identifiers"></tt>,
<tt><ref id="dollar_in_identifiers" name="dollar_in_identifiers"></tt> and
<tt><ref id="leading_dot_in_identifiers" name="leading_dot_in_identifiers"></tt>)
other characters may be present. Use of identifiers consisting of a single
character will not work in all cases, because some of these identifiers are
reserved keywords (for example "A" is not a valid identifier for a label,
because it is the keyword for the accumulator).

The assembler allows you to use symbols instead of naked values to make
the source more readable. There are a lot of different ways to define and
use symbols and labels, giving a lot of flexibility.

<sect1>Numeric constants<p>

Numeric constants are defined using the equal sign or the label assignment
operator. After doing

<tscreen><verb>
      two = 2
</verb></tscreen>

may use the symbol "two" in every place where a number is expected, and it is
evaluated to the value 2 in this context. The label assignment operator causes
the same, but causes the symbol to be marked as a label, which may cause a
different handling in the debugger:

<tscreen><verb>
      io := $d000
</verb></tscreen>

The right side can of course be an expression:

<tscreen><verb>
      four = two * two
</verb></tscreen>


<sect1>Standard labels<p>

A label is defined by writing the name of the label at the start of the line
(before any instruction mnemonic, macro or pseudo directive), followed by a
colon. This will declare a symbol with the given name and the value of the
current program counter.


<sect1>Local labels and symbols<p>

Using the <tt><ref id=".PROC" name=".PROC"></tt> directive, it is possible to
create regions of code where the names of labels and symbols are local to this
region. They are not known outside of this region and cannot be accessed from
there. Such regions may be nested like PROCEDUREs in Pascal.

See the description of the <tt><ref id=".PROC" name=".PROC"></tt>
directive for more information.


<sect1>Cheap local labels<p>

Cheap local labels are defined like standard labels, but the name of the
label must begin with a special symbol (usually '@', but this can be
changed by the <tt><ref id=".LOCALCHAR" name=".LOCALCHAR"></tt>
directive).

Cheap local labels are visible only between two non cheap labels. As soon as a
standard symbol is encountered (this may also be a local symbol if inside a
region defined with the <tt><ref id=".PROC" name=".PROC"></tt> directive), the
cheap local symbol goes out of scope.

You may use cheap local labels as an easy way to reuse common label
names like "Loop". Here is an example:

<tscreen><verb>
        Clear:  lda    #$00             ; Global label
                ldy    #$20
        @Loop:  sta    Mem,y            ; Local label
                dey
                bne    @Loop            ; Ok
                rts
        Sub:    ...                     ; New global label
                bne    @Loop            ; ERROR: Unknown identifier!
</verb></tscreen>

<sect1>Unnamed labels<p>

If you really want to write messy code, there are also unnamed labels. These
labels do not have a name (you guessed that already, didn't you?). A colon is
used to mark the absence of the name.

Unnamed labels may be accessed by using the colon plus several minus or plus
characters as a label designator. Using the '-' characters will create a back
reference (use the n'th label backwards), using '+' will create a forward
reference (use the n'th label in forward direction). An example will help to
understand this:

<tscreen><verb>
        :       lda     (ptr1),y        ; #1
                cmp     (ptr2),y
                bne     :+              ; -> #2
                tax
                beq     :+++            ; -> #4
                iny
                bne     :-              ; -> #1
                inc     ptr1+1
                inc     ptr2+1
                bne     :-              ; -> #1

        :       bcs     :+              ; #2 -> #3
                ldx     #$FF
                rts

        :       ldx     #$01            ; #3
        :       rts                     ; #4
</verb></tscreen>

As you can see from the example, unnamed labels will make even short
sections of code hard to understand, because you have to count labels
to find branch targets (this is the reason why I for my part do
prefer the "cheap" local labels). Nevertheless, unnamed labels are
convenient in some situations, so it's your decision.


<sect1>Using macros to define labels and constants<p>

While there are drawbacks with this approach, it may be handy in some
situations. Using <tt><ref id=".DEFINE" name=".DEFINE"></tt>, it is
possible to define symbols or constants that may be used elsewhere. Since
the macro facility works on a very low level, there is no scoping. On the
other side, you may also define string constants this way (this is not
possible with the other symbol types).

Example:

<tscreen><verb>
        .DEFINE two     2
        .DEFINE version "SOS V2.3"

        four = two * two        ; Ok
        .byte   version         ; Ok

        .PROC                   ; Start local scope
        two = 3                 ; Will give "2 = 3" - invalid!
        .ENDPROC
</verb></tscreen>


<sect1>Symbols and <tt>.DEBUGINFO</tt><p>

If <tt><ref id=".DEBUGINFO" name=".DEBUGINFO"></tt> is enabled (or <ref
id="option-g" name="-g"> is given on the command line), global, local and
cheap local labels are written to the object file and will be available in the
symbol file via the linker. Unnamed labels are not written to the object file,
because they don't have a name which would allow to access them.



<sect>Scopes<label id="scopes"><p>

ca65 implements several sorts of scopes for symbols.

<sect1>Global scope<p>

All (non cheap local) symbols that are declared outside of any nested scopes
are in global scope.


<sect1>Cheap locals<p>

A special scope is the scope for cheap local symbols. It lasts from one non
local symbol to the next one, without any provisions made by the programmer.
All other scopes differ in usage but use the same concept internally.


<sect1>Generic nested scopes<p>

A nested scoped for generic use is started with <tt/<ref id=".SCOPE"
name=".SCOPE">/ and closed with <tt/<ref id=".ENDSCOPE" name=".ENDSCOPE">/.
The scope can have a name, in which case it is accessible from the outside by
using <ref id="scopesyntax" name="explicit scopes">. If the scope does not
have a name, all symbols created within the scope are local to the scope, and
aren't accessible from the outside.

A nested scope can access symbols from the local or from enclosing scopes by
name without using explicit scope names. In some cases there may be
ambiguities, for example if there is a reference to a local symbol that is not
yet defined, but a symbol with the same name exists in outer scopes:

<tscreen><verb>
        .scope  outer
                foo     = 2
                .scope  inner
                        lda     #foo
                        foo     = 3
                .endscope
        .endscope
</verb></tscreen>

In the example above, the <tt/lda/ instruction will load the value 3 into the
accumulator, because <tt/foo/ is redefined in the scope. However:

<tscreen><verb>
        .scope  outer
                foo     = $1234
                .scope  inner
                        lda     foo,x
                        foo     = $12
                .endscope
        .endscope
</verb></tscreen>

Here, <tt/lda/ will still load from <tt/$12,x/, but since it is unknown to the
assembler that <tt/foo/ is a zeropage symbol when translating the instruction,
absolute mode is used instead. In fact, the assembler will not use absolute
mode by default, but it will search through the enclosing scopes for a symbol
with the given name. If one is found, the address size of this symbol is used.
This may lead to errors:

<tscreen><verb>
        .scope  outer
                foo     = $12
                .scope  inner
                        lda     foo,x
                        foo     = $1234
                .endscope
        .endscope
</verb></tscreen>

In this case, when the assembler sees the symbol <tt/foo/ in the <tt/lda/
instruction, it will search for an already defined symbol <tt/foo/. It will
find <tt/foo/ in scope <tt/outer/, and a close look reveals that it is a
zeropage symbol. So the assembler will use zeropage addressing mode. If
<tt/foo/ is redefined later in scope <tt/inner/, the assembler tries to change
the address in the <tt/lda/ instruction already translated, but since the new
value needs absolute addressing mode, this fails, and an error message "Range
error" is output.

Of course the most simple solution for the problem is to move the definition
of <tt/foo/ in scope <tt/inner/ upwards, so it precedes its use. There may be
rare cases when this cannot be done. In these cases, you can use one of the
address size override operators:

<tscreen><verb>
        .scope  outer
                foo     = $12
                .scope  inner
                        lda     a:foo,x
                        foo     = $1234
                .endscope
        .endscope
</verb></tscreen>

This will cause the <tt/lda/ instruction to be translated using absolute
addressing mode, which means changing the symbol reference later does not
cause any errors.


<sect1>Nested procedures<p>

A nested procedure is created by use of <tt/<ref id=".PROC" name=".PROC">/. It
differs from a <tt/<ref id=".SCOPE" name=".SCOPE">/ in that it must have a
name, and a it will introduce a symbol with this name in the enclosing scope.
So

<tscreen><verb>
        .proc   foo
                ...
        .endproc
</verb></tscreen>

is actually the same as

<tscreen><verb>
        foo:
        .scope  foo
                ...
        .endscope
</verb></tscreen>

This is the reason why a procedure must have a name. If you want a scope
without a name, use <tt/<ref id=".SCOPE" name=".SCOPE">/.

<bf/Note:/ As you can see from the example above, scopes and symbols live in
different namespaces. There can be a symbol named <tt/foo/ and a scope named
<tt/foo/ without any conflicts (but see the section titled <ref
id="scopesearch" name="&quot;Scope search order&quot;">).


<sect1>Structs, unions and enums<p>

Structs, unions and enums are explained in a <ref id="structs" name="separate
section">, I do only cover them here, because if they are declared with a
name, they open a nested scope, similar to <tt/<ref id=".SCOPE"
name=".SCOPE">/. However, when no name is specified, the behaviour is
different: In this case, no new scope will be opened, symbols declared within
a struct, union, or enum declaration will then be added to the enclosing scope
instead.


<sect1>Explicit scope specification<label id="scopesyntax"><p>

Accessing symbols from other scopes is possible by using an explicit scope
specification, provided that the scope where the symbol lives in has a name.
The namespace token (<tt/::/) is used to access other scopes:

<tscreen><verb>
        .scope  foo
        bar:    .word   0
        .endscope

                ...
                lda     foo::bar        ; Access foo in scope bar
</verb></tscreen>

The only way to deny access to a scope from the outside is to declare a scope
without a name (using the <tt/<ref id=".SCOPE" name=".SCOPE">/ command).

A special syntax is used to specify the global scope: If a symbol or scope is
preceded by the namespace token, the global scope is searched:

<tscreen><verb>
        bar     = 3

        .scope  foo
                bar     = 2
                lda     #::bar  ; Access the global bar (which is 3)
        .endscope
</verb></tscreen>


<sect1>Scope search order<label id="scopesearch"><p>

The assembler searches for a scope in a similar way as for a symbol. First, it
looks in the current scope, and then it walks up the enclosing scopes until
the scope is found.

However, one important thing to note when using explicit scope syntax is, that
a symbol may be accessed before it is defined, but a scope may <bf/not/ be
used without a preceding definition. This means that in the following
example:

<tscreen><verb>
        .scope  foo
                bar     = 3
        .endscope

        .scope  outer
                lda     #foo::bar  ; Will load 3, not 2!
                .scope  foo
                        bar     = 2
                .endscope
        .endscope
</verb></tscreen>

the reference to the scope <tt/foo/ will use the global scope, and not the
local one, because the local one is not visible at the point where it is
referenced.

Things get more complex if a complete chain of scopes is specified:

<tscreen><verb>
        .scope  foo
                .scope  outer
                        .scope  inner
                                bar = 1
                        .endscope
                .endscope
                .scope  another
                        .scope  nested
                                lda     #outer::inner::bar      ; 1
                        .endscope
                .endscope
        .endscope

        .scope  outer
                .scope  inner
                        bar = 2
                .endscope
        .endscope
</verb></tscreen>

When <tt/outer::inner::bar/ is referenced in the <tt/lda/ instruction, the
assembler will first search in the local scope for a scope named <tt/outer/.
Since none is found, the enclosing scope (<tt/another/) is checked. There is
still no scope named <tt/outer/, so scope <tt/foo/ is checked, and finally
scope <tt/outer/ is found. Within this scope, <tt/inner/ is searched, and in
this scope, the assembler looks for a symbol named <tt/bar/.

Please note that once the anchor scope is found, all following scopes
(<tt/inner/ in this case) are expected to be found exactly in this scope. The
assembler will search the scope tree only for the first scope (if it is not
anchored in the root scope). Starting from there on, there is no flexibility,
so if the scope named <tt/outer/ found by the assembler does not contain a
scope named <tt/inner/, this would be an error, even if such a pair does exist
(one level up in global scope).

Ambiguities that may be introduced by this search algorithm may be removed by
anchoring the scope specification in the global scope. In the example above,
if you want to access the "other" symbol <tt/bar/, you would have to write:

<tscreen><verb>
        .scope  foo
                .scope  outer
                        .scope  inner
                                bar = 1
                        .endscope
                .endscope
                .scope  another
                        .scope  nested
                                lda     #::outer::inner::bar    ; 2
                        .endscope
                .endscope
        .endscope

        .scope  outer
                .scope  inner
                        bar = 2
                .endscope
        .endscope
</verb></tscreen>


<sect>Address sizes and memory models<label id="address-sizes"><p>

<sect1>Address sizes<p>

ca65 assigns each segment and each symbol an address size. This is true, even
if the symbol is not used as an address. You may also think of a value range
of the symbol instead of an address size.

Possible address sizes are:

<itemize>
<item>Zeropage or direct (8 bits)
<item>Absolute (16 bits)
<item>Far (24 bits)
<item>Long (32 bits)
</itemize>

Since the assembler uses default address sizes for the segments and symbols,
it is usually not necessary to override the default behaviour. In cases, where
it is necessary, the following keywords may be used to specify address sizes:

<itemize>
<item>DIRECT, ZEROPAGE or ZP for zeropage addressing (8 bits).
<item>ABSOLUTE, ABS or NEAR for absolute addressing (16 bits).
<item>FAR for far addressing (24 bits).
<item>LONG or DWORD for long addressing (32 bits).
</itemize>


<sect1>Address sizes of segments<p>

The assembler assigns an address size to each segment. Since the
representation of a label within this segment is "segment start + offset",
labels will inherit the address size of the segment they are declared in.

The address size of a segment may be changed, by using an optional address
size modifier. See the <tt/<ref id=".SEGMENT" name="segment directive">/ for
an explanation on how this is done.


<sect1>Address sizes of symbols<p>




<sect1>Memory models<p>

The default address size of a segment depends on the memory model used. Since
labels inherit the address size from the segment they are declared in,
changing the memory model is an easy way to change the address size of many
symbols at once.




<sect>Pseudo variables<label id="pseudo-variables"><p>

Pseudo variables are readable in all cases, and in some special cases also
writable.

<sect1><tt>*</tt><p>

  Reading this pseudo variable will return the program counter at the start
  of the current input line.

  Assignment to this variable is possible when <tt/<ref id=".FEATURE"
  name=".FEATURE pc_assignment">/ is used. Note: You should not use
  assignments to <tt/*/, use <tt/<ref id=".ORG" name=".ORG">/ instead.


<sect1><tt>.CPU</tt><label id=".CPU"><p>

  Reading this pseudo variable will give a constant integer value that
  tells which CPU is currently enabled. It can also tell which instruction
  set the CPU is able to translate. The value read from the pseudo variable
  should be further examined by using one of the constants defined by the
  "cpu" macro package (see <tt/<ref id=".MACPACK" name=".MACPACK">/).

  It may be used to replace the .IFPxx pseudo instructions or to construct
  even more complex expressions.

  Example:

  <tscreen><verb>
        .macpack        cpu
        .if     (.cpu .bitand CPU_ISET_65816)
                phx
                phy
        .else
                txa
                pha
                tya
                pha
        .endif
  </verb></tscreen>


<sect1><tt>.PARAMCOUNT</tt><label id=".PARAMCOUNT"><p>

  This builtin pseudo variable is only available in macros. It is replaced by
  the actual number of parameters that were given in the macro invocation.

  Example:

  <tscreen><verb>
        .macro  foo     arg1, arg2, arg3
        .if     .paramcount <> 3
        .error  "Too few parameters for macro foo"
        .endif
        ...
        .endmacro
  </verb></tscreen>

  See section <ref id="macros" name="Macros">.


<sect1><tt>.TIME</tt><label id=".TIME"><p>

  Reading this pseudo variable will give a constant integer value that
  represents the current time in POSIX standard (as seconds since the
  Epoch).

  It may be used to encode the time of translation somewhere in the created
  code.

  Example:

  <tscreen><verb>
        .dword  .time   ; Place time here
  </verb></tscreen>


<sect1><tt>.VERSION</tt><label id=".VERSION"><p>

  Reading this pseudo variable will give the assembler version according to
  the following formula:

        VER_MAJOR*$100 + VER_MINOR*$10 + VER_PATCH

  It may be used to encode the assembler version or check the assembler for
  special features not available with older versions.

  Example:

  Version 2.11.1 of the assembler will return $2B1 as numerical constant when
  reading the pseudo variable <tt/.VERSION/.



<sect>Pseudo functions<label id="pseudo-functions"><p>

Pseudo functions expect their arguments in parenthesis, and they have a result,
either a string or an expression.


<sect1><tt>.BANKBYTE</tt><label id=".BANKBYTE"><p>

  The function returns the bank byte (that is, bits 16-23) of its argument.
  It works identical to the '^' operator.

  See: <tt><ref id=".HIBYTE" name=".HIBYTE"></tt>,
       <tt><ref id=".LOBYTE" name=".LOBYTE"></tt>


<sect1><tt>.BLANK</tt><label id=".BLANK"><p>

  Builtin function. The function evaluates its argument in braces and yields
  "false" if the argument is non blank (there is an argument), and "true" if
  there is no argument.  The token list that makes up the function argument
  may optionally be enclosed in curly braces. This allows the inclusion of
  tokens that would otherwise terminate the list (the closing right
  parenthesis). The curly braces are not considered part of the list, a list
  just consisting of curly braces is considered to be empty.

  As an example, the <tt/.IFBLANK/ statement may be replaced by

  <tscreen><verb>
        .if     .blank({arg})
  </verb></tscreen>



<sect1><tt>.CONCAT</tt><label id=".CONCAT"><p>

  Builtin string function. The function allows to concatenate a list of string
  constants separated by commas. The result is a string constant that is the
  concatenation of all arguments. This function is most useful in macros and
  when used together with the <tt/.STRING/ builtin function. The function may
  be used in any case where a string constant is expected.

  Example:

  <tscreen><verb>
        .include        .concat ("myheader", ".", "inc")
  </verb></tscreen>

  This is the same as the command

  <tscreen><verb>
        .include        "myheader.inc"
  </verb></tscreen>


<sect1><tt>.CONST</tt><label id=".CONST"><p>

  Builtin function. The function evaluates its argument in braces and
  yields "true" if the argument is a constant expression (that is, an
  expression that yields a constant value at assembly time) and "false"
  otherwise. As an example, the .IFCONST statement may be replaced by

  <tscreen><verb>
        .if     .const(a + 3)
  </verb></tscreen>


<sect1><tt>.HIBYTE</tt><label id=".HIBYTE"><p>

  The function returns the high byte (that is, bits 8-15) of its argument.
  It works identical to the '>' operator.

  See: <tt><ref id=".LOBYTE" name=".LOBYTE"></tt>,
       <tt><ref id=".BANKBYTE" name=".BANKBYTE"></tt>


<sect1><tt>.HIWORD</tt><label id=".HIWORD"><p>

  The function returns the high word (that is, bits 16-31) of its argument.

  See: <tt><ref id=".LOWORD" name=".LOWORD"></tt>


<sect1><tt>.IDENT</tt><label id=".IDENT"><p>

  The function expects a string as its argument, and converts this argument
  into an identifier. If the string starts with the current <tt/<ref
  id=".LOCALCHAR" name=".LOCALCHAR">/, it will be converted into a cheap local
  identifier, otherwise it will be converted into a normal identifier.

  Example:

  <tscreen><verb>
        .macro  makelabel       arg1, arg2
                .ident (.concat (arg1, arg2)):
        .endmacro

                makelabel       "foo", "bar"

                .word           foobar          ; Valid label
  </verb></tscreen>


<sect1><tt>.LEFT</tt><label id=".LEFT"><p>

  Builtin function. Extracts the left part of a given token list.

  Syntax:

  <tscreen><verb>
        .LEFT (&lt;int expr&gt;, &lt;token list&gt;)
  </verb></tscreen>

  The first integer expression gives the number of tokens to extract from
  the token list. The second argument is the token list itself. The token
  list may optionally be enclosed into curly braces. This allows the
  inclusion of tokens that would otherwise terminate the list (the closing
  right paren in the given case).

  Example:

  To check in a macro if the given argument has a '#' as first token
  (immediate addressing mode), use something like this:

  <tscreen><verb>
        .macro  ldax    arg
                ...
                .if (.match (.left (1, {arg}), #))

                ; ldax called with immediate operand
                ...

                .endif
                ...
        .endmacro
  </verb></tscreen>

  See also the <tt><ref id=".MID" name=".MID"></tt> and <tt><ref id=".RIGHT"
  name=".RIGHT"></tt> builtin functions.


<sect1><tt>.LOBYTE</tt><label id=".LOBYTE"><p>

  The function returns the low byte (that is, bits 0-7) of its argument.
  It works identical to the '<' operator.

  See: <tt><ref id=".HIBYTE" name=".HIBYTE"></tt>,
       <tt><ref id=".BANKBYTE" name=".BANKBYTE"></tt>


<sect1><tt>.LOWORD</tt><label id=".LOWORD"><p>

  The function returns the low word (that is, bits 0-15) of its argument.

  See: <tt><ref id=".HIWORD" name=".HIWORD"></tt>


<sect1><tt>.MATCH</tt><label id=".MATCH"><p>

  Builtin function. Matches two token lists against each other. This is
  most useful within macros, since macros are not stored as strings, but
  as lists of tokens.

  The syntax is

  <tscreen><verb>
        .MATCH(&lt;token list #1&gt;, &lt;token list #2&gt;)
  </verb></tscreen>

  Both token list may contain arbitrary tokens with the exception of the
  terminator token (comma resp. right parenthesis) and

  <itemize>
  <item>end-of-line
  <item>end-of-file
  </itemize>

  The token lists may optionally be enclosed into curly braces. This allows
  the inclusion of tokens that would otherwise terminate the list (the closing
  right paren in the given case). Often a macro parameter is used for any of
  the token lists.

  Please note that the function does only compare tokens, not token
  attributes. So any number is equal to any other number, regardless of the
  actual value. The same is true for strings. If you need to compare tokens
  <em/and/ token attributes, use the <tt><ref id=".XMATCH"
  name=".XMATCH"></tt> function.

  Example:

  Assume the macro <tt/ASR/, that will shift right the accumulator by one,
  while honoring the sign bit. The builtin processor instructions will allow
  an optional "A" for accu addressing for instructions like <tt/ROL/ and
  <tt/ROR/. We will use the <tt><ref id=".MATCH" name=".MATCH"></tt> function
  to check for this and print and error for invalid calls.

  <tscreen><verb>
        .macro  asr     arg

                .if (.not .blank(arg)) .and (.not .match ({arg}, a))
                .error "Syntax error"
                .endif

                cmp     #$80            ; Bit 7 into carry
                lsr     a               ; Shift carry into bit 7

        .endmacro
  </verb></tscreen>

  The macro will only accept no arguments, or one argument that must be the
  reserved keyword "A".

  See: <tt><ref id=".XMATCH" name=".XMATCH"></tt>


<sect1><tt>.MAX</tt><label id=".MAX"><p>

  Builtin function. The result is the larger of two values.

  The syntax is

  <tscreen><verb>
        .MAX (&lt;value #1&gt;, &lt;value #2&gt;)
  </verb></tscreen>

  Example:

  <tscreen><verb>
        ; Reserve space for the larger of two data blocks
        savearea:       .max (.sizeof (foo), .sizeof (bar))
  </verb></tscreen>

  See: <tt><ref id=".MIN" name=".MIN"></tt>


<sect1><tt>.MID</tt><label id=".MID"><p>

  Builtin function. Takes a starting index, a count and a token list as
  arguments. Will return part of the token list.

  Syntax:

  <tscreen><verb>
        .MID (&lt;int expr&gt;, &lt;int expr&gt;, &lt;token list&gt;)
  </verb></tscreen>

  The first integer expression gives the starting token in the list (the first
  token has index 0). The second integer expression gives the number of tokens
  to extract from the token list. The third argument is the token list itself.
  The token list may optionally be enclosed into curly braces. This allows the
  inclusion of tokens that would otherwise terminate the list (the closing
  right paren in the given case).

  Example:

  To check in a macro if the given argument has a '<tt/#/' as first token
  (immediate addressing mode), use something like this:

    <tscreen><verb>
        .macro  ldax    arg
                ...
                .if (.match (.mid (0, 1, {arg}), #))

                ; ldax called with immediate operand
                ...

                .endif
                ...
        .endmacro
  </verb></tscreen>

  See also the <tt><ref id=".LEFT" name=".LEFT"></tt> and <tt><ref id=".RIGHT"
  name=".RIGHT"></tt> builtin functions.


<sect1><tt>.MIN</tt><label id=".MIN"><p>

  Builtin function. The result is the smaller of two values.

  The syntax is

  <tscreen><verb>
        .MIN (&lt;value #1&gt;, &lt;value #2&gt;)
  </verb></tscreen>

  Example:

  <tscreen><verb>
        ; Reserve space for some data, but 256 bytes minimum
        savearea:       .min (.sizeof (foo), 256)
  </verb></tscreen>

  See: <tt><ref id=".MAX" name=".MAX"></tt>


<sect1><tt>.REF, .REFERENCED</tt><label id=".REFERENCED"><p>

  Builtin function. The function expects an identifier as argument in braces.
  The argument is evaluated, and the function yields "true" if the identifier
  is a symbol that has already been referenced somewhere in the source file up
  to the current position. Otherwise the function yields false. As an example,
  the <tt><ref id=".IFREF" name=".IFREF"></tt> statement may be replaced by

  <tscreen><verb>
        .if     .referenced(a)
  </verb></tscreen>

  See: <tt><ref id=".DEFINED" name=".DEFINED"></tt>


<sect1><tt>.RIGHT</tt><label id=".RIGHT"><p>

  Builtin function. Extracts the right part of a given token list.

  Syntax:

  <tscreen><verb>
        .RIGHT (&lt;int expr&gt;, &lt;token list&gt;)
  </verb></tscreen>

  The first integer expression gives the number of tokens to extract from the
  token list. The second argument is the token list itself.  The token list
  may optionally be enclosed into curly braces. This allows the inclusion of
  tokens that would otherwise terminate the list (the closing right paren in
  the given case).

  See also the <tt><ref id=".LEFT" name=".LEFT"></tt> and <tt><ref id=".MID"
  name=".MID"></tt> builtin functions.


<sect1><tt>.SIZEOF</tt><label id=".SIZEOF"><p>

  <tt/.SIZEOF/ is a pseudo function that returns the size of its argument. The
  argument can be a struct/union, a struct member, a procedure, or a label. In
  case of a procedure or label, its size is defined by the amount of data
  placed in the segment where the label is relative to. If a line of code
  switches segments (for example in a macro) data placed in other segments
  does not count for the size.

  Please note that a symbol or scope must exist, before it is used together with
  <tt/.SIZEOF/ (this may get relaxed later, but will always be true for scopes).
  A scope has preference over a symbol with the same name, so if the last part
  of a name represents both, a scope and a symbol, the scope is chosen over the
  symbol.

  After the following code:

  <tscreen><verb>
        .struct Point                   ; Struct size = 4
                xcoord  .word
                xcoord  .word
        .endstruct

        P:      .tag    Point           ; Declare a point
        @P:     .tag    Point           ; Declare another point

        .code
        .proc   Code
                nop
                .proc   Inner
                        nop
                .endproc
                nop
        .endproc

        .proc   Data
        .data                           ; Segment switch!!!
                .res    4
        .endproc
  </verb></tscreen>

  <descrip>
    <tag><tt/.sizeof(Point)/</tag>
    will have the value 4, because this is the size of struct <tt/Point/.

    <tag><tt/.sizeof(Point::xcoord)/</tag>
    will have the value 2, because this is the size of the member <tt/xcoord/
    in struct <tt/Point/.

    <tag><tt/.sizeof(P)/</tag>
    will have the value 4, this is the size of the data declared on the same
    source line as the label <tt/P/, which is in the same segment that <tt/P/
    is relative to.

    <tag><tt/.sizeof(@P)/</tag>
    will have the value 4, see above. The example demonstrates that <tt/.SIZEOF/
    does also work for cheap local symbols.

    <tag><tt/.sizeof(Code)/</tag>
    will have the value 3, since this is amount of data emitted into the code
    segment, the segment that was active when <tt/Code/ was entered. Note that
    this value includes the amount of data emitted in child scopes (in this
    case <tt/Code::Inner/).

    <tag><tt/.sizeof(Code::Inner)/</tag>
    will have the value 1 as expected.

    <tag><tt/.sizeof(Data)/</tag>
    will have the value 0. Data is emitted within the scope <tt/Data/, but since
    the segment is switched after entry, this data is emitted into another
    segment.
  </descrip>


<sect1><tt>.STRAT</tt><label id=".STRAT"><p>

  Builtin function. The function accepts a string and an index as
  arguments and returns the value of the character at the given position
  as an integer value. The index is zero based.

  Example:

  <tscreen><verb>
        .macro  M       Arg
                ; Check if the argument string starts with '#'
                .if (.strat (Arg, 0) = '#')
                ...
                .endif
        .endmacro
  </verb></tscreen>


<sect1><tt>.SPRINTF</tt><label id=".SPRINTF"><p>

  Builtin function. It expects a format string as first argument. The number
  and type of the following arguments depend on the format string. The format
  string is similar to the one of the C <tt/printf/ function. Missing things
  are: Length modifiers, variable width.

  The result of the function is a string.

  Example:

  <tscreen><verb>
        num     = 3

        ; Generate an identifier:
        .ident (.sprintf ("%s%03d", "label", num)):
  </verb></tscreen>


<sect1><tt>.STRING</tt><label id=".STRING"><p>

  Builtin function. The function accepts an argument in braces and converts
  this argument into a string constant. The argument may be an identifier, or
  a constant numeric value.

  Since you can use a string in the first place, the use of the function may
  not be obvious. However, it is useful in macros, or more complex setups.

  Example:

  <tscreen><verb>
        ; Emulate other assemblers:
        .macro  section name
                .segment        .string(name)
        .endmacro
  </verb></tscreen>


<sect1><tt>.STRLEN</tt><label id=".STRLEN"><p>

  Builtin function. The function accepts a string argument in braces and
  evaluates to the length of the string.

  Example:

  The following macro encodes a string as a pascal style string with
  a leading length byte.

  <tscreen><verb>
        .macro  PString Arg
                .byte   .strlen(Arg), Arg
        .endmacro
  </verb></tscreen>


<sect1><tt>.TCOUNT</tt><label id=".TCOUNT"><p>

  Builtin function. The function accepts a token list in braces. The function
  result is the number of tokens given as argument. The token list may
  optionally be enclosed into curly braces which are not considered part of
  the list and not counted. Enclosement in curly braces allows the inclusion
  of tokens that would otherwise terminate the list (the closing right paren
  in the given case).

  Example:

  The <tt/ldax/ macro accepts the '#' token to denote immediate addressing (as
  with the normal 6502 instructions). To translate it into two separate 8 bit
  load instructions, the '#' token has to get stripped from the argument:

  <tscreen><verb>
        .macro  ldax    arg
                .if (.match (.mid (0, 1, {arg}), #))
                ; ldax called with immediate operand
                lda     #<(.right (.tcount ({arg})-1, {arg}))
                ldx     #>(.right (.tcount ({arg})-1, {arg}))
                .else
                ...
                .endif
        .endmacro
  </verb></tscreen>


<sect1><tt>.XMATCH</tt><label id=".XMATCH"><p>

  Builtin function. Matches two token lists against each other. This is
  most useful within macros, since macros are not stored as strings, but
  as lists of tokens.

  The syntax is

  <tscreen><verb>
        .XMATCH(&lt;token list #1&gt;, &lt;token list #2&gt;)
  </verb></tscreen>

  Both token list may contain arbitrary tokens with the exception of the
  terminator token (comma resp. right parenthesis) and

  <itemize>
  <item>end-of-line
  <item>end-of-file
  </itemize>

  The token lists may optionally be enclosed into curly braces. This allows
  the inclusion of tokens that would otherwise terminate the list (the closing
  right paren in the given case). Often a macro parameter is used for any of
  the token lists.

  The function compares tokens <em/and/ token values. If you need a function
  that just compares the type of tokens, have a look at the <tt><ref
  id=".MATCH" name=".MATCH"></tt> function.

  See: <tt><ref id=".MATCH" name=".MATCH"></tt>



<sect>Control commands<label id="control-commands"><p>

Here's a list of all control commands and a description, what they do:


<sect1><tt>.A16</tt><label id=".A16"><p>

  Valid only in 65816 mode. Switch the accumulator to 16 bit.

  Note: This command will not emit any code, it will tell the assembler to
  create 16 bit operands for immediate accumulator addressing mode.

  See also: <tt><ref id=".SMART" name=".SMART"></tt>


<sect1><tt>.A8</tt><label id=".A8"><p>

  Valid only in 65816 mode. Switch the accumulator to 8 bit.

  Note: This command will not emit any code, it will tell the assembler to
  create 8 bit operands for immediate accu addressing mode.

  See also: <tt><ref id=".SMART" name=".SMART"></tt>


<sect1><tt>.ADDR</tt><label id=".ADDR"><p>

  Define word sized data. In 6502 mode, this is an alias for <tt/.WORD/ and
  may be used for better readability if the data words are address values. In
  65816 mode, the address is forced to be 16 bit wide to fit into the current
  segment. See also <tt><ref id=".FARADDR" name=".FARADDR"></tt>. The command
  must be followed by a sequence of (not necessarily constant) expressions.

  Example:

  <tscreen><verb>
        .addr   $0D00, $AF13, _Clear
  </verb></tscreen>

  See: <tt><ref id=".FARADDR" name=".FARADDR"></tt>, <tt><ref id=".WORD"
       name=".WORD"></tt>


<sect1><tt>.ALIGN</tt><label id=".ALIGN"><p>

  Align data to a given boundary. The command expects a constant integer
  argument that must be a power of two, plus an optional second argument
  in byte range. If there is a second argument, it is used as fill value,
  otherwise the value defined in the linker configuration file is used
  (the default for this value is zero).

  Since alignment depends on the base address of the module, you must
  give the same (or a greater) alignment for the segment when linking.
  The linker will give you a warning, if you don't do that.

  Example:

  <tscreen><verb>
        .align  256
  </verb></tscreen>


<sect1><tt>.ASCIIZ</tt><label id=".ASCIIZ"><p>

  Define a string with a trailing zero.

  Example:

  <tscreen><verb>
        Msg:    .asciiz "Hello world"
  </verb></tscreen>

  This will put the string "Hello world" followed by a binary zero into
  the current segment. There may be more strings separated by commas, but
  the binary zero is only appended once (after the last one).


<sect1><tt>.ASSERT</tt><label id=".ASSERT"><p>

  Add an assertion. The command is followed by an expression, an action
  specifier, and an optional message that is output in case the assertion
  fails. If no message was given, the string "Assertion failed" is used. The
  action specifier may be one of <tt/warning/, <tt/error/, <tt/ldwarning/ or
  <tt/lderror/. In the former two cases, the assertion is evaluated by the
  assembler if possible, and in any case, it's also passed to the linker in
  the object file (if one is generated). The linker will then evaluate the
  expression when segment placement has been done.

  Example:

  <tscreen><verb>
        .assert         * = $8000, error, "Code not at $8000"
  </verb></tscreen>

  The example assertion will check that the current location is at $8000,
  when the output file is written, and abort with an error if this is not
  the case. More complex expressions are possible. The action specifier
  <tt/warning/ outputs a warning, while the <tt/error/ specifier outputs
  an error message. In the latter case, generation of the output file is
  suppressed in both the assembler and linker.


<sect1><tt>.AUTOIMPORT</tt><label id=".AUTOIMPORT"><p>

  Is followed by a plus or a minus character. When switched on (using a
  +), undefined symbols are automatically marked as import instead of
  giving errors. When switched off (which is the default so this does not
  make much sense), this does not happen and an error message is
  displayed. The state of the autoimport flag is evaluated when the
  complete source was translated, before outputting actual code, so it is
  <em/not/ possible to switch this feature on or off for separate sections
  of code. The last setting is used for all symbols.

  You should probably not use this switch because it delays error
  messages about undefined symbols until the link stage. The cc65
  compiler (which is supposed to produce correct assembler code in all
  circumstances, something which is not true for most assembler
  programmers) will insert this command to avoid importing each and every
  routine from the runtime library.

  Example:

  <tscreen><verb>
        .autoimport     +       ; Switch on auto import
  </verb></tscreen>

<sect1><tt>.BANKBYTES</tt><label id=".BANKBYTES"><p>

  Define byte sized data by extracting only the bank byte (that is, bits 16-23) from
  each expression.  This is equivalent to <tt><ref id=".BYTE" name=".BYTE"></tt> with
  the operator '^' prepended to each expression in its list.

  Example:

  <tscreen><verb>
        .define MyTable TableItem0, TableItem1, TableItem2, TableItem3

        TableLookupLo:   .lobytes   MyTable
        TableLookupHi:   .hibytes   MyTable
        TableLookupBank: .bankbytes MyTable
  </verb></tscreen>

  which is equivalent to

  <tscreen><verb>
        TableLookupLo:   .byte &lt;TableItem0, &lt;TableItem1, &lt;TableItem2, &lt;TableItem3
        TableLookupHi:   .byte &gt;TableItem0, &gt;TableItem1, &gt;TableItem2, &gt;TableItem3
        TableLookupBank: .byte ^TableItem0, ^TableItem1, ^TableItem2, ^TableItem3
  </verb></tscreen>

  See also: <tt><ref id=".BYTE" name=".BYTE"></tt>,
            <tt><ref id=".HIBYTES" name=".HIBYTES"></tt>,
            <tt><ref id=".LOBYTES" name=".LOBYTES"></tt>


<sect1><tt>.BSS</tt><label id=".BSS"><p>

  Switch to the BSS segment. The name of the BSS segment is always "BSS",
  so this is a shortcut for

  <tscreen><verb>
        .segment  "BSS"
  </verb></tscreen>

  See also the <tt><ref id=".SEGMENT" name=".SEGMENT"></tt> command.


<sect1><tt>.BYT, .BYTE</tt><label id=".BYTE"><p>

  Define byte sized data. Must be followed by a sequence of (byte ranged)
  expressions or strings.

  Example:

  <tscreen><verb>
        .byte   "Hello "
        .byt    "world", $0D, $00
  </verb></tscreen>


<sect1><tt>.CASE</tt><label id=".CASE"><p>

  Switch on or off case sensitivity on identifiers. The default is off
  (that is, identifiers are case sensitive), but may be changed by the
  -i switch on the command line.
  The command must be followed by a '+' or '-' character to switch the
  option on or off respectively.

  Example:

  <tscreen><verb>
        .case   -               ; Identifiers are not case sensitive
  </verb></tscreen>


<sect1><tt>.CHARMAP</tt><label id=".CHARMAP"><p>

  Apply a custom mapping for characters. The command is followed by two
  numbers in the range 1..255. The first one is the index of the source
  character, the second one is the mapping. The mapping applies to all
  character and string constants when they generate output, and overrides
  a mapping table specified with the <tt><ref id="option-t" name="-t"></tt>
  command line switch.

  Example:

  <tscreen><verb>
        .charmap        $41, $61        ; Map 'A' to 'a'
  </verb></tscreen>


<sect1><tt>.CODE</tt><label id=".CODE"><p>

  Switch to the CODE segment. The name of the CODE segment is always
  "CODE", so this is a shortcut for

  <tscreen><verb>
        .segment  "CODE"
  </verb></tscreen>

  See also the <tt><ref id=".SEGMENT" name=".SEGMENT"></tt> command.


<sect1><tt>.CONDES</tt><label id=".CONDES"><p>

  Export a symbol and mark it in a special way. The linker is able to build
  tables of all such symbols. This may be used to automatically create a list
  of functions needed to initialize linked library modules.

  Note: The linker has a feature to build a table of marked routines, but it
  is your code that must call these routines, so just declaring a symbol with
  <tt/.CONDES/ does nothing by itself.

  All symbols are exported as an absolute (16 bit) symbol. You don't need to
  use an additional <tt><ref id=".EXPORT" name=".EXPORT"></tt> statement, this
  is implied by <tt/.CONDES/.

  <tt/.CONDES/ is followed by the type, which may be <tt/constructor/,
  <tt/destructor/ or a numeric value between 0 and 6 (where 0 is the same as
  specifying <tt/constructor/ and 1 is equal to specifying <tt/destructor/).
  The <tt><ref id=".CONSTRUCTOR" name=".CONSTRUCTOR"></tt>, <tt><ref
  id=".DESTRUCTOR" name=".DESTRUCTOR"></tt> and <tt><ref id=".INTERRUPTOR"
  name=".INTERRUPTOR"></tt> commands are actually shortcuts for <tt/.CONDES/
  with a type of <tt/constructor/ resp. <tt/destructor/ or <tt/interruptor/.

  After the type, an optional priority may be specified. Higher numeric values
  mean higher priority. If no priority is given, the default priority of 7 is
  used. Be careful when assigning priorities to your own module constructors
  so they won't interfere with the ones in the cc65 library.

  Example:

  <tscreen><verb>
        .condes         ModuleInit, constructor
        .condes         ModInit, 0, 16
  </verb></tscreen>

  See the <tt><ref id=".CONSTRUCTOR" name=".CONSTRUCTOR"></tt>, <tt><ref
  id=".DESTRUCTOR" name=".DESTRUCTOR"></tt> and <tt><ref id=".INTERRUPTOR"
  name=".INTERRUPTOR"></tt> commands and the separate section <ref id="condes"
  name="Module constructors/destructors"> explaining the feature in more
  detail.


<sect1><tt>.CONSTRUCTOR</tt><label id=".CONSTRUCTOR"><p>

  Export a symbol and mark it as a module constructor. This may be used
  together with the linker to build a table of constructor subroutines that
  are called by the startup code.

  Note: The linker has a feature to build a table of marked routines, but it
  is your code that must call these routines, so just declaring a symbol as
  constructor does nothing by itself.

  A constructor is always exported as an absolute (16 bit) symbol. You don't
  need to use an additional <tt/.export/ statement, this is implied by
  <tt/.constructor/. It may have an optional priority that is separated by a
  comma. Higher numeric values mean a higher priority. If no priority is
  given, the default priority of 7 is used. Be careful when assigning
  priorities to your own module constructors so they won't interfere with the
  ones in the cc65 library.

  Example:

  <tscreen><verb>
        .constructor    ModuleInit
        .constructor    ModInit, 16
  </verb></tscreen>

  See the <tt><ref id=".CONDES" name=".CONDES"></tt> and <tt><ref
  id=".DESTRUCTOR" name=".DESTRUCTOR"></tt> commands and the separate section
  <ref id="condes" name="Module constructors/destructors"> explaining the
  feature in more detail.


<sect1><tt>.DATA</tt><label id=".DATA"><p>

  Switch to the DATA segment. The name of the DATA segment is always
  "DATA", so this is a shortcut for

  <tscreen><verb>
        .segment  "DATA"
  </verb></tscreen>

  See also the <tt><ref id=".SEGMENT" name=".SEGMENT"></tt> command.


<sect1><tt>.DBYT</tt><label id=".DBYT"><p>

  Define word sized data with the hi and lo bytes swapped (use <tt/.WORD/ to
  create word sized data in native 65XX format). Must be followed by a
  sequence of (word ranged) expressions.

  Example:

  <tscreen><verb>
        .dbyt   $1234, $4512
  </verb></tscreen>

  This will emit the bytes

  <tscreen><verb>
        $12 $34 $45 $12
  </verb></tscreen>

  into the current segment in that order.


<sect1><tt>.DEBUGINFO</tt><label id=".DEBUGINFO"><p>

  Switch on or off debug info generation. The default is off (that is,
  the object file will not contain debug infos), but may be changed by the
  -g switch on the command line.
  The command must be followed by a '+' or '-' character to switch the
  option on or off respectively.

  Example:

  <tscreen><verb>
        .debuginfo      +       ; Generate debug info
  </verb></tscreen>


<sect1><tt>.DEFINE</tt><label id=".DEFINE"><p>

  Start a define style macro definition. The command is followed by an
  identifier (the macro name) and optionally by a list of formal arguments
  in braces.
  See section <ref id="macros" name="Macros">.


<sect1><tt>.DEF, .DEFINED</tt><label id=".DEFINED"><p>

  Builtin function. The function expects an identifier as argument in braces.
  The argument is evaluated, and the function yields "true" if the identifier
  is a symbol that is already defined somewhere in the source file up to the
  current position. Otherwise the function yields false. As an example, the
  <tt><ref id=".IFDEF" name=".IFDEF"></tt> statement may be replaced by

  <tscreen><verb>
        .if     .defined(a)
  </verb></tscreen>


<sect1><tt>.DESTRUCTOR</tt><label id=".DESTRUCTOR"><p>

  Export a symbol and mark it as a module destructor. This may be used
  together with the linker to build a table of destructor subroutines that
  are called by the startup code.

  Note: The linker has a feature to build a table of marked routines, but it
  is your code that must call these routines, so just declaring a symbol as
  constructor does nothing by itself.

  A destructor is always exported as an absolute (16 bit) symbol. You don't
  need to use an additional <tt/.export/ statement, this is implied by
  <tt/.destructor/. It may have an optional priority that is separated by a
  comma. Higher numerical values mean a higher priority. If no priority is
  given, the default priority of 7 is used. Be careful when assigning
  priorities to your own module destructors so they won't interfere with the
  ones in the cc65 library.

  Example:

  <tscreen><verb>
        .destructor     ModuleDone
        .destructor     ModDone, 16
  </verb></tscreen>

  See the <tt><ref id=".CONDES" name=".CONDES"></tt> and <tt><ref
  id=".CONSTRUCTOR" name=".CONSTRUCTOR"></tt> commands and the separate
  section <ref id="condes" name="Module constructors/destructors"> explaining
  the feature in more detail.


<sect1><tt>.DWORD</tt><label id=".DWORD"><p>

  Define dword sized data (4 bytes) Must be followed by a sequence of
  expressions.

  Example:

  <tscreen><verb>
        .dword  $12344512, $12FA489
  </verb></tscreen>


<sect1><tt>.ELSE</tt><label id=".ELSE"><p>

  Conditional assembly: Reverse the current condition.


<sect1><tt>.ELSEIF</tt><label id=".ELSEIF"><p>

  Conditional assembly: Reverse current condition and test a new one.


<sect1><tt>.END</tt><label id=".END"><p>

  Forced end of assembly. Assembly stops at this point, even if the command
  is read from an include file.


<sect1><tt>.ENDENUM</tt><label id=".ENDENUM"><p>

  End a <tt><ref id=".ENUM" name=".ENUM"></tt> declaration.


<sect1><tt>.ENDIF</tt><label id=".ENDIF"><p>

  Conditional assembly: Close a <tt><ref id=".IF" name=".IF..."></tt> or
  <tt><ref id=".ELSE" name=".ELSE"></tt> branch.


<sect1><tt>.ENDMAC, .ENDMACRO</tt><label id=".ENDMACRO"><p>

  End of macro definition (see section <ref id="macros" name="Macros">).


<sect1><tt>.ENDPROC</tt><label id=".ENDPROC"><p>

  End of local lexical level (see <tt><ref id=".PROC" name=".PROC"></tt>).


<sect1><tt>.ENDREP, .ENDREPEAT</tt><label id=".ENDREPEAT"><p>

  End a <tt><ref id=".REPEAT" name=".REPEAT"></tt> block.


<sect1><tt>.ENDSCOPE</tt><label id=".ENDSCOPE"><p>

  End of local lexical level (see <tt/<ref id=".SCOPE" name=".SCOPE">/).


<sect1><tt>.ENDSTRUCT</tt><label id=".ENDSTRUCT"><p>

  Ends a struct definition. See the <tt/<ref id=".STRUCT" name=".STRUCT">/
  command and the separate section named <ref id="structs" name="&quot;Structs
  and unions&quot;">.


<sect1><tt>.ENUM</tt><label id=".ENUM"><p>

  Start an enumeration. This directive is very similar to the C <tt/enum/
  keyword. If a name is given, a new scope is created for the enumeration,
  otherwise the enumeration members are placed in the enclosing scope.

  In the enumeration body, symbols are declared. The first symbol has a value
  of zero, and each following symbol will get the value of the preceding plus
  one. This behaviour may be overridden by an explicit assignment. Two symbols
  may have the same value.

  Example:

  <tscreen><verb>
        .enum   errorcodes
                no_error
                file_error
                parse_error
        .endenum
  </verb></tscreen>

  Above example will create a new scope named <tt/errorcodes/ with three
  symbols in it that get the values 0, 1 and 2 respectively. Another way
  to write this would have been:

  <tscreen><verb>
        .scope  errorcodes
                no_error        = 0
                file_error      = 1
                parse_error     = 2
        .endscope
  </verb></tscreen>

  Please note that explicit scoping must be used to access the identifiers:

  <tscreen><verb>
        .word   errorcodes::no_error
  </verb></tscreen>

  A more complex example:

  <tscreen><verb>
        .enum
                EUNKNOWN        = -1
                EOK
                EFILE
                EBUSY
                EAGAIN
                EWOULDBLOCK     = EAGAIN
        .endenum
  </verb></tscreen>

  In this example, the enumeration does not have a name, which means that the
  members will be visible in the enclosing scope and can be used in this scope
  without explicit scoping. The first member (<tt/EUNKNOWN/) has the value -1.
  The value for the following members is incremented by one, so <tt/EOK/ would
  be zero and so on. <tt/EWOULDBLOCK/ is an alias for <tt/EGAIN/, so it has an
  override for the value using an already defined symbol.


<sect1><tt>.ERROR</tt><label id=".ERROR"><p>

  Force an assembly error. The assembler will output an error message
  preceded by "User error" and will <em/not/ produce an object file.

  This command may be used to check for initial conditions that must be
  set before assembling a source file.

  Example:

  <tscreen><verb>
        .if     foo = 1
        ...
        .elseif bar = 1
        ...
        .else
        .error  "Must define foo or bar!"
        .endif
  </verb></tscreen>

  See also the <tt><ref id=".WARNING" name=".WARNING"></tt> and <tt><ref
  id=".OUT" name=".OUT"></tt> directives.


<sect1><tt>.EXITMAC, .EXITMACRO</tt><label id=".EXITMACRO"><p>

  Abort a macro expansion immediately. This command is often useful in
  recursive macros. See separate section <ref id="macros" name="Macros">.


<sect1><tt>.EXPORT</tt><label id=".EXPORT"><p>

  Make symbols accessible from other modules. Must be followed by a comma
  separated list of symbols to export, with each one optionally followed by an
  address specification and (also optional) an assignment. Using an additional
  assignment in the export statement allows to define and export a symbol in
  one statement. The default is to export the symbol with the address size it
  actually has. The assembler will issue a warning, if the symbol is exported
  with an address size smaller than the actual address size.

  Examples:

  <tscreen><verb>
        .export foo
        .export bar: far
        .export foobar: far = foo * bar
        .export baz := foobar, zap: far = baz - bar
  </verb></tscreen>

  As with constant definitions, using <tt/:=/ instead of <tt/=/ marks the
  symbols as a label.

  See: <tt><ref id=".EXPORTZP" name=".EXPORTZP"></tt>


<sect1><tt>.EXPORTZP</tt><label id=".EXPORTZP"><p>

  Make symbols accessible from other modules. Must be followed by a comma
  separated list of symbols to export. The exported symbols are explicitly
  marked as zero page symbols. An assignment may be included in the
  <tt/.EXPORTZP/ statement. This allows to define and export a symbol in one
  statement.

  Examples:

  <tscreen><verb>
        .exportzp  foo, bar
        .exportzp  baz := &dollar;02
  </verb></tscreen>

  See: <tt><ref id=".EXPORT" name=".EXPORT"></tt>


<sect1><tt>.FARADDR</tt><label id=".FARADDR"><p>

  Define far (24 bit) address data. The command must be followed by a
  sequence of (not necessarily constant) expressions.

  Example:

  <tscreen><verb>
        .faraddr        DrawCircle, DrawRectangle, DrawHexagon
  </verb></tscreen>

  See: <tt><ref id=".ADDR" name=".ADDR"></tt>


<sect1><tt>.FEATURE</tt><label id=".FEATURE"><p>

  This directive may be used to enable one or more compatibility features
  of the assembler. While the use of <tt/.FEATURE/ should be avoided when
  possible, it may be useful when porting sources written for other
  assemblers. There is no way to switch a feature off, once you have
  enabled it, so using

  <tscreen><verb>
        .FEATURE        xxx
  </verb></tscreen>

  will enable the feature until end of assembly is reached.

  The following features are available:

  <descrip>

  <tag><tt>at_in_identifiers</tt><label id="at_in_identifiers"></tag>

    Accept the at character (`@') as a valid character in identifiers. The
    at character is not allowed to start an identifier, even with this
    feature enabled.

  <tag><tt>c_comments</tt></tag>

    Allow C like comments using <tt>/*</tt> and <tt>*/</tt> as left and right
    comment terminators. Note that C comments may not be nested. There's also a
    pitfall when using C like comments: All statements must be terminated by
    "end-of-line". Using C like comments, it is possible to hide the newline,
    which results in error messages. See the following non working example:

    <tscreen><verb>
        lda     #$00  /* This comment hides the newline
*/      sta     $82
    </verb></tscreen>

  <tag><tt>dollar_in_identifiers</tt><label id="dollar_in_identifiers"></tag>

    Accept the dollar sign (`&dollar;') as a valid character in identifiers. The
    dollar character is not allowed to start an identifier, even with this
    feature enabled.

  <tag><tt>dollar_is_pc</tt></tag>

    The dollar sign may be used as an alias for the star (`*'), which
    gives the value of the current PC in expressions.
    Note: Assignment to the pseudo variable is not allowed.

  <tag><tt>labels_without_colons</tt></tag>

    Allow labels without a trailing colon. These labels are only accepted,
    if they start at the beginning of a line (no leading white space).

  <tag><tt>leading_dot_in_identifiers</tt><label id="leading_dot_in_identifiers"></tag>

    Accept the dot (`.') as the first character of an identifier. This may be
    used for example to create macro names that start with a dot emulating
    control directives of other assemblers. Note however, that none of the
    reserved keywords built into the assembler, that starts with a dot, may be
    overridden. When using this feature, you may also get into trouble if
    later versions of the assembler define new keywords starting with a dot.

  <tag><tt>loose_char_term</tt></tag>

    Accept single quotes as well as double quotes as terminators for char
    constants.

  <tag><tt>loose_string_term</tt></tag>

    Accept single quotes as well as double quotes as terminators for string
    constants.

  <tag><tt>missing_char_term</tt></tag>

    Accept single quoted character constants where the terminating quote is
    missing.
    <tscreen><verb>
        lda     #'a
    </verb></tscreen>
    <bf/Note:/ This does not work in conjunction with <tt/.FEATURE
    loose_string_term/, since in this case the input would be ambiguous.

  <tag><tt>org_per_seg</tt><label id="org_per_seg"></tag>

    This feature makes relocatable/absolute mode local to the current segment.
    Using <tt><ref id=".ORG" name=".ORG"></tt> when <tt/org_per_seg/ is in
    effect will only enable absolute mode for the current segment. Dito for
    <tt><ref id=".RELOC" name=".RELOC"></tt>.

  <tag><tt>pc_assignment</tt></tag>

    Allow assignments to the PC symbol (`*' or `&dollar;' if <tt/dollar_is_pc/
    is enabled). Such an assignment is handled identical to the <tt><ref
    id=".ORG" name=".ORG"></tt> command (which is usually not needed, so just
    removing the lines with the assignments may also be an option when porting
    code written for older assemblers).

  <tag><tt>ubiquitous_idents</tt></tag>

    Allow the use of instructions names as names for macros and symbols. This
    makes it possible to "overload" instructions by defining a macro with the
    same name. This does also make it possible to introduce hard to find errors
    in your code, so be careful!

  </descrip>

  It is also possible to specify features on the command line using the
  <tt><ref id="option--feature" name="--feature"></tt> command line option.
  This is useful when translating sources written for older assemblers, when
  you don't want to change the source code.

  As an example, to translate sources written for Andre Fachats xa65
  assembler, the features

  <verb>
        labels_without_colons, pc_assignment, loose_char_term
  </verb>

  may be helpful. They do not make ca65 completely compatible, so you may not
  be able to translate the sources without changes, even when enabling these
  features. However, I have found several sources that translate without
  problems when enabling these features on the command line.


<sect1><tt>.FILEOPT, .FOPT</tt><label id=".FOPT"><p>

  Insert an option string into the object file. There are two forms of
  this command, one specifies the option by a keyword, the second
  specifies it as a number. Since usage of the second one needs knowledge
  of the internal encoding, its use is not recommended and I will only
  describe the first form here.

  The command is followed by one of the keywords

  <tscreen><verb>
        author
        comment
        compiler
  </verb></tscreen>

  a comma and a string. The option is written into the object file
  together with the string value. This is currently unidirectional and
  there is no way to actually use these options once they are in the
  object file.

  Examples:

  <tscreen><verb>
        .fileopt        comment, "Code stolen from my brother"
        .fileopt        compiler, "BASIC 2.0"
        .fopt           author, "J. R. User"
  </verb></tscreen>


<sect1><tt>.FORCEIMPORT</tt><label id=".FORCEIMPORT"><p>

  Import an absolute symbol from another module. The command is followed by a
  comma separated list of symbols to import. The command is similar to <tt>
  <ref id=".IMPORT" name=".IMPORT"></tt>, but the import reference is always
  written to the generated object file, even if the symbol is never referenced
  (<tt><ref id=".IMPORT" name=".IMPORT"></tt> will not generate import
  references for unused symbols).

  Example:

  <tscreen><verb>
        .forceimport    needthisone, needthistoo
  </verb></tscreen>

  See: <tt><ref id=".IMPORT" name=".IMPORT"></tt>


<sect1><tt>.GLOBAL</tt><label id=".GLOBAL"><p>

  Declare symbols as global. Must be followed by a comma separated list of
  symbols to declare. Symbols from the list, that are defined somewhere in the
  source, are exported, all others are imported. Additional <tt><ref
  id=".IMPORT" name=".IMPORT"></tt> or <tt><ref id=".EXPORT"
  name=".EXPORT"></tt> commands for the same symbol are allowed.

  Example:

  <tscreen><verb>
        .global foo, bar
  </verb></tscreen>


<sect1><tt>.GLOBALZP</tt><label id=".GLOBALZP"><p>

  Declare symbols as global. Must be followed by a comma separated list of
  symbols to declare. Symbols from the list, that are defined somewhere in the
  source, are exported, all others are imported. Additional <tt><ref
  id=".IMPORTZP" name=".IMPORTZP"></tt> or <tt><ref id=".EXPORTZP"
  name=".EXPORTZP"></tt> commands for the same symbol are allowed. The symbols
  in the list are explicitly marked as zero page symbols.

  Example:

  <tscreen><verb>
        .globalzp foo, bar
  </verb></tscreen>

<sect1><tt>.HIBYTES</tt><label id=".HIBYTES"><p>

  Define byte sized data by extracting only the high byte (that is, bits 8-15) from
  each expression.  This is equivalent to <tt><ref id=".BYTE" name=".BYTE"></tt> with
  the operator '>' prepended to each expression in its list.

  Example:

  <tscreen><verb>
        .lobytes         $1234, $2345, $3456, $4567
        .hibytes         $fedc, $edcb, $dcba, $cba9
  </verb></tscreen>

  which is equivalent to

  <tscreen><verb>
        .byte            $34, $45, $56, $67
        .byte            $fe, $ed, $dc, $cb
  </verb></tscreen>

  Example:

  <tscreen><verb>
        .define MyTable TableItem0, TableItem1, TableItem2, TableItem3

        TableLookupLo:   .lobytes MyTable
        TableLookupHi:   .hibytes MyTable
  </verb></tscreen>

  which is equivalent to

  <tscreen><verb>
        TableLookupLo:   .byte &lt;TableItem0, &lt;TableItem1, &lt;TableItem2, &lt;TableItem3
        TableLookupHi:   .byte &gt;TableItem0, &gt;TableItem1, &gt;TableItem2, &gt;TableItem3
  </verb></tscreen>

  See also: <tt><ref id=".BYTE" name=".BYTE"></tt>,
            <tt><ref id=".LOBYTES" name=".LOBYTES"></tt>,
            <tt><ref id=".BANKBYTES" name=".BANKBYTES"></tt>


<sect1><tt>.I16</tt><label id=".I16"><p>

  Valid only in 65816 mode. Switch the index registers to 16 bit.

  Note: This command will not emit any code, it will tell the assembler to
  create 16 bit operands for immediate operands.

  See also the <tt><ref id=".I8" name=".I8"></tt> and <tt><ref id=".SMART"
  name=".SMART"></tt> commands.


<sect1><tt>.I8</tt><label id=".I8"><p>

  Valid only in 65816 mode. Switch the index registers to 8 bit.

  Note: This command will not emit any code, it will tell the assembler to
  create 8 bit operands for immediate operands.

  See also the <tt><ref id=".I16" name=".I16"></tt> and <tt><ref id=".SMART"
  name=".SMART"></tt> commands.


<sect1><tt>.IF</tt><label id=".IF"><p>

  Conditional assembly: Evaluate an expression and switch assembler output
  on or off depending on the expression. The expression must be a constant
  expression, that is, all operands must be defined.

  A expression value of zero evaluates to FALSE, any other value evaluates
  to TRUE.


<sect1><tt>.IFBLANK</tt><label id=".IFBLANK"><p>

  Conditional assembly: Check if there are any remaining tokens in this line,
  and evaluate to FALSE if this is the case, and to TRUE otherwise. If the
  condition is not true, further lines are not assembled until an <tt><ref
  id=".ELSE" name=".ESLE"></tt>, <tt><ref id=".ELSEIF" name=".ELSEIF"></tt> or
  <tt><ref id=".ENDIF" name=".ENDIF"></tt> directive.

  This command is often used to check if a macro parameter was given. Since an
  empty macro parameter will evaluate to nothing, the condition will evaluate
  to FALSE if an empty parameter was given.

  Example:

  <tscreen><verb>
        .macro     arg1, arg2
        .ifblank   arg2
                   lda     #arg1
        .else
                   lda     #arg2
        .endif
        .endmacro
  </verb></tscreen>

  See also: <tt><ref id=".BLANK" name=".BLANK"></tt>


<sect1><tt>.IFCONST</tt><label id=".IFCONST"><p>

  Conditional assembly: Evaluate an expression and switch assembler output
  on or off depending on the constness of the expression.

  A const expression evaluates to to TRUE, a non const expression (one
  containing an imported or currently undefined symbol) evaluates to
  FALSE.

  See also: <tt><ref id=".CONST" name=".CONST"></tt>


<sect1><tt>.IFDEF</tt><label id=".IFDEF"><p>

  Conditional assembly: Check if a symbol is defined. Must be followed by
  a symbol name. The condition is true if the the given symbol is already
  defined, and false otherwise.

  See also: <tt><ref id=".DEFINED" name=".DEFINED"></tt>


<sect1><tt>.IFNBLANK</tt><label id=".IFNBLANK"><p>

  Conditional assembly: Check if there are any remaining tokens in this line,
  and evaluate to TRUE if this is the case, and to FALSE otherwise. If the
  condition is not true, further lines are not assembled until an <tt><ref
  id=".ELSE" name=".ELSE"></tt>, <tt><ref id=".ELSEIF" name=".ELSEIF"></tt> or
  <tt><ref id=".ENDIF" name=".ENDIF"></tt> directive.

  This command is often used to check if a macro parameter was given.
  Since an empty macro parameter will evaluate to nothing, the condition
  will evaluate to FALSE if an empty parameter was given.

  Example:

  <tscreen><verb>
        .macro     arg1, arg2
                   lda     #arg1
        .ifnblank  arg2
                   lda     #arg2
        .endif
        .endmacro
  </verb></tscreen>

  See also: <tt><ref id=".BLANK" name=".BLANK"></tt>


<sect1><tt>.IFNDEF</tt><label id=".IFNDEF"><p>

  Conditional assembly: Check if a symbol is defined. Must be followed by
  a symbol name. The condition is true if the the given symbol is not
  defined, and false otherwise.

  See also: <tt><ref id=".DEFINED" name=".DEFINED"></tt>


<sect1><tt>.IFNREF</tt><label id=".IFNREF"><p>

  Conditional assembly: Check if a symbol is referenced. Must be followed
  by a symbol name. The condition is true if if the the given symbol was
  not referenced before, and false otherwise.

  See also: <tt><ref id=".REFERENCED" name=".REFERENCED"></tt>


<sect1><tt>.IFP02</tt><label id=".IFP02"><p>

  Conditional assembly: Check if the assembler is currently in 6502 mode
  (see <tt><ref id=".P02" name=".P02"></tt> command).


<sect1><tt>.IFP816</tt><label id=".IFP816"><p>

  Conditional assembly: Check if the assembler is currently in 65816 mode
  (see <tt><ref id=".P816" name=".P816"></tt> command).


<sect1><tt>.IFPC02</tt><label id=".IFPC02"><p>

  Conditional assembly: Check if the assembler is currently in 65C02 mode
  (see <tt><ref id=".PC02" name=".PC02"></tt> command).


<sect1><tt>.IFPSC02</tt><label id=".IFPSC02"><p>

  Conditional assembly: Check if the assembler is currently in 65SC02 mode
  (see <tt><ref id=".PSC02" name=".PSC02"></tt> command).


<sect1><tt>.IFREF</tt><label id=".IFREF"><p>

  Conditional assembly: Check if a symbol is referenced. Must be followed
  by a symbol name. The condition is true if if the the given symbol was
  referenced before, and false otherwise.

  This command may be used to build subroutine libraries in include files
  (you may use separate object modules for this purpose too).

  Example:

  <tscreen><verb>
        .ifref  ToHex                   ; If someone used this subroutine
        ToHex:  tay                     ; Define subroutine
                lda     HexTab,y
                rts
        .endif
  </verb></tscreen>

  See also: <tt><ref id=".REFERENCED" name=".REFERENCED"></tt>


<sect1><tt>.IMPORT</tt><label id=".IMPORT"><p>

  Import a symbol from another module. The command is followed by a comma
  separated list of symbols to import, with each one optionally followed by
  an address specification.

  Example:

  <tscreen><verb>
        .import foo
        .import bar: zeropage
  </verb></tscreen>

  See: <tt><ref id=".IMPORTZP" name=".IMPORTZP"></tt>


<sect1><tt>.IMPORTZP</tt><label id=".IMPORTZP"><p>

  Import a symbol from another module. The command is followed by a comma
  separated list of symbols to import. The symbols are explicitly imported
  as zero page symbols (that is, symbols with values in byte range).

  Example:

  <tscreen><verb>
        .importzp       foo, bar
  </verb></tscreen>

  See: <tt><ref id=".IMPORT" name=".IMPORT"></tt>


<sect1><tt>.INCBIN</tt><label id=".INCBIN"><p>

  Include a file as binary data. The command expects a string argument
  that is the name of a file to include literally in the current segment.
  In addition to that, a start offset and a size value may be specified,
  separated by commas. If no size is specified, all of the file from the
  start offset to end-of-file is used. If no start position is specified
  either, zero is assumed (which means that the whole file is inserted).

  Example:

  <tscreen><verb>
        ; Include whole file
        .incbin         "sprites.dat"

        ; Include file starting at offset 256
        .incbin         "music.dat", $100

        ; Read 100 bytes starting at offset 200
        .incbin         "graphics.dat", 200, 100
  </verb></tscreen>


<sect1><tt>.INCLUDE</tt><label id=".INCLUDE"><p>

  Include another file. Include files may be nested up to a depth of 16.

  Example:

  <tscreen><verb>
        .include        "subs.inc"
  </verb></tscreen>


<sect1><tt>.INTERRUPTOR</tt><label id=".INTERRUPTOR"><p>

  Export a symbol and mark it as an interruptor. This may be used together
  with the linker to build a table of interruptor subroutines that are called
  in an interrupt.

  Note: The linker has a feature to build a table of marked routines, but it
  is your code that must call these routines, so just declaring a symbol as
  interruptor does nothing by itself.

  An interruptor is always exported as an absolute (16 bit) symbol. You don't
  need to use an additional <tt/.export/ statement, this is implied by
  <tt/.interruptor/. It may have an optional priority that is separated by a
  comma. Higher numeric values mean a higher priority. If no priority is
  given, the default priority of 7 is used. Be careful when assigning
  priorities to your own module constructors so they won't interfere with the
  ones in the cc65 library.

  Example:

  <tscreen><verb>
        .interruptor    IrqHandler
        .interruptor    Handler, 16
  </verb></tscreen>

  See the <tt><ref id=".CONDES" name=".CONDES"></tt> command and the separate
  section <ref id="condes" name="Module constructors/destructors"> explaining
  the feature in more detail.


<sect1><tt>.LINECONT</tt><label id=".LINECONT"><p>

  Switch on or off line continuations using the backslash character
  before a newline. The option is off by default.
  Note: Line continuations do not work in a comment. A backslash at the
  end of a comment is treated as part of the comment and does not trigger
  line continuation.
  The command must be followed by a '+' or '-' character to switch the
  option on or off respectively.

  Example:

  <tscreen><verb>
        .linecont       +               ; Allow line continuations

        lda     \
                #$20                    ; This is legal now
  </verb></tscreen>


<sect1><tt>.LIST</tt><label id=".LIST"><p>

  Enable output to the listing. The command must be followed by a boolean
  switch ("on", "off", "+" or "-") and will enable or disable listing
  output.
  The option has no effect if the listing is not enabled by the command line
  switch -l. If -l is used, an internal counter is set to 1. Lines are output
  to the listing file, if the counter is greater than zero, and suppressed if
  the counter is zero. Each use of <tt/.LIST/ will increment or decrement the
  counter.

  Example:

  <tscreen><verb>
        .list   on              ; Enable listing output
  </verb></tscreen>


<sect1><tt>.LISTBYTES</tt><label id=".LISTBYTES"><p>

  Set, how many bytes are shown in the listing for one source line. The
  default is 12, so the listing will show only the first 12 bytes for any
  source line that generates more than 12 bytes of code or data.
  The directive needs an argument, which is either "unlimited", or an
  integer constant in the range 4..255.

  Examples:

  <tscreen><verb>
        .listbytes      unlimited       ; List all bytes
        .listbytes      12              ; List the first 12 bytes
        .incbin         "data.bin"      ; Include large binary file
  </verb></tscreen>


<sect1><tt>.LOBYTES</tt><label id=".LOBYTES"><p>

  Define byte sized data by extracting only the low byte (that is, bits 0-7) from
  each expression.  This is equivalent to <tt><ref id=".BYTE" name=".BYTE"></tt> with
  the operator '<' prepended to each expression in its list.

  Example:

  <tscreen><verb>
        .lobytes         $1234, $2345, $3456, $4567
        .hibytes         $fedc, $edcb, $dcba, $cba9
  </verb></tscreen>

  which is equivalent to

  <tscreen><verb>
        .byte            $34, $45, $56, $67
        .byte            $fe, $ed, $dc, $cb
  </verb></tscreen>

  Example:

  <tscreen><verb>
        .define MyTable TableItem0, TableItem1, TableItem2, TableItem3

        TableLookupLo:   .lobytes MyTable
        TableLookupHi:   .hibytes MyTable
  </verb></tscreen>

  which is equivalent to

  <tscreen><verb>
        TableLookupLo:   .byte &lt;TableItem0, &lt;TableItem1, &lt;TableItem2, &lt;TableItem3
        TableLookupHi:   .byte &gt;TableItem0, &gt;TableItem1, &gt;TableItem2, &gt;TableItem3
  </verb></tscreen>

  See also: <tt><ref id=".BYTE" name=".BYTE"></tt>,
            <tt><ref id=".HIBYTES" name=".HIBYTES"></tt>,
            <tt><ref id=".BANKBYTES" name=".BANKBYTES"></tt>


<sect1><tt>.LOCAL</tt><label id=".LOCAL"><p>

  This command may only be used inside a macro definition. It declares a
  list of identifiers as local to the macro expansion.

  A problem when using macros are labels: Since they don't change their name,
  you get a "duplicate symbol" error if the macro is expanded the second time.
  Labels declared with <tt><ref id=".LOCAL" name=".LOCAL"></tt> have their
  name mapped to an internal unique name (<tt/___ABCD__/) with each macro
  invocation.

  Some other assemblers start a new lexical block inside a macro expansion.
  This has some drawbacks however, since that will not allow <em/any/ symbol
  to be visible outside a macro, a feature that is sometimes useful. The
  <tt><ref id=".LOCAL" name=".LOCAL"></tt> command is in my eyes a better way
  to address the problem.

  You get an error when using <tt><ref id=".LOCAL" name=".LOCAL"></tt> outside
  a macro.


<sect1><tt>.LOCALCHAR</tt><label id=".LOCALCHAR"><p>

  Defines the character that start "cheap" local labels. You may use one
  of '@' and '?' as start character. The default is '@'.

  Cheap local labels are labels that are visible only between two non
  cheap labels. This way you can reuse identifiers like "<tt/loop/" without
  using explicit lexical nesting.

  Example:

  <tscreen><verb>
        .localchar      '?'

        Clear:  lda     #$00            ; Global label
        ?Loop:  sta     Mem,y           ; Local label
                dey
                bne     ?Loop           ; Ok
                rts
        Sub:    ...                     ; New global label
                bne     ?Loop           ; ERROR: Unknown identifier!
  </verb></tscreen>


<sect1><tt>.MACPACK</tt><label id=".MACPACK"><p>

  Insert a predefined macro package. The command is followed by an
  identifier specifying the macro package to insert. Available macro
  packages are:

  <tscreen><verb>
        atari           Defines the scrcode macro.
        cbm             Defines the scrcode macro.
        cpu             Defines constants for the .CPU variable.
        generic         Defines generic macros like add and sub.
        longbranch      Defines conditional long jump macros.
  </verb></tscreen>

  Including a macro package twice, or including a macro package that
  redefines already existing macros will lead to an error.

  Example:

  <tscreen><verb>
        .macpack        longbranch      ; Include macro package

                cmp     #$20            ; Set condition codes
                jne     Label           ; Jump long on condition
  </verb></tscreen>

  Macro packages are explained in more detail in section <ref
  id="macropackages" name="Macro packages">.


<sect1><tt>.MAC, .MACRO</tt><label id=".MAC"><p>

  Start a classic macro definition. The command is followed by an identifier
  (the macro name) and optionally by a comma separated list of identifiers
  that are macro parameters.

  See section <ref id="macros" name="Macros">.


<sect1><tt>.ORG</tt><label id=".ORG"><p>

  Start a section of absolute code. The command is followed by a constant
  expression that gives the new PC counter location for which the code is
  assembled. Use <tt><ref id=".RELOC" name=".RELOC"></tt> to switch back to
  relocatable code.

  By default, absolute/relocatable mode is global (valid even when switching
  segments). Using <tt>.FEATURE <ref id="org_per_seg" name="org_per_seg"></tt>
  it can be made segment local.

  Please note that you <em/do not need/ <tt/.ORG/ in most cases. Placing
  code at a specific address is the job of the linker, not the assembler, so
  there is usually no reason to assemble code to a specific address.

  Example:

  <tscreen><verb>
        .org    $7FF            ; Emit code starting at $7FF
  </verb></tscreen>


<sect1><tt>.OUT</tt><label id=".OUT"><p>

  Output a string to the console without producing an error. This command
  is similar to <tt/.ERROR/, however, it does not force an assembler error
  that prevents the creation of an object file.

  Example:

  <tscreen><verb>
        .out    "This code was written by the codebuster(tm)"
  </verb></tscreen>

  See also the <tt><ref id=".WARNING" name=".WARNING"></tt> and <tt><ref
  id=".ERROR" name=".ERROR"></tt> directives.


<sect1><tt>.P02</tt><label id=".P02"><p>

  Enable the 6502 instruction set, disable 65SC02, 65C02 and 65816
  instructions. This is the default if not overridden by the
  <tt><ref id="option--cpu" name="--cpu"></tt> command line option.

  See: <tt><ref id=".PC02" name=".PC02"></tt>, <tt><ref id=".PSC02"
  name=".PSC02"></tt> and <tt><ref id=".P816" name=".P816"></tt>


<sect1><tt>.P816</tt><label id=".P816"><p>

  Enable the 65816 instruction set. This is a superset of the 65SC02 and
  6502 instruction sets.

  See: <tt><ref id=".P02" name=".P02"></tt>, <tt><ref id=".PSC02"
  name=".PSC02"></tt> and <tt><ref id=".PC02" name=".PC02"></tt>


<sect1><tt>.PAGELEN, .PAGELENGTH</tt><label id=".PAGELENGTH"><p>

  Set the page length for the listing. Must be followed by an integer
  constant. The value may be "unlimited", or in the range 32 to 127. The
  statement has no effect if no listing is generated. The default value is -1
  (unlimited) but may be overridden by the <tt/--pagelength/ command line
  option. Beware: Since ca65 is a one pass assembler, the listing is generated
  after assembly is complete, you cannot use multiple line lengths with one
  source. Instead, the value set with the last <tt/.PAGELENGTH/ is used.

  Examples:

  <tscreen><verb>
        .pagelength     66              ; Use 66 lines per listing page

        .pagelength     unlimited       ; Unlimited page length
  </verb></tscreen>


<sect1><tt>.PC02</tt><label id=".PC02"><p>

  Enable the 65C02 instructions set. This instruction set includes all
  6502 and 65SC02 instructions.

  See: <tt><ref id=".P02" name=".P02"></tt>, <tt><ref id=".PSC02"
  name=".PSC02"></tt> and <tt><ref id=".P816" name=".P816"></tt>


<sect1><tt>.POPCPU</tt><label id=".POPCPU"><p>

  Pop the last CPU setting from the stack, and activate it.

  This command will switch back to the CPU that was last pushed onto the CPU
  stack using the <tt><ref id=".PUSHCPU" name=".PUSHCPU"></tt> command, and
  remove this entry from the stack.

  The assembler will print an error message if the CPU stack is empty when
  this command is issued.

  See: <tt><ref id=".CPU" name=".CPU"></tt>, <tt><ref id=".PUSHCPU"
  name=".PUSHCPU"></tt>, <tt><ref id=".SETCPU" name=".SETCPU"></tt>


<sect1><tt>.POPSEG</tt><label id=".POPSEG"><p>

  Pop the last pushed segment from the stack, and set it.

  This command will switch back to the segment that was last pushed onto the
  segment stack using the <tt><ref id=".PUSHSEG" name=".PUSHSEG"></tt>
  command, and remove this entry from the stack.

  The assembler will print an error message if the segment stack is empty
  when this command is issued.

  See: <tt><ref id=".PUSHSEG" name=".PUSHSEG"></tt>


<sect1><tt>.PROC</tt><label id=".PROC"><p>

  Start a nested lexical level with the given name and adds a symbol with this
  name to the enclosing scope. All new symbols from now on are in the local
  lexical level and are accessible from outside only via <ref id="scopesyntax"
  name="explicit scope specification">. Symbols defined outside this local
  level may be accessed as long as their names are not used for new symbols
  inside the level. Symbols names in other lexical levels do not clash, so you
  may use the same names for identifiers. The lexical level ends when the
  <tt><ref id=".ENDPROC" name=".ENDPROC"></tt> command is read. Lexical levels
  may be nested up to a depth of 16 (this is an artificial limit to protect
  against errors in the source).

  Note: Macro names are always in the global level and in a separate name
  space. There is no special reason for this, it's just that I've never
  had any need for local macro definitions.

  Example:

  <tscreen><verb>
        .proc   Clear           ; Define Clear subroutine, start new level
                lda     #$00
        L1:     sta     Mem,y   ; L1 is local and does not cause a
                                ; duplicate symbol error if used in other
                                ; places
                dey
                bne     L1      ; Reference local symbol
                rts
        .endproc                ; Leave lexical level
  </verb></tscreen>

  See: <tt/<ref id=".ENDPROC" name=".ENDPROC">/ and <tt/<ref id=".SCOPE"
  name=".SCOPE">/


<sect1><tt>.PSC02</tt><label id=".PSC02"><p>

  Enable the 65SC02 instructions set. This instruction set includes all
  6502 instructions.

  See: <tt><ref id=".P02" name=".P02"></tt>, <tt><ref id=".PC02"
  name=".PC02"></tt> and <tt><ref id=".P816" name=".P816"></tt>


<sect1><tt>.PUSHCPU</tt><label id=".PUSHCPU"><p>

  Push the currently active CPU onto a stack. The stack has a size of 8
  entries.

  <tt/.PUSHCPU/ allows together with <tt><ref id=".POPCPU"
  name=".POPCPU"></tt> to switch to another CPU and to restore the old CPU
  later, without knowledge of the current CPU setting.

  The assembler will print an error message if the CPU stack is already full,
  when this command is issued.

  See: <tt><ref id=".CPU" name=".CPU"></tt>, <tt><ref id=".POPCPU"
  name=".POPCPU"></tt>, <tt><ref id=".SETCPU" name=".SETCPU"></tt>


<sect1><tt>.PUSHSEG</tt><label id=".PUSHSEG"><p>

  Push the currently active segment onto a stack. The entries on the stack
  include the name of the segment and the segment type. The stack has a size
  of 16 entries.

  <tt/.PUSHSEG/ allows together with <tt><ref id=".POPSEG" name=".POPSEG"></tt>
  to switch to another segment and to restore the old segment later, without
  even knowing the name and type of the current segment.

  The assembler will print an error message if the segment stack is already
  full, when this command is issued.

  See: <tt><ref id=".POPSEG" name=".POPSEG"></tt>


<sect1><tt>.RELOC</tt><label id=".RELOC"><p>

  Switch back to relocatable mode. See the <tt><ref id=".ORG"
  name=".ORG"></tt> command.


<sect1><tt>.REPEAT</tt><label id=".REPEAT"><p>

  Repeat all commands between <tt/.REPEAT/ and <tt><ref id=".ENDREPEAT"
  name=".ENDREPEAT"></tt> constant number of times. The command is followed by
  a constant expression that tells how many times the commands in the body
  should get repeated. Optionally, a comma and an identifier may be specified.
  If this identifier is found in the body of the repeat statement, it is
  replaced by the current repeat count (starting with zero for the first time
  the body is repeated).

  <tt/.REPEAT/ statements may be nested. If you use the same repeat count
  identifier for a nested <tt/.REPEAT/ statement, the one from the inner
  level will be used, not the one from the outer level.

  Example:

  The following macro will emit a string that is "encrypted" in that all
  characters of the string are XORed by the value $55.

  <tscreen><verb>
        .macro  Crypt   Arg
                .repeat .strlen(Arg), I
                .byte   .strat(Arg, I) ^ $55
                .endrep
        .endmacro
  </verb></tscreen>

  See: <tt><ref id=".ENDREPEAT" name=".ENDREPEAT"></tt>


<sect1><tt>.RES</tt><label id=".RES"><p>

  Reserve storage. The command is followed by one or two constant
  expressions. The first one is mandatory and defines, how many bytes of
  storage should be defined. The second, optional expression must by a
  constant byte value that will be used as value of the data. If there
  is no fill value given, the linker will use the value defined in the
  linker configuration file (default: zero).

  Example:

  <tscreen><verb>
        ; Reserve 12 bytes of memory with value $AA
        .res    12, $AA
  </verb></tscreen>


<sect1><tt>.RODATA</tt><label id=".RODATA"><p>

  Switch to the RODATA segment. The name of the RODATA segment is always
  "RODATA", so this is a shortcut for

  <tscreen><verb>
        .segment  "RODATA"
  </verb></tscreen>

  The RODATA segment is a segment that is used by the compiler for
  readonly data like string constants.

  See also the <tt><ref id=".SEGMENT" name=".SEGMENT"></tt> command.


<sect1><tt>.SCOPE</tt><label id=".SCOPE"><p>

  Start a nested lexical level with the given name. All new symbols from now
  on are in the local lexical level and are accessible from outside only via
  <ref id="scopesyntax" name="explicit scope specification">. Symbols defined
  outside this local level may be accessed as long as their names are not used
  for new symbols inside the level. Symbols names in other lexical levels do
  not clash, so you may use the same names for identifiers. The lexical level
  ends when the <tt><ref id=".ENDSCOPE" name=".ENDSCOPE"></tt> command is
  read. Lexical levels may be nested up to a depth of 16 (this is an
  artificial limit to protect against errors in the source).

  Note: Macro names are always in the global level and in a separate name
  space. There is no special reason for this, it's just that I've never
  had any need for local macro definitions.

  Example:

  <tscreen><verb>
        .scope  Error                   ; Start new scope named Error
                None = 0                ; No error
                File = 1                ; File error
                Parse = 2               ; Parse error
        .endscope                       ; Close lexical level

                ...
                lda #Error::File        ; Use symbol from scope Error
  </verb></tscreen>

  See: <tt/<ref id=".ENDSCOPE" name=".ENDSCOPE">/ and <tt/<ref id=".PROC"
  name=".PROC">/


<sect1><tt>.SEGMENT</tt><label id=".SEGMENT"><p>

  Switch to another segment. Code and data is always emitted into a
  segment, that is, a named section of data. The default segment is
  "CODE". There may be up to 254 different segments per object file
  (and up to 65534 per executable). There are shortcut commands for
  the most common segments ("CODE", "DATA" and "BSS").

  The command is followed by a string containing the segment name (there are
  some constraints for the name - as a rule of thumb use only those segment
  names that would also be valid identifiers). There may also be an optional
  address size separated by a colon. See the section covering <tt/<ref
  id="address-sizes" name="address sizes">/ for more information.

  The default address size for a segment depends on the memory model specified
  on the command line. The default is "absolute", which means that you don't
  have to use an address size modifier in most cases.

  "absolute" means that the is a segment with 16 bit (absolute) addressing.
  That is, the segment will reside somewhere in core memory outside the zero
  page. "zeropage" (8 bit) means that the segment will be placed in the zero
  page and direct (short) addressing is possible for data in this segment.

  Beware: Only labels in a segment with the zeropage attribute are marked
  as reachable by short addressing. The `*' (PC counter) operator will
  work as in other segments and will create absolute variable values.

  Please note that a segment cannot have two different address sizes. A
  segment specified as zeropage cannot be declared as being absolute later.

  Examples:

  <tscreen><verb>
        .segment "ROM2"                 ; Switch to ROM2 segment
        .segment "ZP2": zeropage        ; New direct segment
        .segment "ZP2"                  ; Ok, will use last attribute
        .segment "ZP2": absolute        ; Error, redecl mismatch
  </verb></tscreen>

  See: <tt><ref id=".BSS" name=".BSS"></tt>, <tt><ref id=".CODE"
  name=".CODE"></tt>, <tt><ref id=".DATA" name=".DATA"></tt> and <tt><ref
  id=".RODATA" name=".RODATA"></tt>


<sect1><tt>.SETCPU</tt><label id=".SETCPU"><p>

  Switch the CPU instruction set. The command is followed by a string that
  specifies the CPU. Possible values are those that can also be supplied to
  the <tt><ref id="option--cpu" name="--cpu"></tt> command line option,
  namely: 6502, 6502X, 65SC02, 65C02, 65816, sunplus and HuC6280. Please
  note that support for the sunplus CPU is not available in the freeware
  version, because the instruction set of the sunplus CPU is "proprietary
  and confidential".

  See: <tt><ref id=".CPU" name=".CPU"></tt>,
       <tt><ref id=".IFP02" name=".IFP02"></tt>,
       <tt><ref id=".IFP816" name=".IFP816"></tt>,
       <tt><ref id=".IFPC02" name=".IFPC02"></tt>,
       <tt><ref id=".IFPSC02" name=".IFPSC02"></tt>,
       <tt><ref id=".P02" name=".P02"></tt>,
       <tt><ref id=".P816" name=".P816"></tt>,
       <tt><ref id=".PC02" name=".PC02"></tt>,
       <tt><ref id=".PSC02" name=".PSC02"></tt>


<sect1><tt>.SMART</tt><label id=".SMART"><p>

  Switch on or off smart mode. The command must be followed by a '+' or '-'
  character to switch the option on or off respectively. The default is off
  (that is, the assembler doesn't try to be smart), but this default may be
  changed by the -s switch on the command line.

  In smart mode the assembler will do the following:

  <itemize>
  <item>Track usage of the <tt/REP/ and <tt/SEP/ instructions in 65816 mode
        and update the operand sizes accordingly. If the operand of such an
        instruction cannot be evaluated by the assembler (for example, because
        the operand is an imported symbol), a warning is issued. Beware: Since
        the assembler cannot trace the execution flow this may lead to false
        results in some cases. If in doubt, use the <tt/.Inn/ and <tt/.Ann/
        instructions to tell the assembler about the current settings.
  <item>In 65816 mode, replace a <tt/RTS/ instruction by <tt/RTL/ if it is
        used within a procedure declared as <tt/far/, or if the procedure has
        no explicit address specification, but it is <tt/far/ because of the
        memory model used.
  </itemize>

  Example:

  <tscreen><verb>
        .smart                          ; Be smart
        .smart  -                       ; Stop being smart
  </verb></tscreen>

  See: <tt><ref id=".A16" name=".A16"></tt>,
       <tt><ref id=".A8" name=".A8"></tt>,
       <tt><ref id=".I16" name=".I16"></tt>,
       <tt><ref id=".I8" name=".I8"></tt>


<sect1><tt>.STRUCT</tt><label id=".STRUCT"><p>

  Starts a struct definition. Structs are covered in a separate section named
  <ref id="structs" name="&quot;Structs and unions&quot;">.

  See: <tt><ref id=".ENDSTRUCT" name=".ENDSTRUCT"></tt>


<sect1><tt>.SUNPLUS</tt><label id=".SUNPLUS"><p>

  Enable the SunPlus instructions set. This command will not work in the
  freeware version of the assembler, because the instruction set is
  "proprietary and confidential".

  See: <tt><ref id=".P02" name=".P02"></tt>, <tt><ref id=".PSC02"
  name=".PSC02"></tt>, <tt><ref id=".PC02" name=".PC02"></tt>, and
  <tt><ref id=".P816" name=".P816"></tt>


<sect1><tt>.TAG</tt><label id=".TAG"><p>

  Allocate space for a struct or union.

  Example:

  <tscreen><verb>
        .struct Point
                xcoord  .word
                ycoord  .word
        .endstruct

        .bss
                .tag    Point           ; Allocate 4 bytes
  </verb></tscreen>


<sect1><tt>.WARNING</tt><label id=".WARNING"><p>

  Force an assembly warning. The assembler will output a warning message
  preceded by "User warning". This warning will always be output, even if
  other warnings are disabled with the <tt><ref id="option-W" name="-W0"></tt>
  command line option.

  This command may be used to output possible problems when assembling
  the source file.

  Example:

  <tscreen><verb>
        .macro  jne     target
                .local L1
                .ifndef target
                .warning "Forward jump in jne, cannot optimize!"
                beq     L1
                jmp     target
        L1:
                .else
                ...
                .endif
        .endmacro
  </verb></tscreen>

  See also the <tt><ref id=".ERROR" name=".ERROR"></tt> and <tt><ref id=".OUT"
  name=".OUT"></tt> directives.


<sect1><tt>.WORD</tt><label id=".WORD"><p>

  Define word sized data. Must be followed by a sequence of (word ranged,
  but not necessarily constant) expressions.

  Example:

  <tscreen><verb>
        .word   $0D00, $AF13, _Clear
  </verb></tscreen>


<sect1><tt>.ZEROPAGE</tt><label id=".ZEROPAGE"><p>

  Switch to the ZEROPAGE segment and mark it as direct (zeropage) segment.
  The name of the ZEROPAGE segment is always "ZEROPAGE", so this is a
  shortcut for

  <tscreen><verb>
        .segment  "ZEROPAGE", zeropage
  </verb></tscreen>

  Because of the "zeropage" attribute, labels declared in this segment are
  addressed using direct addressing mode if possible. You <em/must/ instruct
  the linker to place this segment somewhere in the address range 0..$FF
  otherwise you will get errors.

  See: <tt><ref id=".SEGMENT" name=".SEGMENT"></tt>



<sect>Macros<label id="macros"><p>


<sect1>Introduction<p>

Macros may be thought of as "parametrized super instructions". Macros are
sequences of tokens that have a name. If that name is used in the source
file, the macro is "expanded", that is, it is replaced by the tokens that
were specified when the macro was defined.


<sect1>Macros without parameters<p>

In it's simplest form, a macro does not have parameters. Here's an
example:

<tscreen><verb>
        .macro  asr             ; Arithmetic shift right
                cmp     #$80    ; Put bit 7 into carry
                ror             ; Rotate right with carry
        .endmacro
</verb></tscreen>

The macro above consists of two real instructions, that are inserted into
the code, whenever the macro is expanded. Macro expansion is simply done
by using the name, like this:

<tscreen><verb>
        lda     $2010
        asr
        sta     $2010
</verb></tscreen>


<sect1>Parametrized macros<p>

When using macro parameters, macros can be even more useful:

<tscreen><verb>
        .macro  inc16   addr
                clc
                lda     addr
                adc     #$01
                sta     addr
                lda     addr+1
                adc     #$00
                sta     addr+1
        .endmacro
</verb></tscreen>

When calling the macro, you may give a parameter, and each occurrence of
the name "addr" in the macro definition will be replaced by the given
parameter. So

<tscreen><verb>
        inc16   $1000
</verb></tscreen>

will be expanded to

<tscreen><verb>
                clc
                lda     $1000
                adc     #$01
                sta     $1000
                lda     $1000+1
                adc     #$00
                sta     $1000+1
</verb></tscreen>

A macro may have more than one parameter, in this case, the parameters
are separated by commas. You are free to give less parameters than the
macro actually takes in the definition. You may also leave intermediate
parameters empty. Empty parameters are replaced by empty space (that is,
they are removed when the macro is expanded). If you have a look at our
macro definition above, you will see, that replacing the "addr" parameter
by nothing will lead to wrong code in most lines. To help you, writing
macros with a variable parameter list, there are some control commands:

<tt><ref id=".IFBLANK" name=".IFBLANK"></tt> tests the rest of the line and
returns true, if there are any tokens on the remainder of the line. Since
empty parameters are replaced by nothing, this may be used to test if a given
parameter is empty. <tt><ref id=".IFNBLANK" name=".IFNBLANK"></tt> tests the
opposite.

Look at this example:

<tscreen><verb>
        .macro  ldaxy   a, x, y
        .ifnblank       a
                lda     #a
        .endif
        .ifnblank       x
                ldx     #x
        .endif
        .ifnblank       y
                ldy     #y
        .endif
        .endmacro
</verb></tscreen>

This macro may be called as follows:

<tscreen><verb>
        ldaxy   1, 2, 3         ; Load all three registers

        ldaxy   1, , 3          ; Load only a and y

        ldaxy   , , 3           ; Load y only
</verb></tscreen>

There's another helper command for determining, which macro parameters are
valid: <tt><ref id=".PARAMCOUNT" name=".PARAMCOUNT"></tt> This command is
replaced by the parameter count given, <em/including/ intermediate empty macro
parameters:

<tscreen><verb>
        ldaxy   1               ; .PARAMCOUNT = 1
        ldaxy   1,,3            ; .PARAMCOUNT = 3
        ldaxy   1,2             ; .PARAMCOUNT = 2
        ldaxy   1,              ; .PARAMCOUNT = 2
        ldaxy   1,2,3           ; .PARAMCOUNT = 3
</verb></tscreen>

Macro parameters may optionally be enclosed into curly braces. This allows the
inclusion of tokens that would otherwise terminate the parameter (the comma in
case of a macro parameter).

<tscreen><verb>
        .macro  foo     arg1, arg2
                ...
        .endmacro

                foo     ($00,x)         ; Two parameters passed
                foo     {($00,x)}       ; One parameter passed
</verb></tscreen>

In the first case, the macro is called with two parameters: '<tt/(&dollar;00/'
and 'x)'. The comma is not passed to the macro, since it is part of the
calling sequence, not the parameters.

In the second case, '(&dollar;00,x)' is passed to the macro, this time
including the comma.


<sect1>Detecting parameter types<p>

Sometimes it is nice to write a macro that acts differently depending on the
type of the argument supplied. An example would be a macro that loads a 16 bit
value from either an immediate operand, or from memory. The <tt/<ref
id=".MATCH" name=".MATCH">/ and <tt/<ref id=".XMATCH" name=".XMATCH">/
functions will allow you to do exactly this:

<tscreen><verb>
        .macro  ldax    arg
                .if (.match (.left (1, {arg}), #))
                    ; immediate mode
                    lda     #<(.right (.tcount ({arg})-1, {arg}))
                    ldx     #>(.right (.tcount ({arg})-1, {arg}))
                .else
                    ; assume absolute or zero page
                    lda     arg
                    ldx     1+(arg)
                .endif
        .endmacro
</verb></tscreen>

Using the <tt/<ref id=".MATCH" name=".MATCH">/ function, the macro is able to
check if its argument begins with a hash mark. If so, two immediate loads are
emitted, Otherwise a load from an absolute zero page memory location is
assumed. Please note how the curly braces are used to enclose parameters to
pseudo functions handling token lists. This is necessary, because the token
lists may include commas or parens, which would be treated by the assembler
as end-of-list.

The macro can be used as

<tscreen><verb>
        foo:    .word   $5678
        ...
                ldax    #$1234          ; X=$12, A=$34
        ...
                ldax    foo             ; X=$56, A=$78
</verb></tscreen>


<sect1>Recursive macros<p>

Macros may be used recursively:

<tscreen><verb>
        .macro  push    r1, r2, r3
                lda     r1
                pha
        .if     .paramcount > 1
                push    r2, r3
        .endif
        .endmacro
</verb></tscreen>

There's also a special macro to help writing recursive macros: <tt><ref
id=".EXITMACRO" name=".EXITMACRO"></tt> This command will stop macro expansion
immediately:

<tscreen><verb>
        .macro  push    r1, r2, r3, r4, r5, r6, r7
        .ifblank        r1
                ; First parameter is empty
                .exitmacro
        .else
                lda     r1
                pha
        .endif
                push    r2, r3, r4, r5, r6, r7
        .endmacro
</verb></tscreen>

When expanding this macro, the expansion will push all given parameters
until an empty one is encountered. The macro may be called like this:

<tscreen><verb>
        push    $20, $21, $32           ; Push 3 ZP locations
        push    $21                     ; Push one ZP location
</verb></tscreen>


<sect1>Local symbols inside macros<p>

Now, with recursive macros, <tt><ref id=".IFBLANK" name=".IFBLANK"></tt> and
<tt><ref id=".PARAMCOUNT" name=".PARAMCOUNT"></tt>, what else do you need?
Have a look at the inc16 macro above. Here is it again:

<tscreen><verb>
        .macro  inc16   addr
                clc
                lda     addr
                adc     #$01
                sta     addr
                lda     addr+1
                adc     #$00
                sta     addr+1
        .endmacro
</verb></tscreen>

If you have a closer look at the code, you will notice, that it could be
written more efficiently, like this:

<tscreen><verb>
        .macro  inc16   addr
                inc     addr
                bne     Skip
                inc     addr+1
        Skip:
        .endmacro
</verb></tscreen>

But imagine what happens, if you use this macro twice? Since the label
"Skip" has the same name both times, you get a "duplicate symbol" error.
Without a way to circumvent this problem, macros are not as useful, as
they could be. One solution is, to start a new lexical block inside the
macro:

<tscreen><verb>
        .macro  inc16   addr
        .proc
                inc     addr
                bne     Skip
                inc     addr+1
        Skip:
        .endproc
        .endmacro
</verb></tscreen>

Now the label is local to the block and not visible outside. However,
sometimes you want a label inside the macro to be visible outside. To make
that possible, there's a new command that's only usable inside a macro
definition: <tt><ref id=".LOCAL" name=".LOCAL"></tt>. <tt/.LOCAL/ declares one
or more symbols as local to the macro expansion. The names of local variables
are replaced by a unique name in each separate macro expansion. So we could
also solve the problem above by using <tt/.LOCAL/:

<tscreen><verb>
        .macro  inc16   addr
                .local  Skip            ; Make Skip a local symbol
                clc
                lda     addr
                adc     #$01
                sta     addr
                bcc     Skip
                inc     addr+1
        Skip:                           ; Not visible outside
        .endmacro
</verb></tscreen>


<sect1>C style macros<p>

Starting with version 2.5 of the assembler, there is a second macro type
available: C style macros using the <tt/.DEFINE/ directive. These macros are
similar to the classic macro type described above, but behaviour is sometimes
different:

<itemize>

<item>  Macros defined with <tt><ref id=".DEFINE" name=".DEFINE"></tt> may not
        span more than a line. You may use line continuation (see <tt><ref
        id=".LINECONT" name=".LINECONT"></tt>) to spread the definition over
        more than one line for increased readability, but the macro itself
        may not contain an end-of-line token.

<item>  Macros defined with <tt><ref id=".DEFINE" name=".DEFINE"></tt> share
        the name space with classic macros, but they are detected and replaced
        at the scanner level. While classic macros may be used in every place,
        where a mnemonic or other directive is allowed, <tt><ref id=".DEFINE"
        name=".DEFINE"></tt> style macros are allowed anywhere in a line. So
        they are more versatile in some situations.

<item>  <tt><ref id=".DEFINE" name=".DEFINE"></tt> style macros may take
        parameters. While classic macros may have empty parameters, this is
        not true for <tt><ref id=".DEFINE" name=".DEFINE"></tt> style macros.
        For this macro type, the number of actual parameters must match
        exactly the number of formal parameters.

        To make this possible, formal parameters are enclosed in braces when
        defining the macro. If there are no parameters, the empty braces may
        be omitted.

<item>  Since <tt><ref id=".DEFINE" name=".DEFINE"></tt> style macros may not
        contain end-of-line tokens, there are things that cannot be done. They
        may not contain several processor instructions for example. So, while
        some things may be done with both macro types, each type has special
        usages. The types complement each other.

</itemize>

Let's look at a few examples to make the advantages and disadvantages
clear.

To emulate assemblers that use "<tt/EQU/" instead of "<tt/=/" you may use the
following <tt/.DEFINE/:

<tscreen><verb>
        .define EQU     =

        foo     EQU     $1234           ; This is accepted now
</verb></tscreen>

You may use the directive to define string constants used elsewhere:

<tscreen><verb>
        ; Define the version number
        .define VERSION         "12.3a"

        ; ... and use it
        .asciiz VERSION
</verb></tscreen>

Macros with parameters may also be useful:

<tscreen><verb>
        .define DEBUG(message)  .out    message

        DEBUG   "Assembling include file #3"
</verb></tscreen>

Note that, while formal parameters have to be placed in braces, this is
not true for the actual parameters. Beware: Since the assembler cannot
detect the end of one parameter, only the first token is used. If you
don't like that, use classic macros instead:

<tscreen><verb>
        .macro  message
                .out    message
        .endmacro
</verb></tscreen>

(This is an example where a problem can be solved with both macro types).


<sect1>Characters in macros<p>

When using the <ref id="option-t" name="-t"> option, characters are translated
into the target character set of the specific machine. However, this happens
as late as possible. This means that strings are translated if they are part
of a <tt><ref id=".BYTE" name=".BYTE"></tt> or <tt><ref id=".ASCIIZ"
name=".ASCIIZ"></tt> command. Characters are translated as soon as they are
used as part of an expression.

This behaviour is very intuitive outside of macros but may be confusing when
doing more complex macros. If you compare characters against numeric values,
be sure to take the translation into account.




<sect>Macro packages<label id="macropackages"><p>

Using the <tt><ref id=".MACPACK" name=".MACPACK"></tt> directive, predefined
macro packages may be included with just one command. Available macro packages
are:


<sect1><tt>.MACPACK generic</tt><p>

This macro package defines macros that are useful in almost any program.
Currently, two macros are defined:

<tscreen><verb>
        .macro  add     Arg
                clc
                adc     Arg
        .endmacro

        .macro  sub     Arg
                sec
                sbc     Arg
        .endmacro
</verb></tscreen>


<sect1><tt>.MACPACK longbranch</tt><p>

This macro package defines long conditional jumps. They are named like the
short counterpart but with the 'b' replaced by a 'j'. Here is a sample
definition for the "<tt/jeq/" macro, the other macros are built using the same
scheme:

<tscreen><verb>
        .macro  jeq     Target
                .if     .def(Target) .and ((*+2)-(Target) <= 127)
                beq     Target
                .else
                bne     *+5
                jmp     Target
                .endif
        .endmacro
</verb></tscreen>

All macros expand to a short branch, if the label is already defined (back
jump) and is reachable with a short jump. Otherwise the macro expands to a
conditional branch with the branch condition inverted, followed by an absolute
jump to the actual branch target.

The package defines the following macros:

<tscreen><verb>
        jeq, jne, jmi, jpl, jcs, jcc, jvs, jvc
</verb></tscreen>



<sect1><tt>.MACPACK cbm</tt><p>

The cbm macro package will define a macro named <tt/scrcode/. It takes a
string as argument and places this string into memory translated into screen
codes.


<sect1><tt>.MACPACK cpu</tt><p>

This macro package does not define any macros but constants used to examine
the value read from the <tt/<ref id=".CPU" name=".CPU">/ pseudo variable. For
each supported CPU a constant similar to

<tscreen><verb>
    CPU_6502
    CPU_65SC02
    CPU_65C02
    CPU_65816
    CPU_SUNPLUS
    CPU_SWEET16
    CPU_HUC6280
</verb></tscreen>

is defined. These constants may be used to determine the exact type of the
currently enabled CPU. In addition to that, for each CPU instruction set,
another constant is defined:

<tscreen><verb>
    CPU_ISET_6502
    CPU_ISET_65SC02
    CPU_ISET_65C02
    CPU_ISET_65816
    CPU_ISET_SUNPLUS
    CPU_ISET_SWEET16
    CPU_ISET_HUC6280
</verb></tscreen>

The value read from the <tt/<ref id=".CPU" name=".CPU">/ pseudo variable may
be checked with <tt/<ref id="operators" name=".BITAND">/ to determine if the
currently enabled CPU supports a specific instruction set. For example the
65C02 supports all instructions of the 65SC02 CPU, so it has the
<tt/CPU_ISET_65SC02/ bit set in addition to its native <tt/CPU_ISET_65C02/
bit. Using

<tscreen><verb>
        .if (.cpu .bitand CPU_ISET_65SC02)
                lda     (sp)
        .else
                ldy     #$00
                lda     (sp),y
        .endif
</verb></tscreen>

it is possible to determine if the

<tscreen><verb>
                lda     (sp)
</verb></tscreen>

instruction is supported, which is the case for the 65SC02, 65C02 and 65816
CPUs (the latter two are upwards compatible to the 65SC02).



<sect>Predefined constants<label id="predefined-constants"><p>

For better orthogonality, the assembler defines similar symbols as the
compiler, depending on the target system selected:

<itemize>
<item><tt/__APPLE2__/ - Target system is <tt/apple2/
<item><tt/__APPLE2ENH__/ - Target system is <tt/apple2enh/
<item><tt/__ATARI__/ - Target system is <tt/atari/
<item><tt/__ATMOS__/ - Target system is <tt/atmos/
<item><tt/__BBC__/ - Target system is <tt/bbc/
<item><tt/__C128__/ - Target system is <tt/c128/
<item><tt/__C16__/ - Target system is <tt/c16/
<item><tt/__C64__/ - Target system is <tt/c64/
<item><tt/__CBM__/ - Target is a Commodore system
<item><tt/__CBM510__/ - Target system is <tt/cbm510/
<item><tt/__CBM610__/ - Target system is <tt/cbm610/
<item><tt/__GEOS__/ - Target system is <tt/geos/
<item><tt/__LUNIX__/ - Target system is <tt/lunix/
<item><tt/__NES__/ - Target system is <tt/nes/
<item><tt/__PET__/ - Target system is <tt/pet/
<item><tt/__PLUS4__/ - Target system is <tt/plus4/
<item><tt/__SUPERVISION__/ - Target system is <tt/supervision/
<item><tt/__VIC20__/ - Target system is <tt/vic20/
</itemize>


<sect>Structs and unions<label id="structs"><p>

<sect1>Structs and unions Overview<p>

Structs and unions are special forms of <ref id="scopes" name="scopes">.  They
are to some degree comparable to their C counterparts. Both have a list of
members. Each member allocates storage and may optionally have a name, which,
in case of a struct, is the offset from the beginning and, in case of a union,
is always zero.


<sect1>Declaration<p>

Here is an example for a very simple struct with two members and a total size
of 4 bytes:

<tscreen><verb>
      .struct Point
              xcoord  .word
              ycoord  .word
      .endstruct
</verb></tscreen>

A union shares the total space between all its members, its size is the same
as that of the largest member.

A struct or union must not necessarily have a name. If it is anonymous, no
local scope is opened, the identifiers used to name the members are placed
into the current scope instead.

A struct may contain unnamed members and definitions of local structs. The
storage allocators may contain a multiplier, as in the example below:

<tscreen><verb>
      .struct Circle
              .struct Point
                      .word   2         ; Allocate two words
              .endstruct
              Radius  .word
      .endstruct
</verb></tscreen>


<sect1>The <tt/.TAG/ keyword<p>

Using the <ref id=".TAG" name=".TAG"> keyword, it is possible to reserve space
for an already defined struct or unions within another struct:

<tscreen><verb>
      .struct Point
              xcoord  .word
              ycoord  .word
      .endstruct

      .struct Circle
              Origin  .tag    Point
              Radius  .byte
      .endstruct
</verb></tscreen>

Space for a struct or union may be allocated using the <ref id=".TAG"
name=".TAG"> directive.

<tscreen><verb>
        C:      .tag    Circle
</verb></tscreen>

Currently, members are just offsets from the start of the struct or union. To
access a field of a struct, the member offset has to be added to the address
of the struct itself:

<tscreen><verb>
        lda     C+Circle::Radius        ; Load circle radius into A
</verb></tscreen>

This may change in a future version of the assembler.


<sect1>Limitations<p>

Structs and unions are currently implemented as nested symbol tables (in fact,
they were a by-product of the improved scoping rules). Currently, the
assembler has no idea of types. This means that the <ref id=".TAG"
name=".TAG"> keyword will only allocate space. You won't be able to initialize
variables declared with <ref id=".TAG" name=".TAG">, and adding an embedded
structure to another structure with <ref id=".TAG" name=".TAG"> will not make
this structure accessible by using the '::' operator.



<sect>Module constructors/destructors<label id="condes"><p>

<em>Note:</em> This section applies mostly to C programs, so the explanation
below uses examples from the C libraries. However, the feature may also be
useful for assembler programs.


<sect1>Module constructors/destructors Overview<p>

Using the <tt><ref id=".CONSTRUCTOR" name=".CONSTRUCTOR"></tt>, <tt><ref
id=".DESTRUCTOR" name=".DESTRUCTOR"></tt> and <tt><ref id=".INTERRUPTOR"
name=".INTERRUPTOR"></tt> keywords it it possible to export functions in a
special way. The linker is able to generate tables with all functions of a
specific type. Such a table will <em>only</em> include symbols from object
files that are linked into a specific executable. This may be used to add
initialization and cleanup code for library modules, or a table of interrupt
handler functions.

The C heap functions are an example where module initialization code is used.
All heap functions (<tt>malloc</tt>, <tt>free</tt>, ...) work with a few
variables that contain the start and the end of the heap, pointers to the free
list and so on. Since the end of the heap depends on the size and start of the
stack, it must be initialized at runtime. However, initializing these
variables for programs that do not use the heap are a waste of time and
memory.

So the central module defines a function that contains initialization code and
exports this function using the <tt/.CONSTRUCTOR/ statement. If (and only if)
this module is added to an executable by the linker, the initialization
function will be placed into the table of constructors by the linker. The C
startup code will call all constructors before <tt/main/ and all destructors
after <tt/main/, so without any further work, the heap initialization code is
called once the module is linked in.

While it would be possible to add explicit calls to initialization functions
in the startup code, the new approach has several advantages:

<enum>
<item>
If a module is not included, the initialization code is not linked in and not
called. So you don't pay for things you don't need.

<item>
Adding another library that needs initialization does not mean that the
startup code has to be changed. Before we had module constructors and
destructors, the startup code for all systems had to be adjusted to call the
new initialization code.

<item>
The feature saves memory: Each additional initialization function needs just
two bytes in the table (a pointer to the function).

</enum>


<sect1>Calling order<p>

The symbols are sorted in increasing priority order by the linker when using
one of the builtin linker configurations, so the functions with lower
priorities come first and are followed by those with higher priorities. The C
library runtime subroutine that walks over the function tables calls the
functions starting from the top of the table - which means that functions with
a high priority are called first.

So when using the C runtime, functions are called with high priority functions
first, followed by low priority functions.


<sect1>Pitfalls<p>

When using these special symbols, please take care of the following:

<itemize>

<item>
The linker will only generate function tables, it will not generate code to
call these functions. If you're using the feature in some other than the
existing C environments, you have to write code to call all functions in a
linker generated table yourself. See the <tt/condes/ and <tt/callirq/ modules
in the C runtime for an example on how to do this.

<item>
The linker will only add addresses of functions that are in modules linked to
the executable. This means that you have to be careful where to place the
condes functions. If initialization or an irq handler is needed for a group of
functions, be sure to place the function into a module that is linked in
regardless of which function is called by the user.

<item>
The linker will generate the tables only when requested to do so by the
<tt/FEATURE CONDES/ statement in the linker config file. Each table has to
be requested separately.

<item>
Constructors and destructors may have priorities. These priorities determine
the order of the functions in the table. If your initialization or cleanup code
does depend on other initialization or cleanup code, you have to choose the
priority for the functions accordingly.

<item>
Besides the <tt><ref id=".CONSTRUCTOR" name=".CONSTRUCTOR"></tt>, <tt><ref
id=".DESTRUCTOR" name=".DESTRUCTOR"></tt> and <tt><ref id=".INTERRUPTOR"
name=".INTERRUPTOR"></tt> statements, there is also a more generic command:
<tt><ref id=".CONDES" name=".CONDES"></tt>. This allows to specify an
additional type. Predefined types are 0 (constructor), 1 (destructor) and 2
(interruptor). The linker generates a separate table for each type on request.

</itemize>


<sect>Porting sources from other assemblers<p>

Sometimes it is necessary to port code written for older assemblers to ca65.
In some cases, this can be done without any changes to the source code by
using the emulation features of ca65 (see <tt><ref id=".FEATURE"
name=".FEATURE"></tt>). In other cases, it is necessary to make changes to the
source code.

Probably the biggest difference is the handling of the <tt><ref id=".ORG"
name=".ORG"></tt> directive. ca65 generates relocatable code, and placement is
done by the linker. Most other assemblers generate absolute code, placement is
done within the assembler and there is no external linker.

In general it is not a good idea to write new code using the emulation
features of the assembler, but there may be situations where even this rule is
not valid.

<sect1>TASS<p>

You need to use some of the ca65 emulation features to simulate the behaviour
of such simple assemblers.

<enum>
<item>Prepare your sourcecode like this:

<tscreen><verb>
        ; if you want TASS style labels without colons
        .feature labels_without_colons

        ; if you want TASS style character constants
        ; ("a" instead of the default 'a')
        .feature loose_char_term

                .word *+2       ; the cbm load address

                [yourcode here]
</verb></tscreen>

notice that the two emulation features are mostly useful for porting
sources originally written in/for TASS, they are not needed for the
actual "simple assembler operation" and are not recommended if you are
writing new code from scratch.

<item>Replace all program counter assignments (which are not possible in ca65
by default, and the respective emulation feature works different from what
you'd expect) by another way to skip to memory locations, for example the
<tt><ref id=".RES" name=".RES"></tt> directive.

<tscreen><verb>
        ; *=$2000
        .res $2000-*    ; reserve memory up to $2000
</verb></tscreen>

Please note that other than the original TASS, ca65 can never move the program
counter backwards - think of it as if you are assembling to disk with TASS.

<item>Conditional assembly (<tt/.ifeq//<tt/.endif//<tt/.goto/ etc.) must be
rewritten to match ca65 syntax. Most importantly notice that due to the lack
of <tt/.goto/, everything involving loops must be replaced by
<tt><ref id=".REPEAT" name=".REPEAT"></tt>.

<item>To assemble code to a different address than it is executed at, use the
<tt><ref id=".ORG" name=".ORG"></tt> directive instead of
<tt/.offs/-constructs.

<tscreen><verb>
        .org $1800

        [floppy code here]

        .reloc  ; back to normal
</verb></tscreen>

<item>Then assemble like this:

<tscreen><verb>
        cl65 --start-addr 0x0ffe -t none myprog.s -o myprog.prg
</verb></tscreen>

Note that you need to use the actual start address minus two, since two bytes
are used for the cbm load address.

</enum>


<sect>Bugs/Feedback<p>

If you have problems using the assembler, if you find any bugs, or if
you're doing something interesting with the assembler, I would be glad to
hear from you. Feel free to contact me by email
(<htmlurl url="mailto:uz@cc65.org" name="uz@cc65.org">).



<sect>Copyright<p>

ca65 (and all cc65 binutils) are (C) Copyright 1998-2003 Ullrich von
Bassewitz. For usage of the binaries and/or sources the following
conditions do apply:

This software is provided 'as-is', without any expressed or implied
warranty.  In no event will the authors be held liable for any damages
arising from the use of this software.

Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it
freely, subject to the following restrictions:

<enum>
<item>  The origin of this software must not be misrepresented; you must not
        claim that you wrote the original software. If you use this software
        in a product, an acknowledgment in the product documentation would be
        appreciated but is not required.
<item>  Altered source versions must be plainly marked as such, and must not
        be misrepresented as being the original software.
<item>  This notice may not be removed or altered from any source
        distribution.
</enum>



</article>