Documented the new push/pop factility for #pragmas.

[cc65] / doc / cc65.sgml

<!doctype linuxdoc system>

<article>
<title>cc65 Users Guide
<author>Ullrich von Bassewitz, <htmlurl url="mailto:uz@cc65.org" name="uz@cc65.org">
<date>03.09.2000, 02.10.2001

<abstract>
cc65 is a C compiler for 6502 targets. It supports several 6502 based home
computers like the Commodore and Atari machines, but it is easily retargetable.
</abstract>

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

<!-- Begin the document -->


<sect>Overview<p>

cc65 was originally a C compiler for the Atari 8-bit machines written by
John R. Dunning. In prior releases I've described the compiler by listing
up the changes made by me. I have made many more changes in the meantime
(and rewritten major parts of the compiler), so I will no longer do that,
since the list would be too large and of no use to anyone. Instead I will
describe the compiler in respect to the ANSI/ISO C standard. In fact, I'm
planning a complete rewrite (that is, a complete new compiler) for the
next release, since there are too many limitations in the current code,
and removing these limitations would mean a rewrite of many more parts of
the compiler.

There is a separate document named "library.txt" that covers the library
available for the compiler. If you know C and are interested in doing
actual programming, the library documentation is probably of much more use
than this document.

If you need some hints for getting the best code out of the compiler, you
may have a look at "coding.txt" which covers some code generation issues.



<sect>Usage<p>

The compiler translates C files into files containing assembler code that
may be translated by the ca65 macroassembler (for more information about
the assembler, have a look at ca65.txt).


<sect1>Command line option overview<p>

The compiler may be called as follows:

<tscreen><verb>
---------------------------------------------------------------------------
Usage: cc65 [options] file
Short options:
  -A                    Strict ANSI mode
  -Cl                   Make local variables static
  -Dsym[=defn]          Define a symbol
  -I dir                Set an include directory search path
  -O                    Optimize code
  -Oi                   Optimize code, inline more code
  -Or                   Enable register variables
  -Os                   Inline some known functions
  -T                    Include source as comment
  -V                    Print the compiler version number
  -W                    Suppress warnings
  -d                    Debug mode
  -g                    Add debug info to object file
  -h                    Help (this text)
  -j                    Default characters are signed
  -o name               Name the output file
  -r                    Enable register variables
  -t sys                Set the target system
  -v                    Increase verbosity

Long options:
  --add-source          Include source as comment
  --ansi                Strict ANSI mode
  --bss-name seg        Set the name of the BSS segment
  --check-stack         Generate stack overflow checks
  --code-name seg       Set the name of the CODE segment
  --codesize x          Accept larger code by factor x
  --cpu type            Set cpu type
  --create-dep          Create a make dependency file
  --data-name seg       Set the name of the DATA segment
  --debug               Debug mode
  --debug-info          Add debug info to object file
  --help                Help (this text)
  --include-dir dir     Set an include directory search path
  --register-space b    Set space available for register variables
  --register-vars       Enable register variables
  --rodata-name seg     Set the name of the RODATA segment
  --signed-chars        Default characters are signed
  --static-locals       Make local variables static
  --target sys          Set the target system
  --verbose             Increase verbosity
  --version             Print the compiler version number
---------------------------------------------------------------------------
</verb></tscreen>


<sect1>Command line options in detail<p>

Here is a description of all the command line options:

<descrip>

  <label id="option-A">
  <tag><tt>-A, --ansi</tt></tag>

  This option disables any compiler exensions. Have a look at section 5
  for a discussion of compiler extensions. In addition, the macro
  <tt/__STRICT_ANSI__/ is defined, when using one of these options.


  <tag><tt>--bss-name seg</tt></tag>

  Set the name of the bss segment.


  <tag><tt>--check-stack</tt></tag>

  Tells the compiler to generate code that checks for stack overflows. See
  <tt><ref id="pragma-checkstack" name="#pragma checkstack"></tt> for an
  explanation of this feature.


  <tag><tt>--code-name seg</tt></tag>

  Set the name of the code segment.


  <tag><tt>--codesize x</tt></tag>

  This options allows finer control about speed vs. size decisions in the
  code generation phase. It gives the allowed size increase factor (in
  percent). The default is 100 when not using <tt/-Oi/ and 200 when using
  <tt/-Oi/ (<tt/-Oi/ is the same as <tt/--codesize 200/).


  <tag><tt>--cpu CPU</tt></tag>

  A new, still experimental option. You may specify "6502" or "65C02" as
  the CPU. 6502 is the default, so this will not change anything.
  Specifying 65C02 will use a few 65C02 instructions when generating code.
  Don't expect too much from this option: It is still new (and may have
  bugs), and the additional instructions for the 65C02 are not that
  overwhelming.


  <tag><tt>--create-dep</tt></tag>

  Tells the compiler to generate a file containing the dependency list for
  the compiled module in makefile syntax. The file is named as the C input
  file with the extension replaced by <tt/.u/.


  <tag><tt>-d, --debug</tt></tag>

  Enables debug mode, something that should not be needed for mere
  mortals:-)


  <tag><tt>-D sym[=definition]</tt></tag>

  Define a macro on the command line. If no definition is given, the macro
  is defined to the value "1".


  <tag><tt>-g, --debug-info</tt></tag>

  This will cause the compiler to insert a <tt/.DEBUGINFO/ command into the
  generated assembler code. This will cause the assembler to include all
  symbols in a special section in the object file.


  <tag><tt>-h, --help</tt></tag>

  Print the short option summary shown above.


  <tag><tt>-o name</tt></tag>

  Specify the name of the output file. If you don't specify a name, the
  name of the C input file is used, with the extension replaced by ".s".


  <tag><tt>-r, --register-vars</tt></tag>

  <tt/-r/ will make the compiler honor the <tt/register/ keyword. Local
  variables may be placed in registers (which are actually zero page
  locations). There is some overhead involved with register variables, since
  the old contents of the registers must be saved and restored. Since register
  variables are of limited use without the optimizer, there is also a combined
  switch: <tt/-Or/ will enable both, the optmizer and register variables.

  For more information about register variables see <ref id="regvars"
  name="register variables">. 

  The compiler setting can also be changed within the source file by using
  <tt><ref id="pragma-regvars" name="#pragma regvars"></tt>.


  <tag><tt>--register-space</tt></tag>

  This option takes a numeric parameter and is used to specify, how much
  zero page register space is available. Please note that just giving this
  option will not increase or decrease by itself, it will just tell the
  compiler about the available space. You will have to allocate that space
  yourself using an assembler module with the necessary allocations, and a
  linker configuration that matches the assembler module. The default value
  for this option is 6 (bytes).

  If you don't know what all this means, please don't use this option.


  <tag><tt>--rodata-name seg</tt></tag>

  Set the name of the rodata segment (the segment used for readonly data).


  <tag><tt>-j, --signed-chars</tt></tag>

  Using this option, you can make the default characters signed. Since the
  6502 has no provisions for sign extending characters (which is needed on
  almost any load operation), this will make the code larger and slower. A
  better way is to declare characters explicitly as "signed" if needed. You
  can also use <tt><ref id="pragma-signedchars" name="#pragma
  signedchars"></tt> for better control of this option.


  <tag><tt>-t target, --target target</tt></tag>

  This option is used to set the target system. The target system
  determines things like the character set that is used for strings and
  character constants. The following target systems are supported:

  <itemize>
  <item>none
  <item>apple2
  <item>atari
  <item>atmos
  <item>c16 (works also for the c116 with memory up to 32K)
  <item>c64
  <item>c128
  <item>plus4
  <item>cbm510 (CBM-II series with 40 column video)
  <item>cbm610 (all CBM-II II computers with 80 column video)
  <item>pet (all CBM PET systems except the 2001)
  <item>geos
  </itemize>

  <tag><tt>-v, --verbose</tt></tag>

  Using this option, the compiler will be somewhat more verbose if errors
  or warnings are encountered.


  <tag><tt>-Cl, --static-locals</tt></tag>

  Use static storage for local variables instead of storage on the stack.
  Since the stack is emulated in software, this gives shorter and usually
  faster code, but the code is no longer reentrant. The difference between
  <tt/-Cl/ and declaring local variables as static yourself is, that
  initializer code is executed each time, the function is entered. So when
  using

  <tscreen><verb>
        void f (void)
        {
            unsigned a = 1;
            ...
        }
  </verb></tscreen>

  the variable a will always have the value 1 when entering the function
  and using <tt/-Cl/, while in

  <tscreen><verb>
        void f (void)
        {
            static unsigned a = 1;
            ....
        }
  </verb></tscreen>

  the variable a will have the value 1 only the first time, the function
  is entered, and will keep the old value from one call of the function to
  the next.

  You may also use <tt><ref id="pragma-staticlocals" name="#pragma
  staticlocals"></tt> to change this setting in your sources.


  <tag><tt>-I dir, --include-dir dir</tt></tag>

  Set a directory where the compiler searches for include files. You may
  use this option multiple times to add more than one directory to the
  search list.


  <tag><tt>-O, -Oi, -Or, -Os</tt></tag>

  Enable an optimizer run over the produced code.

  Using <tt/-Oi/, the code generator will inline some code where otherwise a
  runtime functions would have been called, even if the generated code is
  larger. This will not only remove the overhead for a function call, but will
  make the code visible for the optimizer. <tt/-Oi/ is an alias for
  <tt/--codesize 200/.

  <tt/-Or/ will make the compiler honor the <tt/register/ keyword. Local
  variables may be placed in registers (which are actually zero page
  locations). There is some overhead involved with register variables, since
  the old contents of the registers must be saved and restored. In addition,
  the current implementation does not make good use of register variables, so
  using <tt/-Or/ may make your program even slower and larger. Use with care!

  Using <tt/-Os/ will force the compiler to inline some known functions from
  the C library like strlen. Note: This has two consequences:
  <p>
  <itemize>
  <item>You may not use names of standard C functions in your own code. If you
        do that, your program is not standard compliant anyway, but using
        <tt/-Os/ will actually break things.
        <p>
  <item>The inlined string and memory functions will not handle strings or
        memory areas larger than 255 bytes. Similar, the inlined <tt/is..()/
        functions will not work with values outside char range.
        <p>
  </itemize>
  <p>
  It is possible to concatenate the modifiers for <tt/-O/. For example, to
  enable register variables and inlining of known functions, you may use
  <tt/-Ors/.


  <tag><tt>-T, --add-source</tt></tag>

  This include the source code as comments in the generated code. This is
  normally not needed.


  <tag><tt>-V, --version</tt></tag>

  Print the version number of the compiler. When submitting a bug report,
  please include the operating system you're using, and the compiler
  version.


  <tag><tt>-W</tt></tag>

  This option will suppress any warnings generated by the compiler. Since
  any source file may be written in a manner that it will not produce
  compiler warnings, using this option is usually not a good idea.

</descrip><p>


<sect>Input and output<p>

The compiler will accept one C file per invocation and create a file with
the same base name, but with the extension replaced by ".s". The output
file contains assembler code suitable for the use with the ca65 macro
assembler.

In addition to the paths named in the <tt/-I/ option on the command line, the
directory named in the environment variable <tt/CC65_INC/ is added to the
search path for include files on startup.



<sect>Differences to the ISO standard<p>

Here is a list of differences between the language, the compiler accepts,
and the one defined by the ISO standard:

<itemize>

<item>  The compiler allows single line comments that start with //. This
        feature is disabled in strict ANSI mode.
        <p>
<item>  The compiler allows unnamed parameters in parameter lists. The
        compiler will not issue warnings about unused parameters that don't
        have a name. This feature is disabled in strict ANSI mode.
        <p>
<item>  The compiler has some additional keywords:
        <p>
        <itemize>
        <item><tt/asm/
        <item><tt/__asm__/
        <item><tt/fastcall/
        <item><tt/__fastcall__/
        <item><tt/__AX__/
        <item><tt/__EAX__/
        <item><tt/__func__/
        <item><tt/__attribute__/
        </itemize>
        <p>
        The keywords without the underlines are disabled in strict ANSI mode.
        <p>
<item>  The datatypes "float" and "double" are not available.
        <p>
<item>  The compiler does not support bit fields.
        <p>
<item>  C Functions may not return structs (or unions), and structs may not
        be passed as parameters by value. However, struct assignment *is*
        possible.
        <p>
<item>  Part of the C library is available only with fastcall calling
        conventions (see below). This means, that you may not mix pointers to
        those functions with pointers to user written functions.
        <p>
</itemize>

There may be some more minor differences, I'm currently not aware off. The
biggest problem is the missing float data type. With this limitation in
mind, you should be able to write fairly portable code.



<sect>Extensions<p>

This cc65 version has some extensions to the ISO C standard.

<itemize>

<item>  The compiler allows // comments (like in C++ and in the proposed C9x
        standard). This feature is disabled by <tt><ref id="option-A"
        name="-A"></tt>.
        <p>

<item>  The compiler allows to insert assembler statements into the output
        file. The syntax is

        <tscreen><verb>
        asm (&lt;string literal&gt;[, optional parameters]) ;
        </verb></tscreen>
        or
        <tscreen><verb>
        __asm__ (&lt;string literal&gt;[, optional parameters]) ;
        </verb></tscreen>

        The first form is in the user namespace and is disabled if the <tt/-A/
        switch is given.

        There is a whole section covering inline assembler statements,
        <ref id="inline-asm" name="see there">.
        <p>

<item>  There is a special calling convention named "fastcall". This calling
        convention is currently only usable for functions written in
        assembler. The syntax for a function declaration using fastcall is

        <tscreen><verb>
        &lt;return type&gt; fastcall &lt;function name&gt; (&lt;parameter list&gt;)
        </verb></tscreen>
        or
        <tscreen><verb>
        &lt;return type&gt; __fastcall__ &lt;function name&gt; (&lt;parameter list&gt;)
        </verb></tscreen>
        An example would be
        <tscreen><verb>
        void __fastcall__ f (unsigned char c)
        </verb></tscreen>
        The first form of the fastcall keyword is in the user namespace and is
        therefore disabled in strict ANSI mode.

        For functions declared as <tt/fastcall/, the rightmost parameter is not
        pushed on the stack but left in the primary register when the function
        is called. This will reduce the cost when calling assembler functions
        significantly, especially when the function itself is rather small.
        <p>

<item>  There are two pseudo variables named <tt/__AX__/ and <tt/__EAX__/.
        Both refer to the primary register that is used by the compiler to
        evaluate expressions or return function results. <tt/__AX__/ is of
        type <tt/unsigned int/ and <tt/__EAX__/ of type <tt/long unsigned int/
        respectively. The pseudo variables may be used as lvalue and rvalue as
        every other variable. They are most useful together with short
        sequences of assembler code. For example, the macro

        <tscreen><verb>
        #define hi(x) (__AX__=(x),asm("\ttxa\n\tldx\t#$00",__AX__)
        </verb></tscreen>

        will give the high byte of any unsigned value.
        <p>

<item>  Inside a function, the identifier <tt/__func__/ gives the name of the
        current function as a string. Outside of functions, <tt/__func__/ is
        undefined.
        Example:

        <tscreen><verb>
        #define PRINT_DEBUG(s)  printf ("%s: %s\n", __func__, s);
        </verb></tscreen>

        The macro will print the name of the current function plus a given
        string.
        <p>

<item>  cc65 allows the initialization of <tt/void/ variables. This may be
        used to create variable structures that are more compatible with
        interfaces written for assembler languages. Here is an example:

        <tscreen><verb>
        void GCmd = {   (char)3, (unsigned)0x2000, (unsigned)0x3000 };
        </verb></tscreen>

        This will be translated as follows:

        <tscreen><verb>
        _GCmd:
                .byte   3
                .word   $2000
                .word   $3000
        </verb></tscreen>

        Since the variable is of type <tt/void/ you may not use it as is.
        However, taking the address of the variable results in a <tt/void*/
        which may be passed to any function expecting a pointer.

        See the <htmlurl url="geos.html" name="GEOS library"> for examples on
        how to use this feature.
        <p>

<item>  cc65 implements flexible array struct members as defined in the C99 ISO
        standard. As an extension, in non ANSI mode, these fields may be
        initialized. There are several exceptions, however (which is probably
        the reason why the standard does not define this feature, because it is
        highly unorthogonal). Flexible array members cannot be initialized...

        <itemize>
        <item>  ...when defining an array of structs with flexible members.
        <item>  ...if such a struct is a member field of another struct which
                is not the last field.
        <item>  If the struct which contains a flexible array member is
                declared as <tt/register/ and the size and compiler settings
                do allow the compiler to actually place the struct into the
                register bank in the zero page.
        </itemize>

        Please note that - as defined in the ISO C standard - the <tt/sizeof/
        operator returns the struct size with the flexible array member having
        size zero, even if it is initialized.
        <p>

</itemize>
<p>


<sect>Predefined macros<p>

The compiler defines several macros at startup:

<descrip>

  <tag><tt>__CC65__</tt></tag>

  This macro is always defined. Its value is the version number of the
  compiler in hex. Version 2.0.1 of the compiler will have this macro defined
  as 0x0201.

  <tag><tt>__APPLE2__</tt></tag>

  This macro is defined if the target is the Apple ][ (-t apple2).

  <tag><tt>__ATARI__</tt></tag>

  This macro is defined if the target is one of the Atari computers
  (400/800/130XL/800XL).

  <tag><tt>__ATMOS__</tt></tag>

  This macro is defined if the target is the Oric Atmos (-t atmos).

  <tag><tt>__CBM__</tt></tag>

  This macro is defined if the target system is one of the CBM targets.

  <tag><tt>__C16__</tt></tag>

  This macro is defined if the target is the c16 (-t c16).

  <tag><tt>__C64__</tt></tag>

  This macro is defined if the target is the c64 (-t c64).

  <tag><tt>__C128__</tt></tag>

  This macro is defined if the target is the c128 (-t c128).

  <tag><tt>__CBM510__</tt></tag>

  This macro is defined if the target is the CBM 500 series of computers.

  <tag><tt>__CBM610__</tt></tag>

  This macro is defined if the target is one of the CBM 600/700 family of
  computers (called B series in the US).

  <tag><tt>__GEOS__</tt></tag>

  This macro is defined if you are compiling for the GEOS system (-t geos).

  <tag><tt>__PET__</tt></tag>

  This macro is defined if the target is the PET family of computers (-t pet).

  <tag><tt>__PLUS4__</tt></tag>

  This macro is defined if the target is the plus/4 (-t plus4).

  <tag><tt>__VIC20__</tt></tag>

  This macro is defined if the target is the vic20 (-t vic20).

  <tag><tt>__FILE__</tt></tag>

  This macro expands to a string containing the name of the C source file.

  <tag><tt>__LINE__</tt></tag>

  This macro expands to the current line number.

  <tag><tt>__STRICT_ANSI__</tt></tag>

  This macro is defined to 1 if the <tt/-A/ compiler option was given, and
  undefined otherwise.

  <tag><tt>__OPT__</tt></tag>

  Is defined if the compiler was called with the <tt/-O/ command line option.

  <tag><tt>__OPT_i__</tt></tag>

  Is defined if the compiler was called with the <tt/-Oi/ command line option.

  <tag><tt>__OPT_r__</tt></tag>

  Is defined if the compiler was called with the <tt/-Or/ command line option.

  <tag><tt>__OPT_s__</tt></tag>

  Is defined if the compiler was called with the <tt/-Os/ command line option.

</descrip>


<sect>#pragmas<label id="pragmas"><p>

The compiler understands some pragmas that may be used to change code
generation and other stuff. Some of these pragmas understand a special form:
If the first parameter is <tt/push/, the old value is saved onto a stack
before changing it. The value may later be restored by using the <tt/pop/
parameter with the <tt/#pragma/.

<sect1><tt>#pragma bssseg (&lt;name&gt;)</tt><p>

  This pragma changes the name used for the BSS segment (the BSS segment
  is used to store uninitialized data). The argument is a string enclosed
  in double quotes.

  Note: The default linker configuration file does only map the standard
  segments. If you use other segments, you have to create a new linker
  configuration file.

  Beware: The startup code will zero only the default BSS segment. If you
  use another BSS segment, you have to do that yourself, otherwise
  uninitialized variables do not have the value zero.

  Example:
  <tscreen><verb>
        #pragma bssseg ("MyBSS")
  </verb></tscreen>


<sect1><tt>#pragma charmap (&lt;index&gt;, &lt;code&gt;)</tt><p>

  Each literal string and each literal character in the source is translated
  by use of a translation table. This translation table is preset when the
  compiler is started depending on the target system, for example to map
  ISO-8859-1 characters into PETSCII if the target is a commodore machine.

  This pragma allows to change entries in the translation table, so the
  translation for individual characters, or even the complete table may be
  adjusted.

  Both arguments are assumed to be unsigned characters with a valid range of
  1-255.

  Beware of two pitfalls:

    <itemize>
    <item>The character index is actually the code of the character in the
          C source, so character mappings do always depend on the source
          character set. This means that <tt/#pragma charmap/ is not portable
          - it depends on the build environment.
    <item>While it is possible to use character literals as indices, the
          result may be somewhat unexpected, since character literals are
          itself translated. For this reason I would suggest to avoid
          character literals and use numeric character codes instead.
    </itemize>

  Example:
  <tscreen><verb>
        /* Use a space wherever an 'a' occurs in ISO-8859-1 source */
        #pragma charmap (0x61, 0x20);
  </verb></tscreen>


<sect1><tt>#pragma checkstack (on|off)</tt><label id="pragma-checkstack"><p>

  Tells the compiler to insert calls to a stack checking subroutine to detect
  stack overflows. The stack checking code will lead to somewhat larger and
  slower programs, so you may want to use this pragma when debugging your
  program and switch it off for the release version. If a stack overflow is
  detected, the program is aborted.

  If the argument is "off", stack checks are disabled (the default), otherwise
  they're enabled.

  The <tt/#pragma/ understands the push and pop parameters as explained above.

<sect1><tt>#pragma codeseg (&lt;name&gt;)</tt><p>

  This pragma changes the name used for the CODE segment (the CODE segment
  is used to store executable code). The argument is a string enclosed in
  double quotes.

  Note: The default linker configuration file does only map the standard
  segments. If you use other segments, you have to create a new linker
  configuration file.

  Example:
  <tscreen><verb>
        #pragma codeseg ("MyCODE")
  </verb></tscreen>


<sect1><tt>#pragma dataseg (&lt;name&gt;)</tt><p>

  This pragma changes the name used for the DATA segment (the DATA segment
  is used to store initialized data). The argument is a string enclosed in
  double quotes.

  Note: The default linker configuration file does only map the standard
  segments. If you use other segments, you have to create a new linker
  configuration file.

  Example:
  <tscreen><verb>
        #pragma dataseg ("MyDATA")
  </verb></tscreen>


<sect1><tt>#pragma rodataseg (&lt;name&gt;)</tt><p>

  This pragma changes the name used for the RODATA segment (the RODATA
  segment is used to store readonly data). The argument is a string
  enclosed in double quotes.

  Note: The default linker configuration file does only map the standard
  segments. If you use other segments, you have to create a new linker
  configuration file.

  Example:
  <tscreen><verb>
        #pragma rodataseg ("MyRODATA")
  </verb></tscreen>


<sect1><tt>#pragma regvaraddr ([push,]on|off)</tt><p>

  The compiler does not allow to take the address of register variables.
  The regvaraddr pragma changes this. Taking the address of a register
  variable is allowed after using this pragma with "on" as argument.
  Using "off" as an argument switches back to the default behaviour.

  Beware: The C standard does not allow taking the address of a variable
  declared as register. So your programs become non-portable if you use
  this pragma. In addition, your program may not work. This is usually the
  case if a subroutine is called with the address of a register variable,
  and this subroutine (or a subroutine called from there) uses itself
  register variables. So be careful with this #pragma.

  The <tt/#pragma/ understands the push and pop parameters as explained above.

  Example:
  <tscreen><verb>
        #pragma regvaraddr(on)  /* Allow taking the address
                                 * of register variables
                                 */
  </verb></tscreen>


<sect1><tt>#pragma regvars ([push,]on|off)</tt><label id="pragma-regvars"><p>

  Enables or disables use of register variables. If register variables are
  disabled (the default), the <tt/register/ keyword is ignored. Register
  variables are explained in more detail in <ref id="regvars" name="a separate
  chapter">.

  The <tt/#pragma/ understands the push and pop parameters as explained above.


<sect1><tt>#pragma signedchars (on|off)</tt><label id="pragma-signedchars"><p>

  Changes the signedness of the default character type. If the argument is
  "on", default characters are signed, otherwise characters are unsigned.
  The compiler default is to make characters unsigned since this creates a
  lot better code. This default may be overridden by the <tt/--signed-chars/
  command line option.

  The <tt/#pragma/ understands the push and pop parameters as explained above.


<sect1><tt>#pragma staticlocals (on|off)</tt><label id="pragma-staticlocals"<p>

  Use variables in the bss segment instead of variables on the stack. This
  pragma changes the default set by the compiler option <tt/-Cl/. If the
  argument is "on", local variables are allocated in the BSS segment,
  leading to shorter and in most cases faster, but non-reentrant code.

  The <tt/#pragma/ understands the push and pop parameters as explained above.


<sect1><tt>#pragma zpsym (&lt;name&gt;)</tt><p>

  Tell the compiler that the - previously as external declared - symbol with
  the given name is a zero page symbol (usually from an assembler file).
  The compiler will create a matching import declaration for the assembler.

  Example:
  <tscreen><verb>
        extern int foo;
        #pragma zpsym ("foo");  /* foo is in the zeropage */
  </verb></tscreen>




<sect>Register variables<label id="regvars"><p>

The runtime for all supported platforms has 6 bytes of zero page space
available for register variables (this could be increased, but I think it's a
good value). So you can declare register variables up to a total size of 6 per
function. The compiler will allocate register space on a "first come, first
served" base and convert any <tt/register/ declarations that exceed the
available register space silently to <tt/auto/. Parameters can also be
declared as <tt/register/, this will in fact give slightly shorter code than
using a register variable.

Since a function must save the current values of the registers on entry and
restore them on exit, there is an overhead associated with register variables,
and this overhead is quite high (about 20 bytes per variable). This means that
just declaring anything as <tt/register/ is not a good idea.

The best use for register variables are pointers, especially those that point
to structures. The magic number here is about 3 uses of a struct field: If the
function contains this number or even more, the generated code will be usually
shorter and faster when using a register variable for the struct pointer. The
reason for this is that the register variable can in many cases be used as a
pointer directly. Having a pointer in an auto variable means that this pointer
must first be copied into a zero page location, before it can be dereferenced.

Second best use for register variables are counters. However, there is not
much difference in the code generated for counters, so you will need at least
100 operations on this variable (for example in a loop) to make it worth the
trouble. The only savings you get here are by the use of a zero page variable
instead of one on the stack or in the data segment.

Register variables must be explicitly enabled by using <tt/-Or/ or <tt/-r/ on
the command line. Register variables are only accepted on function top level,
register variables declared in interior blocks are silently converted to
<tt/auto/. With register variables disabled, all variables declared as
<tt/register/ are actually auto variables.

Please take care when using register variables: While they are helpful and can
lead to a tremendous speedup when used correctly, improper usage will cause
bloated code and a slowdown.



<sect>Inline assembler<label id="inline-asm"><p>

The compiler allows to insert assembler statements into the output file. The
syntax is

<tscreen><verb>
        asm (&lt;string literal&gt;[, optional parameters]) ;
</verb></tscreen>
or
<tscreen><verb>
        __asm__ (&lt;string literal&gt;[, optional parameters]) ;
</verb></tscreen>
<p>

The first form is in the user namespace and is disabled by <tt><ref
id="option-A" name="-A"></tt>.

The asm statement may be used inside a function and on global file level. An
inline assembler statement is a primary expression, so it may also be used as
part of an expression. Please note however that the result of an expression
containing just an inline assembler statement is always of type <tt/void/.

The contents of the string literal are preparsed by the compiler and inserted
into the generated assembly output, so that the can be further processed by
the backend and especially the optimizer. For this reason, the compiler does
only allow regular 6502 opcodes to be used with the inline assembler. Pseudo
instructions (like <tt/.import/, <tt/.byte/ and so on) are <em/not/ allowed,
even if the ca65 assembler (which is used to translate the generated assembler
code) would accept them. The builtin inline assembler is not a replacement for
the full blown macro assembler which comes with the compiler.

Note: Inline assembler statements are subject to all optimizations done by the
compiler. There is currently no way to protect an inline assembler statement
from being moved or removed completely by the optimizer. If in doubt, check
the generated assembler output, or disable optimizations.

The string literal may contain format specifiers from the following list. For
each format specifier, an argument is expected which is inserted instead of
the format specifier before passing the assembly code line to the backend.

<itemize>
  <item><tt/%b/ - Numerical 8 bit value
  <item><tt/%w/ - Numerical 16 bit value
  <item><tt/%l/ - Numerical 32 bit value
  <item><tt/%v/ - Assembler name of a (global) variable or function
  <item><tt/%o/ - Stack offset of a (local) variable
  <item><tt/%s/ - The argument is converted to a string
  <item><tt/%%/ - The % sign itself
</itemize><p>

Using these format specifiers, you can access C <tt/#defines/, variables or
similar stuff from the inline assembler. For example, to load the value of
a C <tt/#define/ into the Y register, one would use

<tscreen><verb>
        #define OFFS  23
        __asm__ ("ldy #%b", OFFS);
</verb></tscreen>

Or, to access a struct member of a static variable:

<tscreen><verb>
        typedef struct {
            unsigned char x;
            unsigned char y;
            unsigned char color;
        } pixel_t;
        static pixel_t pixel;
        __asm__ ("ldy #%b", offsetof(pixel_t, color));
        __asm__ ("lda %v,y", pixel);
</verb></tscreen>
<p>

Note: Do not embedd the assembler labels that are used as names of global
variables or functions into your asm statements. Code like this

<tscreen><verb>
        int foo;
        int bar () { return 1; }
        __asm__ ("lda _foo");   /* DON'T DO THAT! */
        ...
        __asm__ ("jsr _bar");   /* DON'T DO THAT EITHER! */
</verb></tscreen>
<p>

may stop working if the way, the compiler generates these names is changed in
a future version. Instead use the format specifiers from the table above.
<p>


<sect>Bugs/Feedback<p>

If you have problems using the compiler, if you find any bugs, or if you're
doing something interesting with it, 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>

This is the original compiler copyright:

<tscreen><verb>
--------------------------------------------------------------------------
  -*- Mode: Text -*-

     This is the copyright notice for RA65, LINK65, LIBR65, and other
  Atari 8-bit programs.  Said programs are Copyright 1989, by John R.
  Dunning.  All rights reserved, with the following exceptions:

      Anyone may copy or redistribute these programs, provided that:

  1:  You don't charge anything for the copy.  It is permissable to
      charge a nominal fee for media, etc.

  2:  All source code and documentation for the programs is made
      available as part of the distribution.

  3:  This copyright notice is preserved verbatim, and included in
      the distribution.

      You are allowed to modify these programs, and redistribute the
  modified versions, provided that the modifications are clearly noted.

      There is NO WARRANTY with this software, it comes as is, and is
  distributed in the hope that it may be useful.

      This copyright notice applies to any program which contains
  this text, or the refers to this file.

      This copyright notice is based on the one published by the Free
  Software Foundation, sometimes known as the GNU project.  The idea
  is the same as theirs, ie the software is free, and is intended to
  stay that way.  Everybody has the right to copy, modify, and re-
  distribute this software.  Nobody has the right to prevent anyone
  else from copying, modifying or redistributing it.

--------------------------------------------------------------------------
</verb></tscreen>

In acknowledgment of this copyright, I will place my own changes to the
compiler under the same copyright. Please note however, that the library
and all binutils are covered by another copyright, and that I'm planning
to do a complete rewrite of the compiler, after which the compiler
copyright will also change.

For the list of changes requested by this copyright see newvers.txt.


</article>