[cc65] / doc / intro.sgml

<!doctype linuxdoc system>

<article>

<title>cc65 compiler intro
<author>Ullrich von Bassewitz, <htmlurl url="mailto:uz@cc65.org" name="uz@cc65.org"> and CyberX
<date>06.13.2002

<abstract>
How to use the cc65 C compiler - an introduction.
</abstract>

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

<!-- Begin the document -->

<sect>Overview<p>

This is a short intro of how to use the compiler and the binutils. It contains a
step-by-step example of how to build a complete application from one C and one
assembler module. This file does <em/not/ contain a complete reference for the
tools used in the process. There are separate files describing these tools in
detail.

You are assumed to have downloaded and extracted the executables and the
target specific files. For example, for Windows users targeting C64, you need
cc65-win32-2.8.0.zip and cc65-c64-2.8.0.zip extracted to the same directory.
If you received the files as a bzip2 archive (extension *.bz2), you will need
to get the <htmlurl url="http://sources.redhat.com/bzip2/#bzip2-latest"
name="bzip2 package"> to decompress it.

<bf>Note</bf>: There is a much simpler way to compile this example using the
cl65 compiler and link utility. However, it makes sense to understand how the
separate steps work. How to do the example with the cl65 utility is described
<ref id="using-cl65" name="later">.


<sect1>Before we start<p>

You will find a copy of the sample modules used in the next section in the
samples/tutorial directory. Windows users will also find a small batch file in
this directory named "cc65setup.bat". Be sure to examine and understand the
commands in this file, then adjust them for your setup, and execute the file.


<sect1>The sample modules<p>

To explain the development flow, I will use the following example modules:

hello.c:

<tscreen><code>
        #include <stdio.h>
        #include <stdlib.h>

        extern const char text[];       /* In text.s */

        int main (void)
        {
            printf ("%s\n", text);
            return EXIT_SUCCESS;
        }
</code></tscreen>

text.s:
<tscreen><code>
        .export _text
        _text:  .asciiz "Hello world!"
</code></tscreen>


<sect1>Translation phases<p>

We assume that the target file should be named "hello", and the target system
is the C64.

<tscreen><verb>
    +---------+
    | hello.c |
    +---------+
         |
        cc65
         \/
    +---------+       +---------+
    | hello.s |       | text.s  |
    +---------+       +---------+
         |                 |
        ca65              ca65
         \/                \/
    +---------+       +---------+       +----------+       +---------+
    | hello.o |       | text.o  |       |  c64.o   |       | c64.lib |
    +---------+       +---------+       +----------+       +---------+
         |                     \         /                      |
         |                      \       /                       |
         |                       \     /                        |
         +----------------------->ld65<-------------------------+
                                   \/
                                  hello
</verb></tscreen>

<tt/c64.o/ (the startup code) and <tt/c64.lib/ (the c64 version of the runtime
and C library) are provided in binary form in the cc65 package.



<sect>The compiler<p>

The compiler translates one C source into one assembler source for each
invocation. It does <em/not/ create object files directly, and it is <em/not/
able to translate more than one file per run.

In the example above, we would use the following command line, to translate
<tt/hello.c/ into <tt/hello.s/:

<tscreen><verb>
        cc65 -O -I ../include -t c64 hello.c
</verb></tscreen>

The <tt/-O/ switch tells the compiler to do an additional optimizer run, which
is usually a good idea, since it makes the code smaller. If you don't care
about the size, but want to have slightly faster code, use <tt/-Oi/ to inline
some runtime functions.

The <tt/-I/ switch gives a search path for the include files. You may also set
the environment variable CC65_INC to the search path.

The <tt/-t/ switch is followed by the target system.

If the compiler does not complain about errors in our hello world, we will
have a file named "<tt/hello.s/" in our directory that contains the assembler
source for the hello module.

For more information about the compiler see <htmlurl url="cc65.html"
name="cc65.html">.



<sect>The assembler<p>

The assembler translates one assembler source into an object file for each
invocation. The assembler is <tt/not/ able to translate more than one source
file per run.

Let's translate the hello.s and text.s files from our example:

<tscreen><verb>
        ca65 hello.s
        ca65 -t c64 text.s
</verb></tscreen>

The <tt/-t/ switch is needed when translating the <tt/text.s/ file, so the
text is converted from the input character set (usually ISO-8859-1) into the
target character set (PETSCII) by the assembler. The compiler generated file
<tt/hello.s/ does not contain any character constants, so specification of a
target is not necessary (it wouldn't do any harm, however).

If the assembler does not complain, we should now have two object files (named
<tt/hello.o/ and <tt/text.o/) in the current directory.

For more information about the assembler see <htmlurl url="ca65.html"
name="ca65.html">.



<sect>The linker<p>

The linker combines several object and library file into one output file. ld65
is very configurable, but fortunately has a builtin configuration for the C64,
so we don't need to mess with configuration files here.

The compiler uses small functions to do things that cannot be done inline
without big impact on code size. These runtime functions, together with the C
library are in an object file archive named after the system, in this case
"<tt/c64.lib/". We have to specify this file on the command line so that the
linker can resolve these functions.

A second file (this time an object file) needed, is the startup code that
prepares the grounds for the C program to run. The startup file must be
executed first, so it must be the first file on the linker command line.

Let's link our files to get the final executable:

<tscreen><verb>
        ld65 -t c64 -o hello c64.o hello.o text.o c64.lib
</verb></tscreen>

The argument after <tt/-o/ specifies the name of the output file, the argument
after <tt/-t/ gives the target system. As discussed, the startup file must be
the first file on the command line (you may have to add a path here, if
<tt/c64.o/ is not in your current directory). Since the library resolves
imports in <tt/hello.o/ and <tt/text.o/, it must be specified <em/after/ these
files.

After a successful linker run, we have a file named "<tt/hello/", ready for
our C64!

For more information about the linker see <htmlurl url="ld65.html"
name="ld65.html">.



<sect>The easy way (using the cl65 utility)<label id="using-cl65"><p>

The cl65 utility is able to do all of the steps described above in just one
call, and it has defaults for some options that are very well suited for our
example.

To compile both files into one executable enter

<tscreen><verb>
        cl65 -O -I ../include hello.c text.s
</verb></tscreen>

(The <tt/-I/ switch is not needed if you are working under Linux with the
include files in the default path, or the <tt/CC65_INC/ environment variable
is set correctly).

The cl65 utility knows, how to translate C files into object files (it will
call the compiler and then the assembler). It does also know how to create
object files from assembler files (it will call the assember for that). It
knows how to build an executable (it will pass all object files to the
linker). And, finally, it has the C64 as a default target and will supply the
correct startup file and runtime library names to the linker, so you don't
have to care about that.

The one-liner above should give you a C64 executable named "<tt/hello/" in the
current directory.

For more information about the compile &amp; link utility see <htmlurl
url="cl65.html" name="cl65.html">.

<sect>Running The Executable<p>

<bf>Note: this section is incomplete!</bf>

Depending on the target, the compiler chooses several methods of making a
program available for execution.  Here we list sample emulators and
instructions for running the program.  Unless noted, similar instructions
would also apply to a real machine.

<sect1>Apple<p>

<bf>AppleWin 1.10.4</bf>: Emulates Apple II+/IIe computer, with sound, video,
joysticks, serial port, and disk images. Roms and dos disk included. Includes
monitor. Only for Windows. Unfortunately we were unable to find documentation
on running programs.  Please help.

<sect1>Atari<p>

<bf>Atari800Win Plus 3.0</bf> (available at
<url url="http://www.a800win.atari-area.prv.pl">): Emulates Atari
400/800/65XE/130XE/800XL/1200XL/5200, with stereo sound, disk images, scanline
exact NTSC/PAL video, joysticks, mouse, cartridges and ram expansions.
Includes monitor. Unfortunately only for Windows. You will need the emulator
only.  Optionally you will need atarixl.rom and/or atariosb.rom/ataribas.rom
and dos25.xfd files (not supplied).

Compile the tutorial with

<tscreen><verb>
cl65 -O -t atari hello.c text.s
</verb></tscreen>

Start the emulator, choose File>Autoboot image or File>Load executable, and point
to the hello executable.  It is customary to rename executables of this type to
hello.xex.  The file has a 7 byte header meant to be loaded directly from Atari
DOS 2/2.5 or compatibles.

On a real Atari, you would need a disk drive and Atari Dos 2.5 or compatible.
Turn on the computer, type

<tscreen><verb>
DOS
</verb></tscreen>

at the basic prompt, then choose N. CREATE MEM.SAV then choose L. BINARY LOAD
and enter HELLO.

The emulation also supports this method.  Look at Atari>Settings and check
Enable H: Patch for Hard Disk Devices, then Atari>Hard disks and set the path
of H1: to your executables directory, then use H0:HELLO.XEX in the above
proceedure to access your hardrive directly.

<bf>Note:</bf> There is no delay after the program exits.  Your C program
should wait for a keypress if you want to see any output.

<sect1>Commodore<p>

<bf>Vice 1.9</bf> (available at
<url url="ftp://ftp.funet.fi/pub/cbm/crossplatform/emulators/VICE/">):
Emulates Commodore 64/128/Vic 20/PET/CBM II computers. Missing is the Plus/4
and Commodore 16. Supports printer, serial port, stereo sound, disk drives and
images, ram expansions, cartridges, cycle exact NTSC/PAL video, mice,
joysticks. Includes monitor. Runs on Win9x/NT/2000/XP/ME/OS2/MSDOS, Beos x86,
Acorn RISC OS, and many Unixes.

Start the desired version of the emulator, choose File>Autoboot disk/tape
image, and choose your executable.  The file has a 14 byte header which
corresponds to a PRG format BASIC program, consisting of a single line;

<tscreen><code>
1000 sys2061
</code></tscreen>

On a real Commodore with attached disk drive, you would type:

<tscreen><verb>
LOAD "HELLO",8
</verb></tscreen>

for Vic 20/C64, or

<tscreen><verb>
DLOAD "0:HELLO"
</verb></tscreen>

on PET/CBM II/C128, then type

<tscreen><verb>
RUN
</verb></tscreen>

We need your help! Recommended emulators and instructions for other machines
are missing. We suggest an emulator with good compatibility. Also, being able
to run all computers in the target series is good for target compatibility
testing. A machine language monitor is almost essential for debugging, but a
native debugger could be used as well.

Finally, emulators which run on Unix/Windows would help reach a wider audience.

</article>