[ << ] | [ >> ] | [] | [] | [] | [ ? ] |
This chapter describes the module-independent part of the assembler. It documents the options and extensions which are not specific to a certain target, syntax or output driver. Be sure to also read the chapters on the cpu-backend, syntax- and output-module you are using. They will likely contain important additional information like data-representation or additional options.
vasm
is run from the command line using the following syntax:
vasm<target>_<syntax> [options] [sourcefile]
When the source file name is missing the assembler reads the source
text from ‘stdin’ until encountering EOF
(end-of-file,
which is CTRL-D
in Unix shells, or CTRL-\
in AmigaOS
shells). Note, that most debugging formats (DWARF, etc.) no longer
function with such temporary source texts.
The following options are supported by the machine independent part
of vasm
. The most important ones are:
Defines a symbol with the name <name>
and assigns the value of the
expression when given. The assigned value defaults to 1 otherwise.
Use module <fmt>
as output driver. See the chapter on output
drivers for available formats and options.
Define another include path. They are searched in the order of occurrence on the command line, and always before any include paths defined in the source.
Enables generation of a listing file and directs the output into
the file <listfile>
.
Disables case-sensitivity for everything - identifiers, directives and instructions. Note that directives and instructions may already be case-insensitive by default in some modules.
Strips all local symbols from the output file and doesn’t include any other symbols than those which are required for external linkage.
Write the generated assembler output to <ofile>
rather than
‘a.out’.
Do not print the copyright notice and the final statistics.
Show an error message, when referencing an undefined symbol. The default behaviour is to declare this symbol as externally defined.
Other options:
Issues a warning when a label matches a mnemonic or directive name in either upper or lower case.
Print all dependencies while assembling the source with the given
options. No output is generated. <type>
may be the word ‘list’
for printing one file name in each new line, or ‘make’ for
printing a sequence of file names on a single line, suitable for
Makefiles.
When the output file name is given by ‘-o outname’ then
vasm
will also print outname:
in front of it.
Note that in contrast to option ‘-dependall’ only relative
include file dependencies will be listed (which is the common case).
Prints dependencies in the same way as ‘-depend’, but also prints all include files with absolute paths.
Used together with ‘-depend’ or ‘-dependall’ and instructs vasm to output all dependencies into a new file, instead of stdout. Additionally, code will be generated in parallel to the dependencies output.
Automatically generate DWARF debugging sections, suitable for
source level debugging. When the version specification is missing,
DWARF V3 will be emitted. The only difference to V2 is that it
creates a .debug_ranges
section, with address ranges for all
sections, instead of using a workaround by specifying
DW_AT_low_pc=0
and DW_AT_high_pc=~0
.
Note, that when you build vasm from source, you may have to set
your host operating system with -Dname
in the Makefile to
include the appropriate code which can determine the current
work directory. Otherwise the default would be to set the current
work directory to an empty string. Currently supported are:
AMIGA
, ATARI
, MSDOS
, UNIX
,
_WIN32
.
Enable escape character sequences. This will make vasm treat the escape character \ in string constants similar as in the C language.
Use big-endian order when reading target-bytes with more than 8 bits per byte from the host’s file system (default).
When the same file is included multiple times, using the same path, this is silently ignored, causing the file to be processed only once. Note, that you can still include the same file twice when using different paths to access it.
Use little-endian order when reading target-bytes with more than 8 bits per byte from the host’s file system.
List all symbols, including unused equates. Default is to list all labels and all used expressions only.
Set the maximum number of bytes per line in a listing file to <n>
.
Defaults to 8 (fmt=wide
).
Set the listing file format to <fmt>
. Defaults to wide.
Available are: wide
, old
.
Show only program labels in the sorted symbol listing. Default is to list all symbols, including absolute expressions.
Do not show included source files in the listing file (fmt=wide
).
Do not include symbols in the listing file (fmt=wide
).
Sets the maximum number of errors to display before assembly
is aborted. When <n>
is 0 then there is no limit. Defaults to 5.
Defines the maximum number of recursion levels within a macro. Defaults to 1000.
Adjusts the maximum number of passes while resolving a section. Defaults to 1500.
Do not search for include files relative to the compile directory (where the main input source is located).
No escape character sequences. This will make vasm treat the escape character \ as any other character. Might be useful for compatibility.
Perform no automatic alignment for instructions. Note that unaligned instructions can make your code crash when executed! Only set when you know what you are doing!
Disable the informational message <n>
. <n>
has to be the number
of a valid informational message, like an optimization message.
Disable warning message <n>
. <n>
has to be the number of a valid
warning message, otherwise an error is generated.
Write target-bytes with more than 8 bits per byte in big-endian order to the host’s file system (default).
Write target-bytes with more than 8 bits per byte in little-endian order to the host’s file system.
The given padding value can be one or multiple bytes (up to the
cpu-backend’s address size). It is used for alignment purposes
and to fill gaps between absolute ORG
sections in the
binary output module. Defaults to a zero-byte.
Try to generate position independent code. Every relocation position is flagged by an error message.
Do not interpret a source path starting with ’/’ or ’\’, or including a colon, as absolute, but always attach it relative to defined include paths first.
Add a leading underscore in front of all imported and exported (also common, weak) symbol names, just before writing the output file.
Sections are no longer distinguished by their name, but only by their attributes. This has the effect that when defining a second section with a different name but same attributes as a first one, it will switch to the first, instead of starting a new section. Is enabled automatically, when using an output-module which doesn’t support section names. For example: aout, tos, xfile.
The shift-right operator (>>
) treats the value to shift as
unsigned, which has the effect that only 0 bits are inserted on the
left side. The number of bits in a value depend on the target
address type (refer to the appropriate cpu module documentation).
Uninitialized memory regions, declared by "space" directives
(.space
in std-syntax, ds
in mot-syntax, etc.)
are filled with the given value. Defaults to zero.
Hide all warning messages.
The return code of vasm will no longer be 0 (success), when there was a warning. Errors always make the return code non-zero (failure).
Print version and copyright messages from the assembler and all its modules, then exit.
Note, that while most options allow an argument without any separating blank, some do require it (e.g. ‘-o’ and ‘-L’).
Standard expressions are usually evaluated by vasm’s core routines rather than by one of the modules (unless this is necessary).
All expressions evaluated by the frontend are calculated in terms of target address values, i.e. the range depends on the backend. Constants which exceed the target address range may be supported by some backends up to 128 bits.
Backends also have the option to support floating point constants directly and convert them to a backend-specific format which is described in the backend’s documentation.
Warning: Be aware that the quality and precision of the backend’s floating point output depends on the combination of host- and backend-format! If you need absolute precision, encode the floating point constants yourself in binary.
The available operators include all those which are common in assembler as well as in C expressions.
C like operators:
+ - ! ~
+ - * / % << >>
& | ^
&& ||
< > <= >= == !=
Assembler like operators:
+ - ~
+ - * / // << >>
& ! ~
< > <= >= = <>
Up to version 1.4b the operators had the same precedence and associativity as in the C language. Newer versions have changed the operator priorities to comply with common assembler behaviour. The expression evaluation priorities, from highest to lowest, are:
+ - ! ~
(unary +/- sign, not, complement)
<< >>
(shift left, shift right)
&
(bitwise and)
^ ~
(two forms of bitwise exclusive-or)
| !
(two forms bitwise inclusive-or)
* / % //
(multiply, divide, modulo, modulo)
+ -
(plus, minus)
< > <= >=
(less, greater, less or equal, greater or equal)
== = != <>
(two forms of equality and inequality)
&&
(logical and)
||
(logical or)
Operands are integral values of the target address type. They can either be
specified as integer constants of different bases (see the documentation
on the syntax module to see how the base is specified) or character
constants. Character constants are introduced by '
or "
and have to be terminated by the same character that started them.
Multiple characters are allowed and a constant is built according to the endianness of the target.
When the ‘-esc’ option was specified, or automatically enabled by a syntax module, vasm interprets escape character sequences as in the C language:
\\
Produces a single \
.
\b
The bell character.
\f
Form feed.
\n
Line feed.
\r
Carriage return.
\t
Tabulator.
\"
Produces a single "
.
\'
Produces a single '
.
\e
Escape character (27).
\<octal-digits>
One character with the code specified by the digits as octal value.
\x<hexadecimal-digits>
One character with the code specified by the digits as hexadecimal value.
\X<hexadecimal-digits>
Same as \x
.
Note, that the default behaviour of vasm has changed since V1.7! Escape
sequence handling has been the default in older versions. This was
changed to improve compatibility with other assemblers. Use ‘-esc’
to assemble sources with escape character sequences. It is still the
default in the std
syntax module, though.
You can define as many symbols as your available memory permits. A symbol may have any length and can be of global or local scope. Internally, there are three types of symbols:
Expression
These symbols are usually not visible outside the source, unless they are explicitly exported.
Label
Labels are always addresses within a program section. By default they have local scope for the linker.
Imported
These symbols are externally defined and must be resolved by the linker.
Beginning with vasm V1.5c at least one expression symbol is always defined
to allow conditional assembly depending on the assembler being used:
__VASM
. Its value depends on the selected cpu module.
Since V1.8i there may be a second internal symbol which reflects the format of the paths in the host file system. Currently there may be one of:
__UNIXFS
Host file system uses Unix-style paths.
__MSDOSFS
Host file system uses MS-DOS-, Windows-, Atari-style paths.
__AMIGAFS
Host file system uses AmigaDOS-style paths.
Note that such a path-style symbol must be configured by a -D
option
while compiling vasm from source. Refer to the section about
building vasm (Interface chapter) for a listing of all supported host
OS options.
There may be other internal symbols, which are defined by the syntax- or by the cpu module.
Vasm supports include files and defining include paths. Whether this functionality is available depends on the syntax module, which has to provide the appropriate directives.
On startup vasm defines one or two default include paths: the current work directory and, when the main source is not located there, the compile directory.
Include paths are searched in the following order:
Additionally, all the relative paths, defined by ‘-I’ or directives, are first appended to the current work directory name, then to the compile directory name, while searching for an include file.
Searching for include files in paths based on the compile directory can be completely disabled by ‘-nocompdir’.
Macros are supported by vasm, but the directives for defining them have
to be implemented in the syntax module. The assembler core supports 9
macro arguments by default to be passed in the operand field,
which can be extended to any number by the syntax module.
They can be referenced within the macro either by name (\name
) or by
number (\1
to \9
), or both, depending on the syntax module.
Recursions and early exits are supported.
Refer to the selected syntax module for more details.
Vasm supports structures, but the directives for defining them have to be implemented in the syntax module.
Has to be provided completely by the syntax module.
Some known module-independent problems of vasm
at the moment:
All those who wrote parts of the vasm
distribution, made suggestions,
answered my questions, tested vasm
, reported errors or were otherwise
involved in the development of vasm
(in descending alphabetical order,
under work, not complete):
The frontend has the following error messages:
[ << ] | [ >> ] | [] | [] | [] | [ ? ] |