Assembler Basics

When you program in assembly language, you're directly programming the "bare metal" hardware. This means that many of the compile-time and run-time checks, error messages, and diagnostics are not available. The computer will follow your instructions exactly, even if they are completely wrong (like executing data), and when something goes wrong, your program won't terminate until it tries to do something that's not permitted, such as execute an invalid opcode or attempt to access a protected or unmapped region of memory. When that happens, the CPU will signal an exception, and in most cases the operating system will shut down the offending process.

Format of an Assembly Language program

The traditional extension for assembly-language source files is .s (e.g., example.s)

An assembly-language program consists of:

1. Symbols that are constants which correspond to memory addresses, registers, or other values.
2. Instructions - Mnemonics for an operation followed by zero or more arguments.
3. Data - Values used by the program for things such as initial variable values and string or numeric constants.

Assembler directives are used to control the assembly of the code, by specifying output file sections (such as .text or .data in an ELF file) and data formats (such as word size for numeric values), and defining macros.

Consider this x86_64 assembly language "Hello World" program:

.text
.global  _start

stdout = 1

_start:
        mov    $len,%rdx                       /* message length */
        mov    $msg,%rsi                       /* message location */
        mov    $stdout,%rdi                    /* file descriptor stdout */
        mov    $1,%rax                         /* syscall sys_write */
        syscall

        mov    $0,%rdi                         /* exit status */
        mov    $60,%rax                        /* syscall sys_exit */
        syscall

.data

msg:    .ascii      "Hello, world!\n"
.set len . - msg

In this program, which was written using GNU Assembler (gas) syntax, text is coloured according to its type:

• directives
• instructions
• symbols
• expressions
• other syntax elements, such as prefixes and delimiters

A symbol may be set in one of three ways:

1. Using the .set directive (in the example above, len line)
2. Using an assignment using an equals sign (in the example above, stdout line), which the GNU assembler treats as an alternative syntax for the .set directive
3. As a label (such as _start or msg in the example above). A label set in this way is identified by the trailing colon, and is set to the current memory location in the instruction or data sequence. Labels may be used for loading/storing information, or as the target of branches/jumps.

In the program above:

• .start is a directive (equivalent to the longer directive ".section .start") which specifies that the following instructions/data should be placed in the ".start" section of the output ELF file.
• .data is a similar directive which specifies that the following instructions/data should be placed in the .data section of the output ELF file. In the case of this program, they could alternately be placed in the .rodata section, which is for read-only data (data which is write-protected in memory), because the string is not modified by the program.
• .global (or .globl) is a directive which makes the following symbol visible to the linker. Otherwise, symbols are normally lost by link time. In this case, the linker needs to know the value of the special symbol _start in order to know where execution is to begin in the program (which is not always at the start of the .text section).
• _start is a label which is equivalent to the memory location of the first instruction in the program.
• msg is a label which is equivalent to the memory location of the first byte of the string "Hello, World!\n"
• .set is a directive which sets a symbol (len) equal to the value of an expression (in this example, ". - msg" meaning the current memory location minus the value of the label "msg"). Note that the GNU assembler accepts a=1 as equivalent to .set a 1 -- both are counted as directives regardless of the presence of the .set keyword.

Note that symbols are not variables - they are constants that are calculated at compile-time. However, they may contain an address at which a variable is stored.

Note also that the syntax will vary from assembler to assembler and from architecture to architecture.

Instruction Set Architecture Information

To get started with specific instruction set architectures, see:

• x86_64 Register and Instruction Quick Start
• aarch64 Register and Instruction Quick Start

Resources

• Using as, The GNU Assembler (GAS manual)