Machine code

Machine code is the fundamental data structure used to control a computer's central processing unit (CPU). It consists of encoded sequences that the CPU interprets directly through its programmable interface. Because the CPU reads this language without needing a translator, machine code is considered native to its host processor. Most computer programs are primarily composed of these machine-code instructions. Without this low-level language, the hardware would have no way to execute the tasks required by software.

Each machine instruction is a specific pattern of bits that triggers the computer's digital logic. These instructions act on various hardware components to perform precise tasks. For example, an instruction might load a word from memory into a CPU register. It might also command the arithmetic logic unit (ALU) to perform a calculation on specific data. Other instructions allow the CPU to jump or skip to a different instruction in the sequence. This process allows the processor to control registers, the bus, and memory to execute complex programs.

Instruction sets are organized into different formats and types. Instructions can be general-purpose or special-purpose. General-purpose instructions handle common tasks like data movement, comparisons, and loop control. Special-purpose instructions exploit unique features found only in a specific computer architecture. Furthermore, instructions may vary in length. Some sets use instructions that are all the same size, while others use variable lengths. Designers must also decide how many bits to allocate to the address field. A shorter address field can increase execution speed, but it might limit the amount of physical memory the CPU can access.

To make programming easier, humans often use assembly language instead of writing raw machine code. Assembly language provides a direct mapping to machine code by using mnemonics and labels. For instance, in an x86 processor, the assembly mnemonic NOP represents the opcode 0x90. While possible, writing programs directly in machine code is considered tedious and error-prone. Most modern software is instead written in high-level programming languages. These languages are then translated down into the machine code that the hardware understands.

An instruction set architecture (ISA) defines the specific interface for a CPU. Different families of CPU design, such as x86 or ARM, use different ISAs. Generally, machine code designed for one family will not work on another. However, some architectures provide optional support for other instruction sets. The VAX architecture can support the PDP-11 set, and the IA-64 architecture can support the IA-32 set. The PowerPC 615 is another example, as it can natively process both PowerPC and x86 instructions.

In certain systems, an even deeper layer called microcode exists. Microcode provides a common machine language interface across different models of a computer family. This helps developers port machine language programs between models with different dataflows. The IBM System/360 family is a notable example of this approach. Additionally, some architectures use a technique called overlapping instructions. In the 1970s and 1980s, this was used to save memory space. Today, it is rarely used except for extreme optimization, such as in boot loaders, or for code obfuscation to prevent tampering.

Understanding machine code is difficult because it is not human-readable. The writer Douglas Hofstadter compared examining machine code to looking at the atoms of a DNA molecule. To help programmers, several specialized tools are used. A disassembler decodes machine code back into assembly language. A decompiler attempts to convert it into a high-level language, though the result can be hard to read. Debuggers allow programmers to interact with the code to find and fix errors. These tools often use debug symbols to map the machine code back to the original source code.