# Instruction parsing (ivss)

**Instruction parsing** identifies a mnemonic and checks the shape of its operands before encoding. ivss describes accepted forms in an instruction table. Its entry for `add` is:

```c
{ "add", I_ADD, ALU_ADD, 2, I_ARGS_REG_REG },
```

The fields select the instruction kind, arithmetic operation, expected argument count, and operand format. For the source statement:

```asm
add %r2, %r1
```

The parser splits the arguments at the comma and interprets them as source and destination registers. The destination is also an input to the addition, so the statement means `%r1 = %r1 + %r2`.

The outer line parser separates an optional label from a directive or instruction. Operand-specific functions then distinguish registers, literals, symbols, and memory forms. For example, the `ld` entry accepts a data operand followed by a register, rather than the two-register form required here.

Parsing establishes which operation was requested. [[ivss-instruction-encoding|Encoding]] turns that operation into bytes, and [[ivss-pseudoinstructions|expansion]] may produce several machine instructions for one accepted statement.

Source: [asembler.c](https://github.com/yanevskiv/ivss/blob/main/src/asembler.c).