Alignment directives

Description

num_bytes

This specifies the number of bytes that must be aligned to. This must be a power of 2.

exponent

This specifies the alignment boundary as an exponent. The actual alignment boundary is 2exponent.

fill_value

The value to fill any inserted padding bytes with. This value is optional.

Operation

The alignment directives align the current location in the file to a specified boundary. The unused space between the previous and the new current location are filled with:

• Copies of fill_value, if it is specified. The width of fill_value can be controlled with the w and l suffixes, see below.
• NOP instructions appropriate to the current instruction set, if all the following conditions are specified:
  • The fill_value argument is not specified.
  • The w or l suffix is not specified.
  • The alignment directive follows an instruction.
• Zeroes otherwise.

The .balign directive takes an absolute number of bytes as its first argument, and the .p2align directive takes a power of 2. For example, the following directives align the current location to the next multiple of 16 bytes:

• .balign 16
• .p2align 4
• .align 4

The w and l suffixes modify the width of the padding value that will be inserted.

• By default, the fill_value is a 1-byte value.
• If the w suffix is specified, the fill_value is a 2-byte value.
• If the l suffix is specified, the fill_value is a 4-byte value.

If either of these suffixes are specified, the padding values are emitted as data (defaulting to a value of zero), even if following an instruction.

The .align directive is an alias for .p2align, but it does not accept the w and l suffixes.

Alignment is relative to the start of the section in which the directive occurs. If the current alignment of the section is lower than the alignment requested by the directive, the alignment of the section will be increased.

Usage

Use the alignment directives to ensure that your data and code are aligned to appropriate boundaries. This is typically required in the following circumstances:

• In T32 code, the ADR instruction and the PC-relative version of the LDR instruction can only reference addresses that are 4-byte aligned, but a label within T32 code might only be 2-byte aligned. Use .balign 4 to ensure 4-byte alignment of an address within T32 code.
• Use alignment directives to take advantage of caches on some Arm processors. For example, many processors have an instruction cache with 16-byte lines. Use .p2align 4 or .balign 16 to align function entry points on 16-byte boundaries to maximize the efficiency of the cache.

Examples

Aligning a constant pool value to a 4-byte boundary in T32 code:

    get_val:
      ldr r0, value
      adds r0, #1
      bx lr
      // The above code is 6 bytes in size. 
      // Therefore the data defined by the .word directive below must be manually aligned
      // to a 4-byte boundary to be able to use the LDR instruction.
      .p2align 2
    value:
      .word 42

Ensuring that the entry points to functions are on 16-byte boundaries, to better utilize caches:

      .p2align 4
      .type func1, "function"   
    func1:
      // code

      .p2align 4
      .type func2, "function"   
    func2:
      // code