Endianness and Alignment at the Hardware Level

This chapter explains endianness (byte order) and alignment as hardware properties, and how they surface in C code.

Why This Matters

Byte order determines how multi-byte integers are stored in memory, which matters for network protocols, file formats, and FFI. Alignment determines where objects may live and whether a pointer access is valid. Both are frequent sources of subtle portability bugs.

Prerequisites

• c.cpu.1 — CPU registers and memory access.

Core Concept

Endianness

A multi-byte integer is stored in consecutive bytes. The order is either:

• Little-endian: least significant byte at the lowest address (x86, most

ARM in practice).

• Big-endian: most significant byte at the lowest address (some network

protocols, some embedded SoCs).

For 0x01020304 at address A:

Alignment

An object of type T must be stored at an address divisible by its alignment (_Alignof(T)). Misaligned access is UB in C, and on some hardware it faults or is very slow.

Examples

Detecting endianness at run time (portably enough)

#include <stdint.h>
#include <stdio.h>

int main(void)
{
    uint16_t x = 0x0102;
    unsigned char *b = (unsigned char *)&x;
    if (b[0] == 0x02)
        puts("little-endian");
    else
        puts("big-endian");
    return 0;
}

Reading the representation through unsigned char is well-defined.

Alignment demonstration

#include <stdio.h>
#include <stdalign.h>

int main(void)
{
    printf("alignof(int)   = %zu\n", alignof(int));
    printf("alignof(double)= %zu\n", alignof(double));
    return 0;
}

How It Works

Endianness is a property of how the CPU stores/loads multi-byte values. Alignment is a hardware constraint: the CPU loads aligned words efficiently (or at all). The C compiler honors the platform's alignment by padding structs and aligning stack/heap objects.

Variations

Network byte order

Network protocols conventionally use big-endian ("network byte order"), which is why htons/ntohl conversions exist (c.net.4).

Bi-endian

Some architectures can switch endianness (e.g., ARM in some modes), but a given process is one or the other.

Common Mistakes

• Writing multi-byte integers directly to a file or socket and assuming the

reader shares the same endianness.

• Dereferencing a pointer cast to a wider type from an unaligned byte buffer.
• Assuming all ARM is little-endian (it is configurable; network/embedded

parts can be big-endian).

Undefined Behavior

• Dereferencing a misaligned pointer. VERIFIED
• Reading a multi-byte value through an lvalue of an incompatible type (strict

aliasing) — use unsigned char or memcpy.

Portability

• Endianness is implementation-defined (hardware). Alignment is

implementation-defined. Use explicit byte-wise serialization for portability.

Under the Hood

Little-endian CPUs have byte-swap instructions (bswap on x86) for converting to/from network order. Misaligned loads are handled by hardware (x86, slowly) or fault (some ARM/RISC-V), or by the compiler via multiple byte loads.

Practical Usage

• Serialize integers byte-by-byte (or use explicit endian conversion) for

network/file data.

• Keep buffers aligned; use memcpy to read unaligned data safely.
• Use htons/ntohl/htonl/ntohs for network order.

Exercises

1. Write a program that detects your platform's endianness. 2. Convert a uint32_t to and from big-endian bytes using shifts. 3. Demonstrate that memcpy can safely read an unaligned int from a byte buffer.

Deep Challenge

Write a portable read_u32_be(const unsigned char *p) and read_u32_le that decode big-endian and little-endian integers from a byte buffer without alignment or aliasing UB, and explain why they are portable.

Related Concepts

• c.net.4 — byte order and endian conversion.
• c.obj.alignment — alignment.
• c.obj.representation — object representation.

References

• CPU architecture manuals; ISO C §6.2.6 (representations).

Verification

• Endianness and alignment are hardware/implementation-defined. HARDWARE
• Misaligned dereference is UB. VERIFIED
• No example was executed during generation unless noted. `Execution not

verified.`

Progress

• [ ] Read
• [ ] Understand
• [ ] Complete examples
• [ ] Complete exercises
• [ ] Complete deep challenge

Concept checkboxes

• [ ] Little vs. big endian
• [ ] Network byte order
• [ ] Hardware alignment
• [ ] Misaligned access UB

Mastery levels