Little-endian and Big-endian

I was watching the prequel series to Casey Muratori’s excellent Handmade Hero last weekend and there was a small segment about endianness. I had heard about this in passing before, but the Wikipedia article was too dry for me to tackle at the time.

Basically, endianness refers to how bytes for numerical values are formatted in memory.

Casey explained it in very simple terms: an architecture is little-endian if the lower order bit comes first in memory. It’s big-endian if it’s vice versa.

Example

Say that we have a number 511 that we want to store in memory. Let’s assume that each address in our memory holds a value that is a byte (8-bits) large.

To store 511, we’ll need 2 bytes: 1 byte to hold 255 and another to hold 256.

255 + 256 = 511

Let’s say that we store that number at contiguous addresses: 0x9000 and 0x9001. If our architecture is little-endian, the memory will look like this:

In big-endian architecture, it will look like this:

What does this mean for me?

This is only a problem when a binary created on a certain endianness is used on a machine with conflicting endianness, because the program will expect its native endianness when manipulating memory.

For the most part, most machines nowadays are on little-endian architecture because of the pervalence of Intel processors (x86, x64) and ARM chips, so endianness doesn’t really affect us.

Interestingly, big-endian is more common in networking, which is why it is also known as network byte order.