"Word" is an outdated concept we should try to get rid of.
You're right. To be consistent with bytes we should call it a snack.
Appeasing that attitude is what prevented Microsoft from migrating to LP64. Would have been an easier task if their 32-bit LONG type never existed, they stuck with DWORD, and told the RISC platforms to live with it.
It's very useful on hardware that is not an x86 CPU.
As an abstraction on the size of a CPU register, it really turned out to be more confusing than useful.
On RISC machines, it can be very useful to have the concept of "words," because that indicates things about how the computer loads and stores data, as well as the native instruction size. In DSPs and custom hardware, it can indicate the only available datatype.
The land of x86 goes to great pains to eliminate the concept of a word at a silicon cost.
Fortunately we have `register_t` these days.
Is it 32 or 64 bits on ARM64? Why not both?
such as...?
How exactly ? How else do you suggest CPUs do addressing ?
Or are you suggesting to increase the size of a byte until it's the same size as a word, and merge both concepts ?
I'm saying the term "Word" abstracting the number of bytes a CPU can process in a single operation is an outdated concept. We don't really talk about word-sized values anymore. Instead we mostly explicit on the size of value in bits. Even the idea of a CPU having just one relevant word size is a bit outdated.