The CPU (fetch-decode-execute, the ISA, assembly)

The loop every chip runs, drawn as a circle: read the next instruction, work out what it wants, do it, then go back for the next one, with the one arrow that jumps out of the circle instead of moving to the next stop.

Before you start. Logic From Switches (transistors → gates → adder → memory) built an ALU, a circuit of gates that adds and compares numbers, and flip-flops, single bits of memory that hold a value until told to change. This lesson wires both into a chip that repeats one small loop forever, and by the end you can read the actual bytes that loop is running.

The idea in 30 seconds

A CPU repeats one loop forever: read the next instruction, work out what it wants, do it, go get the next one.
Registers are a handful of storage slots inside the chip; work happens on them, not straight on memory, because memory is far slower to reach.
A program is a long list of instructions; assembly is that list in human-readable form.
The same code becomes different instructions on different chip families, because each family has its own vocabulary.
A pipeline lets several instructions be in progress at once, one per stage, so a chip finishes roughly one instruction a tick instead of one every few ticks.
A wrong guess about which way a jump goes is costly: in one real test, unsorted data ran ten times slower than sorted data.

Read the story · Skip to the numbers