How does a CPU process data?

1the loop

A processor knows no photos, no words, no Python. It repeats one loop: fetch an instruction, decode it, execute it, store the result. At four gigahertz, that is four billion loops a second.

2four instructions

Take x equals a plus b. The compiler turns it into four instructions: load a, load b, add, store x. Instructions and data are both just numbers in memory.

3the machine

Here is the machine. Memory on the left holds instructions and data. Inside the chip, a program counter tracks the next instruction, and a few registers hold the numbers being used.

4load

Instructions one and two load the numbers: seven and five are copied from memory into registers r1 and r2. Registers are the only storage the arithmetic unit touches.

5the ALU

Instruction three, add, goes to the arithmetic logic unit, the A-L-U: a circuit of logic gates that takes two inputs and produces their sum.

6write back

The sum, twelve, lands in register r3. Instruction four stores it back to memory as x. Then the program counter steps on and the loop begins again.

7memory is slow

Here is the catch. A register answers in about a quarter nanosecond, the closest cache in about one, but main memory takes roughly a hundred. That is hundreds of cycles for one read.

8caches

So chips keep recently used data in small, fast caches, betting you will reuse it. A hit costs a few cycles. A miss stalls the core while it waits.

9the pipeline

Doing the loop one instruction at a time would leave most of the hardware idle. A pipeline overlaps the stages: while one executes, the next decodes and the one after is fetched.

10going wide

Modern cores go further. They run several pipelines side by side, execute out of order, guess which way branches go, and a chip packs many cores, each with its own loop.

11the whole idea

So, the whole idea: every program is that loop of fetch, decode, execute and write back, with fast caches and overlapped stages hiding the slow trip to memory.