A single-cycle processor fetches, decodes and executes each instruction within one clock cycle. Everything happens between two edges.
What follows from that:
- CPI = 1. Every instruction takes exactly one cycle.
- The period is long. It must fit the slowest instruction's whole journey: memory read, decode, register read, ALU, another memory access, write-back. Fast instructions waste the rest of the cycle.
- No hardware reuse. Anything needed twice in one instruction must be duplicated. Fetching the instruction and reading its data both need memory in the same cycle, so there must be two memories (or two ports): a harvard architecture split. Likewise, PC + 1 needs its own adder, separate from the ALU.
Compare a multi cycle processor like Tiny8, which splits each instruction into short steps. Which is faster depends on the numbers: compute instructions × CPI × period for both, using cpu execution time.