Pipelining (OCR A Level Computer Science): Revision Note

Pipelining

What is pipelining?

• In A Level Computer Science, pipelining is the process of carrying out multiple instructions concurrently

• Each instruction will be at a different stage of the fetch-decode-execute cycle

• One instruction can be fetched while the previous one is being decoded and the one before is being executed

• This table shows which stage each instruction is at during each step:

• While one instruction is being executed, the next instruction will be decoded and the following instruction will be fetched

How does pipelining improve processor performance?

• Pipelining increases instruction throughput by overlapping stages, while latency per instruction remains unchanged

• The CPU is not idle while waiting for the next instruction which increases the speed of execution

• The next instruction is fetched while the current one is decoded/executed

• All parts of the processor can be used at any instant in time

What happens to the pipeline when a branch is taken?

• Pipelining works best when instructions run in sequence, because the next instructions are fetched before the current one has been executed

• If a branch instruction is executed and the branch is taken, the instructions already fetched and decoded behind it are the wrong ones

• These instructions are discarded: the pipeline is flushed

• The processor then fetches from the new address in the PC, so the pipeline has to refill and some stages sit idle

• This table shows instruction A as a branch to instruction X:

• Instructions B and C are flushed, so no instruction is executed in steps 4 and 5

• Programs with many branches therefore gain less from pipelining than programs that run mostly in sequence