7.2.2 Pipelined Circuits

TL;DR

Pipelining a combinational logic circuit can increase system throughput at the cost of a slightly longer latency.

Transcript

Okay, now let's apply all this analysis to improving the performance of our circuits. The latency of a combinational logic circuit is simply its propagation delay t_PD. And the throughput is just 1/t_PD since we start processing the next input only after finishing the computation on the current input. Consider a combinational system with three comp... Read More

Key Insights

• ❓ Latency of a combinational logic circuit is determined by its propagation delay.
• 🗂️ Throughput can be increased by using registers and dividing the processing into stages.
• 🏍️ Pipelining involves capturing outputs from one stage and using them as inputs for the next stage, progressing one stage per clock cycle.
• 🐢 The clock period should be determined by the slowest processing stage in order to ensure correct operation.
• ⏲️ The latency of a pipelined system is determined by the number of stages times the clock period.
• 🛩️ Pipelined systems have better throughput at the cost of a small increase in latency.
• 🆘 Pipeline diagrams help visualize the operation of a pipelined system.

Questions & Answers

Q: What determines the latency of a combinational logic circuit?

The latency is determined by the propagation delays of the individual components in the circuit.

Q: Can we increase the throughput of a combinational logic circuit?

Yes, by using registers to divide the processing into stages and allowing parallel computation, we can increase system throughput.

Q: How does pipelining work?

Pipelining involves capturing outputs from one stage using registers and using them as inputs for the next stage. Each stage progresses one step per clock cycle.

Q: What is the trade-off of pipelining?

Pipelining increases system throughput but can lead to a slightly longer latency compared to an unpipelined system.

Summary & Key Takeaways

• Combinational logic circuits have a latency determined by their propagation delay.

• By dividing the processing into stages and using registers, we can increase the throughput of the system.

• Pipelining involves capturing outputs from one stage and using them as inputs for the next stage, progressing one stage per clock cycle.