Timing optimization for nano-meter VLSI designs.

摘要

In nanometer technology, designers face many challenges to meet timing constraints. Three of those critical challenges are power noise, process variation, and interconnect delay. In this dissertation, we will review these challenges and also propose our solutions.; Power noise leads to IR-drop, which can affect chip timing and even functionality. To overcome the power noise problem, we propose a power-noise-reduction flow. There are two steps in the flow---prediction and correction. Prediction is done before placement to estimate decoupling-capacitance (decap) requirements for each cell. Correction is a new post-layout gate-sizing algorithm that considers power noise and timing constraints.; Because of the lithography challenges and ever-shrinking device sizes, process variation leads to uncertainty of chip performance. Some ways to adjust the chip performance after fabrication will be very useful to reduce the process variation effects. We propose a new clock architecture for FPGAs with the advantageous feature that clock skew is field-programmable. We can adjust the clock skew to achieve our timing constraints. We also propose algorithms that can conduct clock scheduling efficiently.; Although the gate delay shrinks as the technology advances, the interconnect delay does not shrink accordingly. The increased coupling capacitance and interconnect resistance keep the interconnect delay high. To tackle this interconnect-delay problem we need a better delay-budgeting algorithm that can explore various means for improvement. We propose an algorithm that combines the advantage of retiming with delay budgeting to help us achieve timing goals.; The challenges for nanometer designs are daunting. Many nanometer effects have great impact on circuit timing. We need to rely on continuous innovation to keep in step with the rapid advances. In this dissertation, we propose three techniques to answer some nanometer challenges. Our experiments have shown that these techniques are efficient and produce good results.