PERFORMANCE AND POWER OPTIMIZATION IN VLSI PHYSICAL DESIGN A Thesis

As VLSI technology enters the nanoscale regime, a great amount of efforts have been made to reduce interconnect delay. Among them, buffer insertion stands out as an effective technique for timing optimization. A dramatic rise in on-chip buffer density has been witnessed. For example, in two recent IBM ASIC designs, 25% gates are buffers. In this thesis, three buffer insertion algorithms are presented for the procedure of performance and power optimization. The second chapter focuses on improving circuit performance under inductance effect. The new algorithm works under the dynamic programming framework and runs in provably linear time for multiple buffer types due to two novel techniques: restrictive cost bucketing and efficient delay update. The experimental results demonstrate that our linear time algorithm consistently outperforms all known RLC buffering algorithms in terms of both solution quality and runtime. That is, the new algorithm uses fewer buffers, runs in shorter time and the buffered tree has better timing. The third chapter presents a method to guarantee a high fidelity signal transmission in global bus. It proposes a new redundant via insertion technique to reduce via variation and signal distortion in twisted differential line. In addition, a new buffer insertion technique is proposed to synchronize the transmitted signals, thus further improving the effectiveness of the twisted differential line. Experimental re-iv sults demonstrate a 6GHz signal can be transmitted with high fidelity using the new approaches. In contrast, only a 100MHz signal can be reliably transmitted using a single-end bus with power/ground shielding. Compared to conventional twisted differential line structure, our new techniques can reduce the magnitude of noise by 45% as witnessed in our simulation. The fourth chapter proposes a buffer insertion and gate sizing algorithm for million plus gates. The algorithm takes a combinational circuit as input instead of individual nets and greatly reduces the buffer and gate cost of the entire circuit. The algorithm has two main features: 1) A circuit partition technique based on the criticality of the primary inputs, which provides the scalability for the algorithm, and 2) A linear programming formulation of non-linear delay versus cost tradeoff, which formulates the simultaneous buffer insertion and gate sizing into linear programming problem. Experimental results on ISCAS85 circuits show that even without the circuit partition technique, the new algorithm achieves 17X speedup compared with path based algorithm. In the meantime, the new algorithm saves 16.0% buffer cost, 4.9% gate cost, 5.8% …