| Summary: | 碩士 === 國立交通大學 === 資訊科學系所 === 94 === In this work, we have proposed crosstalk-driven and RLC-bounded track assignment algorithms. First, RBTA first applies LEA to produces a utilization-driven TA, and then transforms the crosstalk minimization problem into finding a MWHP by the proposed new track overlap graph. Experimental results show that RBTA algorithm reduced crosstalk 32.33% and produced fewer failed nets than previous works. In second, we proposed an enhanced finding clique heuristic. For each clique in IRoute OLG in the decreasing order of clique size, RLC-bounded TA first apply initial assignment to produce a capacitance-free TA result, and then if inductance coupling occurred, we use Tabu search to reduce inductance coupling more. Experimental results show that Tabu search phase reduced inductance coupling 66.9%, 62.80% and 55.44% when the sensitive rate is 25%, 33% and 50%, respectively. In the future, we hope to develop a row-based RLC-bounded track assignment algorithm.
|
|---|
