Gate sizing under uncertainty

Abstract

We present a gate sizing approach to efficiently utilize gate switching activity (SA) and gate input vector control leakage (IVC) uncertainty factors in the objective function in order enable more efficient power and speed yield trade-offs. Our algorithm conducts iterative gate freezing and unlocking with cut-based search for the most beneficial gate sizes under delay constraints. In an iterative flow, we interchangeably conduct gate sizing and IVC refinement to adapt to new circuit configurations. We evaluate our approach on benchmarks in 45 nm technology and demonstrate up to 62% (29% avg.) energy savings compared to a traditional objective function that does not consider SA and IVC. We further adapt our approach to optimize yield objectives by addressing processing variation (PV). Significant improvements were achieved under identical timing yield targets of up to 84% max (55% avg.) and 74% max (25% avg.) mean-power savings for selected ISCAS-85 and ITC-99 benchmarks, respectively.