A new algorithm for simultaneous gate sizing and threshold voltage assignment

Abstract

Gate sizing and threshold voltage (V t ) assignment are popular techniques for circuit timing and power optimization. Existing methods, by and large, are either sensitivity-driven heuristics or based on rounding continuous optimization solutions. Sensitivity-driven heuristics are easily trapped in local optimum and the rounding is subject to remarkable errors. In this paper, we propose a systematic combinatorial approach for simultaneous gate sizing and V t assignment. The core ideas of this approach include consistency relaxation and coupled bi-directional search. Our algorithm is compared with a state-of-the-art previous work on benchmark circuits. The results from our algorithm can lead to about 24% less power dissipation subject to the same timing constraints.

Collaborative gate implementation selection and adaptivity assignment for robust combinational circuits

“…That is, if a block is assigned with adaptivity, our implementation selection must be performed with anticipation of performancepower changes due to the adaptivity. We make such sophisticated enhancement over a previous work of deterministic gate implementation selection [14]. Moreover, we propose a new technique to avoid redundant counting when candidate solutions are propagated in circuit traversals.…”

“…The problem of finding λ to maximize the optimal solution to the subproblem is the Lagrangian dual problem. Like in [14], [16], we solve the dual problem using subgradient method guided by SSTA. By doing so, the Lagrangian subproblem is allowed to use simple, less accurate timing and variability models.…”

“…We solve the reduced subproblem L µ ( ξ) by Joint Relaxation and Restriction (JRR) [14]. The candidate solution propagations on a DAG faces another problem.…”

Lagrangian relaxation-based Discrete Gate Sizing for leakage power minimization

“…In [9] Chen used LR for gate sizing, solving LRS by a greedy heuristic and LDP by the subgradient method. An important contribution of this work was the application of the Karush-Kuhn-Tucker (KKT) conditions to simplify the LRS, which was also employed by a number of other works (e.g., [10], [11], [12], [13] and [14]). Tennakoon [10] proposed a technique to estimate the initial values to speedup LDP convergence.…”

“…Rahman [12] proposed a continuous sizing algorithm along with a branch-and-bound discretization procedure. Zhou [13] improved [11] by identifying some limitations on their delay modeling and used LR for timing minimization. Huang [14] focused on some convergence issues on the subgradient method to solve the LDP and proposed a projection-based method.…”