NBTI-aware adaptive minimum leakage vector selection using a linear programming approach

• An NBTI-aware adaptive minimum leakage vector selection method has been proposed.
• The impact of the NBTI effect on leakage power and MLV selection has been analyzed.
• SVR is used to build the model between threshold voltage change and leakage power.
• MLV is updated according to aging condition by ILP approach to reduce leakage power.
• The impact of MLV update on NBTI-induced path delay increase has been analyzed.

Due to the circuit aging effect, the minimum leakage vector (MLV) found by the traditional input vector control method may not obtain the optimal leakage power reduction result when the circuit begins to degrade. To solve this problem, we present an adaptive MLV selection strategy based on a linear programming approach. The method divides the total lifetime of the circuit into a succession of time intervals, and the MLV used in each interval is periodically updated according to the transistor׳s threshold voltage degradation so that the best overall power reduction result can be achieved. Experimental results on various benchmark circuits show the effectiveness of our method.