PROJECT TITLE :

A Selected Inversion Approach for Locality Driven Vectorless Power Grid Verification

ABSTRACT:

Vectorless power grid verification may be a sensible approach for early stage safety check without input current patterns. The ability grid is usually formulated as a linear system and needs intensive matrix inversion and various linear programming (LP), that is very time-consuming for massive-scale power grid verification. During this paper, the facility grid is represented in the style of domain-decomposition approach, and we tend to propose a specific inversion technique to scale back the computation price of matrix inversion for vectorless verification. The locality existence among power grids is exploited to make your mind up which blocks of matrix inversion ought to be computed whereas remaining blocks don't seem to be necessary. The vectorless verification might be purposefully performed by this way of selected inversion, whereas previous direct approaches are needed to perform full matrix inversion and then discard tiny entries to cut back the complexity of LP. Meanwhile, constraint locality is proposed to boost the verification accuracy. Additionally, an idea of quasi-Poisson block is introduced to use grid locality among realistic power grids and a scheme of pad-aware partitioning is proposed to enable the chosen inversion approach available for practical use. Experimental results show that the proposed approach might achieve significant speedups compared with previous approaches while still guaranteeing the quality of solution accuracy.