论文信息 - A parallel sparse linear system solver for large-scale circuit simulation based on Schur Complement

A parallel sparse linear system solver for large-scale circuit simulation based on Schur Complement

In this paper, the parallelism of large sparse matrix solve in circuit simulation is studied. We present an efficient matrix ordering strategy to permute the circuit conductance matrix into Bordered Block Diagonal (BBD) form with a small connection matrix. Target on multi-core and shared memory architecture, the sparse linear system is solved in parallel using Schur Complement algorithm. The suitable matrix partition dimension is investigated for better performance. The experimental results suggest an average 6.8x speedup when our method is in comparison with KLU.

[1] J. Gilbert,et al. Sparse Partial Pivoting in Time Proportional to Arithmetic Operations , 1986 .

[2] Timothy A. Davis,et al. Algorithm 832: UMFPACK V4.3---an unsymmetric-pattern multifrontal method , 2004, TOMS.

[3] Timothy A. Davis,et al. The university of Florida sparse matrix collection , 2011, TOMS.

[4] Wayne B. Hayes,et al. Algorithm 908 , 2010 .

[5] Sachin S. Sapatnekar,et al. Proceedings - Design Automation Conference , 2004, DAC 2004.

[6] Alan F. Murray,et al. Proceedings - IEEE International Symposium on Circuits and Systems , 2006 .

[7] L. Dagum,et al. OpenMP: an industry standard API for shared-memory programming , 1998 .

[8] Timothy A. Davis,et al. Algorithm 837: AMD, an approximate minimum degree ordering algorithm , 2004, TOMS.

[9] James Demmel,et al. A Supernodal Approach to Sparse Partial Pivoting , 1999, SIAM J. Matrix Anal. Appl..

[10] Uwe Jaekel,et al. Parallel iterative solvers for sparse linear systems in circuit simulation , 2005, Future Gener. Comput. Syst..

[11] Iain S. Duff,et al. Algorithm 575: Permutations for a Zero-Free Diagonal [F1] , 1981, TOMS.