Fast and memory-efficient GPU implementations of krylov subspace methods for efficient power grid analysis

Authors:
Takumi Morishita;Hiroshi Tsutsui;Hiroyuki Ochi;Takashi Sato
Affiliations:
Kyoto University, Kyoto, Japan;Kyoto University, Kyoto, Japan;Kyoto University, Kyoto, Japan;Kyoto University, Kyoto, Japan
Venue:
Proceedings of the 23rd ACM international conference on Great lakes symposium on VLSI
Year:
2013

Citing 6
Cited 0

GMRES: a generalized minimal residual algorithm for solving nonsymmetric linear systems

SIAM Journal on Scientific and Statistical Computing
Power grid analysis benchmarks

Proceedings of the 2008 Asia and South Pacific Design Automation Conference
Multigrid on GPU: tackling power grid analysis on parallel SIMT platforms

Proceedings of the 2008 IEEE/ACM International Conference on Computer-Aided Design
GPU friendly fast Poisson solver for structured power grid network analysis

Proceedings of the 46th Annual Design Automation Conference
On the preconditioner of conjugate gradient method: a power grid simulation perspective

Proceedings of the International Conference on Computer-Aided Design
Robust Parallel Preconditioned Power Grid Simulation on GPU With Adaptive Runtime Performance Modeling and Optimization

IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

Power grid analysis for modern LSI is computationally challenging in terms of both runtime and memory usage. In this paper, we implement Krylov subspace based linear circuit solvers on a graphics processing unit (GPU) to realize fast power grid analysis. Efficiencies of memory space and access performance are pursued by improving a data structure that stores elements of large sparse matrices. Experimental results on benchmark circuits show that the proposed data structures are more suitable than widely used compressed sparse row (CSR) format and our GPU implementations can achieve up to 17x speedup over CPU implementations.