Parallel iterative methods for dense linear systems in inductance extraction
Introduction
A key component in the design of high-end microprocessors is the estimation of signal delay of the VLSI circuit. The signal delay depends on several factors including parasitic resistance, capacitance, and inductance due to the on-chip interconnect. At higher frequencies, it is critical to estimate the effect of the inductive coupling between the interconnect segments quickly and accurately.
This paper presents a class of parallel algorithms for solving the linear systems of equations that arise in inductance extraction of VLSI circuits. To obtain the parasitic impedance at a particular frequency, the interconnect segments are discretized by a uniform mesh of filaments. Each filament carries an unknown amount of current which depends on the potential difference between the end nodes of the filament. The net flow of current into any node must be zero. These conditions give rise to linear systems with large coefficient matrices that have both sparse and dense submatrices. The sparse submatrices represent the constraints on current flow at each node. The dense submatrices represent the potential drop across each filament due to the inductive effect of current in all other filaments as well as due to its own resistance.
For accurate impedance estimation, discretizations with millions of filaments may be necessary. Since the computation and storage of the associated dense submatrix is not feasible, one must use iterative methods that rely on the ability to compute matrix–vector products with the coefficient matrix. The nature of the inductive coupling between filaments permits use of hierarchical multipole-based techniques such as the fast multipole method (FMM) [3], [8] to compute approximations to the matrix–vector products. The non-availability of the coefficient matrix makes it difficult to construct preconditioners that can be used to increase the rate of convergence of the iterative methods. The constraints imposed on the flow of current leads to an indefinite coefficient matrix, further complicating efforts to design robust preconditioners.
The approach presented in the paper uses a solenoidal basis method to restrict the current to the subspace where Kirchoff’s current law is satisfied, and solves the resulting reduced system by a preconditioned iterative method. The preconditioners are constructed from the inductive coupling of solenoidal functions of the basis, and the preconditioning step is implemented using hierarchical multipole-based approximations. The description of the inductance problem and the solution methodology in 2 Inductance extraction problem, 3 A preconditioned iterative solver follows the presentation in [7]. Section 4 describes a parallelization scheme for the hierarchical approximation algorithms used to compute matrix–vector products with the dense coefficient submatrix as well as the preconditioner. Benchmark experiments presented in Section 5 demonstrate that the algorithm can achieve high parallel efficiency on shared-memory multiprocessors. Concluding remarks are presented in Section 6.
Section snippets
Inductance extraction problem
Given a set of s conductors, the inductance extraction problem consists of finding an s×s matrix that summarizes the impedance among the conductors. The (k,l) element of this impedance matrix is equal to the potential difference generated across the kth conductor in response to a unit current source applied to the lth conductor. An entire column of the matrix can be computed by solving a linear system of equations. The impedance matrix is determined by solving s instances of the same linear
A preconditioned iterative solver
An alternate approach to solving the linear system in (7) uses a basis for the subspace of current vectors that satisfy the constraint . For instance, given a basis for the null space of , the current vector satisfies the constraint for arbitrary x. By restricting the current to the null space of , the linear system in (7) is transformed to the following system:The unknown vector can be eliminated by multiplying the above equation with , resulting in the
Parallelism
To develop efficient parallel formulations of the iterative solver, it is necessary to understand the structure of the dense matrix–vector products. The product with is viewed as the calculation of the potential difference across each filament due to the inductive effect of current in all other filaments. The integral in (2) is approximated by a weighted sum over a set of discrete points within each filament. The mid-point of a filament is sufficient for calculating the mutual inductance
Experiments
The preconditioned iterative solver outlined earlier can be implemented efficiently on a shared-memory multiprocessor machine. This section presents the results of experiments to study the parallel performance of a multiprocessor implementation of the algorithm. The experiments are organized into three sets. The first set of experiments presents the parallel efficiency of the code on three benchmark problems. The second set of experiments provides an insight into the scalability of the
Conclusions
In this paper, we presented a preconditioned iterative method for solving the dense linear systems that arise in the inductance extraction problem in VLSI circuit design. The approach uses discrete solenoidal basis functions to obtain an equivalent reduced system which is solved by the preconditioned GMRES method. Matrix–vector products with the dense coefficient matrices as well as preconditioners are computed via hierarchical approximations. We outlined parallelization schemes for the
Acknowledgements
This work has been supported in part by NSF under the grants NSF-CCR 9984400 and NSF-CCR 0113668, and by the Texas Advanced Technology Program grant 000512-0266-2001.
References (10)
Rapid solution of integral equations of classical potential theory
Journal of Computational Physics
(1985)- et al.
Load balancing and data locality in hierarchical N-body methods
Journal of Parallel and Distributed Computing
(1995) An efficient program for many-body simulation
SIAM Journal on Scientific and Statistical Computing
(1985)- et al.
A hierarchical O(nlogn) force calculation algorithm
Nature
(1986) The Rapid Evaluation of Potential Fields in Particle Systems
(1988)
Cited by (8)
The antitriangular factorization of saddle point matrices
2014, SIAM Journal on Matrix Analysis and ApplicationsParallel computing applied to inductance calculation for flexible inductors
2013, COMPEL - The International Journal for Computation and Mathematics in Electrical and Electronic EngineeringPreconditioned iterative solvers for inductance extraction of VLSI circuits
2007, SIAM Journal on Scientific ComputingParallel algorithms for inductance extraction of VLSI circuits
2006, 20th International Parallel and Distributed Processing Symposium, IPDPS 2006Numerical solution of saddle point problems
2005, Acta Numerica