Fast Sparse Cholesky Decomposition and Inversion using Nested Dissection Matrix Reordering

被引:15
|
作者
Brandhorst, Kai [1 ]
Head-Gordon, Martin [1 ,2 ]
机构
[1] Univ Calif Berkeley, Dept Chem, Berkeley, CA 94720 USA
[2] Lawrence Berkeley Natl Lab, Div Chem Sci, Berkeley, CA USA
关键词
ELECTRONIC-STRUCTURE CALCULATIONS; PLESSET CORRELATION-ENERGY; DEGREE ORDERING ALGORITHM; DENSITY-MATRIX; NULL SPACE; DIRECT OPTIMIZATION; TENSOR FORMULATION; ELIMINATION TREES; MOLECULAR-SYSTEMS; HARTREE-FOCK;
D O I
10.1021/ct100618s
中图分类号
O64 [物理化学(理论化学)、化学物理学];
学科分类号
070304 ; 081704 ;
摘要
Here we present an efficient, yet nonlinear scaling, algorithm for the computation of Cholesky factors of sparse symmetric positive definite matrices and their inverses. The key feature of this implementation is the separation of the task into an algebraic and a numeric part. The algebraic part of the algorithm attempts to find a reordering of the rows and columns which preserves at least some degree of sparsity and afterward determines the exact nonzero structure of both the Cholesky factor and its corresponding inverse. It is based on graph theory and does not involve any kind of numerical thresholding. This preprocessing then allows for a very efficient implementation of the numerical factorization step. Furthermore this approach even allows use of highly optimized dense linear algebra kernels which leads to yet another performance boost. We will show some illustrative timings of our sparse code and compare it to the standard library implementation and a recent sparse implementation using thresholding. We conclude with some comments on how to deal with positive semidefinite matrices.
引用
收藏
页码:351 / 368
页数:18
相关论文
共 50 条
  • [21] A Nested Dissection Partitioning Method for Parallel Sparse Matrix-Vector Multiplication
    Boman, Erik G.
    Wolf, Michael M.
    2013 IEEE CONFERENCE ON HIGH PERFORMANCE EXTREME COMPUTING (HPEC), 2013,
  • [22] Fast variable matrix algorithm for sparse decomposition based on PSO
    Han, N. (haning1103@163.com), 1600, Chinese Institute of Electronics (34):
  • [23] Hybridizing nested dissection and halo approximate minimum degree for efficient sparse matrix ordering
    Pellegrini, François
    Roman, Jean
    Amestoy, Patrick
    Concurrency Practice and Experience, 2000, 12 (02): : 69 - 84
  • [24] Hybridizing nested dissection and halo approximate minimum degree for efficient sparse matrix ordering
    Pellegrini, F
    Roman, J
    Amestoy, P
    CONCURRENCY-PRACTICE AND EXPERIENCE, 2000, 12 (2-3): : 69 - 84
  • [25] Optimization of sparse matrix-vector multiplication using reordering techniques on GPUs
    Pichel, Juan C.
    Rivera, Francisco F.
    Fernandez, Marcos
    Rodriguez, Aurelio
    MICROPROCESSORS AND MICROSYSTEMS, 2012, 36 (02) : 65 - 77
  • [26] MatRIS: Addressing the Challenges for Portability and Heterogeneity Using Tasking for Matrix Decomposition (Cholesky)
    Monil, Mohammad Alaul Haque
    Miniskar, Narasinga Rao
    Valero-Lara, Pedro
    Teranishi, Keita
    Vetter, Jeffrey S.
    ASYNCHRONOUS MANY-TASK SYSTEMS AND APPLICATIONS, WAMTA 2024, 2024, 14626 : 59 - 70
  • [27] Regularized estimation of Kronecker structured covariance matrix using modified Cholesky decomposition
    Dai, Deliang
    Hao, Chengcheng
    Jin, Shaobo
    Liang, Yuli
    JOURNAL OF STATISTICAL COMPUTATION AND SIMULATION, 2023,
  • [28] Fast and accurate pseudoinverse with sparse matrix reordering and incremental approach (vol 109, pg 1, 2020)
    Jung, Jinhong
    Sael, Lee
    MACHINE LEARNING, 2021, 110 (02) : 449 - 449
  • [29] Fast and accurate pseudoinverse with sparse matrix reordering and incremental approach (vol 109, pg 2333, 2020)
    Jung, Jinhong
    Sael, Lee
    MACHINE LEARNING, 2021, 110 (03) : 619 - 620
  • [30] FAST EXACT STATE-ESTIMATION ALGORITHM BASED ON QUADRATIC EQUATIONS USING SPARSE-MATRIX INVERSION
    ALLAM, MF
    INTERNATIONAL JOURNAL OF ELECTRICAL POWER & ENERGY SYSTEMS, 1984, 6 (02) : 79 - 82