The LU factorization of very large sparse matrices requires a significant amount of computing resources, including memory and broadband communication. A hybrid MPI + OpenMP + CUDA algorithm named SuperLU3D can efficiently compute the LU factorization with GPU acceleration. However, this algorithm faces difficulties when dealing with very large sparse matrices with limited GPU resources. Factorizing very large matrices involves a vast amount of nonblocking communication between processes, often leading to a break in SuperLU3D calculation due to the overflow of cluster communication buffers. In this paper, we present an improved GPU-accelerated algorithm named SuperLU3D_Alternate for the LU factorization of very large sparse matrices with fewer GPU resources. The basic idea is “divide and conquer”, which means dividing a very large matrix into multiple submatrices, performing LU factorization on each submatrix, and then assembling the factorized results of all submatrices into two complete matrices L and U. In detail, according to the number of available GPUs, a very large matrix is first divided into multiple submatrices using the elimination tree. Then, the LU factorization of each submatrix is alternately computed with limited GPU resources, and its intermediate LU factors from GPUs are saved to the host memory or hard disk. Finally, after finishing the LU factorization of all submatrices, these factorized submatrices are assembled into a complete lower triangular matrix L and a complete upper triangular matrix U, respectively. The SuperLU3D_Alternate algorithm is suitable for hybrid CPU/GPU cluster systems, especially for a subset of nodes without GPUs. To accommodate different hardware resources in various clusters, we designed the algorithm to run in the following three cases: sufficient memory for GPU nodes, insufficient memory for GPU nodes, and insufficient memory for the entire cluster. The results from LU factorization test on different matrices in various cases show that the larger the matrix is, the more efficient this algorithm is under the same GPU memory consumption. In our numerical experiments, SuperLU3D_Alternate achieves speeds of up to 8× that of SuperLU3D (CPU only) and 2.5× that of SuperLU3D (CPU + GPU) on the hybrid cluster with six Tesla V100S GPUs. Furthermore, when the matrix is too big to be handled by SuperLU3D, SuperLU3D_Alternate can still utilize the cluster’s host memory or hard disk to solve it. By reducing the amount of data exchange to prevent exceeding the buffer’s limit of the cluster MPI nonblocking communication, our algorithm enhances the stability of the program.

More from our Archive

• DOI: 10.3390/sym15071324 2023

  The Analysis of Bifurcation, Quasi-Periodic and Solitons Patterns to the New Form of the Generalized q-Deformed Sinh-Gordon Equation

  Syeda Sarwat Kazmi, Adil Jhangeer, Nauman Raza, Haifa I. Alrebdi, Abdel-Haleem Abdel-Aty, Hichem Eleuch

• DOI: 10.3390/math11143069 2023

  An Overview of Forecast Analysis with ARIMA Models during the COVID-19 Pandemic: Methodology and Case Study in Brazil

  Raydonal Ospina, João A. M. Gondim, Víctor Leiva, Cecilia Castro

• DOI: 10.1098/rsta.2022.0217 2023

  Analysing the dynamic structure of warm dense matter in the imaginary-time domain: theoretical models and simulations

  Tobias Dornheim, Jan Vorberger, Zhandos A. Moldabekov, Maximilian Böhme

• DOI: 10.3390/math11143133 2023

  Dynamic RNA Coding Color Image Cipher Based on Chain Feedback Structure

  Heping Wen, Shenghao Kang, Zhuxi Wu, Yiting Lin, Yiming Huang

• DOI: 10.3390/math11143110 2023

  Almost Sure Exponential Stability of Uncertain Stochastic Hopfield Neural Networks Based on Subadditive Measures

  Zhifu Jia, Cunlin Li

• DOI: 10.3390/math11143052 2023

  Greenhouse Micro-Climate Prediction Based on Fixed Sensor Placements: A Machine Learning Approach

  Oladayo S. Ajani, Member Joy Usigbe, Esther Aboyeji, Daniel Dooyum Uyeh, Yushin Ha, Tusan Park, Rammohan Mallipeddi

• DOI: 10.3390/math11133022 2023

  Mathematical Modelling of Fused Deposition Modeling (FDM) 3D Printing of Poly Vinyl Alcohol Parts through Statistical Design of Experiments Approach

  Mahmoud Moradi, Mojtaba Karamimoghadam, Saleh Meiabadi, Giuseppe Casalino, Mohammad Ghaleeh, Bobymon Baby, Harikrishna Ganapathi, Jomal Jose, Muhammed Shahzad Abdulla, Paul Tallon, Mahmoud Shamsborhan, Mohammad Rezayat, Satyam Paul, Davood Khodadad

• DOI: 10.3390/math11133008 2023

  Prognostics and Health Management of Rotating Machinery of Industrial Robot with Deep Learning Applications—A Review

  Prashant Kumar, Salman Khalid, Heung Kim

• DOI: 10.3390/sym15071338 2023

  Nonlinear Dynamics of a Piecewise Modified ABC Fractional-Order Leukemia Model with Symmetric Numerical Simulations

  Hasib Khan, Jehad Alzabut, Wafa F. Alfwzan, Haseena Gulzar

• DOI: 10.1002/mma.9474 2023

  Existence and multiplicity of solutions for a class of critical anisotropic elliptic equations of Schrödinger–Kirchhoff‐type

  Nabil Chems Eddine, Phuong Duc Nguyen, Maria Alessandra Ragusa