Pipelined iterative solvers with kernel fusion for graphics processing units

DOI10.1145/2907944MaRDI QIDQ5270772zbMATH OpenOpenAlexWikidataFDO

Authors Karl Rupp, Josef Weinbub, Ansgar Jüngel, Tibor Grasser

Publication date 30 June 2017

Published in ACM Transactions on Mathematical Software (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1410.4054

zbMATH Keywords

BiCGStab methodconjugate gradient methodCUDAGMRES methodGPUiterative solversOpenCL

Mathematics Subject Classification ID

Iterative numerical methods for linear systems (65F10) Numerical algorithms for specific classes of architectures (65Y10) Packaged methods for numerical algorithms (65Y15)

Abstract: We revisit the implementation of iterative solvers on discrete graphics processing units and demonstrate the benefit of implementations using extensive kernel fusion for pipelined formulations over conventional implementations of classical formulations. The proposed implementations with both CUDA and OpenCL are freely available in ViennaCL and are shown to be competitive with or even superior to other solver packages for graphics processing units. Highest performance gains are obtained for small to medium-sized systems, while our implementations are on par with vendor-tuned implementations for very large systems. Our results are especially beneficial for transient problems, where many small to medium-sized systems instead of a single big system need to be solved.

Recommendations

• Accelerating iterative linear solvers using multiple graphical processing units
• A factored sparse approximate inverse preconditioned conjugate gradient solver on graphics processing units
• Accelerating preconditioned iterative linear solvers on GPU
• Out-of-core solution of linear systems on graphics processors
• Stencil-aware GPU optimization of iterative solvers

Cites work

• A GPU accelerated aggregation algebraic multigrid method
• A simpler GMRES
• A unified sparse matrix data format for efficient general sparse matrix-vector multiplication on modern processors with wide SIMD units
• Bi-CGSTAB: A Fast and Smoothly Converging Variant of Bi-CG for the Solution of Nonsymmetric Linear Systems
• Biomolecular electrostatics using a fast multipole BEM on up to 512 GPUs and a billion unknowns
• Exposing fine-grained parallelism in algebraic multigrid methods
• GMRES: A Generalized Minimal Residual Algorithm for Solving Nonsymmetric Linear Systems
• Hiding global communication latency in the GMRES algorithm on massively parallel machines
• scientific article; zbMATH DE number 3969995 (Why is no real title available?)
• scientific article; zbMATH DE number 1953444 (Why is no real title available?)
• scientific article; zbMATH DE number 814821 (Why is no real title available?)
• scientific article; zbMATH DE number 1424576 (Why is no real title available?)
• Methods of conjugate gradients for solving linear systems
• Practical Use of Polynomial Preconditionings for the Conjugate Gradient Method
• s-step iterative methods for symmetric linear systems

Cited in

(7)

• GPU-based matrix-free finite element solver exploiting symmetry of elemental matrices
• Revisiting performance of biCGStab methods for solving systems with multiple right-hand sides
• Pipelined, flexible Krylov subspace methods
• ViennaCL-linear algebra library for multi- and many-core architectures
• Stencil-aware GPU optimization of iterative solvers
• Accelerating iterative linear solvers using multiple graphical processing units
• scientific article; zbMATH DE number 2222876 (Why is no real title available?)

Describes a project that uses

Uses Software

• CUDA
• ViennaCL
• OpenCL
• CU2CL
• Swan

This page was built for publication: Pipelined iterative solvers with kernel fusion for graphics processing units

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5270772)