Parallel GMRES implementation for solving sparse linear systems on GPU clusters

نویسندگان

  • Jacques M. Bahi
  • Raphaël Couturier
  • Lilia Ziane Khodja
چکیده

In this paper, we propose an efficient parallel implementation of the GMRES method for GPU clusters. This implementation requires us to parallelize the GMRES algorithm between CPUs of the cluster. Hence, all parallel and intensive computations on local data are performed on GPUs and reduction operations to compute global results are carried out by CPUs. The performances of our parallel GMRES solver are evaluated on test matrices of sizes exceeding 10 7 rows. They show that solving large and sparse linear systems on a GPU cluster is faster than that performed on its CPU counterpart. It is noticed that a cluster of 12 GPUs is about 8 times faster than a cluster of 12 CPUs and about 5 times faster than a cluster of 24 CPUs.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Preconditioned Generalized Minimal Residual Method for Solving Fractional Advection-Diffusion Equation

Introduction Fractional differential equations (FDEs)  have  attracted much attention and have been widely used in the fields of finance, physics, image processing, and biology, etc. It is not always possible to find an analytical solution for such equations. The approximate solution or numerical scheme  may be a good approach, particularly, the schemes in numerical linear algebra for solving ...

متن کامل

Parallel Sparse Linear Algebra for Homotopy Methods

Globally convergent homotopy methods are used to solve difficult nonlinear systems of equations by tracking the zero curve of a homotopy map. Homotopy curve tracking involves solving a sequence of linear systems, which often vary greatly in difficulty. In this research, a popular iterative solution tool, GMRES(k), is adapted to deal with the sequence of such systems. The proposed adaptive strat...

متن کامل

Accelerating Preconditioned Iterative Linear Solvers on Gpu

Linear systems are required to solve in many scientific applications and the solution of these systems often dominates the total running time. In this paper, we introduce our work on developing parallel linear solvers and preconditioners for solving large sparse linear systems using NVIDIA GPUs. We develop a new sparse matrix-vector multiplication kernel and a sparse BLAS library for GPUs. Base...

متن کامل

High Performance Manycore Solvers for Reservoir Simulation

SUMMARY The forthcoming generation of many-core architectures compels a paradigm shift in algorithmic design to effectively unlock its full potential for maximum performance. In this paper, we discuss a novel approach for solving large sparse linear systems arising in realistic black oil and compositional flow simulations. A flexible variant of GMRES (FGMRES) is implemented using the CUDA progr...

متن کامل

Large Scale Linear System Global Computing Resolution with GMRES Method

Grid computing attains high throughput computing by making use of a very large amount of unexploited computing resources. We present a typical parallel method GMRES to solve large sparse linear systems by the use of a lightweight GRID system XtremWeb. This global computing platform, just as many popular GRID systems, is mainly devoted to multi-parameters generic applications. We have implemente...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2011