CUDA projects: (still under constructing)

CUDA based GMRES(m) Solver for Large Scale Sparse Linear Systems

Abstract: As one popular iterative method to solve linear equations, restarted generalized minimal residual method(GMRES(m)) has the advantages of fast convergence and good stability. In this paper we implement GMRES with CUDA in parallel to solve large scale linear problems. With GPU these problems could be solved quickly. On the hardware with GeForce GTX260, the algorithm could be speed up by average 40 times compared with the according one only with Intel Core 2 Quad CPU Q9400@2.66GHz, average 20 times than the one only with Intel Core i7 CPU 920@2.67GHz.

Source Code (Sponsored by NVIDIA)

Test Set of Matrices from The University of Florida Sparse Matrix Collection

Updated in 8/25/2010