The Research On Optimizing Sparse Matrix Computation Based On GPU

Large-scale sparse matrix solver is a common problem in high-performance computing,widely exists in engineering practice, particularly in computer simulation field. Usingconventional methods for solving sparse matrix will waste a lot of computing resources.At present, both at home and abroad, research on sparse matrix computation in generalpurpose GPU computing is less. Existing research is major on sparse matrix and vectormultiplication.Sparse matrix vector multiplication on the GPU is achieved and optimized. In order tosolve the problem caused by uneven distribution for non-zero elements in sparse matrixand the problem that threads in a same warp can’t visit GPU memory in a merged way, theSC-CSR sparse matrix vector multiplication method on GPU is proposed. To solve threadwaiting issue caused by load imbalance of thread computing in a warp and the memoryaccess issue as a result of the thread does not meet the merger of the global memoryaccess requirements, a sparse matrix vector multiplication approach based on VAB sparsematrix storage format is proposed. The optimization of global memory access and the useof texture memory and constant memory is proposed for the above two methods. Linearequation solver on the GPU sparse matrix with Jacobi iteration and Generalized MinimumResidual method is achieved and optimized. The optimization method proposed can beextended to all the iterative method on GPU for solving sparse matrix linear equation ofuniversal significance. Finally, the sparse matrix equation solving is accelerated by thehost device communication and shared memory access optimization.Experiments show that the sparse linear equation computation speed increasessignificantly in relation to serial code on CPU with a speed up ranging from10.3to74.0.