Based on norm-minimization technique, a parallel sparse approximate inversepreconditioning method (PPAT method) is present for the unsymmetric sparselinear systems. The sparsity pattern of the approximate inverse is t...Based on norm-minimization technique, a parallel sparse approximate inversepreconditioning method (PPAT method) is present for the unsymmetric sparselinear systems. The sparsity pattern of the approximate inverse is the same as thatof the transpose of the coefficient matrix. This keeps the amount of work and theneed of storage small. The computation of the preconditioner is inherently parallel.Some numerical experiments show that PPAT preconditioners can accelerate theconvergence.展开更多
由于SSOR预条件共轭梯度算法中预条件方程求解需要前推和回代,导致算法迁移到GPU平台上并行效率不高.为此,基于诺依曼多项式分解技术,提出了一种GPU加速的SSOR稀疏近似逆预条件子(GSSORSAI).它不仅保持了原线性系统系数矩阵的稀疏和对...由于SSOR预条件共轭梯度算法中预条件方程求解需要前推和回代,导致算法迁移到GPU平台上并行效率不高.为此,基于诺依曼多项式分解技术,提出了一种GPU加速的SSOR稀疏近似逆预条件子(GSSORSAI).它不仅保持了原线性系统系数矩阵的稀疏和对称正定特性,而且预条件方程求解仅需一次稀疏矩阵矢量乘运算,避免了前推和回代过程.实验结果表明:在NVIDIA Tesla C2050GPU上,对比使用Python在单个CPU上SSOR稀疏近似逆预条件子实现方法,GSSORSAI平均快将近100倍;应用到并行的PCG算法中,相比无预条件的CG算法,平均提高了算法的3倍的收敛速度.展开更多
文摘Based on norm-minimization technique, a parallel sparse approximate inversepreconditioning method (PPAT method) is present for the unsymmetric sparselinear systems. The sparsity pattern of the approximate inverse is the same as thatof the transpose of the coefficient matrix. This keeps the amount of work and theneed of storage small. The computation of the preconditioner is inherently parallel.Some numerical experiments show that PPAT preconditioners can accelerate theconvergence.
文摘由于SSOR预条件共轭梯度算法中预条件方程求解需要前推和回代,导致算法迁移到GPU平台上并行效率不高.为此,基于诺依曼多项式分解技术,提出了一种GPU加速的SSOR稀疏近似逆预条件子(GSSORSAI).它不仅保持了原线性系统系数矩阵的稀疏和对称正定特性,而且预条件方程求解仅需一次稀疏矩阵矢量乘运算,避免了前推和回代过程.实验结果表明:在NVIDIA Tesla C2050GPU上,对比使用Python在单个CPU上SSOR稀疏近似逆预条件子实现方法,GSSORSAI平均快将近100倍;应用到并行的PCG算法中,相比无预条件的CG算法,平均提高了算法的3倍的收敛速度.
