P000034R1
Isogeometric Analysis on Intel Single Chip Cloud System
Licai Guo (School of Computer Science and Technology, University of Science and Technology of China, Hefei, Anhui, China, 230027)
*Zhangjin Huang (School of Computer Science and Technology, University of Science and Technology of China, Hefei, Anhui, China, 230027)
Isogeometric analysis is a numerical method to solve partial differential equations. In this paper, a novel isogeometric analysis framework on Intel Single Chip Cloud (SCC) system is presented, which is simple, efficient and easy to implement. The framework consists of two parts. The first is a scalable many-core sparse solver, which solves sparse linear systems on SCC system. The second is a simple method to construct the stiffness matrix and right hand side vector. The stiffness matrix and right hand side vector are uniformly divided into a list of blocks, each block is constructed by 1 processor. The sparse solver for isogeometric analysis uses the same data allocation scheme. So there is no need to move data between processors. The experiments show that the framework solves the poisson equation correctly and efficiently. The tests show that the construction of the linear system scales very well, achieves nearly linear speed up. The sparse solver does not scale well. As more and more processors are used, the construction time decreases linearly or even faster. The sparse solver is limited by the memory bandwidth.