GPU accelerated higher-order moment method on NURBS model

A novel higher-order moment method (HOMM) on NURBS (non-uniform rational B-splines) model enhanced by the graphics processing unit (GPU) programming technique is presented. The NURBS-HOMM can greatly reduce the number of unknowns and peak memory usage by using higher-order basis functions and larger NURBS patches. However, filling in the impedance matrix of the NURBS-HOMM always consumes long CPU time because of a four-dimensional integral consisting in each element. The GPU is a massively parallel processor and very suitable for compute-intensive and highly parallel computation. So, the GPU programming technique is applied to NURBS-HOMM to significantly reduce computation time. Numerical results show good efficiency of the proposed technique.