针对传统解算器未能实现 GPU 上运行网格自适应加密过程,造成 GPU 与 CPU 之间繁琐的数据交换的问题,本文发展优化了一种 GPU 加速的基于非结构自适应加密网格的解算器 VA2DG。利用加密网格表的方式实现网格自适应加密过程在 GPU 上高速运行,并通过原子操作并行生成网格加密表,对废弃的网格及时回收,节约存储空间,加快运行速度。%Aiming at the problem that the traditional solution can not implement the adaptive encryption of GPU,which results in the complex data exchange between CPU and GPU,this paper develops a kind of speed GPU based on unstructured adaptive encryption mesh VA2DG.It is to realize that the grid adaptive encryption process can be fast running on CPU by using encryption grid table,and the grid encryption table can be generated by atomic operations.The waste is collected in time to save the storage space and increase the running speed.
展开▼