首页> 外文期刊>Computer physics communications >HYBRID ORDERED PARTICLE SIMULATION (HOPS) CODE FOR PLASMA MODELLING ON VECTOR-SERIAL, VECTOR-PARALLEL, AND MASSIVELY PARALLEL COMPUTERS8
【24h】

HYBRID ORDERED PARTICLE SIMULATION (HOPS) CODE FOR PLASMA MODELLING ON VECTOR-SERIAL, VECTOR-PARALLEL, AND MASSIVELY PARALLEL COMPUTERS8

机译:矢量串行,矢量并行和大规模并行计算机上等离子建模的混合有序粒子模拟(HOPS)代码8

获取原文
获取原文并翻译 | 示例
           

摘要

The Hybrid Ordered Particle Simulation (HOPS) code has been developed to provide a more efficient method for carrying out particle-in-cell (PIG) simulations of plasma phenomena. Conventional PIC methods store the particle attributes in tables that tend to have a random spatial order, which was appropriate for older serial-scalar computers. Problems associated with excessive accessing of memory, indirect indexing, and with many-to-one mappings in the deposition phase can make these codes inefficient on vector-serial, vector-parallel, and massively parallel machines. In the HOPS code we employ a sorting scheme to keep the particles ordered with respect to their spatial positions. By doing so, we have reduced memory accesses, recovered substantial direct indexing, and most importantly removed the many-to-one mapping problem. A low overhead sorting and reordering scheme is presented which allows HOPS to be most efficient on vector-serial machines and which scales linearly tc, various kinds of parallel computers. This paper focuses on the Gray C-90 vector-parallel computer but also discusses aspects of a massively parallel implementation. [References: 11]
机译:已开发出混合有序粒子模拟(HOPS)代码,以提供一种更有效的方法来执行等离子体现象的单元内粒子(PIG)模拟。传统的PIC方法将粒子属性存储在趋于具有随机空间顺序的表中,这适用于较旧的串行标量计算机。与内存的过度访问,间接索引以及沉积阶段的多对一映射相关的问题可能使这些代码在矢量串行,矢量并行和大规模并行机上效率低下。在HOPS代码中,我们采用排序方案以使粒子相对于其空间位置保持有序。这样,我们减少了内存访问,恢复了大量的直接索引,最重要的是消除了多对一映射问题。提出了一种低开销的排序和重新排序方案,该方案允许HOPS在矢量串行计算机上最高效,并且可以线性缩放tc和各种并行计算机。本文着重于Gray C-90向量并行计算机,但同时也讨论了大规模并行实现的各个方面。 [参考:11]

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号