Collaborative Research: Supernova Simulation With CHIMERA

OCI-0749242 MesserOCI 0749248BlondinOCI 0749204 Marronetti千万亿级计算的出现带来了大幅提高物理保真度以解决一系列科学问题的前景，但千万亿级计算机的架构特征将需要相当大的创新才能有效使用。这个由三部分组成的合作项目解决了Chimera的几个最直接的开发需求，Chimera是一个大规模并行的多物理代码，专门为模拟核心坍塌的超新星而设计。这些恒星爆炸产生了宇宙中的大部分元素，是中微子、引力波和光子的巨大来源，并导致了中子星和黑洞的形成。将Chimera扩展到千万亿级平台将包括：1)用于流体动力学的完整的、经过测试的三维空间区域分解，包括在球极坐标中处理坐标奇异性的策略；2)用于引力场的可扩展的三维泊松解算器，包括牛顿和广义相对论；3)用于并行I/O的灵活的模块化系统；4)用于主导热核动力学和中微子传输的线性系统的多核解决方案；5)研究用于重叠计算和通信的分区全球地址空间(PGAS)语言；6)改进物理模型和算法。在这个项目下培训的学生和博士后将精通计算科学的几乎所有重要方面，包括数值线性代数、计算流体力学、辐射传输和Stiff系统解决方案，并将获得管理、分析和可视化大数据集以及伴随着拍级计算的硬件和软件环境的实际经验。推广的方法将适用于其他领域，而天体物理学的广受欢迎的吸引力使其在超级计算的公共推广中特别有效。该团队预计将继续进行关于计算天体物理学的旅行、讲座和演讲，每年接触到400多名K-12学生和教师、200多名大学本科生和数百名其他非科学家。

OCI-0749242MesserOCI 0749248BlondinOCI 0749204MarronettiThe advent of petascale computing brings with it the promise of substantial increases in physical fidelity for a host of scientific problems, but the architectural features of petascale machines will require considerable innovation for effective use. This three-part collaborative project addresses several of the most immediate development needs for CHIMERA, a massively parallel multi-physics code designed expressly to simulate core-collapse supernovae. These stellar explosions produce most of the elements in the Universe; are prodigious sources of neutrinos, gravitational waves, and photons, and lead to the formation of neutron stars and black holes. The extension of CHIMERA to petascale platforms will proceed through: 1) a complete, tested, three-dimensional spatial domain decomposition for the hydrodynamics, including a strategy to handle coordinate singularities in spherical polar coordinates; 2) a scalable three-dimensional Poisson solver for the gravitational field, both Newtonian and general relativistic; 3) a flexible, modular system for parallel I/O; 4) multi-core solutions for the linear systems dominating the thermonuclear kinetics and neutrino transport; 5) investigating Partitioned Global Address Space (PGAS) languages to overlap computation and communication; and 6) improvements to the physical models and algorithms.Students and postdocs trained under this project will be well versed in virtually every important aspect of computational science, including numerical linear algebra, computational fluid dynamics, radiation transport, and stiff system solution, and will gain real world experience with the management, analysis, and visualization of large data sets, and the attendant hardware and software environments of petascale computing. The generalized methods will be applicable in other fields, and the popular allure of astrophysics makes it especially effective in public outreach for supercomputing. The team expects to continue with tours, talks, and presentations on computational astrophysics, reaching each year over 400 K-12 students and teachers, more than 200 college undergraduates, and hundreds of other non-scientists.