Massively Parallel Particle Hydrodynamics for Engineering and Astrophysics

SPH（光滑粒子流体动力学）和拉格朗日方法是解决流体力学问题的一种有效方法。在该方案中，流体由大量随流动而移动的粒子表示。该方案不需要一个预定义的网格，使其非常适合于跟踪流动与移动边界，特别是流动与自由表面，和问题，涉及流动与物理活动的元素或大的动态范围。该方法的应用范围正在迅速扩大，并被越来越多的商业公司采用，包括空中客车、联合利华、壳牌、法国电力公司、米其林和雷诺。SPH的广泛使用及其在广泛科学领域的应用潜力使其成为Excalibur项目的优先用例。具有数十亿到数千亿粒子的大规模并行模拟有可能彻底改变我们对宇宙的理解，并将使工程应用达到前所未有的规模，从端到端的瞬态仿真（如鸟撞）在喷气发动机的模拟海啸波超过-运行一系列的防御墙。工作组将确定一条通往艾级计算挑战的道路。该小组拥有工程和天体物理学方面的专业知识，使我们能够开发出满足广泛社区需求的方法。该小组将从最近的两个守则开始，这些守则已经突出了关键问题，并将作为工作组的起点。SWIFT（SPH with Interdependent Fine-grained Tasking）实现了基于任务的并行性的尖端方法。将问题分解为一系列相互依赖的任务，可以在调度方面具有很大的灵活性，并允许通信任务与通信完全重叠。该代码使用时间步长层次结构来集中计算工作在最需要响应问题的地方。DualSPHysics通过有效使用GPU加速器在大量相同粒子上执行SPH操作来提高速度。这允许代码从异常并行执行中获益。挑战在于如何在大量互连的计算节点上有效地连接多个GPU。工作组将在这些代码的基础上确定在exa规模系统上进行大规模并行执行的最佳方法。该项目将受益于与达勒姆的Excalibur硬件试点工作组的密切联系，推动代码和硬件的共同设计。我们将解决的特殊挑战是：-Exascale性能的最佳算法。特别是，我们将解决最好的方法，自适应时间步进和时间积分，自适应区域分解。第一个允许不同的空间区域在时间上最佳地向前集成，第二个允许区域在硬件上最佳地分布。模块化和关注点分离。未来的代码需要灵活和模块化，以便在集成例程，任务调度和物理模块之间实现分离。这将使代码面向未来，易于适应新的科学领域要求和计算硬件。CPU/GPU性能优化。下一代硬件将需要开发特定的（可能是新颖的）技术，以优化SPH方案中的粒子。我们将在DualSPHysics中获得的编程专业知识的基础上，实现跨多个节点的高效GPU使用。通信性能优化。分离的计算区域需要在其边界处交换信息。这可以异步完成，这样通信的时间延迟不会减慢计算。虽然这已经在当前系统上得到证明，但Excalibur的规模将使当前子系统过载，因此需要新的解决方案。

SPH (smoothed particle hydrodynamics), and Lagrangian approaches to hydrodynamics in general, are a powerful approach to hydrodynamics problems. In this scheme, the fluid is represented by a large number of particles, moving with the flow. The scheme does not require a predefined grid making it very suitable for tracking flows with moving boundaries, particularly flows with free surfaces, and problems that involve flows with physically active elements or large dynamic range. The range of applications of the method is growing rapidly and is being adopted by a rapidly growing range of commercial companies including Airbus, Unilever, Shell, EDF, Michelin and Renault.The widespread use of SPH, and its potential for adoption across a wide range of science domains, make it a priority use case for the Excalibur project. Massively parallel simulations with billion to hundreds of billions of particles have the potential for revolutionising our understanding of the Universe and will empower engineering applications of unprecedented scale, ranging from the end-to-end simulation of transients (such as a bird strike) in jet engines to the simulation of tsunami waves over-running a series of defensive walls.The working group will identify a path to the exascale computing challenge. The group has expertise across both Engineering and Astrophysics allowing us to develop an approach that satisfies the needs of a wide community. The group will start from two recent codes that already highlight the key issues and will act as the working group's starting point.- SWIFT (SPH with Interdependent Fine-grained Tasking) implements a cutting-edge approach to task-based parallelism. Breaking the problem into a series of inter-dependent tasks allows for great flexibility in scheduling, and allows communication tasks to be entirely overlapped with communication. The code uses a timestep hierarchy to focus computational effort where is most need in response to the problems.- DualSPHysics draws its speed from effective use of GPU accelerators to execute the SPH operations on large groups of identical particles. This allows the code to gain from exceptional parallel execution. The challenge is to effectively connect multiple GPUs across large numbers of inter-connected computing nodes.The working group will build on these codes to identify the optimal approach to massively parallel execution on exa-scale systems. The project will benefit from close connections to the Excalibur Hardware Pilot working group in Durham, driving the co-design of code and hardware. The particular challenges that we will address are:- Optimal algorithms for Exascale performance. In particular, we will address the best approaches to the adaptive time-stepping and out-of-time integration, and adaptive domain decomposition. The first allows different spatial regions to be integrated forward in time optimally, the second allows the regions to be optimally distributed over the hardware.- Modularisation and Separation of Concerns. Future codes need to be flexible and modularised, so that a separation can be achieved between integration routines, task scheduling and physics modules. This will make the code future-proof and easy to adapt to new science domain requirements and computing hardware.- CPU/GPU performance optimisation. Next generation hardware will require specific (and possibly novel) techniques to be developed to optimally advance particles in the SPH scheme. We will build on the programming expertise gain in DualSPHysics to allow efficient GPU use across multiple nodes.- Communication performance optimisation. Separated computational regions need to exchange information at their boundaries. This can be done asynchronously, so that the time-lag of communication does not slow computation. While this has been demonstrated on current systems, the scale of Excalibur will overload current subsystems, and a new solution is needed.