Highly Asynchronous and Scalable Algorithms for Distributed-Memory Adaptive Mesh Refinement at Extreme Scales
نویسندگان
چکیده
In this paper, we present our developments of a novel approach for distributed memory Adaptive Mesh Refinement (AMR). Our approach is highly asynchronous and fully distributed that makes it suitable for extreme-level scaling. It takes negligible memory to store mesh structure as compared to the traditional approaches which are not scalable. It accomplishes adaptive mesh restructuring in just 1 global collective call as compared to O(d) calls in the traditional approaches. A new distributed load balancer has been developed that led to an improvement in performance by 18%. We present our scaling results on up to 131, 072 cores of BG/Q supercomputer.
منابع مشابه
An Adaptive Cartesian Detonation Solver for Fluid-Structure Interaction Simulation on Distributed Memory Computers
Time-accurate fluid-structure interaction simulations of strong shock and detonation waves impinging on deforming solid structures benefit significantly from the application of dynamic mesh adaptation in the fluid. A patch-based parallel fluid solver with adaptive mesh refinement in space and time tailored for this problem class is presented; special attention is given to the robustness of the ...
متن کاملAdaptive Parallel Computation of a Grand-Challenge Problem: Prediction of the Path of a Solar-Coronal Mass Ejection
One of the ways that the Sun interacts with the Earth is through the solar wind, which is an ionized multicomponent fluid that emanates from the Sun and travels radially outward at hundreds of kilometers per second. Solar-wind transients, such as Coronal Mass Ejections (CME’s), can be particularly important. In rare cases, CME’s have affected the lower atmosphere of the Earth, causing regional ...
متن کاملDetonation Simulation with the AMROC Framework
Numerical simulations can be the key to the thorough understanding of the multi-dimensional nature of transient detonation waves. But the accurate approximation of realistic detonations is extremely demanding, because a wide range of different scales need to be resolved. This paper describes an efficient simulation strategy based on a generic implementation of a blockstructured dynamically adap...
متن کاملExtreme-Scale Block-Structured Adaptive Mesh Refinement
In this article, we present a novel approach for block-structured adaptive mesh refinement (AMR) that is suitable for extreme-scale parallelism. All data structures are designed such that the size of the meta data in each distributed processor memory remains bounded independent of the processor number. In all stages of the AMR process, we use only distributed algorithms. No central resources su...
متن کاملGRACOS: Scalable and Load Balanced PM Cosmological N-body Code
We present a parallel implementation of the particle-particle/particle-mesh (PM) algorithm for distributed memory clusters. The GRACOS (GRAvitational COSmology) code uses a hybrid method for both computation and domain decomposition. Long-range forces are computed using a Fourier transform gravity solver on a regular mesh; the mesh is distributed across parallel processes using a static one-dim...
متن کامل