Large-scale particle-based fluid simulations using dynamic load balance on GPU supercomputer

Tsuzuki, Satori; Aoki, Takayuki

doi:10.1299/jsmecmd.2014.27.674

Cited by 7 publications

(8 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Marrone et al [8] and Leffe et al [9] could also get the high parallel efficiency using SPH of an explicit algorithm. In researches of the discrete element method (DEM) by Horner et al [10] and Tsuzuki et al [11,12] and the non-smooth contact dynamics (NSCD) method by Balzer et al [13], the distributed memory parallelization was developed successfully. However, these all distributed memory parallel particle methods have shown relative values like parallel scalabilities not but achieved FLOPS.…”

Section: Introductionmentioning

confidence: 99%

Development of Hierarchical Domain Decomposition Explicit MPS Method and Application to Large-scale Tsunami Analysis with Floating Objects

Murotani

Koshizuka

Tamai

et al. 2014

JASSE

View full text Add to dashboard Cite

In this research, a novel distributed memory parallel algorithm for the explicit moving particle simulation (MPS) method is presented. In this method, the analysis area is divided for distributed memory parallel computation using ParMETIS. The development offers two improved points from the distributed parallel explicit particle method of single bucket-based domain decomposition [1]. The first improved point is the hierarchical domain decomposition of two levels. The second improved point is that particle data are managed by buckets. Based on the improvements, we are efficiently able to run a large-scale run-up tsunami analysis.

show abstract

Section: Introductionmentioning

confidence: 99%

Development of Hierarchical Domain Decomposition Explicit MPS Method and Application to Large-scale Tsunami Analysis with Floating Objects

Murotani

Koshizuka

Tamai

et al. 2014

JASSE

View full text Add to dashboard Cite

show abstract

“…Communications between cores is achieved using MPI [17]. Although computations with up to a few tens of millions of particles are emerging with GPU-based implementations [10,11,15,12], simulations with up to a few billions of particles can be envisioned with CPU-based implementations, provided computational practitioners have access to large supercomputers with many thousands of cores [6,7,8]. The forthcoming new GPU technology is likely to offer similar parallel computing capabilities as the CPU technology, either by improving inter-GPU communications without using CPUs or by speeding up data exchange between GPUs and CPUs.…”

Section: Cpu-based Dem Codes Generally Exhibit No Limit In Number Of mentioning

confidence: 99%

“…The latter involves dividing the work load between different computing units and hence using distributed computing. Nowadays, there are two (potentially complementary) technologies for DEM distributed computing: CPU [4,5,6,7,8,9] vs GPU [10,11,12,13,14,15,16,9].…”

Section: Introductionmentioning

confidence: 99%

Grains3D, a flexible DEM approach for particles of arbitrary convex shape - Part II: Parallel implementation and scalable performance

Rakotonirina

Wachs

2018

Powder Technology

View full text Add to dashboard Cite

In [1] we suggested an original Discrete Element Method that offers the capability to consider non-spherical particles of arbitrary convex shape. We elaborated on the foundations of our numerical method and validated it on assorted test cases. However, the implementation was serial and impeded to examine large systems. Here we extend our method to parallel computing using a classical domain decomposition approach and inter-domain MPI communication. The code is implemented in C++ for multi-CPU architecture. Although object-oriented C++ offers high-level programming concepts that enhance the versatility required to treat multi-shape and multi-size granular systems, particular care has to be devoted to memory management on multi-core architecture to achieve reasonable computing efficiency. The parallel performance of our code Grains3D is assessed on various granular flow configurations comprising both spherical and angular particles. We show that our parallel granular solver is able to compute systems with up to a few hundreds of millions of particles. This opens up new perspectives in the study of granular material dynamics.

show abstract

“…The slice grid method has the simplest implementation among the dynamic load-balancing methods. In previous studies, the two-dimensional slice grid method has mainly been used (Tsuzuki and Aoki, 2014;Furuichi et al, 2017;Nishiura et al, 2019). However, this system has a large spatial spread, and it is expected that the three-dimensional method is more suitable.…”

Section: Introductionmentioning

confidence: 99%

Parallelization of DEM simulation on distributed-memory computer via three-dimensional slice grid method

Yokoo

Kishida

Yamamoto

2020

Mechanical Engineering Letters

View full text Add to dashboard Cite

Large-scale particle-based fluid simulations using dynamic load balance on GPU supercomputer

Cited by 7 publications

References 0 publications

Development of Hierarchical Domain Decomposition Explicit MPS Method and Application to Large-scale Tsunami Analysis with Floating Objects

Development of Hierarchical Domain Decomposition Explicit MPS Method and Application to Large-scale Tsunami Analysis with Floating Objects

Grains3D, a flexible DEM approach for particles of arbitrary convex shape - Part II: Parallel implementation and scalable performance

Parallelization of DEM simulation on distributed-memory computer via three-dimensional slice grid method

Contact Info

Product

Resources

About