Multi-scale Continuum-Particle Simulation on CPU–GPU Hybrid Supercomputer

Ge, Wei; Xu, Jianzhong; Xiong, Qingang; Wang, Xiaowei; Chen, Feiguo; Wang, Limin; Hou, Chaofeng; Xu, Minyi; Li, Jinghai

doi:10.1007/978-3-642-16405-7_8

Cited by 5 publications

(3 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In regard to large-scale computational models, the solution time of equations becomes particularly important [46]. We plan to adopt a more efficient parallel method to accelerate the solution speed of the equations on the GPU [47,48].…”

Section: Outlook and Future Workmentioning

confidence: 99%

juSFEM: A Julia-based open-source package of parallel Smoothed Finite Element Method (S-FEM) for elastic problems

Huo

Mei

2021

Computers & Mathematics with Applications

View full text Add to dashboard Cite

The Smoothed Finite Element Method (S-FEM) proposed by Liu G.R. can achieve more accurate results than the conventional FEM. Currently, much commercial software and many open-source packages have been developed to analyze various science and engineering problems using the FEM. However, there is little work focusing on designing and developing software or packages for the S-FEM. In this paper, we design and implement an open-source package of the parallel S-FEM for elastic problems by utilizing the Julia language on multi-core CPU. The Julia language is a fast, easy-to-use, and open-source programming language that was originally designed for high-performance computing. We term our package as juSFEM. To the best of the authors knowledge, juSFEM is the first package of parallel S-FEM developed with the Julia language. To verify the correctness and evaluate the efficiency of juSFEM, two groups of benchmark tests are conducted. The benchmark results show that (1) juSFEM can achieve accurate results when compared to commercial FEM software ABAQUS, and (2) juSFEM only requires 543 seconds to calculate the displacements of a 3D elastic cantilever beam model which is composed of approximately 2 million tetrahedral elements, while in contrast the commercial FEM software needs 930 seconds for the same calculation model; (3) the parallel juSFEM executed on the 24-core CPU is approximately 20× faster than the corresponding serial version. Moreover, the structure and function of juSFEM are easily modularized, and the code in juSFEM is clear and readable, which is convenient for further development.is to theoretical redesign the algorithm structure in combination with the characteristics of the S-FEM to improve the computational efficiency. The other is to utilize the parallelism on multicore CPU or many-core GPU. As a variation of the S-FEM, the edge-based smooth element method (ES-FEM) is parallelized on the GPU [9].Because the process of the S-FEM is quite similar to the FEM, most current S-FEM programs are modified on the basis of the FEM [10]. Reference [11] introduced an effective algorithm for establishing the smoothing domains, realizing the automation of 3D entity calculation and adaptive analysis. However, the current S-FEM program, which uses a programming language such as C/C++ or Fortran, usually requires high demand of programming skill, and the code is difficult to read and modify. For programs written in high-level languages such as Python and MATLAB, the code is readable but computationally inefficient [12].In summary, quite few efforts are dedicated to developing the open-source packages of the S-FEM by comprehensively considering the accuracy, efficiency, readability, and ease-ofdevelopment. To balance the program execution efficiency and the ease of implementing the program, in this paper, we design and implement an open-source package of parallel S-FEM for elastic problems by using the Julia language [13]. We term our package as juSFEM. To the best of the authors knowledge, juSFEM is the first package of p...

show abstract

Section: Outlook and Future Workmentioning

confidence: 99%

juSFEM: A Julia-based open-source package of parallel Smoothed Finite Element Method (S-FEM) for elastic problems

Huo

Mei

2021

Computers & Mathematics with Applications

View full text Add to dashboard Cite

show abstract

“…Moreover, these changes facilitate scaling to multiple GPUs across separate machines with MPI. Such holistic approaches are recommended to achieve scalability in heterogeneous environments, 14 15 . We further utilize a variety of CPU and GPU optimizations.…”

Section: Introductionmentioning

confidence: 99%

Enabling simulation of high‐dimensional micro‐macro biophysical models through hybrid CPU and multi‐GPU parallelism

Cook

Shinar

2021

Concurrency and Computation

View full text Add to dashboard Cite

Micro‐macro models provide a powerful tool to study the relationship between microscale mechanisms and emergent macroscopic behavior. However, the detailed microscopic modeling requires tracking and evolving a potentially high‐dimensional configuration space at high computational cost. In this work, we present a novel parallel algorithm for simulating a high‐dimensional micro‐macro model of a gliding motility assay. We utilize a holistic approach aligning the data residency and simulation scales with the hybrid CPU and multi‐GPU hardware. Our novel approach achieves a speedup factor of 9.25× over previous GPU‐accelerated micro‐macro methods on the same hardware. Furthermore, by decoupling dependencies in the microstructure update, we are able to efficiently distribute the microstructure over multiple GPUs with minimal overhead. We test on up to four GPUs and observe excellent scaling, suggesting that significant further speedups are achievable with additional GPUs. Our approach enables micro‐macro simulations of higher complexity and resolution than would otherwise be feasible.

show abstract

“…Key to our approach is moving the microscopic scale and related tasks, which are smaller scale in both space and time, to the GPU and limiting CPU-GPU communications to the longer timescale of the filaments and fluid. Such holistic approaches are recommended to achieve scalability in heterogeneous environments [15], [6]. We note that our method does not suffer from common GPU simulation challenges encountered in various other approaches such as building adjacency lists [10], reordering storage based on cell location [21], dynamic, irregular data accesses [22], thread divergence [5], or neighbor exchanges of halo regions.…”

Section: Introductionmentioning

confidence: 99%

Enabling Simulation of High-Dimensional Micro-Macro Biophysical Models through Hybrid CPU and Multi-GPU Parallelism

Cook¹,

Shinar²

2019

Preprint

View full text Add to dashboard Cite

Micro-macro models provide a powerful tool to study the relationship between microscale mechanisms and emergent macroscopic behavior. However, the detailed microscopic modeling may require tracking and evolving a high-dimensional configuration space at high computational cost. In this work, we present a parallel algorithm for simulation a high-dimensional micro-macro model of a gliding motility assay. We utilize a holistic approach aligning the data residency and simulation scales with the hybrid CPU and multi-GPU hardware. With a combination of algorithmic modifications, GPU optimizations, and scaling to multiple GPUs, we achieve speedup factors of up to 27 over our previous hybrid CPU-GPU implementation and up to 540 over our single-threaded implementation. This approach enables micro-macro simulations of higher complexity and resolution than would otherwise be feasible.

show abstract

Multi-scale Continuum-Particle Simulation on CPU–GPU Hybrid Supercomputer

Cited by 5 publications

References 31 publications

juSFEM: A Julia-based open-source package of parallel Smoothed Finite Element Method (S-FEM) for elastic problems

juSFEM: A Julia-based open-source package of parallel Smoothed Finite Element Method (S-FEM) for elastic problems

Enabling simulation of high‐dimensional micro‐macro biophysical models through hybrid CPU and multi‐GPU parallelism

Enabling Simulation of High-Dimensional Micro-Macro Biophysical Models through Hybrid CPU and Multi-GPU Parallelism

Contact Info

Product

Resources

About