Performance benchmarks for a next generation numerical dynamo model

被引：69

作者：

Matsui, Hiroaki ^{[1
,2
]}

Heien, Eric ^{[1
,2
]}

Aubert, Julien ^{[3
]}

Aurnou, Jonathan M. ^{[1
,4
]}

Avery, Margaret ^{[5
]}

Brown, Ben ^{[1
,6
]}

Buffett, Bruce A. ^{[1
,7
]}

Busse, Friedrich ^{[8
]}

Christensen, Ulrich R. ^{[9
]}

Davies, Christopher J. ^{[5
]}

Featherstone, Nicholas ^{[1
,10
]}

Gastine, Thomas ^{[9
]}

Glatzmaier, Gary A. ^{[1
,11
]}

Gubbins, David ^{[12
]}

Guermond, Jean-Luc ^{[13
]}

Hayashi, Yoshi-Yuki ^{[14
]}

Hollerbach, Rainer ^{[15
]}

Hwang, Lorraine J. ^{[1
,2
]}

Jackson, Andrew ^{[16
]}

Jones, Chris A. ^{[15
]}

Jiang, Weiyuan ^{[17
]}

Kellogg, Louise H. ^{[1
,2
]}

Kuang, Weijia ^{[17
]}

Landeau, Maylis ^{[18
]}

Marti, Philippe ^{[10
]}

Olson, Peter ^{[1
,18
]}

Ribeiro, Adolfo ^{[4
]}

Sasaki, Youhei ^{[19
]}

Schaeffer, Nathanael ^{[20
]}

Simitev, Radostin D. ^{[21
]}

Sheyko, Andrey ^{[16
]}

Silva, Luis ^{[21
]}

Stanley, Sabine ^{[1
,22
]}

Takahashi, Futoshi ^{[23
]}

Takehiro, Shin-ichi ^{[24
]}

Wicht, Johannes ^{[9
]}

Willis, Ashley P. ^{[25
]}

机构：

[1] Univ Calif Davis, Computat Infrastruct Geodynam Dynamo Working Grp, Davis, CA 95616 USA

[2] Univ Calif Davis, Dept Earth & Planetary Sci, Davis, CA 95616 USA

[3] Univ Paris Diderot, CNRS, Inst Phys Globe Paris, Sorbonne Paris Cite, Paris, France

[4] Univ Calif Los Angeles, Dept Earth & Space Sci, Los Angeles, CA 90024 USA

[5] Univ Calif San Diego, Scripps Inst Oceanog, La Jolla, CA 92093 USA

[6] Univ Colorado, Dept Astrophys & Planetary Sci, Boulder, CO 80309 USA

[7] Univ Calif Berkeley, Dept Earth & Planetary Sci, Berkeley, CA 94720 USA

[8] Univ Bayreuth, Inst Phys, Bayreuth, Germany

[9] Max Planck Inst Sonnensyst Forsch, Gottingen, Germany

[10] Univ Colorado, Dept Appl Math, Boulder, CO 80309 USA

[11] Univ Calif Santa Cruz, Dept Earth & Planetary Sci, Santa Cruz, CA 95064 USA

[12] Univ Leeds, Sch Earth Environm, Leeds, W Yorkshire, England

[13] Texas A&M Univ, Dept Math, College Stn, TX 77843 USA

[14] Kobe Univ, Dept Earth & Planetary Sci, Ctr Planetary Sci, Kobe, Hyogo 657, Japan

[15] Univ Leeds, Dept Appl Math, Leeds LS2 9JT, W Yorkshire, England

[16] Swiss Fed Inst Technol, Inst Geophys, Zurich, Switzerland

[17] NASA Goddard Space Flight Ctr, Planetary Geodynam Lab, Greenbelt, MD USA

[18] Johns Hopkins Univ, Dept Earth & Planetary Sci, Baltimore, MD 21218 USA

[19] Kyoto Univ, Dept Math, Kyoto 606, Japan

[20] Univ Grenoble Alpes, CNRS, ISTerre, Grenoble, France

[21] Univ Glasgow, Sch Math & Stat, Glasgow, Lanark, Scotland

[22] Univ Toronto, Dept Phys, Toronto, ON, Canada

[23] Kyushu Univ, Dept Earth & Planetary Sci, Fukuoka, Japan

[24] Kyoto Univ, Math Sci Res Inst, Kyoto, Japan

[25] Univ Sheffield, Sch Math & Stat, Sheffield, S Yorkshire, England

来源：

GEOCHEMISTRY GEOPHYSICS GEOSYSTEMS | 2016年 / 17卷 / 05期

基金：

美国国家科学基金会;

关键词：

geodynamo; magnetohydrodynamics; benchmark; high-performance computing; FINITE-ELEMENT-METHOD; ROTATING SPHERICAL-SHELL; GEOFEM PLATFORM; MAGNETIC-FIELD; FLUID SHELLS; CONVECTION; SIMULATION; CORE; EQUATIONS; BOUNDARY;

D O I：

10.1002/2015GC006159

中图分类号：

P3 [地球物理学]; P59 [地球化学];

学科分类号：

0708 ; 070902 ;

摘要：

Numerical simulations of the geodynamo have successfully represented many observable characteristics of the geomagnetic field, yielding insight into the fundamental processes that generate magnetic fields in the Earth's core. Because of limited spatial resolution, however, the diffusivities in numerical dynamo models are much larger than those in the Earth's core, and consequently, questions remain about how realistic these models are. The typical strategy used to address this issue has been to continue to increase the resolution of these quasi-laminar models with increasing computational resources, thus pushing them toward more realistic parameter regimes. We assess which methods are most promising for the next generation of supercomputers, which will offer access to O(10(6)) processor cores for large problems. Here we report performance and accuracy benchmarks from 15 dynamo codes that employ a range of numerical and parallelization methods. Computational performance is assessed on the basis of weak and strong scaling behavior up to 16,384 processor cores. Extrapolations of our weak-scaling results indicate that dynamo codes that employ two-dimensional or three-dimensional domain decompositions can perform efficiently on up to approximate to 10(6) processor cores, paving the way for more realistic simulations in the next model generation.

引用

页码：1586 / 1607

页数：22

共 39 条

[1] A non-linear, 3-D spherical α2 dynamo using a finite element method [J].