Asynchronous fine-grain parallel implicit smoother in multigrid solvers for compressible flow

Aditya Kashi; Syam Vangara; Siva Nadarajah; Patrice Castonguay

doi:10.1016/j.compfluid.2019.104255

Asynchronous fine-grain parallel implicit smoother in multigrid solvers for compressible flow

Aditya Kashi, Syam Vangara, Siva Nadarajah, Patrice Castonguay

Research output: Contribution to journal › Article › peer-review

2 Scopus citations

Abstract

New massively parallel hardware for scientific computation promise new levels of performance for computational fluid dynamics codes. However, for leveraging such hardware for implicit solvers, fine-grain parallel algorithms are needed. We develop an asynchronous iteration suitable for parallel solution of systems of equations arising from pseudo-time implicit discretization of the compressible Reynolds-averaged Navier-Stokes equations. We perform numerical experiments to demonstrate the smoothing property and parallel scalability of the developed algorithm as a smoother in nonlinear and linear multigrid solvers. Several cases of external aerodynamics, differing in computational grid size and flow complexity, are used for the study. We also show the scalability of the proposed smoother on Intel's Xeon Phi Knights Landing many-core processor, and investigate the impact of some features of this processor on the parallel scalability of the asynchronous smoother.

Original language	English
Article number	104255
Journal	Computers and Fluids
Volume	198
DOIs	https://doi.org/10.1016/j.compfluid.2019.104255
State	Published - Feb 15 2020
Externally published	Yes

Funding

We thank Dr Hong Yang of Bombardier Aerospace for his help with the contour plots shown on this paper. We gratefully acknowledge funding from the Natural Sciences and Engineering Research Council (NSERC) of Canada, Consortium for Research and Innovation in Aerospace in Canada (CARIC) and Québec (CRIAQ), Bombardier Aerospace and Cray Inc. We also thank Calcul Québec for their help with the Xeon Phi Knights Landing. We thank Dr Hong Yang of Bombardier Aerospace for his help with the contour plots shown on this paper. We gratefully acknowledge funding from the Natural Sciences and Engineering Research Council (NSERC) of Canada, Consortium for Research and Innovation in Aerospace in Canada (CARIC) and Qu?bec (CRIAQ), Bombardier Aerospace and Cray Inc. We also thank Calcul Qu?bec for their help with the Xeon Phi Knights Landing.

Funders	Funder number
Bombardier Aerospace and Cray Inc
Bombardier Aerospace and Cray Inc.
CARIC
CRIAQ
Consortium for Research and Innovation in Aerospace in Canada
Natural Sciences and Engineering Research Council of Canada

Access to Document

10.1016/j.compfluid.2019.104255

Cite this

@article{54a8c0525fc1454494b293b02b540de6,

title = "Asynchronous fine-grain parallel implicit smoother in multigrid solvers for compressible flow",

abstract = "New massively parallel hardware for scientific computation promise new levels of performance for computational fluid dynamics codes. However, for leveraging such hardware for implicit solvers, fine-grain parallel algorithms are needed. We develop an asynchronous iteration suitable for parallel solution of systems of equations arising from pseudo-time implicit discretization of the compressible Reynolds-averaged Navier-Stokes equations. We perform numerical experiments to demonstrate the smoothing property and parallel scalability of the developed algorithm as a smoother in nonlinear and linear multigrid solvers. Several cases of external aerodynamics, differing in computational grid size and flow complexity, are used for the study. We also show the scalability of the proposed smoother on Intel's Xeon Phi Knights Landing many-core processor, and investigate the impact of some features of this processor on the parallel scalability of the asynchronous smoother.",

author = "Aditya Kashi and Syam Vangara and Siva Nadarajah and Patrice Castonguay",

note = "Publisher Copyright: {\textcopyright} 2019 Elsevier Ltd",

year = "2020",

month = feb,

day = "15",

doi = "10.1016/j.compfluid.2019.104255",

language = "English",

volume = "198",

journal = "Computers and Fluids",

issn = "0045-7930",

publisher = "Elsevier",

}

TY - JOUR

T1 - Asynchronous fine-grain parallel implicit smoother in multigrid solvers for compressible flow

AU - Kashi, Aditya

AU - Vangara, Syam

AU - Nadarajah, Siva

AU - Castonguay, Patrice

PY - 2020/2/15

Y1 - 2020/2/15

N2 - New massively parallel hardware for scientific computation promise new levels of performance for computational fluid dynamics codes. However, for leveraging such hardware for implicit solvers, fine-grain parallel algorithms are needed. We develop an asynchronous iteration suitable for parallel solution of systems of equations arising from pseudo-time implicit discretization of the compressible Reynolds-averaged Navier-Stokes equations. We perform numerical experiments to demonstrate the smoothing property and parallel scalability of the developed algorithm as a smoother in nonlinear and linear multigrid solvers. Several cases of external aerodynamics, differing in computational grid size and flow complexity, are used for the study. We also show the scalability of the proposed smoother on Intel's Xeon Phi Knights Landing many-core processor, and investigate the impact of some features of this processor on the parallel scalability of the asynchronous smoother.

AB - New massively parallel hardware for scientific computation promise new levels of performance for computational fluid dynamics codes. However, for leveraging such hardware for implicit solvers, fine-grain parallel algorithms are needed. We develop an asynchronous iteration suitable for parallel solution of systems of equations arising from pseudo-time implicit discretization of the compressible Reynolds-averaged Navier-Stokes equations. We perform numerical experiments to demonstrate the smoothing property and parallel scalability of the developed algorithm as a smoother in nonlinear and linear multigrid solvers. Several cases of external aerodynamics, differing in computational grid size and flow complexity, are used for the study. We also show the scalability of the proposed smoother on Intel's Xeon Phi Knights Landing many-core processor, and investigate the impact of some features of this processor on the parallel scalability of the asynchronous smoother.

UR - http://www.scopus.com/inward/record.url?scp=85076260145&partnerID=8YFLogxK

U2 - 10.1016/j.compfluid.2019.104255

DO - 10.1016/j.compfluid.2019.104255

M3 - Article

AN - SCOPUS:85076260145

SN - 0045-7930

VL - 198

JO - Computers and Fluids

JF - Computers and Fluids

M1 - 104255

ER -

Asynchronous fine-grain parallel implicit smoother in multigrid solvers for compressible flow

Abstract

Funding

Access to Document

Other files and links

Fingerprint

Cite this