High rayleigh number mantle convection on GPU

David A. Sanchez; Christopher Gonzalez; David A. Yuen; Grady B. Wright; Gregory A. Barnett

doi:10.1007/978-3-642-16405-7_22

High rayleigh number mantle convection on GPU

David A. Sanchez, Christopher Gonzalez, David A. Yuen, Grady B. Wright, Gregory A. Barnett

Earth and Environmental Sciences-Twin Cities

Research output: Chapter in Book/Report/Conference proceeding › Chapter

Abstract

We implemented two- and three-dimensional Rayleigh–Benard convection on Nvidia GPUs by utilizing a 2nd-order finite difference method. By exploiting the massive parallelism of GPU using both CUDA for C and optimized CUBLAS routines, we have on a single Fermi GPU run simulations of Rayleigh number up to 6 × 10¹⁰ (on a mesh of 2000 × 4000 uniform grid points) in two dimensions and up to 10⁷ (on a mesh of 450 × 450 × 225 uniform grid points) for three dimensions. On Nvidia Tesla C2070 GPUs, these implementations enjoy single-precision performance of 535 GFLOP/s and 100 GFLOP/s respectively, and double-precision performance of 230 GFLOP/s and 70 GFLOP/s respectively.

Original language	English (US)
Title of host publication	Lecture Notes in Earth System Sciences
Publisher	Springer International Publishing
Pages	335-352
Number of pages	18
Edition	9783642164040
DOIs	https://doi.org/10.1007/978-3-642-16405-7_22
State	Published - 2013

Publication series

Name	Lecture Notes in Earth System Sciences
Number	9783642164040
Volume	0
ISSN (Print)	2193-8571
ISSN (Electronic)	2193-858X

Bibliographical note

Publisher Copyright:
© Springer-Verlag Berlin Heidelberg 2013.

Publisher link

10.1007/978-3-642-16405-7_22

Find It

Find It at U of M Libraries

Cite this

@inbook{c0f8bc2562c14609bcc59a6de214bef2,

title = "High rayleigh number mantle convection on GPU",

abstract = "We implemented two- and three-dimensional Rayleigh–Benard convection on Nvidia GPUs by utilizing a 2nd-order finite difference method. By exploiting the massive parallelism of GPU using both CUDA for C and optimized CUBLAS routines, we have on a single Fermi GPU run simulations of Rayleigh number up to 6 × 1010 (on a mesh of 2000 × 4000 uniform grid points) in two dimensions and up to 107 (on a mesh of 450 × 450 × 225 uniform grid points) for three dimensions. On Nvidia Tesla C2070 GPUs, these implementations enjoy single-precision performance of 535 GFLOP/s and 100 GFLOP/s respectively, and double-precision performance of 230 GFLOP/s and 70 GFLOP/s respectively.",

author = "Sanchez, {David A.} and Christopher Gonzalez and Yuen, {David A.} and Wright, {Grady B.} and Barnett, {Gregory A.}",

note = "Publisher Copyright: {\textcopyright} Springer-Verlag Berlin Heidelberg 2013.",

year = "2013",

doi = "10.1007/978-3-642-16405-7_22",

language = "English (US)",

series = "Lecture Notes in Earth System Sciences",

publisher = "Springer International Publishing",

number = "9783642164040",

pages = "335--352",

booktitle = "Lecture Notes in Earth System Sciences",

edition = "9783642164040",

}

TY - CHAP

T1 - High rayleigh number mantle convection on GPU

AU - Sanchez, David A.

AU - Gonzalez, Christopher

AU - Yuen, David A.

AU - Wright, Grady B.

AU - Barnett, Gregory A.

PY - 2013

Y1 - 2013

N2 - We implemented two- and three-dimensional Rayleigh–Benard convection on Nvidia GPUs by utilizing a 2nd-order finite difference method. By exploiting the massive parallelism of GPU using both CUDA for C and optimized CUBLAS routines, we have on a single Fermi GPU run simulations of Rayleigh number up to 6 × 1010 (on a mesh of 2000 × 4000 uniform grid points) in two dimensions and up to 107 (on a mesh of 450 × 450 × 225 uniform grid points) for three dimensions. On Nvidia Tesla C2070 GPUs, these implementations enjoy single-precision performance of 535 GFLOP/s and 100 GFLOP/s respectively, and double-precision performance of 230 GFLOP/s and 70 GFLOP/s respectively.

AB - We implemented two- and three-dimensional Rayleigh–Benard convection on Nvidia GPUs by utilizing a 2nd-order finite difference method. By exploiting the massive parallelism of GPU using both CUDA for C and optimized CUBLAS routines, we have on a single Fermi GPU run simulations of Rayleigh number up to 6 × 1010 (on a mesh of 2000 × 4000 uniform grid points) in two dimensions and up to 107 (on a mesh of 450 × 450 × 225 uniform grid points) for three dimensions. On Nvidia Tesla C2070 GPUs, these implementations enjoy single-precision performance of 535 GFLOP/s and 100 GFLOP/s respectively, and double-precision performance of 230 GFLOP/s and 70 GFLOP/s respectively.

UR - http://www.scopus.com/inward/record.url?scp=85038082044&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85038082044&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-16405-7_22

DO - 10.1007/978-3-642-16405-7_22

M3 - Chapter

AN - SCOPUS:85038082044

T3 - Lecture Notes in Earth System Sciences

SP - 335

EP - 352

BT - Lecture Notes in Earth System Sciences

PB - Springer International Publishing

ER -

High rayleigh number mantle convection on GPU

Abstract

Publication series

Bibliographical note

Publisher link

Find It

Other files and links

Fingerprint

Cite this