TY - CHAP
T1 - Parallel computation of turbulent fluid flows with the piecewise-parabolic method
AU - Woodward, Paul R.
AU - Porter, David H.
AU - Anderson, Sarah E.
AU - Edgar, B. Kevin
AU - Puthenveetil, Amitkumar
AU - Fuchs, Tyler
PY - 2006
Y1 - 2006
N2 - The chapter uses the piecewise-parabolic method (PPM) gas dynamics code on large NSF TeraGrid parallel systems as well as on SMP and cluster systems in a lab, to simulate turbulent flows at grid resolutions of 10243 and 20483 cells. The chapter discusses both single- and two-fluid flows with statistically homogeneous and unstable shear layer initial conditions. The purpose of these simulations is to capture detailed datasets that allow design and validation of sub grid-scale turbulence models. The parallel code implementation manages a set of shared data objects, each describing a sub domain of the problem. These data objects can be instantiated as either disk files or in-memory objects on a designated set of network nodes that serve them to computational processes over a cluster network. When implemented as disk files, large problems can be run, at no loss in performance, on systems that do not have sufficient memory to hold a complete description of the problem state.
AB - The chapter uses the piecewise-parabolic method (PPM) gas dynamics code on large NSF TeraGrid parallel systems as well as on SMP and cluster systems in a lab, to simulate turbulent flows at grid resolutions of 10243 and 20483 cells. The chapter discusses both single- and two-fluid flows with statistically homogeneous and unstable shear layer initial conditions. The purpose of these simulations is to capture detailed datasets that allow design and validation of sub grid-scale turbulence models. The parallel code implementation manages a set of shared data objects, each describing a sub domain of the problem. These data objects can be instantiated as either disk files or in-memory objects on a designated set of network nodes that serve them to computational processes over a cluster network. When implemented as disk files, large problems can be run, at no loss in performance, on systems that do not have sufficient memory to hold a complete description of the problem state.
UR - http://www.scopus.com/inward/record.url?scp=77953622671&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=77953622671&partnerID=8YFLogxK
U2 - 10.1016/B978-044452206-1/50010-0
DO - 10.1016/B978-044452206-1/50010-0
M3 - Chapter
AN - SCOPUS:77953622671
SN - 9780444522061
SP - 93
EP - 100
BT - Parallel Computational Fluid Dynamics 2005
PB - Elsevier
ER -