We present a "multipatch" infrastructure for numerical simulation of fluid
problems in which sub-regions require different gridscales, different grid
geometries, different physical equations, or different reference frames. Its
key element is a sophisticated client-router-server framework for efficiently
linking processors supporting different regions ("patches") that must exchange
boundary data. This infrastructure may be used with a wide variety of fluid
dynamics codes; the only requirement is that their primary dependent variables
be the same in all patches, e.g., fluid mass density, internal energy density,
and velocity. Its structure can accommodate either Newtonian or relativistic
dynamics. The overhead imposed by this system is both problem- and computer
cluster architecture-dependent. Compared to a conventional simulation using the
same number of cells and processors, the increase in runtime can be anywhere
from negligible to a factor of a few; however, one of the infrastructure's
advantages is that it can lead to a very large reduction in the total number of
zone-updates.
Description
[1701.05610] PATCHWORK: A Multipatch Infrastructure for Multiphysics/Multiscale/Multiframe Fluid Simulations
%0 Generic
%1 shiokawa2017patchwork
%A Shiokawa, Hotaka
%A Cheng, Roseanne M.
%A Noble, Scott C.
%A Krolik, Julian H.
%D 2017
%K code patches simulation
%T PATCHWORK: A Multipatch Infrastructure for
Multiphysics/Multiscale/Multiframe Fluid Simulations
%U http://arxiv.org/abs/1701.05610
%X We present a "multipatch" infrastructure for numerical simulation of fluid
problems in which sub-regions require different gridscales, different grid
geometries, different physical equations, or different reference frames. Its
key element is a sophisticated client-router-server framework for efficiently
linking processors supporting different regions ("patches") that must exchange
boundary data. This infrastructure may be used with a wide variety of fluid
dynamics codes; the only requirement is that their primary dependent variables
be the same in all patches, e.g., fluid mass density, internal energy density,
and velocity. Its structure can accommodate either Newtonian or relativistic
dynamics. The overhead imposed by this system is both problem- and computer
cluster architecture-dependent. Compared to a conventional simulation using the
same number of cells and processors, the increase in runtime can be anywhere
from negligible to a factor of a few; however, one of the infrastructure's
advantages is that it can lead to a very large reduction in the total number of
zone-updates.
@misc{shiokawa2017patchwork,
abstract = {We present a "multipatch" infrastructure for numerical simulation of fluid
problems in which sub-regions require different gridscales, different grid
geometries, different physical equations, or different reference frames. Its
key element is a sophisticated client-router-server framework for efficiently
linking processors supporting different regions ("patches") that must exchange
boundary data. This infrastructure may be used with a wide variety of fluid
dynamics codes; the only requirement is that their primary dependent variables
be the same in all patches, e.g., fluid mass density, internal energy density,
and velocity. Its structure can accommodate either Newtonian or relativistic
dynamics. The overhead imposed by this system is both problem- and computer
cluster architecture-dependent. Compared to a conventional simulation using the
same number of cells and processors, the increase in runtime can be anywhere
from negligible to a factor of a few; however, one of the infrastructure's
advantages is that it can lead to a very large reduction in the total number of
zone-updates.},
added-at = {2017-01-23T10:10:02.000+0100},
author = {Shiokawa, Hotaka and Cheng, Roseanne M. and Noble, Scott C. and Krolik, Julian H.},
biburl = {https://www.bibsonomy.org/bibtex/23992ff93646c7e0d53b70ab2c9182c67/miki},
description = {[1701.05610] PATCHWORK: A Multipatch Infrastructure for Multiphysics/Multiscale/Multiframe Fluid Simulations},
interhash = {5ee82828f1123758388f41cecf6ea64e},
intrahash = {3992ff93646c7e0d53b70ab2c9182c67},
keywords = {code patches simulation},
note = {cite arxiv:1701.05610Comment: 17 pages, 9 figures, submitted to ApJ},
timestamp = {2017-01-23T10:10:02.000+0100},
title = {PATCHWORK: A Multipatch Infrastructure for
Multiphysics/Multiscale/Multiframe Fluid Simulations},
url = {http://arxiv.org/abs/1701.05610},
year = 2017
}