Reduced MHD Simulations of the Solar Corona using GPU Accelerated Machines

ruralrompSoftware and s/w Development

Dec 2, 2013 (3 years and 10 months ago)

106 views

C. S. Ng

Geophysical Institute, University of Alaska Fairbanks, Fairbanks, Alaska, USA

Reduced MHD Simulations of the Solar Corona using GPU Accelerated Machines


We have recently carried out a computational campaign

[Ng et al., ApJ
747
, 109, 2012]

to inve
stigate a model of coronal heating
and current
-
sheet formation
in three
-
dimensions
using reduced magnetohydrodynamics (RMHD).
W
e have obtained scaling results of
heating rate versus Lundquist number

(
S
)

based on a series of runs in which random
photospheri
c motions are imposed for hundreds to thousands of Alfvén time in order to
obtain converged statistical values. The heating rate found in these simulations tends to a
level that is independent of

S

and is supported by an analysis based on the random walk
n
ature of the footpoint motions. Various statistical properties, including distributions of
heating events (nanoflares) and current sheet dimensions have been analyzed recently.
Due to long integration time requirement, o
ur code
has been
port
ed

to Nvidia CU
DA
(Compute Unified Device Architecture) for hardware acceleration using graphics
processing units (GPUs).
C
ode performance

has been studied

on several machines that
span a variety of hardware configurations and capabilities
, including

a workstation
equipp
ed with four Nvidia C2050 GPUs, as well as several large
-
scale GPU accelerated
distributed memory machines: Lincoln/NCSA, Dirac/NERSC, and Keeneland/NICS.

Collaborators: L. Lin, A. Bhattacharjee (University of New Hampshire, USA)