Description of our 3D MHD Computations
Our final prediction was run on a 3D mesh with 221 x
221 x 432 mesh
points (about 21 million cells) in spherical coordinates
(r,θ,φ).
Various versions of this run with different parameters were run on two
supercomputers,
Ranger,
a massively parallel supercomputer at the
Texas Advanced
Computing Center (TACC), and Pleiades,
the new massively parallel supercomputer at NASA's Advanced Supercomputing Divison
(NAS).
Pleiades
is a high performance
computer
with 51,200 CPUs and a theoretical peak speed of 609 teraflops/s.
In July 2009 it was ranked as the
fourth fastest supercomputer in the
world.
Pleiades is an SGI Altix that uses quad-core Intel Xeon E5472
(Harpertown) 3.0 GHz processors and Intel Xeon X5355 (Clovertown) 2.66
GHz processors, with an Infiniband interconnect. The superb
system run
by NAS allowed us to run long jobs with minimal queue wait times,
allowing us to successfully run our code under the real-time constrains
required for a prediction.
Ranger is a high performance
computer
with 62,976 CPUs and a theoretical peak speed of 580 teraflops/s.
In July 2009 it was ranked as the eighth
fastest supercomputer in the
world.
Ranger is a SunBlade Linux Cluster that uses quad-core AMD
Opteron
(Barcelona) 2.3 GHz processors with an Infiniband interconnect.
The staff at TACC
graciously agreed to give us dedicated time to run our
simulations without interruption in order to make a timely prediction
for the eclipse, and were very helpful in making this happen.
We used our spherical 3D (magnetohydrodynamic) MHD
code MAS, which integrates
the MHD equations using semi-implicit (Alfvén and sound waves),
fully implicit (diffusive terms), and explicit (flow terms)
schemes.
We solve the very large sparse matrix equations generated by these
algorithms using a preconditioned iterative conjugate gradient
solver. We set as a boundary condition the radial component of
the magnetic
field at the base of the corona. This field is deduced from MDI magnetograms
aboard the SOHO spacecraft,
which measure the line
of sight component of the photospheric magnetic field from space.
Our code is written in Fortran
90 and uses the Message
Passing
Interface (MPI) for interprocessor communication. Our code
scales
very well on many high-performance computer systems. We have
shown
essentially linear scaling with processor number up to
about 4096 processors.
Our calculation ran for about 85,000 time steps,
relaxing the corona
in time (for about 1.8 days of solar time) towards a steady state,
thereby approximating the state of the solar corona. Our time
step in the computation was about 1.9 seconds. The
final eclipse prediction runs used 4992 processors on Ranger
and 2000 processors on Pleiades, and ran continuously for about 4
days. We are very grateful for the assistance provided to us by
the dedicated staff at TACC and NAS. Our prediction would not
have been possible without these resources.