GPU Acceleration of an Established Solar MHD Code using OpenACC. (arXiv:1811.02605v1 [physics.comp-ph])
<a href="http://arxiv.org/find/physics/1/au:+Caplan_R/0/1/0/all/0/1">R. M. Caplan</a>, <a href="http://arxiv.org/find/physics/1/au:+Linker_J/0/1/0/all/0/1">J. A. Linker</a>, <a href="http://arxiv.org/find/physics/1/au:+Mikic_Z/0/1/0/all/0/1">Z. Mikić</a>, <a href="http://arxiv.org/find/physics/1/au:+Downs_C/0/1/0/all/0/1">C. Downs</a>, <a href="http://arxiv.org/find/physics/1/au:+Torok_T/0/1/0/all/0/1">T. Török</a>, <a href="http://arxiv.org/find/physics/1/au:+Titov_V/0/1/0/all/0/1">V. S. Titov</a>
GPU accelerators have had a notable impact on high-performance computing
across many disciplines. They provide high performance with low cost/power, and
therefore have become a primary compute resource on many of the largest
supercomputers. Here, we implement multi-GPU acceleration into our Solar MHD
code (MAS) using OpenACC in a fully portable, single-source manner. Our
preliminary implementation is focused on MAS running in a reduced physics
“zero-beta” mode. While valuable on its own, our main goal is to pave the way
for a full physics, thermodynamic MHD implementation. We describe the OpenACC
implementation methodology and challenges. “Time-to-solution” performance
results of a production-level flux rope eruption simulation on multi-CPU and
multi-GPU systems are shown. We find that the GPU-accelerated MAS code has the
ability to run “zero-beta” simulations on a single multi-GPU server at speeds
previously requiring multiple CPU server-nodes of a supercomputer.
GPU accelerators have had a notable impact on high-performance computing
across many disciplines. They provide high performance with low cost/power, and
therefore have become a primary compute resource on many of the largest
supercomputers. Here, we implement multi-GPU acceleration into our Solar MHD
code (MAS) using OpenACC in a fully portable, single-source manner. Our
preliminary implementation is focused on MAS running in a reduced physics
“zero-beta” mode. While valuable on its own, our main goal is to pave the way
for a full physics, thermodynamic MHD implementation. We describe the OpenACC
implementation methodology and challenges. “Time-to-solution” performance
results of a production-level flux rope eruption simulation on multi-CPU and
multi-GPU systems are shown. We find that the GPU-accelerated MAS code has the
ability to run “zero-beta” simulations on a single multi-GPU server at speeds
previously requiring multiple CPU server-nodes of a supercomputer.
http://arxiv.org/icons/sfx.gif
