I am interested in high performance computing (HPC) infrastructure
and the applications that drive its development. Large networks
of realistic neuron models constitute an interesting and
challenging arena for studying architectural bottlenecks
in HPC environments. Simulations of such systems press on
all aspects of HPC – computational cycles, memory hierarchy
parameters, data motion, programming models, interactive
visualization, and fault tolerance. In most of these areas,
large neural network simulations present a substantially
different footprint than other application classes. Furthermore,
I am compelled by the potential for simulations on large-scale
computational platforms to significantly advance our understanding
of brain function and dysfunction.
Further information on Mark Hereld's research.