Parallel Adaptive Mesh Refinement for Mantle Convection Simulation

Omar Ghattas

Joint work with Carsten Burstedde, Michael Gurnis
(Caltech), Georg Stadler, Eh Tan (Caltech), Tiankai Tu
Lucas Wilcox, and Shijie Zhong (CU-Boulder)

Mantle convection is the principal control on the thermal and
geological evolution of the Earth. Mantle convection modeling involves
solution of the mass, momentum, and energy equations for a viscous,
creeping, incompressible non-Newtonian fluid at high Rayleigh and
Peclet numbers. Our goal is to conduct global mantle convection
simulations that can resolve faulted plate boundaries, down to 1 km
scales. Uniform resolution leads to trillion element meshes, which are
intractable even on petascale supercomputers. Thus parallel mesh
adaptivity is essential.

We present Rhea, a new generation mantle convection code designed to
scale to hundreds of thousands of cores. Rhea is built on ALPS, a
parallel octree-based adaptive finite element library that supports
new distributed data structures and parallel algorithms for dynamic
coarsening, refinement, rebalancing, and repartitioning of the
mesh. Using TACC's Ranger system, we demonstrate excellent weak and
strong scalability on up to 32,768 cores and 4.3 billion elements.