Telescopic Approach for Extreme-scale Parallel Mesh Generation for CFD Applications

We address two challenges related to Extreme-scale Mesh Generation Environments: (1) Design a multilayered algorithmic and software framework for 3D tetrahedral parallel mesh generation using state-of-theart functionality supported by our telescopic approach for parallel mesh generation. Our approach explores concurrency at all hardware layers using abstractions at (a) medium-grain level for many cores within a single chip and (b) coarse-grain level, i.e., sub-region and sub-domain level using proper error metricand application-specific discrete data and continuous decomposition methods. We anticipate that the telescoping approach will be capable of sustaining a billion-way concurrency the next 15 years by (i) leveraging concurrency at different granularity levels and (ii) carefully mapping work units to take advantage of the memory and network hierarchies. (2) Design a Parallel Runtime System for extreme-scale anisotropic mesh computations to target modules like mesh generation and CFD solvers. The runtime system will provide support for: (a) One-sided explicit message passing, (b) Global name-space, (c) multithreaded programming model for inter-layer interactions, (d) automatic and preemptive load balancing, (e) customizable data-movement and load-balancing, and (f) domain-specific energy-efficient race-to-halt, concurrency throttling, and component-level (core and memory) power scaling. In addition, we present preliminary results from our work on parallel isotropic 3D Delaunay-based guaranteed quality mesh generation on hundreds of cores and potential extensions of to 4D meshes and 3D anisotropic advancing front methods for CFD applications.