Exascale Computing CODAR
As collaborators in all five co-design centers created by the Exascale Computing Project (ECP), Argonne researchers are helping to solve complex challenges and accelerate discoveries in materials science, precision medicine, national security and numerous other fields, to help pave the way for the creation of exascale supercomputers.
Argonne is leading the Co-design Center for Online Data Analysis and Reduction at the Exascale (CODAR), as summarized below. In addition, Argonne is collaborating in the other four co-design centers as described here.
Co-design Center for Online Data Analysis and Reduction at the Exascale (CODAR)
CODAR will address a growing disparity between simulation speeds and I/O rates rendering it infeasible for HPC and data analytic applications to perform offline analysis. CODAR will target common data analysis and reduction methods (e.g., feature and outlier detection, compression) and methods specific to particular data types and domains (e.g., particles, FEM). In addition, CODAR will focus on deducing application development risk by providing performance tradeoffs for offline vs. inline analyses of simulation results, produce and integrate into applications high-performance products embodying data analysis and reduction methods.
Center for Efficient Exascale Discretizations (CEED)
CEED will focus on the development of finite element (FE) discretization libraries to enable unstructured PDE-based applications to take full advantage of exascale resources and implement complicated FE machinery on coming exascale hardware. Additionally this center will focus on high-order methods for high-fidelity and better performance, with a range of orders providing flexibility in uncertain hardware and software environments, extending to problems with unstructured non-conforming mesh refinement and general curved meshes, including low-order FE discretizations.
Co-design Center for Particle Applications (CoPA)
The CoPA co-design center will be focused on four sub-motifs: short-range particle-particle (e.g., MD and SPH), long-range particle-particle (e.g., electrostatic and gravitational), particle-in-cell, and additional sparse matrix and graph operations of linear-scaling quantum MD. It will serve as a centralized clearinghouse for particle-based methods, and as first users on immature simulators, emulators, and prototype hardware. Deliverables include “Numerical Recipes for Particles” best practices, libraries, and scalable open exascale software platform.
Block-Structured AMR Co-design Center
This co-design center will be a new block-structured Adaptive Mesh Refinement (AMR) framework for systems of nonlinear PDEs, providing basis for temporal and spatial discretization strategy for DOE applications. It will support hierarchical solutions at multiple levels of resolution, with each level of refinement being the union of data containers at that resolution, each of which represents the solution over a logically rectangular subregion of the domain. In addition, it will support conventional representation of field variables on a mesh as well as particle data and embedded boundary representations of complex geometries.
GraphEx Co-design Center
The GraphEx Co-Design Center collaborates with exascale applications that use combinatorial kernels, with key examples being smart power grid, computational biology, computational chemistry, wind energy, and national security. Combinatorial (graph) kernels that play crucial enabling roles in the chosen application areas will be co-designed: graph traversals; graph matching; graph coloring; graph clustering, parallel branch-and-bound and graph partitioning.