ICL Research Profile

Overview

The Development of Exascale Software for Heterogeneous and Interfacial Catalysis (DESC) project focuses on understanding the relationship between algorithms and hardware platforms and how to jointly optimize the software and hardware to achieve efficient implementations for applications in materials science, chemistry, and physics. DESC is a joint effort between ICL/UTK, DOE’s Ames Laboratory, EP Analytics, Inc., Georgia Tech, Old Dominion University, and Virginia Tech and is funded by the DOE Computational Chemical Sciences project.

ICL’s contribution focuses on expressing GAMESS computational chemistry algorithms in the form of a dataflow graph and subsequently mapping the DAG representation of the kernels to the hardware platforms. This representation allows for capturing the essential properties of the algorithms (e.g., data dependencies) and computation at extreme scale by utilizing the hardware components (e.g., CPU or GPU) best suited for the type of computational task under consideration. The dataflow-based form of these algorithms makes them compatible with next-generation task scheduling systems like PaRSEC and StarPU.