Open-source SCFT on graphics processing units
ORAL
Abstract
Self-consistent field theory provides an effective framework for materials discovery, allowing one to target particular morphologies by rapid examination of their phase space. We present a GPU-accelerated self-consistent field solver parallel to the open-source Polymer Self-Consistent Field (PSCF) codebase. The codebase is built from the ground up to utilize highly efficient GPU-acceleration yet remains backwards-compatible with PSCF. Our testing shows up to 60x acceleration in computing time for the largest problems. Improvement in open-source tools enables new opportunities by the community to pursue various problems of interest such as the inverse design of the bulk phases. Increasingly, speed becomes vital in such problems especially in an era of data and machine learning where the manipulation of large data sets is key to the understanding of fundamental polymer physics.
–
Presenters
-
Guo Kang Cheong
University of Minnesota
Authors
-
Guo Kang Cheong
University of Minnesota
-
Anshul Chawla
University of Minnesota
-
David Clark Morse
University of Minnesota
-
Kevin D Dorfman
Chemical Engineering and Materials Science, Universitiy of Minnesota, University of Minnesota