On-line Lecture by Dr. Osni Marques on December 23, 2021

Date & Time：　December 24, 2021 10:00-11:00

Place:　Zoom 
                 For registration, please contact Prof. Katagiri by email. 　katagiri@cc.nagoya-u.ac.jp
            

Speaker: Dr. Osni Marques

Topic:  An Overview of the IDEAS-ECP Project and Minimization Strategies for Solving Eigenvalue Problem

Summary:
This presentation will consist of two parts. In the first part, I will given overview of IDEAS-ECP, a project funded by the US Department of Energy
Exascale Project (ECP) that seeks to promote software quality as essential for quality science. IDEAS-ECP evaluates and disseminates best practices and methodologies to improve developer productivity, software sustainability, and scientific reproducibility. I will summarize the activities carried by IDEAS-ECP, which range from collaborations with individual ECP teams to help in the understand productivity bottlenecks and improve software development practices, to interactions with ECP software developers to help identify training and support needs, and the organization of a variety of events that foster the discussion of software development experiences. In the second part, I will discuss ongoing work on unconstrained minimization schemes for dealing with eigenvalue problems. These schemes employ a preconditioned conjugate gradient approach that avoids an explicit reorthogonalization of the trial eigenvectors, in contrast to typical iterative eigenvalue solvers, therefore reducing communications and becoming a potential alternative for the solution of large problems on massively parallel computers. Our current focus is on problems related to electronic structure calculations. Preliminary results show that the unconstrained formulation, together with an appropriate preconditioner, offers good convergence properties and scales well on a large number of cores.