Abstract: The Computational Framework for Unbiased Studies of Correlated Electron Systems (CompFUSE) SciDAC-4 project seeks to understand, predict, and ultimately control the effects of correlations in quantum materials by developing a computational framework for controlled and unbiased studies of strongly interacting electron systems. In this talk, I will discuss some of the algorithm challenges and improvements in this project.
Speaker’s Bio: Eduardo D’Azevedo has developed and optimized algorithms and application software for advanced high performance computing systems for over 25 years. He is currently the task lead for optimizing the XGC fusion application code for GPU as part of the Center for Accelerated Application Readiness (CAAR) program for ORNL’s Summit system and for Intel Xeon Phi as part of NERSC Exascale Science Applications Program (NESAP). He has developed efficient block update algorithms for QMCPACK, DCA++, and Kronecker product methods in DMRG++. He developed the parallel external memory algorithms for dense matrix computations that take advantage of GPU acceleration. Previously, he has developed mathematical libraries and solvers such as the out-of-core and compact storage in ScaLAPACK library in support of the DOE fusion and materials program. His activities include parallel algorithm design and implementation, application performance optimization, algorithms development, and application performance modeling and prediction.
Host: Eirik Endeve, email@example.com
+1 865-276-6990 United States, Knoxville (Toll)
Conference ID: 872 125 382#
Last Updated: June 5, 2020 - 11:20 am