CRD Staff Share Expertise at SIAM Parallel Processing Conference

Brian Van Straalen, Andrew Canning, and Osni Marques (background) at the SIAMPP2020 meeting.

From organizing committees and sessions to presenting papers on a range of high performance computing and mathematical topics, some two dozen scientists from Berkeley Lab’s Computational Research Division (CRD) participated in the 2020 SIAM Conference on Parallel Processing for Scientific Computing, held Feb. 12-15 in Seattle.

Aydin Buluc was on the meeting’s Organizing Committee and Proceedings Paper Committee; session organizers from CRD included:

• Ann Almgren, “Experiences in Developing GPU Support for Department of Energy Math Libraries”

• Zhe Bai and Zezuan Xu, “Physical-Based Modeling and Machine Learning for Biological and Environmental Sciences”

• Roel Van Beeumen, David B. Williams-Young, and Chao Yang, “Parallel Eigenvalue Algorithms for Physical Simulation:

• Pieter Ghysels and Yang Liu, “Low-Rank Compression-Based Fast Sparse Direct Solvers”

• Osni Marques, “New Approaches for Software Auto-Tuning and Accuracy Assurance”

Ann Almgren discussed one of her favorite topics at the conference, AMR.

In addition, CRD staff were involved in a number of technical presentations during the meeting:

• PETSc’s Accelerator Model and Algebraic Multigrid Work and Data Placement at Extreme-Scale - Mark Adams

• Geometric and Algebraic Multigrid Solvers in PETSc on Many-GPU Supercomputer Architectures - Mark Adams

• AMReX: A Block- Structured AMR Software Framework for the Exascale - Ann Almgren, John Bell, Kevin Gott, Weiqun Zhang, Andrew Myers

• 3D FFTs on HPC ManyCore and Hybrid CPU-GPU Platforms: Applications in Materials and Chemistry Codes - Andrew Canning

• AMReX on GPUs: Strategies, Challenges and Lessons Learned - Kevin Gott

• Infusing Physics and Domain Knowledge into ML and DL Models - Karthik Kashinath, Prabhat (NERSC)

• A Communication-Avoiding Sparse Direct Solver for CPU+GPU Platforms - Sherry Li

Sherry Li gave one of several presentations at SIAMPP2020.

• Clustering Techniques and Hierarchical Matrix Formats for Scalable Kernel Ridge Regression - Sherry Li, Yang Liu

• Parallel Butterfly-Based Sherman-Morrison-Woodbury Inversion - Yang Liu, Ghysels Pieter, Sherry Li

• Parallelizing Exponential Integrators with PFASST - Michael Minion

• Surrogate Optimization for HPC Applications - Juliane Mueller

• Portable Performance for AMR on GPUs: The Proto Approach - Brian Van Straalen,

• Implementation of Fast Eigensolvers on GPUs - Chao Yang

• GPU-Powered Particle-in-Cell Community Frameworks for Laser-Plasma Interaction - WarpX Collaboration

• Leveraging One-Sided Communication for Sparse Triangular Solvers - Nan Ding, Samuel Williams, Yang Liu, Sherry Li

Chao Yang

• An Overview of Particles in AMReX, with Applications to Accelerator Modeling, Cosmology, and Multi-Phase Flow - Andrew Myers

• Parallel Shift-Invert Spectrum Slicing for Symmetric Self- Consistent Eigenvalue Computation - David Williams-Young

• Incorporating Hierarchical Matrix Compression and Butterfly Factorizations in a Multifrontal Lu Solver - Pieter Ghysels, Yang Liu, Sherry Li

• MFIX-Exa: An Exascale CFD-DEM Model for Reactor Design Engineering - Ann Almgren

• Autotuning Exascale Applications - Yang Liu, Sherry Li, Osni Marques

• A Scalable Matrix-Free Eigensolver for Studying Many-Body Localization - Roel Van Beeumen, Chao Yang

• WarpX: Electromagnetic Particle-in-Cell with Adaptive Mesh Refinement for Advanced Particle Accelerators - Weiqun Zhang, Andrew Myers, Ann Almgren, John Bell