08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
›8:00 (30min)
8:00 - 8:30 (30min)
Welcome coffee
›8:30 (10min)
8:30 - 8:40 (10min)
Opening remarks
›8:40 (1h40)
8:40 - 10:20 (1h40)
Session 1
Chair: Valeria Simoncini
› Avoiding breakdown in incomplete factorizations in low precision arithmetic
- Jennifer Scott, STFC-Rutherford Appleton Laboratory
08:40-09:00 (20min)
› Mixed precision strategies for preconditioned GMRES
- Bastien Vieublé, Chinese Academy of Sciences
09:00-09:20 (20min)
› On the numerical stability of the preconditioned conjugate gradient algorithm using mixed precision
- Thomas Bake, Charles University
09:40-10:00 (20min)
›10:20 (40min)
10:20 - 11:00 (40min)
Coffee break
›11:00 (1h20)
11:00 - 12:20 (1h20)
Session 2
Chair: Jean-Michel Muller
› Is Mixed Precision Computing really the Top Priority?
- Hartwig Anzt, Technical University of Munich
11:00-11:20 (20min)
› Emulating High-precision Matrix Operations on Low-precision Matrix Engines
- Rio Yokota, Institute of Science Tokyo
11:20-11:40 (20min)
› Computing Floating-Point Matrix Products via Low-Precision Integer Arithmetic
- Massimiliano Fasi, University of Leeds
11:40-12:00 (20min)
› How Tensor Core-Accelerated Mixed-Precision Computing and Floating-Point Emulation Can Improve Linear Algebra Performance and Energy Efficiency
- Harun Bayraktar, NVIDIA
12:00-12:20 (20min)
›12:20 (1h40)
12:20 - 14:00 (1h40)
Lunch
›14:00 (1h40)
14:00 - 15:40 (1h40)
Session 3
Chair: Paolo Bientinesi
› A practical, fully parallel implementation of the (H-)Tucker decomposition via randomization
- Martina Iannacito, Università di Bologna
14:00-14:20 (20min)
› Recent advances in mixed-precision low-rank tensor approximations
- Eda Oktay, Max Planck Institute for Dynamics of Complex Technical Systems
14:20-14:40 (20min)
› TT-FEAST: A High-Dimensional Eigenvalue Solver using Contour Integration and Tensor Train format
- Mohammad Issa, CEA LIST
14:40-15:00 (20min)
› Error Controlled Compression of Scientific Data with Low-Rank Approximations
- George Turkiyyah, KAUST
15:00-15:20 (20min)
›15:40 (40min)
15:40 - 16:20 (40min)
Coffee break
›16:20 (1h40)
16:20 - 18:00 (1h40)
Session 4
Chair: Hartwig Anzt
› Surrogate techniques for matrix-free sparse linear solvers
- Ulrich Rüde, CERFACS, VSB-Technical University of Ostrava, FAU Erlangen-Nürnberg
16:20-16:40 (20min)
› Multigrid with Linear Storage Complexity
- Daniel Bauer, Friedrich-Alexander-Universität Erlangen-Nürnberg
16:40-17:00 (20min)
› Mixed Precision Domain Decomposition Preconditioners: Perturbation Analysis and Practical Guidelines
- Tom Caruso, LJLL
17:20-17:40 (20min)
› Two-level asynchronous substructuring domain decomposition solvers
- Frédéric Magoulès, CentraleSupéléc
17:40-18:00 (20min)
