Parallel-in-time algorithms for exascale applications

Home > Workshops > 2025 > Parallel-in-time algorithms for exascale applications

Parallel-in-time algorithms for exascale applications

 07 - 11 Jul 2025

ICMS, Bayes Centre, Edinburgh

 Enquiries

Scientific Organisers

  • David Acreman, Hewlett Packard Enterprise
  • Jemma Shipton, University of Exeter
  • Beth Wingate, University of Exeter
  • Hiroe Yamazaki, Imperial College London

Keynote Speakers:

  • Giancarlo Antonucci, UK Atomic Energy Authority
  • Rob Falgout, Lawrence Livermore National Laboratory
  • Josh Hope-Collins, Imperial College London
  • Felix Kwok, Université Laval
  • Juliane Rosemeier, Freie Universität Berlin

About:

Efficient computation of solutions to partial differential equations on modern supercomputers is crucial for many scientific applications such as weather and climate prediction, fusion modelling, molecular dynamics and engineering applications. As we have reached the limit of Dennard scaling, which previously guaranteed faster processors due to smaller components, modern supercomputers now have orders of magnitude more processors than previous generations. Exascale supercomputers are capable of 10^18 floating point operations per second but making full use of this hardware for the solution of partial differential equations requires a deep understanding of the mathematical properties of the governing equations and how to discretise them in a way that enables calculations to be performed concurrently, that is in parallel.

Parallel-in-time algorithms, the topic of this workshop, enable parallel computation in the time domain, in addition to more traditional parallelism through spatial domain decomposition. Time-parallelism is significantly more challenging than spatial parallelism due to causality, since the state of the system at future times must depend on that at past times. However, the advent of the exascale era has inspired recent research in to these algorithms, leading to an explosive increase in the number of different algorithms available. The efficiency of these algorithms depends on tailoring the algorithm to the equations being solved and this requires interdisciplinary collaboration both with the application experts and with mathematicians working in adjacent areas.

This workshop will bring together such an interdisciplinary group of people to advance our understanding of these algorithms and open up new areas of collaboration and research.

 

Programme

MONDAY 7 JULY 2025
10.00 – 10.30 Registration
10.30 - 12.00 Training session (optional)
12.00 - 13.30 Lunch + registration if not attending training session
13.30 - 14.30 Josh Hope-Collins, Imperial College London ParaDiag preconditioners for nonlinear and variable-coefficient problems
14.30 - 15.00 Ryo Yoda, University of Wuppertal Toward efficient solvers using block-epsilon circulant preconditioning on modern integrated CPU-GPU systems
15.00 - 15.30 Refreshments
15.30 – 16.00 Andreas Schafelner, Johannes Kepler University Linz A parallel-in-time solver for nonlinear degenerate time-periodic parabolic problems
16.00 - 16.30 Alexander Schell, University of Münster Towards Vectorised Block Krylov Parallel in Time Methods
16.30 – 17.30 Welcome Reception
TUESDAY 8 JULY 2025
09.15 – 10.15 Juliane Rosemeier, Freie Universität Berlin Multilevel Parareal Methods and Standard Form Transformations for Weakly Nonlinear Problems
10.15 – 10.45 Werner Bauer, University of Surrey Accurate solutions of highly oscillatory systems under large time steps using higher-order phase averages
10.45 – 11.15 Refreshments
11.15 - 11.45 Martin Gander, Université de Genève Parareal for hyperbolic problems just does not work, or does it?
11.45 - 12.15 Stephan Rave, University of Münster A Parareal Algorithm with Spectral Coarse Solver
12.15 - 12.45 Joao Guilherme Caldas Steinstraesser, University of Sao Paulo Stability and numerical study of Parareal and MGRIT applied to the shallow water equations on the rotating sphere
12.45 – 14.15 Lunch
14.15 – 14.45 Daniel Ruprecht, Hamburg University of Technology Machine learning based coarse propagators for Parareal
14.45 - 15.15 Massimiliano Tamborrino, University of Warwick ProbParareal: A Probabilistic Numerical Parallel-in-Time Solver for Differential Equations
15.15 – 15.45 Refreshments
15.45 - 16.15 Abdelouahed Ouardghi, Jülich Supercomputing Centre/ Forschungszentrum Jülich Space-Time Parallelism using Spectral Deferred Corrections and Finite Elements for Incompressible Navier–Stokes Equations
16.15 - 16.45 Thomas Baumann, Forschungszentrum Juelich Diagonal Spectral Deferred Correction for 3D Rayleigh-Benard convection
16.45 - 17.15 Alex Brown, Met Office/ University of Exeter A comparison of time-parallel "across the method" deferred correction schemes for atmospheric modelling
WEDNESDAY 9 JULY 2025
09.15 – 10.15 Giancarlo Antonino Antonucci, UKAEA Time-parallel algorithms for chaotic systems
10.15-10.45 Sriramkrishnan Muralikrishnan, Jülich Supercomputing Centre/ Forschungszentrum Jülich GmbH A Massive Space-Time Parallel Particle-In-Fourier Framework for Kinetic Plasma Simulations
10.45 – 11.15 Refreshments
11.15 - 11.45 Thibaut Lunet, Hamburg University of Technology Spectral Deferred Correction: from theoretical analysis to design of new time-parallel algorithms
11.45 - 12.15 Lisa Wimmer, Bergische Universität Wuppertal On the application of spectral deferred corrections to differential-algebraic equations
12.15 - 12.45 Hans Johansen, Lawrence Berkeley National Laboratory New approaches to space-time splitting with higher accuracy
12.45 Lunch and free afternoon
THURSDAY 10 JULY 2025
09.15 – 10.15 Rob Falgout, Lawrence Livermore National Laboratory Parallel-in-Time Solution of Hyperbolic PDE Systems via Characteristic-Variable Block Preconditioning
10.15 – 10.45 David Vargas, Sandia National Laboratories Parallel multigrid in time for chaos with timescale-independent convergence
10.45 – 11.15 Refreshments
11.15 - 11.45 Arne Naaegel, Goethe University Frankfurt Scalable parallel-in-time solvers for linear poroelasticty
11.45 – 12.15 Wiebke Drews, TU Dortmund University Stabilized Finite Element Multigrid Techniques for Space-Time Parallelism in Convection-Diffusion Problems
12.15 – 12.45 Julius Ehigie, University of Lagos Parallelizing Internal Stages in High-Order Two-Derivative DIRK Methods with applications
12.45 – 14.15 Lunch
14.15 – 15.15 Felix Kwok, Université Laval Optimized Schwarz methods in time for discrete transport control
15.15 – 15.45 Sebastian Götschel, Hamburg University of Technology Parallelization in time for inverse problems
15.45 - 16.15 Refreshments
16.15 – 16.45 Nick Janssens, KU Leuven Parallel-in-time multiple shooting using large-eddy simulation for flow reconstruction in the atmospheric boundary layer
16.45 – 17.15 Bernhard Heinzelreiter, University of Edinburgh A Diagonalization-Based Parallel-in-Time Preconditioner for Instationary Flow Control Problems
19.00 onwards Workshop Dinner at Apex Grassmarket Hotel The Heights Room, Apex Grassmarket Hotel, 31-35 Grassmarket, Edinburgh EH1 2HS
FRIDAY 11 JULY 2025
09.15 – 09.45 Magnus Appel, University of Southern Denmark Towards Fast Topology Optimisation of Transient Heat Conduction Using Space-time Multigrid Methods
09.45 – 10.15 Sean Hon, Hong Kong Baptist University An optimal parallel-in-time preconditioner for parabolic optimal control problems
10.15 – 10.45 Mahadevan Ganesh, Colorado School of Mines Parallel-in-time-and-space simulation for a class of models with non-local operators
10.45 – 11.15 Refreshments
11.15 - 12.30 Closing discussion
12.30 Lunch