Minisymposia and Special Sessions

Minisymposia

 

MS1 - Modelling and Intensive Computation for Aquifer Simulations

Organisers: J. Erhel (INRIA/SAGE, France) , D. Tromeur-Dervout (U. Lyon 1, France)

Numerical modelling is an important key for the management and remediation of groundwater resources. Natural geological formations are highly heterogeneous, leading to preferential flow paths and stagnant regions. The contaminant migration is strongly affected by these irregular water velocity distributions. In order to account for the limited knowledge of the geological characteristics and for the natural heterogeneity, Uncertainty Quantification methods must be used such as classical Monte-Carlo method, with random permeability fields and random Discrete Fracture networks. For such stochastic approach, numerical simulations consist in computing the velocity field over large spatial domains and solving solute transport over large temporal scales. Two main difficulties must be overcome, memory size and runtime, in order to solve very large linear systems and to simulate over a large number of time steps. High performance computing is thus necessary
to carry out these large scale simulations. This minisymposium focuses on some results obtained in the ANR-CIS07-MICAS project.

 

MS2 - Design and implementation of scalable Parallel Algebraical Preconditioners

Organiser: Francois Pacull (FLUOREM, France)

This minisymposium highlights some current research on parallel solvers for sparse linear systems arising from design optimization problems in industrial CFD. While CFD deal with non-linear coupled systems of partial differential equations, the sparse linear systems investigated here are issued from a linearization of the discrete equations, for instance the Reynolds-Averaged Navier-Stokes equations for compressible flows of Newtonian fluids, discretized using the finite-volume method. The Jacobian matrix of a flow is generated here using automatic differentiation. The solutions from these systems are then used to build flow databases, plugged into meta-models in order to parametrize the flow with respect to the geometry or to the boundary conditions. This minisymposium mainly focuses on parallel algebraic preconditoners, but also presents some issues regarding the associated iterative solvers and applications of the resulting flow parametrization approach. 

 

MS3 - Scalable Lattice Boltzmann Method Applications: Best practices on CPU and GPU clusters

Organisers: Gerhard Wellein (Erlangen Regional Computing Center, Germany),

Thomas Seiser (University Erlangen, Germany)

In the past decade the lattice Boltzmann method (LBM) has been established as a promising alternative for the numerical simulation of time-dependent incompressible flows. Its broad applicability ranging from flows in porous media to basic studies of turbulent flows in combination with a simple baseline algorithm makes LBM for a good candidate for efficient parallel computing with both CPUs and GPUs. This minisymposium is a forum for best practice reports on establishing scalable LBM solvers and solving large scale flow problems with LBM solvers on parallel computers based on muticore-CPUs and GPUs.

 

MS4 - High Performance Biomedical Fluid Mechanics 

Organisers: M. Vázquez (BSC, Spain), J. Cebral (GMU, USA), N. Perumal (Swansea Univ., UK)

Physiological systems are amongst the most complex "machines". Obtaining the proper geometries from medical images, with all their variability, personalizing the models adapted to each patient, enrich the Physiological modeling to cover the multiple scales and Physics involved, simulating highly transient problems, validating through scarce experimental results... All these features make Biomedical applications a territory of large scale parallel simulations. This MS welcomes contributions in Parallel CFD in Biomedical applications. 

 

MS5 - Environmental and Aeronautical problems on parallel HPC Environments

Organiser: J. Periaux (CIMNE, Spain)

This MS aims to provide answers to the question "which innovative methods and tools are needed to solve in the future large scale simulation Environmental and Aeronautical problems on Parallel HPC environments?" Large scale simulation with new multi Physics software is more and more often used in Aerospace Engineering Design and Environmental problems. This approach is becoming now and in the near future a critical digital design tool for Industry and Society. The size and complexity of these challenging problems require not only new parallel HPC environments but also new parallel methods to reach a much higher level of efficiency. The objective of this session is to present some of these new parallel concepts in simulation and optimization software implementing new highly parallel methods and to demonstrate their potential on some critical challenging aircraft design applications or greener technologies.

 

Special Sessions

 

SS1 - Efficient programming of multi-core and GPU-based systems: A Lattice Boltzmann Method case study

Organisers: Gerhard Wellein (Erlangen Regional Computing Center, Germany),

Thomas Seiser (University Erlangen, Germany)

Computer architectures have considerably changed over the past years. Trading high single process performance for (massive) on-chip parallelism is the major technology trend which is just about to accelerate. To further benefit from the still exponential progress in semiconductor technology (“Moore’s law”) modern CFD solvers need to be efficient, flexible and scalable. Using the Lattice Boltzmann method (LBM) as an easy to understand prototype, the tutorial addresses the most important topics to build sustainable applications for the years to come. The methods and technologies presented are to a large extent not LBM specific but apply to a broad range of CFD applications.

X
Loading