Charles Moulinec

Towards Petascale Computing with Parallel CFD codes

Many world leading high-end computing (HEC) facilities are now offering over 100 Teraflops/s of performance and several initiatives have begun to look forward to Petascale computing5 (1015 flop/s). Los Alamos National Laboratory and Oak Ridge National Laboratory (ORNL) already have Petascale systems, which are leading the current (Nov 2008) TOP500 list [1]. Computing at the Petascale raises a number of significant challenges for parallel computational fluid dynamics codes. Most significantly, further improvements to the performance of individual processors will be limited and therefore Petascale systems are likely to contain 100,000+ processors. Thus a critical aspect for utilising high Terascale and Petascale resources is the scalability of the underlying numerical methods, both with execution time with the number of processors and scaling of time with problem size. In this paper we analyse the performance of several CFD codes for a range of datasets on some of the latest high performance computing architectures. This includes Direct Numerical Simulations (DNS) via the SBLI [2] and SENGA2 [3] codes, and Large Eddy Simulations (LES) using both STREAMS LES [4] and the general purpose open source CFD code Code Saturne [5].

Benchmark Study of a 3d Parallel Code for the Propagation of Large Subduction Earthquakes

Benchmark studies were carried out on a recently optimized parallel 3D seismic wave propagation code that uses finite differences on a staggered grid with 2ndorder operators in time and 4thorder in space. Three dual-core supercomputer platforms were used to run the parallel program using MPI. Efficiencies of 0.91 and 0.48 with 1024 cores were obtained on HECToR (UK) and KanBalam (Mexico), and 0.66 with 8192 cores on HECToR. The 3D velocity field pattern from a simulation of the 1985 Mexico earthquake (that caused the loss of up to 30000 people and about 7 billion US dollars) which has reasonable agreement with the available observations, shows coherent, well developed surface waves propagating towards Mexico City.

Multiple threads and parallel challenges for large simulations to accelerate a general Navier–Stokes CFD code on massively parallel systems

Computational fluid dynamics is an increasingly important application domain for computational scientists. In this paper, we propose and analyze optimizations necessary to run CFD simulations consisting of multibillion‐cell mesh models on large processor systems. Our investigation leverages the general industrial Navier–Stokes CFD application, Code_Saturne, developed by Electricité de France for incompressible and nearly compressible flows. In this paper, we outline the main bottlenecks and challenges for massively parallel systems and emerging processor features such as many‐core, transactional memory, and thread level speculation. We also present an approach based on an octree search algorithm to facilitate the joining of mesh parts and to build complex larger unstructured meshes of several billion grid cells. We describe two parallel strategies of an algebraic multigrid solver and we detail how to introduce new levels of parallelism based on compiler directives with OpenMP, transactional memory and thread level speculation, for finite volume cell‐centered formulation and face‐based loops. A renumbering scheme for mesh faces is proposed to enhance thread‐level parallelism. Copyright

Computation of Flow in a 3D Diffuser Using a Two-Velocity Field Hybrid RANS/LES

The flow inside a three-dimensional diffuser is computed with a two-velocity hybrid RANS/LES model that ensues the separation of dissipative effects of the mean and fluctuating fields to be treated individually; by extracting a running average velocity field from instantaneous quantities. The averaged field is then used to calculate the contribution of the mean shear which is larger than that from the fluctuating one over the near wall region. Results with the hybrid model are presented and compared to those from a DES on the same grid. RANS results obtained with models upon which both these approaches are based are also reported. The RANS results are unable to capture essential characteristics of the flow whereas the hybrid method produces results which generally agree well with the experiment.

HPC for hydraulics and industrial environmental flow simulations

This paper describes the parallelization needs required by the TELEMAC system, a complete hydroinformatics tool developed by EDF R&D. A focus on three codes part of the TELEMAC system, namely Estel-3D, Telemac-3D and Spartacus-3D, is achieved. For each code, performance in terms of speed up is presented, as well as an industrial application.

LES of a Simplified HVAC System Used for Aeroacoustic Predictions

In recent years noise emissions have been strongly regulated in many industries in order to reduce noise pollution and to improve comfort. A typical example of noise generated by a turbulent flow inside a Heat and Ventilation Air Conditioning (HVAC) system of a vehicle is presented.

Sleeve leakage gas impact on fuel assembly temperature distribution

ABSTRACT The flow in a large section (80%) of an Advanced Gas-cooled Reactor fuel assembly is investigated in two configurations, the former dealing with the fuel assembly operating in normal plant conditions and the latter with a cold gas injected sideways at the top of Element 2 of the fuel assembly. Results show that injecting the cold gas decreases the temperature at the top of the assembly and that mixing is almost complete there.

High-Performance Computing and Smoothed Particle Hydrodynamics

Smoothed Particle Hydrodynamics is a gridless numerical method that can be used to simulate highly complex flows with one or more free surfaces. In the context of engineering applications, very few 3-D simulations have been carried out due to prohibitive computational cost since the number of particles required in 3-D is usually too large to be handled by a single processor. In this work, an improved version of a parallel 3-D SPH code, Spartacus-3D, is presented. Modifications to the code, which include a localisation of all the previously global arrays combined with a switch from global communications to local ones where possible, lead to a more efficient parallel code and allow for a substantial increase in the number of particles in a simulation.