Luis Fazendeiro

Performance of distributed multiscale simulations

Multiscale simulations model phenomena across natural scales using monolithic or component-based code, running on local or distributed resources. In this work, we investigate the performance of distributed multiscale computing of component-based models, guided by six multiscale applications with different characteristics and from several disciplines. Three modes of distributed multiscale computing are identified: supplementing local dependencies with large-scale resources, load distribution over multiple resources, and load balancing of small- and large-scale resources. We find that the first mode has the apparent benefit of increasing simulation speed, and the second mode can increase simulation speed if local resources are limited. Depending on resource reservation and model coupling topology, the third mode may result in a reduction of resource consumption.

Real science at the petascale

We describe computational science research that uses petascale resources to achieve scientific results at unprecedented scales and resolution. The applications span a wide range of domains, from investigation of fundamental problems in turbulence through computational materials science research to biomedical applications at the forefront of HIV/AIDS research and cerebrovascular haemodynamics. This work was mainly performed on the US TeraGrid ‘petascale’ resource, Ranger, at Texas Advanced Computing Center, in the first half of 2008 when it was the largest computing system in the world available for open scientific research. We have sought to use this petascale supercomputer optimally across application domains and scales, exploiting the excellent parallel scaling performance found on up to at least 32 768 cores for certain of our codes in the so-called ‘capability computing’ category as well as high-throughput intermediate-scale jobs for ensemble simulations in the 32–512 core range. Furthermore, this activity provides evidence that conventional parallel programming with MPI should be successful at the petascale in the short to medium term. We also report on the parallel performance of some of our codes on up to 65 636 cores on the IBM Blue Gene/P system at the Argonne Leadership Computing Facility, which has recently been named the fastest supercomputer in the world for open science.

Effects of the equilibrium model on impurity transport in tokamaks

Gyrokinetic simulations of ion temperature gradient mode and trapped electron mode driven impurity transport in a realistic tokamak geometry are presented and compared with results using simplified geometries. The gyrokinetic results, obtained with the GENE code in both linear and non-linear modes are compared with data and analysis for a dedicated impurity injection discharge at JET. The impact of several factors on heat and particle transport is discussed, lending special focus to tokamak geometry and rotational shear. To this end, results using s-alpha and concentric circular equilibria are compared with results with magnetic geometry from a JET experiment. To further approach experimental conditions, non-linear gyrokinetic simulations are performed with collisions and a carbon background included. The impurity peaking factors, computed by finding local density gradients corresponding to zero particle flux, are discussed. The impurity peaking factors are seen to be reduced by a factor of ~2 in realistic geometry compared with the simplified geometries, due to a reduction of the convective pinch. It is also seen that collisions reduce the peaking factor for low-Z impurities, while increasing it for high charge numbers, which is attributed to a shift in the transport spectra towards higher wavenumbers with the addition of collisions. With the addition of roto-diffusion, an overall reduction of the peaking factors is observed, but this decrease is not sufficient to explain the flat carbon profiles seen at JET.

New variational principles for locating periodic orbits of differential equations

We present new methods for the determination of periodic orbits of general dynamical systems. Iterative algorithms for finding solutions by these methods, for both the exact continuum case, and for approximate discrete representations suitable for numerical implementation, are discussed. Finally, we describe our approach to the computation of unstable periodic orbits of the driven Navier–Stokes equations, simulated using the lattice Boltzmann equation.

Unstable periodic orbits in the Lorenz attractor

We apply a new method for the determination of periodic orbits of general dynamical systems to the Lorenz equations. The accuracy of the expectation values obtained using this approach is shown to be much larger and have better convergence properties than the more traditional approach of time averaging over a generic orbit. Finally, we discuss the relevance of the present work to the computation of unstable periodic orbits of the driven Navier–Stokes equations, which can be simulated using the lattice Boltzmann method.