The performance of three implementations of the restarted GMRES algorithm with randomized-SVD-based preconditioning has been analyzed. They have been tested with a wide population of matrices of varying properties and compared to the standard ILU(0) preconditioning. The implementations comprise two variants of the preconditioner, one of them implemented for CPU-only and hybrid GPU–CPU executions to better assess the benefits and pitfalls in both contexts. The trade-off between iteration-to-solution and time-to-solution metrics is discussed and it is shown that a competitive convergence rate is attained. In addition, the CPU off-loading to the GPU leads to a significant improvement of the second metric for the largest matrices analyzed. Such an in efficiency is promising considering the various codes and tools that employ the restarted GMRES algorithm.

The performance of a CPU-only implementation of the restarted GMRES algorithm with direct randomized-SVD -based preconditioning has been analyzed. The method has been tested on a set of sparse and dense matrices exhibiting varying spectral properties and compared to the ILU(0) -based preconditioning. This comparison aims to assess the advantages and drawbacks of both approaches. The trade-off between iteration-to-solution and time-to-solution metrics is discussed, demonstrating that the proposed method achieves an improved convergence rate in terms of iterations. Additionally, the method's competitiveness with respect to both metrics is discussed within the context of several relevant scenarios, particularly those where GMRES-based simulation techniques are applicable.

A novel sparse encoding algorithm is developed to detect and study plasma instabilities automatically. This algorithm, called Elastic Random Mode Decomposition, is applied to the Mirnov coil signals of a dataset of 1291 discharges of the TJ-II stellarator, enabling the identification of the Alfv & eacute;nic activity. In the presented approach, each signal is encoded as a collection of basic waveforms called atoms, drawn from a signal dictionary. Then the modes are identified using clustering and correlations with other plasma signals. The performance of the proposed algorithm is dramatically increased by using elastic net regularization and taking advantage of GPU architectures. Therefore the signal size and the number of dictionary elements are no longer limiting factors for encoding complex signals. Once the modes are retrieved from the shots, standard clustering and dimensionality reduction techniques are applied to obtain a 2D map featuring of the physical mode characteristics of this subset of TJ-II shots. The clustering features consider the relationship with the plasma current Ip, the diamagnetic energy W, and inverse square root of electron density 1/n, profiling different subtypes of Alfv & eacute;nic activity. The proposed algorithm can potentially create large databases of labeled modes with unprecedented detail.

We provide an overview of activities carried out at the TJ-II stellarator for improving our understanding of- and developing plasma physics models for particle density profiles in stellarators. Namely, we report on recent progress in turbulent particle transport simulation, validation of pellet deposition models, density profile shaping for performance control and new experimental techniques for edge turbulence and plasma-neutral interaction.

Every computation presents a small chance that an unexpected phenomenon ruins or modifies its output. Computers are prone to errors that, although may be very unlikely, are hard, expensive or simply impossible to avoid. In the exascale, with thousands of processors involved in a single computation, those errors are especially harmful because they can corrupt or distort the results, wasting human and material resources. In the present work, we study the effect of ionizing radiation on several pieces of commercial hardware, very common in modern supercomputers. Aiming to reproduce the natural radiation that could arise, CPUs (Xeon, EPYC) and GPUs (A100, V100, T4) are subject to a known flux of neutrons coming from two radioactive sources, namely ^252 Cf and ^241 Am-Be, in a special irradiation facility. The working hardware is irradiated under supervision to quantify any appearing error. Once the hardware response is characterised, we are able to scale down the radiation intensity and to estimate the effects on standard data centres. This can help administrators and researchers to develop their contingency plans and protocols.

The present investigation summarizes the numerical prediction and DMD analysis of the flow in the vicinity of a Gurney flap set on an airfoil NACA0015 at Re=10,300 tested in a water tunnel with time-resolved Particle Image Velocimetry. A series of two-dimensional U-RANS simulations has been accomplished using an incompressible, time-accurate solver of OpenFOAM. A purely data-based DMD analysis has followed to extract the spatio-temporal coherent structures from the time-evolving flowfields, to mimic the experiments. It is stressed that this work focuses on the analysis of this kind of flap by applying DMD to datasets generated with numerical simulations, which has not been previously investigated, but using experimental datasets. Under this approach, the comparison of the numerics with the experimental counterpart shows a remarkable agreement for both the structures and major frequencies, which stresses the suitability of the incompresible U-RANS approach in this class of studies. Special attention has been focused on the modelling of the experimental facility, and a quantification of the prediction accuracy is provided by assessing the deviation of the spectral content. The simulations clearly show the key role of having a good characterization of the water tunnel to be successful at capturing the spectrum and flow unsteadiness. Furthermore, they reveal quite promising for those industrial oriented simulations, which deal with lift-enhancement devices and lift-to-drag ratio optimization of airfoils since they permit an accurate prediction.

We present a study of the performance of the Weather Research and Forecasting [WRF] code under several hardware configurations in an HPC environment. The WRF code is a standard code for weather prediction, used in several fields of science and industry. The metrics used in this case are the execution time of the run and the energy consumption of the simulation obtained with the LIMITLESS monitor, which is the main novelty of this work. With these results, it is possible to quantify the energy savings of WRF run configurations, which include variations in the number of computing nodes and in the number of processes per node. It is found out that a slight increase in the computing time can drive to a noticeable reduction in the energy consumption of the cluster.

The resilience behavior of three GMRES prototyped implementations (with Incomplete LU, Flexible and randomized-SVD-based preconditioners) has been analyzed with a soft errors injection approach. A low-level fault injector is inserted into the GMRES solvers, which randomly select locations in the program to inject the fault across multiple executions. This fault injection approach combines the configurability of high-level and the accuracy of low-level techniques at the same time, so the effect of faults may be closely emulated. In order to gather enough statistical data, a set of eighteen sparse matrix-based linear systems Ax = b has been solved with these GMRES implementations in the injection experiments and monitored. The results of this prototype-based fault injection suggest an improved error resilience behavior of the randomized-SVD-based preconditioned GMRES version in many of the analyzed matrices, which points out to its interest in supercomputing applications where silent errors are more prominent.

A dynamic version of Cloudbook is presented in this work, a new tool for automatically and unattendedly parallelizing codes which also lately distributes the tasks dynamically. Cloudbook is designed for Python codes and, above all, makes the parallelization in a way in which the number and main characteristics of the available infrastructure is taken into account for optimizing the execution (performance, bandwidth connection, etc.) in a dynamic way. Cloudbook is designed to allow developers to get the technical benefits of automated distribution and parallelization of programs with a very low learning cost. It only requires labelling the original code with a reduced set of pragmas located at function headers. Results of the tests carried out with Cloudbook with several codes on a real infrastructure are presented as well.

The control and mitigation of MHD oscillations modes is an open problem in fusion science because they can contribute to the outward particle/energy flux and can drive the device away from ignition conditions. It is then of general interest to extract the mode information from large experimental databases in a fast and reliable way. We present a software tool based on Deep Learning that can identify these oscillations modes taking Mirnov coil spectrograms as input data. It uses Convolutional Neural Networks that we trained with manually annotated spectrograms from the TJ-II stellarator database. We have tested several detector architectures, resultingin a detector AUC score of 0.99 on the test set. Finally, it is applied to find MHD modes in our spectrograms to show how this new software tool can be used to mine other databases.

BACKGROUND:In the era of big data, networks are becoming a popular factor in the field of data analysis. Networks are part of the main structure of BeGraph software, which is a 3D visualization application dedicated to the analysis of complex networks.OBJECTIVE:The main objective of this research was to visually analyze tendencies of mental health diseases in a region of Spain, using the BeGraph software, in order to make the most appropriate health-related decisions in each case.METHODS:For the study, a database was used with 13,531 records of patients with mental health disorders in three acute medical units from different health care complexes in a region of Spain. For the analysis, BeGraph software was applied. It is a web-based 3D visualization tool that allows the exploration and analysis of data through complex networks.RESULTS:The results obtained with the BeGraph software allowed us to determine the main disease in each of the health care complexes evaluated. We noted 6.50% (463/7118) of admissions involving unspecified paranoid schizophrenia at the University Clinic of Valladolid, 9.62% (397/4128) of admissions involving chronic paranoid schizophrenia with acute exacerbation at the Zamora Hospital, and 8.84% (202/2285) of admissions involving dysthymic disorder at the Rio Hortega Hospital in Valladolid.CONCLUSIONS:The data analysis allowed us to focus on the main diseases detected in the health care complexes evaluated in order to analyze the behavior of disorders and help in diagnosis and treatment.

An adequate confinement of α-particles is fundamental for the operation of future fusion powered reactors. An even more critical situation arises for stellarator devices, whose complex magnetic geometry can substantially increase α-particle losses. A traditional approach to transport evaluation is based on a diffusive paradigm; however, a growing body of literature presents a considerable amount of examples and arguments toward the validity of non-diffusive transport models for fusion plasmas, particularly in cases of turbulent driven transport [R. Sánchez and D. E. Newman, Plasma Phys. Controlled Fusion 57, 123002 (2015)]. Likewise, a recent study of collisionless α-particle transport in quasi-toroidally symmetric stellarators [A. Gogoleva et al., Nucl. Fusion 60, 056009 (2020)] puts the diffusive framework into question. In search of a better transport model, we numerically characterized and quantified the underlying nature of transport of the resulting α-particle trajectories by employing a whole set of tools, imported from the fractional transport theory. The study was carried out for a set of five configurations to establish the relation between the level of the magnetic field toroidal symmetry and the fractional transport coefficients, i.e., the Hurst H, the spatial α, and the temporal β exponents, each being a merit of non-diffusive transport. The results indicate that the α-particle ripple-enhanced transport is non-Gaussian and non-Markovian. Moreover, as the degree of quasi-toroidal symmetry increases, it becomes strongly subdiffusive, although the validity of the fractional model itself becomes doubtful in the limiting high and low symmetry cases.

At present, network analysis based on the graph theory has become a widely used technique in the field of Mental Health. The networks are part of the main structure of BeGraph software, a 3D visualization cloud application that allows the analysis of complex networks. The main objective of this study is to analyze, through the BeGraph software, the behavior of Mental Health prevalent diseases in a region of Spain, in order to make health decisions. The study used a database with a total of 9403 patient’s records with Mental Health diseases, which belong to two hospitals in Castilla and Leon, Spain, and the 3D visualization software, BeGraph. The results obtained allow us to determine the main diseases detected in each hospital included in the study: 6.5% of admissions from the University Clinic of Valladolid with unspecified paranoid schizophrenia and 8.84% of admissions from Rio Hortega Hospital with dysthymic disorder. The analysis of the data allows us to focus on the Mental Health main pathologies detected in the hospitals evaluated, and propose prediction algorithms that help in their diagnosis and treatment.

In ASDEX Upgrade (AUG), the normalised gyroradius ρ⋆ was varied via a hydrogen isotope scan while keeping other dimensionless parameters constant. This was done in L-mode, to minimise the impact of pedestal stability on confinement. Power balance and perturbative transport analyses reveal that the electron heat transport is unaffected by the differences in isotope mass. Nonlinear simulations with the Gene code suggest that these L-mode discharges are ion temperature gradient (ITG) dominated. The different gyroradii due to the isotope mass do not necessarily result in a change of the predicted heat fluxes. This result is used in simulations with the Astra transport code to match the experimental profiles. In these simulations the experimental profiles and confinement times are reproduced with the same transport coefficients for hydrogen and deuterium plasmas. The mass only enters in the energy exchange term between electrons and ions. These numerical observations are supported by additional experiments which show a lower ion energy confinement compared to that of the electrons. Additionally, hydrogen and deuterium plasmas have a similar confinement when the energy exchange time between electrons and ions is matched. This strongly suggests that the observed isotope dependence in L-mode is not dominated by a gyroradius effect, but a consequence of the mass dependence in the collisional energy exchange between electrons and ions.

Due to their capability to reduce turbulent transport in magnetized plasmas, understanding the dynamics of zonal flows is an important problem in the fusion program. Since the pioneering work by Rosenbluth and Hinton in axisymmetric tokamaks, it is known that studying the linear and collisionless relaxation of zonal flow perturbations gives valuable information and physical insight. Recently, the problem has been investigated in stellarators and it has been found that in these devices the relaxation process exhibits a characteristic feature: a damped oscillation. The frequency of this oscillation might be a relevant parameter in the regulation of turbulent transport, and therefore its efficient and accurate calculation is important. Although an analytical expression can be derived for the frequency, its numerical evaluation is not simple and has not been exploited systematically so far. Here, a numerical method for its evaluation is considered, and the results are compared with those obtained by calculating the frequency from gyrokinetic simulations. This ‘semianalytical’ approach for the determination of the zonal-flow frequency is revealed to be accurate and faster than the one based on gyrokinetic simulations.

The effects of 3D geometry are explored in TJ-II from two relevant points of view: neoclassical transport and modification of stability and dispersion relation of waves. Particle fuelling and impurity transport are studied considering the 3D transport properties, paying attention to both neoclassical transport and other possible mechanisms. The effects of the 3D magnetic topology on stability, confinement and Alfven Eigenmodes properties are also explored, showing the possibility of controlling Alfven modes by modifying the configuration; the onset of modes similar to geodesic acoustic modes are driven by fast electrons or fast ions; and the weak effect of magnetic well on confinement. Finally, we show innovative power exhaust scenarios using liquid metals.