There is great need for coordination around standards and best practices in neuroscience to support efforts to make neuroscience a data-centric discipline. Major brain initiatives launched around the world are poised to generate huge stores of neuroscience data. At the same time, neuroscience, like many domains in biomedicine, is confronting the issues of transparency, rigor, and reproducibility. Widely used, validated standards and best practices are key to addressing the challenges in both big and small data science, as they are essential for integrating diverse data and for developing a robust, effective, and sustainable infrastructure to support open and reproducible neuroscience. However, developing community standards and gaining their adoption is difficult. The current landscape is characterized both by a lack of robust, validated standards and a plethora of overlapping, underdeveloped, untested and underutilized standards and best practices. The International Neuroinformatics Coordinating Facility (INCF), an independent organization dedicated to promoting data sharing through the coordination of infrastructure and standards, has recently implemented a formal procedure for evaluating and endorsing community standards and best practices in support of the FAIR principles. By formally serving as a standards organization dedicated to open and FAIR neuroscience, INCF helps evaluate, promulgate, and coordinate standards and best practices across neuroscience. Here, we provide an overview of the process and discuss how neuroscience can benefit from having a dedicated standards body.

The hydrogen fuel cell vehicle market is projected to increase in the coming years, and fuel cell vehicles will operate in an environment where they coexist with combustion engine vehicles. In this context, the PEM fuel cell will be exposed to significant amounts of contaminants on the roads that will decrease its performance and durability. In the present study the PEM fuel cell is exposed to 100 ppm of nitrogen dioxide in the airflow. Different methods for recovery of performance were tested; recovery during constant current load and by subjecting the cell to successive polarization curves. The results showed that the successive polarization curves are the best choice for recovery. However, recovery at low current density and high potential is also a good alternative.

Colon adenocarcinoma (COAD) is a common tumor of the gastrointestinal tract with a high mortality rate. Current research has identified many genes associated with immune infiltration that play a vital role in the development of COAD. In this study, we analysed the prognostic and diagnostic features of such immune-related genes in the context of colonic adenocarcinoma (COAD). We analysed 17 overlapping gene expression profiles of COAD and healthy samples obtained from TCGA-COAD and public single-cell sequencing resources, to identify potential therapeutic COAD targets. We evaluated the abundance of immune infiltration with those genes using the TIMER (Tumor Immune Estimation Resource) deconvolution method. Subsequently, we developed predictive and survival models to assess the prognostic value of these genes. The LGALS4 (Galectin-4) gene was found to be significantly (P<0.05) downregulated in COAD and bladder urothelial carcinoma (BLCA) compared to healthy samples. We identified LGALS4 as a prognostic and diagnostic marker for multiple cancer types, including COAD and BLCA. Our analysis reveals a series of novel candidate drug targets, as well as candidate molecular markers, that may explain the pathogenesis of COAD and BLCA. LGALS4 gene is associated with multiple cancer types and is a possible prognostic, as well as diagnostic, marker of COAD and BLCA.

Coarse-Grained Reconfigurable Arrays (CGRAs) are a class of reconfigurable architectures that inherit the performance of Domain-specific accelerators and the reconfigurability aspects of Field-Programmable Gate Arrays (FPGAs). Historically, CGRAs have been successfully used to accelerate embedded applications and are now considered to accelerate High-Performance Computing (HPC) applications in future supercomputers. However, embedded systems and supercomputers are two vastly different domains with different applications and constraints, and it is today not fully understood what CGRA design decisions adequately cater to the HPC market. One such unknown design decision is regarding the interconnect that facilitates intra-CGRA communication. Our findings show that even the typical king-style mesh-like topology is often under-utilized with a typical HPC workload, leading to inefficiency. This research aims to explore the provisioning of intra-CGRA interconnect for HPC-oriented workloads and, ultimately, recoup the potential performance and efficiency lost by reducing the interconnect complexity. We proposed several reduced interconnect topologies based on the usage statistic. Then we evaluate the tradeoffs regarding hardware cost, routability of DFGs, and computational throughput.

Among the more salient accelerator technologies to continue performance scaling in High-Performance Computing (HPC) are Coarse-Grained Reconfigurable Arrays (CGRAs). However, what benefits CGRAs will bring to HPC workloads and how those benefits will be reaped is an open research question today. In this work, we propose a framework to explore the design space of CGRAs for HPC workloads, which includes a tool flow of compilation and simulation, a CGRA HDL library written in SystemVerilog, and a synthesizable CGRA design as a baseline. Using RTL simulation, we evaluate two well-known computation kernels with the baseline CGRA for multiple different architectural parameters. The simulation results demonstrate both correctness and usefulness of our exploration framework.

Coarse-Grained Reconfigurable Arrays (CGRAs) are a class of reconfigurable architectures that inherit the performance and usability properties of Central Processing Units (CPUs) and the reconfigurability aspects of Field-Programmable Gate Arrays (FPGAs). Historically, CGRAs have been successfully used to accelerate embedded applications and are today also being considered to accelerate High-Performance Computing (HPC) applications in future supercomputers. However, embedded systems and supercomputers are two vastly different domains with different applications and constraints, and it is today not fully understood what CGRA design decisions adequately cater to the HPC market. One such unknown design decision is regarding the interconnect that facilitates intra-CGRA communication. Today, intra-CGRA communication comes in two flavors: using routers closely embedded into the compute units or using discrete routers outside the compute units. The former trades flexibility for a reduction in hardware cost, while the latter has greater flexibility but is more resource hungry. In this paper, we aspire to understand which of both designs best suits the CGRA HPC segment. We extend our previous methodology, which consists of both a parameterized CGRA design and an OpenMP-capable compiler, to accommodate both types of routing designs, including verification tests using RTL simulation. Our results show that the discrete router design can facilitate better use of processing elements (PEs) compared to embedded routers and can achieve up to 79.27% reduction in unnecessary PE occupancy for an aggressively unrolled stencil kernel on a 18 x 16 CGRA at a (estimated) hardware resource overhead cost of 6.3x. This reduction in PE occupancy can be used, for example, to exploit instruction-level parallelism (ILP) through even more aggressive unrolling.

This research aims to explore the tradeoffs between routing flexibility and hardware resource usage, ultimately reducing the resource usage of our CGRA architecture while maintaining compute efficiency. we investigate statistics of connection usages among switch blocks for benchmark DFGs, propose several CGRA architecture with a reduced connection, and evaluate their hardware cost, routability of DFGs, and computational throughput for benchmarks. We found that the topology with horizontal plus diagonal connection saves about 30% of the resource usage while maintaining virtually the same routing flexibility as the full connectivity topology.

Comprehending the performance bottlenecks at the core of the intricate hardware-software inter-actions exhibited by highly parallel programs on HPC clusters is crucial. This paper sheds light on the issue of automatically asynchronous MPI communication in memory-bound parallel programs on multicore clusters and how it can be facilitated. For instance, slowing down MPI processes by deliberate injection of delays can improve performance if certain conditions are met. This leads to the counter-intuitive conclusion that noise, independent of its source, is not always detrimental but can be leveraged for performance improvements. We employ phase-space graphs as a new tool to visualize parallel program dynamics. They are useful in spotting certain patterns in parallel execution that will easily go unnoticed with traditional tracing tools. We investigate five different microbenchmarks and applications on different supercomputer platforms: an MPI-augmented STREAM Triad, two implementations of Lattice-Boltzmann fluid solvers (D3Q19 and SPEChpc D2Q37), the LULESH and HPCG proxy applications.

This paper studies the utility of using data analytics and machine learning techniques for identifying, classifying, and characterizing the dynamics of large-scale parallel (MPI) programs. To this end, we run microbenchmarks and realistic proxy applications with the regular compute-communicate structure on two different supercomputing platforms and choose the per-process performance and MPI time per time step as relevant observables. Using principal component analysis, clustering techniques, correlation functions, and a new “phase space plot,” we show how desynchronization patterns (or lack thereof) can be readily identified from a data set that is much smaller than a full MPI trace. Our methods also lead the way towards a more general classification of parallel program dynamics.

Aberrant insulin-like growth factor 1 (IGF-1) signaling has been proposed as a contributing factor to the development of neurodegenerative disorders including diabetic neuropathy, and delivery of exogenous IGF-1 has been explored as a treatment for Alzheimer's disease and amyotrophic lateral sclerosis. However, the role of autocrine/paracrine IGF-1 in neuroprotection has not been well established. We therefore used in vitro cell culture systems and animal models of diabetic neuropathy to characterize endogenous IGF-1 in sensory neurons and determine the factors regulating IGF-1 expression and/or affecting neuronal health. Single-cell RNA sequencing (scRNA-Seq) and in situ hybridization analyses revealed high expression of endogenous IGF-1 in non-peptidergic neurons and satellite glial cells (SGCs) of dorsal root ganglia (DRG). Brain cortex and DRG had higher IGF-1 gene expression than sciatic nerve. Bidirectional transport of IGF-1 along sensory nerves was observed. Despite no difference in IGF-1 receptor levels, IGF-1 gene expression was significantly (P < 0.05) reduced in liver and DRG from streptozotocin (STZ)-induced type 1 diabetic rats, Zucker diabetic fatty (ZDF) rats, mice on a high-fat/ high-sugar diet and db/db type 2 diabetic mice. Hyperglycemia suppressed IGF-1 gene expression in cultured DRG neurons and this was reversed by exogenous IGF-1 or the aldose reductase inhibitor sorbinil. Transcription factors, such as NFAT1 and CEBP beta, were also less enriched at the IGF-1 promoter in DRG from diabetic rats vs control rats. CEBP beta overexpression promoted neurite outgrowth and mitochondrial respiration, both of which were blunted by knocking down or blocking IGF-1. Suppression of endogenous IGF-1 in diabetes may contribute to neuropathy and its upregulation at the transcriptional level by CEBP beta can be a promising therapeutic approach.

Coronavirus disease 2019 (COVID-19) is caused by a novel virus named Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2). This virus induced a large number of deaths and millions of confirmed cases worldwide, creating a serious danger to public health. However, there are no specific therapies or drugs available for COVID-19 treatment. While new drug discovery is a long process, repurposing available drugs for COVID-19 can help recognize treatments with known clinical profiles. Computational drug repurposing methods can reduce the cost, time, and risk of drug toxicity. In this work, we build a graph as a COVID-19 related biological network. This network is related to virus targets or their associated biological processes. We select essential proteins in the constructed biological network that lead to a major disruption in the network. Our method from these essential proteins chooses 93 proteins related to COVID-19 pathology. Then, we propose multiple informative features based on drug-target and protein-protein interaction information. Through these informative features, we find five appropriate clusters of drugs that contain some candidates as potential COVID-19 treatments. To evaluate our results, we provide statistical and clinical evidence for our candidate drugs. From our proposed candidate drugs, 80% of them were studied in other studies and clinical trials.

High-Performance Computing (HPC) has become an important scientific driver. A wide variety of research ranging for example from drug design to climate modelling is nowadays performed in HPC systems. Furthermore, the tremendous computer power of such HPC systems allows scientists to simulate problems that were unimaginable a few years ago. However, the continuous increase in size and complexity of HPC systems is turning the development of efficient parallel software into a difficult task. Therefore, the use of per- formance monitoring and analysis is a must in order to unveil inefficiencies in parallel software. Nevertheless, performance tools also face challenges as a result of the size of HPC systems, for example, coping with huge amounts of performance data generated.

In this thesis, we propose a new model for performance characterisation of MPI applications that tackles the challenge of big performance data sets. Our approach uses Event Flow Graphs to balance the scalability of profiling techniques (generating performance reports with aggregated metrics) with the richness of information of tracing methods (generating files with sequences of time-stamped events). In other words, graphs allow to encode ordered se- quences of events without storing the whole sequence of such events, and therefore, they need much less memory and disk space, and are more scal- able. We demonstrate in this thesis how our Event Flow Graph model can be used as a trace compression method. Furthermore, we propose a method to automatically detect the structure of MPI applications using our Event Flow Graphs. This knowledge can afterwards be used to collect performance data in a smarter way, reducing for example the amount of redundant data collected. Finally, we demonstrate that our graphs can be used beyond trace compression and automatic analysis of performance data. We propose a new methodology to use Event Flow Graphs in the task of visual performance data exploration.

In addition to the Event Flow Graph model, we also explore in this thesis the design and use of performance data introspection frameworks. Future HPC systems will be very dynamic environments providing extreme levels of parallelism, but with energy constraints, considerable resource sharing, and heterogeneous hardware. Thus, the use of real-time performance data to or- chestrate program execution in such a complex and dynamic environment will be a necessity. This thesis presents two different performance data introspec- tion frameworks that we have implemented. These introspection frameworks are easy to use, and provide performance data in real time with very low overhead. We demonstrate, among other things, how our approach can be used to reduce in real time the energy consumed by the system.

The approaches proposed in this thesis have been validated in different HPC systems using multiple scientific kernels as well as real scientific applica- tions. The experiments show that our approaches in performance character- isation and performance data introspection are not intrusive at all, and can be a valuable contribution to help in the performance monitoring of future HPC systems.

The expected high levels of parallelism together with the heterogeneity and complexity of new computing systems pose many challenges to current software. New programming approaches and runtime systems that can simplify the development of parallel applications are needed. Task-based runtime systems have emerged as a good solution to cope with high levels of parallelism, while providing software portability, and easing program development. However, these runtime systems require real-time information on the state of the system to properly orchestrate program execution and optimise resource utilisation. In this paper, we present a lightweight monitoring infrastructure developed within the AllScale Runtime System, a task-based runtime system for extreme scale. This monitoring component provides real-time introspection capabilities that help the runtime scheduler in its decision-making process and adaptation, while introducing minimum overhead. In addition, the monitoring component provides several post-mortem reports as well as real-time data visualisation that can be of great help in the task of performance debugging.

We design and develop a new Particle-in-Cell (PIC) method for plasma simulations using Deep-Learning (DL) to calculate the electric field from the electron phase space. We train a Multilayer Perceptron (MLP) and a Convolutional Neural Network (CNN) to solve the two-stream instability test. We verify that the DL-based MLP PIC method produces the correct results using the two-stream instability: the DL-based PIC provides the expected growth rate of the two-stream instability. The DL-based PIC does not conserve the total energy and momentum. However, the DL-based PIC method is stable against the cold-beam instability, affecting traditional PIC methods. This work shows that integrating DL technologies into traditional computational methods is a viable approach for developing next-generation PIC algorithms.

We are experiencing a deluge of data in all fields of scientific and business research, particularly in the life sciences, due to the development of better instrumentation and the rapid advancements that have occurred in information technology in recent times. There are major challenges when it comes to handling such large amounts of data. These range from the practicalities of managing these large volumes of data, to understanding the meaning and practical implications of the data.

In this thesis, I present parallel methods to efficiently manage, process, analyse and visualize large sets of data from several life sciences fields at a rapid rate, while building and utilizing various machine learning techniques in a novel way. Most of the work is centred on applying the latest Big Data Analytics frameworks for creating efficient virtual screening strategies while working with large datasets. Virtual screening is a method in cheminformatics used for Drug discovery by searching large libraries of molecule structures. I also present a method for the analysis of large Electroencephalography data in real time. Electroencephalography is one of the main techniques used to measure the brain electrical activity.

First, I evaluate the suitability of Spark, a parallel framework for large datasets, for performing parallel ligand-based virtual screening. As a case study, I classify molecular library using prebuilt classification models to filter out the active molecules. I also demonstrate a strategy to create cloud-ready pipelines for structure-based virtual screening. The major advantages of this strategy are increased productivity and high throughput. In this work, I show that Spark can be applied to virtual screening, and that it is, in general, an appropriate solution for large-scale parallel pipelining. Moreover, I illustrate how Big Data analytics are valuable in working with life sciences datasets.

Secondly, I present a method to further reduce the overall time of the structured-based virtual screening strategy using machine learning and a conformal-prediction-based iterative modelling strategy. The idea is to only dock those molecules that have a better than average chance of being an inhibitor when searching for molecules that could potentially be used as drugs. Using machine learning models from this work, I built a web service to predict the target profile of multiple compounds against ready-made models for a list of targets where 3D structures are available. These target predictions can be used to understand off-target effects, for example in the early stages of drug discovery projects.

Thirdly, I present a method to detect seizures in long term Electroencephalography readings - this method works in real time taking the ongoing readings in as live data streams. The method involves tackling the challenges of real-time decision-making, storing large datasets in memory and updating the prediction model with newly produced data at a rapid rate. The resulting algorithm not only classifies seizures in real time, it also learns the threshold in real time. I also present a new feature "top-k amplitude measure" for classifying which parts of the data correspond to seizures. Furthermore, this feature helps to reduce the amount of data that needs to be processed in the subsequent steps.

Background: Identifying and assessing ligand-target binding is a core component in early drug discovery as one or more unwanted interactions may be associated with safety issues. Contributions: We present an open-source, extendable web service for predicting target profiles with confidence using machine learning for a panel of 7 targets, where models are trained on molecular docking scores from a large virtual library. The method uses conformal prediction to produce valid measures of prediction efficiency for a particular confidence level. The service also offers the possibility to dock chemical structures to the panel of targets with QuickVina on individual compound basis. Results: The docking procedure and resulting models were validated by docking well-known inhibitors for each of the 7 targets using QuickVina. The model predictions showed comparable performance to molecular docking scores against an external validation set. The implementation as publicly available microservices on Kubernetes ensures resilience, scalability, and extensibility.

Background: Docking and scoring large libraries of ligands against target proteins forms the basis of structure-based virtual screening. The problem is trivially parallelizable, and calculations are generally carried out on computer clusters or on large workstations in a brute force manner, by docking and scoring all available ligands. Contribution: In this study we propose a strategy that is based on iteratively docking a set of ligands to form a training set, training a ligand-based model on this set, and predicting the remainder of the ligands to exclude those predicted as 'low-scoring' ligands. Then, another set of ligands are docked, the model is retrained and the process is repeated until a certain model efficiency level is reached. Thereafter, the remaining ligands are docked or excluded based on this model. We use SVM and conformal prediction to deliver valid prediction intervals for ranking the predicted ligands, and Apache Spark to parallelize both the docking and the modeling. Results: We show on 4 different targets that conformal prediction based virtual screening (CPVS) is able to reduce the number of docked molecules by 62.61% while retaining an accuracy for the top 30 hits of 94% on average and a speedup of 3.7. The implementation is available as open source via GitHub (https://github.com/laeeq80/spark-cpvs) and can be run on high-performance computers as well as on cloud resources.

Background Patients with traumatic brain injury are a heterogeneous population, and the most severely injured individuals are often treated in an intensive care unit (ICU). The primary injury at impact, and the harmful secondary events that can occur during the first week of the ICU stay, will affect outcome in this vulnerable group of patients. We aimed to identify clinical variables that might distinguish disease trajectories among patients with traumatic brain injury admitted to the ICU. Methods We used data from the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) prospective observational cohort study. We included patients aged 18 years or older with traumatic brain injury who were admitted to the ICU at one of the 65 CENTER-TBI participating centres, which range from large academic hospitals to small rural hospitals. For every patient, we obtained pre-injury data and injury features, clinical characteristics on admission, demographics, physiological parameters, laboratory features, brain biomarkers (ubiquitin carboxy-terminal hydrolase L1 [UCH-L1], S100 calcium-binding protein B [S100B], tau, neurofilament light [NFL], glial fibrillary acidic protein [GFAP], and neuron-specific enolase [NSE]), and information about intracranial pressure lowering treatments during the first 7 days of ICU stay. To identify clinical variables that might distinguish disease trajectories, we applied a novel clustering method to these data, which was based on a mixture of probabilistic graph models with a Markov chain extension. The relation of clusters to the extended Glasgow Outcome Scale (GOS-E) was investigated. Findings Between Dec 19, 2014, and Dec 17, 2017, 4509 patients with traumatic brain injury were recruited into the CENTER-TBI core dataset, of whom 1728 were eligible for this analysis. Glucose variation (defined as the difference between daily maximum and minimum glucose concentrations) and brain biomarkers (S100B, NSE, NFL, tau, UCH-L1, and GFAP) were consistently found to be the main clinical descriptors of disease trajectories (ie, the leading variables contributing to the distinguishing clusters) in patients with traumatic brain injury in the ICU. The disease trajectory cluster to which a patient was assigned in a model was analysed as a predictor together with variables from the IMPACT model, and prediction of both mortality and unfavourable outcome (dichotomised GOS-E <= 4) was improved. Interpretation First-day ICU admission data are not the only clinical descriptors of disease trajectories in patients with traumatic brain injury. By analysing temporal variables in our study, variation of glucose was identified as the most important clinical descriptor that might distinguish disease trajectories in the ICU, which should direct further research. Biomarkers of brain injury (S100B, NSE, NFL, tau, UCH-L1, and GFAP) were also top clinical descriptors over time, suggesting they might be important in future clinical practice.

One of the main hurdles of a broad distribution of PGAS approaches is the prevalence of MPI, which as a de-facto standard appears in the code basis of many applications. To take advantage of the PGAS APIs like GASPI without a major change in the code basis, interoperability between MPI and PGAS approaches needs to be ensured. In this article, we address this challenge by providing our study and preliminary performance results regarding interoperating GASPI and MPI on the performance crucial parts of the Ludwig and iPIC3D applications. In addition, we draw a strategy for better coupling of both APIs. 

The preservation and presentation of cultural heritage (CH) encompasses many domains and disciplines and ranges from tangible CH, traditionally taking the form of museum exhibits and historical sites that are open to the public to intangible CH, focussing on human and societal aspects of CH, as opposed to physical artefacts. The use of computer graphics (CG) and related techniques such as interactive virtual environments since the 1990s has had a profound impact on the presentation of and public engagement with CH, allowing virtual reconstruction of archaeological/historical sites as well as the virtual (re-)construction of culturally and historically relevant artefacts. These are frequently implemented using bespoke or proprietary systems, often explicitly created with a CH application in mind, which may require specialist expertise or significant investment. There exist, however, alternative approaches that can simplify and improve the uptake of CG for CH. In this chapter we discuss how off-the-shelf CG systems such as developer and artists’ tools for the entertainment industries, which are comparatively inexpensive, usually provide open developer licenses, and sometimes are even available free of charge, or affordable consumer-level hardware, can be used for the preservation and presentation of tangible and intangible CH, the application of which we illustrate with a set of case studies. 

The Discrete Fourier Transform (DFT) and its improved formulations, the Fast Fourier Transforms (FFTs), are vital for scientists and engineers in a range of domains from signal processing to the solution of partial differential equations.  A growing trend in Scientific Computing is heterogeneous computing, where accelerators are used instead or together with CPUs. This has led to problems for developers in unifying portability, performance, and productivity. 

This thesis first motivates this work by showing the importance of having efficient DFT calculations, describes the DFT algorithm and a formulation based on matrix-factorizations which has been developed to formulate FFT algorithms and express their parallelism to exploit modern computer architectures, such as accelerators.

The first paper is a motivating study of the breakdown of the performance and scalability of the high-performance Molecular Dynamics code GROMACS where DFT calculations are a main performance bottleneck. In particular, the long-range interactions are solved with the Particle-Mesh Ewald algorithm which uses a three-dimensional Fast Fourier Transform. 

The two following papers present two approaches to leverage factorization with the help of two different frameworks using Intermediate Representation and compiler technology, for the development of fast and portable code. The second paper presents a front-end and a pipeline for code generation in a domain-specific language based on Multi-Level Intermediate Representation (MLIR) for developing Fast Fourier Transform libraries. The last paper investigates and optimizes an implementation of an important kernel within the matrix-factorization framework: the batched DFT. It is implemented with data-centric programming and a data-centric intermediate representation called Stateful Dataflow multi-graphs (SDFG). The paper evaluates strategies for complex-valued data layout for performance and portability and we show that there is a trade-off between portability and maintainability in using the native complex data type and that an SDFG-level abstraction could be beneficial for developing higher-level applications.

With the rise of new high-performance computing (HPC) accelerators, such as Nvidia and AMD GPUs, the demand for efficient code targeting diverse hardware accelerators poses a critical challenge for HPC application developers. This hardware diversity in the HPC systems necessitates the development of new code tailored to specific architectures, which, in turn, hampers the sustainability of large scientific application development. In this work, we rely on DaCe [1, 2], a data-centric parallel programming framework, to automate the generation of high-performance kernels. DaCe can generate automatic code for multicore processors and various accelerators, alleviating the programmer burden of rewriting code for a new architecture. Our work demonstrates the automatic code generation capabilities of DaCe, applied to a critical high-performance computational kernel for Computational Fluid Dynamics code. Specifically, we focus on the Fortran-based solver, Neko [4] which is based on the Spectral Element Method. This method relies on small-sized matrix multiplications akin to BLAS dgemm operations. We describe the formulation of this computational kernel through DaCe's Stateful Dataflow Multigraph (SDFG) representation. We discuss how this representation facilitates high-performance code generation and detail the workflow for integration of DaCe's automatically generated code into the Neko solver. Initial work on Nvidia GH200. By showcasing the potential of automatic code generation, we highlight the feasibility of supporting the long-term sustainability of large-scale scientific applications by using portable solutions for critical computational kernels of large-scale codes.

This paper presents the design and development of an Anderson Accelerated Preconditioned Modified Hermitian and Skew-Hermitian Splitting (AA-PMHSS) method for solving complex-symmetric linear systems with application to electromagnetics problems, such as wave scattering and eddy currents. While it has been shown that the Anderson acceleration of real linear systems is essentially equivalent to GMRES, we show here that the formulation using Anderson acceleration leads to a more performant method. We show relatively good robustness compared to existing preconditioned GMRES methods and significantly better performance due to the faster evaluation of the preconditioner. In particular, AA-PMHSS can be applied to solve problems and equations arising from complex-valued systems, such as time-harmonic eddy current simulations discretized with the Finite Element Method. We also evaluate three test systems present in previous literature. We show that the method is competitive with two types of preconditioned GMRES, which share the significant advantage of having a convergence rate that is independent of the discretization size.

With the emergence of new computer architectures, portability and performance-portability become significant concerns for developing HPC applications. This work reports our experience and lessons learned using DaCe to create and optimize batched Discrete Fourier Transform (DFT) calculations on different single node computer systems. The batched DFT calculation is an essential component in FFT algorithms and is widely used in computer science, numerical analysis, and signal processing. We implement the batched DFT with three complex-value array data layouts and compare them with the native complex type implementation. We use DaCe, which relies on Stateful DataFlow multiGraphs (SDFG) as an intermediate representation (IR) which can be optimized through transforms and then generates code for different architectures. We present several performance results showcasing the potential of DaCe for expressing HPC applications on different computer systems.

GROMACS is one of the most widely used HPC software packages using the Molecular Dynamics (MD) simulation technique. In this work, we quantify GROMACS parallel performance using different configurations, HPC systems, and FFT libraries (FFTW, Intel MKL FFT, and FFT PACK). We break down the cost of each GROMACS computational phase and identify non-scalable stages, such as MPI communication during the 3D FFT computation when using a large number of processes. We show that the Particle-Mesh Ewald phase and the 3D FFT calculation significantly impact the GROMACS performance. Finally, we discuss performance opportunities with a particular interest in developing GROMACS for the FFT calculations.

Computational models of ion channels represent the building blocks of conductance-based, biologically inspired models of neurons and neural networks. Ion channels are still widely modelled by means of the formalism developed by the seminal work of Hodgkin and Huxley (HH), although the electrophysiological features of the channels are currently known to be better fitted by means of kinetic Markov-type models. The present study is aimed at showing why simplified Markov-type kinetic models are more suitable for ion channels modelling as compared to HH ones, and how a manual optimization process can be rationally carried out for both. Previously published experimental data of an illustrative ion channel (Na(V)1.5) are exploited to develop a step by step optimization of the two models in close comparison. A conflicting practical limitation is recognized for the HH model, which only supplies one parameter to model two distinct electrophysiological behaviours. In addition, a step by step procedure is provided to correctly optimize the kinetic Markov-type model. Simplified Markov-type kinetic models are currently the best option to closely approximate the known complexity of the macroscopic currents of ion channels. Their optimization can be achieved through a rationally guided procedure, and allows to obtain models with a computational burden that is comparable with HH models one.

The cameras in modern gaze-tracking systems suffer from fundamental bandwidth and power limitations, constraining data acquisition speed to 300 Hz realistically. This obstructs the use of mobile eye trackers to perform, e.g., low latency predictive rendering, or to study quick and subtle eye motions like microsaccades using head-mounted devices in the wild. Here, we propose a hybrid frame-event-based near-eye gaze tracking system offering update rates beyond 10,000 Hz with an accuracy that matches that of high-end desktop-mounted commercial trackers when evaluated in the same conditions. Our system, previewed in Figure 1, builds on emerging event cameras that simultaneously acquire regularly sampled frames and adaptively sampled events. We develop an online 2D pupil fitting method that updates a parametric model every one or few events. Moreover, we propose a polynomial regressor for estimating the point of gaze from the parametric pupil model in real time. Using the first event-based gaze dataset, we demonstrate that our system achieves accuracies of 0.45 degrees -1.75 degrees for fields of view from 45 degrees to 98 degrees. With this technology, we hope to enable a new generation of ultra-low-latency gaze-contingent rendering and display techniques for virtual and augmented reality.

Traditional HPC scientific workloads are tightly coupled, while emerging scientific workflows exhibit even more complex patterns, consisting of multiple characteristically different stages that may be IO-intensive, compute-intensive, or memory-intensive. New high-performance computer systems are evolving to adapt to these new requirements and are motivated by the need for performance and efficiency in resource usage. On the other hand, cloud workloads are loosely coupled, and their systems have matured technologies under different constraints from HPC.

In this thesis, the use of cloud technologies designed for loosely coupled dynamic and elastic workloads is explored, repurposed, and examined in the landscape of HPC in three major parts. The first part deals with the deployment of HPC workloads in cloud-native environments through the use of containers and analyses the feasibility and trade-offs of elastic scaling. The second part relates to the use of workflow management systems in HPC workflows; in particular, a molecular docking workflow executed through Airflow is discussed. Finally, object storage systems, a cost-effective and scalable solution widely used in the cloud, and their usage in HPC applications through MPI I/O are discussed in the third part of this thesis. 

The new emerging scientific workloads to be executed in the upcoming exascale supercomputers face major challenges in terms of storage, given their extreme volume of data. In particular, intelligent data placement, instrumentation, and workflow handling are central to application performance. The IO-SEA project developed multiple solutions to aid the scientific community in adressing these challenges: a Workflow Manager, a hierarchical storage management system, and a semantic API for storage. All of these major products incorporate additional minor products that support their mission. In this paper, we discuss both the roles of all these products and how they can assist the scientific community in achieving exascale performance.

Recently, federated HPC and cloud resources are becoming increasingly strategic for providing diversified and geographically available computing resources. However, accessing data stores across HPC and cloud storage systems is challenging. Many cloud providers use object storage systems to support their clients in storing and retrieving data over the internet. One popular method is REST APIs atop the HTTP protocol, with Amazon's S3 APIs being supported by most vendors. In contrast, HPC systems are contained within their networks and tend to use parallel file systems with POSIX-like interfaces. This work addresses the challenge of diverse data stores on HPC and cloud systems by providing native object storage support through the unified MPI I/O interface in HPC applications. In particular, we provide a prototype library called LibCOS that transparently enables MPI applications running on HPC systems to access object storage on remote cloud systems. We evaluated LibCOS on a Ceph object storage system and a traditional HPC system. In addition, we conducted performance characterization of core S3 operations that enable individual and collective MPI I/O. Our evaluation in HACC, IOR, and BigSort shows that enabling diverse data stores on HPC and Cloud storage is feasible and can be transparently achieved through the widely adopted MPI I/O. Also, we show that a native object storage system like Ceph could improve the scalability of I/O operations in parallel applications.

Complex workflows play a critical role in accelerating scientific discovery. In many scientific domains, efficient workflow management can lead to faster scientific output and broader user groups. Workflows that can leverage resources across the boundary between cloud and HPC are a strong driver for the convergence of HPC and cloud. This study investigates the transition and deployment of a GPU-accelerated molecular docking workflow that was designed for HPC systems onto a cloud-native environment with Kubernetes and Apache Airflow. The case study focuses on state-of-of-the-art molecular docking software for drug discovery. We provide a DAG-based implementation in Apache Airflow and technical details for GPU-accelerated deployment. We evaluated the workflow using the SWEETLEAD bioinformatics dataset and executed it in a Cloud environment with heterogeneous computing resources. Our workflow can effectively overlap different stages when mapped onto different computing resources.

The conventional model of resource allocation in HPC systems is static. Thus, a job cannot leverage newly available resources in the system or release underutilized resources during the execution. In this paper, we present Kub, a methodology that enables elastic execution of HPC workloads on Kubernetes so that the resources allocated to a job can be dynamically scaled during the execution. One main optimization of our method is to maximize the reuse of the originally allocated resources so that the disruption to the running job can be minimized. The scaling procedure is coordinated among nodes through remote procedure calls on Kubernetes for deploying workloads in the cloud. We evaluate our approach using one synthetic benchmark and two production-level MPI-based HPC applications - GRO-MACS and CM1. Our results demonstrate that the benefits of adapting the allocated resources depend on the workload characteristics. In the tested cases, a properly chosen scaling point for increasing resources during execution achieved up to 2x speedup. Also, the overhead of checkpointing and data reshuffling significantly influences the selection of optimal scaling points and requires application-specific knowledge.

This article presents the design and evaluation of visualization concepts supporting After Action Review (AAR) in simulator mission training of fast-jet fighter pilots. The visualization concepts were designed based on three key characteristics of representations: re-representation, graphical constraining, and computational offloading. The visualization concepts represent combined parameters of missile launch and threat range, the former meant to elicit discussions about the prerequisites for launching missiles, and the latter to present details of what threats a certain aircraft is facing at a specific moment. The visualization concepts were designed to: 1) perceptually and cognitively offload mental workload from participants in support of determining relevant situations to discuss; 2) re-represent parameters in a format that facilitates reading-off of crucial information; and 3) graphically constrain plausible interpretations. Through a series of workshop iterations, two visualization concepts were developed and evaluated with 11 pilots and instructors. All pilots were unanimous in their opinion that the visualization concepts should be implemented as part of the AAR. Offloading, in terms of finding interesting events in the dynamic and unique training sessions, was the most important guiding concept, while re-representation and graphical constraining enabled a more structured and grounded collaboration during the AAR.

Magnetic particle imaging (MPI) is a rapidly developing medical imaging modality that exploits the non- linear response of magnetic nanoparticles (MNPs). Color MPI widens the functionality of MPI, empowering it with the capability to distinguish differentMNPs and/orMNP environments. The system function approach for color MPI relies on extensive calibrations that capture the differences in the harmonic responses of the MNPs. An alternative calibration-free x-space-basedmethod called TAURUS estimates amap of the relaxation time constant, tau, by recovering the underlyingmirror symmetry in the MPI signal. However, TAURUS requires a back and forth scanning of a given region, restricting its usage to slow trajectories with constant or piecewise constant focus fields (FFs). In this work, we propose a novel technique to increase the performance of TAURUS and enable tau map estimation for rapid andmultidimensional trajectories. The proposed technique is based on correcting the distortions on mirror symmetry induced by time-varying FFs. We demonstrate via simulations and experiments in our in-house MPI scanner that the proposed method successfully estimates high-fidelity tau maps for rapid trajectories that provide orders of magnitude reduction in scanning time (over 300 fold for simulations and over 8 fold for experiments) while preserving the calibration-free property of TAURUS.

In-situ visualization on HPC systems allows us to analyze simulation results that would otherwise be impossible, given the size of the simulation data sets and offline post-processing execution time. We design and develop in-situ visualization with Paraview Catalyst in Nek5000, a massively parallel Fortran and C code for computational fluid dynamics applications. We perform strong scalability tests up to 2,048 cores on KTH's Beskow Cray XC40 supercomputer and assess in-situ visualization's impact on the Nek5000 performance. In our study case, a high-fidelity simulation of turbulent flow, we observe that in-situ operations significantly limit the strong scalability of the code, reducing the relative parallel efficiency to only ~21\% on 2,048 cores (the relative efficiency of Nek5000 without in-situ operations is ~99\%). Through profiling with Arm MAP, we identified a bottleneck in the image composition step (that uses Radix-kr algorithm) where a majority of the time is spent on MPI communication. We also identified an imbalance of in-situ processing time between rank 0 and all other ranks. Better scaling and load-balancing in the parallel image composition would considerably improve the performance and scalability of Nek5000 with in-situ capabilities in large-scale simulation.

In situ visualization on high-performance computing systems allows us to analyze simulation results that would otherwise be impossible, given the size of the simulation data sets and offline post-processing execution time. We develop an in situ adaptor for Paraview Catalyst and Nek5000, a massively parallel Fortran and C code for computational fluid dynamics. We perform a strong scalability test up to 2048 cores on KTH’s Beskow Cray XC40 supercomputer and assess in situ visualization’s impact on the Nek5000 performance. In our study case, a high-fidelity simulation of turbulent flow, we observe that in situ operations significantly limit the strong scalability of the code, reducing the relative parallel efficiency to only ≈ 21 % on 2048 cores (the relative efficiency of Nek5000 without in situ operations is ≈ 99 %). Through profiling with Arm MAP, we identified a bottleneck in the image composition step (that uses the Radix-kr algorithm) where a majority of the time is spent on MPI communication. We also identified an imbalance of in situ processing time between rank 0 and all other ranks. In our case, better scaling and load-balancing in the parallel image composition would considerably improve the performance of Nek5000 with in situ capabilities. In general, the result of this study highlights the technical challenges posed by the integration of high-performance simulation codes and data-analysis libraries and their practical use in complex cases, even when efficient algorithms already exist for a certain application scenario.

This paper is about quantum heat defined as the change in energy of a bath during a process. The presentation takes into account recent developments in classical strong-coupling thermodynamics and addresses a version of quantum heat that satisfies quantum-classical correspondence. The characteristic function and the full counting statistics of quantum heat are shown to be formally similar. The paper further shows that the method can be extended to more than one bath, e.g., two baths at different temperatures, which opens up the prospect of studying correlations and heat flow. The paper extends earlier results on the expected quantum heat in the setting of one bath [E. Aurell and R. Eichhorn, New .J Phys. 17, 065007 (2015); E. Aurell, Entropy 19, 595 (2017)].

This paper revisits the classical problem of representing a thermal bath interacting with a system as a large collection of harmonic oscillators initially in thermal equilibrium. As is well known, the system then obeys an equation, which in the bulk and in the suitable limit tends to the Kramers-Langevin equation of physical kinetics. I consider time-dependent system-bath coupling and show that this leads to an additional harmonic force acting on the system. When the coupling is switched on and switched off rapidly, the force has delta-function support at the initial and final time. I further show that the work and heat functionals as recently defined in stochastic thermodynamics at strong coupling contain additional terms depending on the time derivative of the system-bath coupling. I discuss these terms and show that while they can be very large if the system-bath coupling changes quickly, they only give a finite contribution to the work that enters in Jarzynski's equality. I also discuss that these corrections to standard work and heat functionals provide an explanation for non-standard terms in the change of the von Neumann entropy of a quantum bath interacting with a quantum system found in an earlier contribution (Aurell and Eichhorn, 2015).

We introduce an alternative solution to Glauber multispin dynamics on random graphs. The solution is based on the recently introduced cavity master equation (CME), a time-closure turning the, in principle, exact dynamic cavity method into a practical method of analysis and of fast simulation. Running CME once is of comparable computational complexity as one Monte Carlo run on the same problem. We show that CME correctly models the ferromagnetic p-spin Glauber dynamics from high temperatures down to and below the spinoidal transition. We also show that CME allows an alternative exploration of the low-temperature spin-glass phase of the model.

We study the heat current flowing between two baths consisting of harmonic oscillators interacting with a qubit through a spin-boson coupling. An explicit expression for the generating function of the total heat flowing between the right and left baths is derived by evaluating the corresponding Feynman-Vernon path integral by performing the noninteracting blip approximation (NIBA). We recover the known expression, obtained by using the polaron transform. This generating function satisfies the Gallavotti-Cohen fluctuation theorem, both before and after performing the NIBA. We also verify that the heat conductance is proportional to the variance of the heat current, retrieving the well-known fluctuation dissipation relation. Finally, we present numerical results for the heat current.

In 1974 Steven Hawking showed that black holes emit thermal radiation, which eventually causes them to evaporate. The problem of the fate of information in this process is known as the "black hole information paradox". Two main types of resolution postulate either a fundamental loss of information in Nature - hence the breakdown of quantum mechanics - or some sort of new physics, e.g. quantum gravity, which guarantee the global preservation of unitarity. Here we explore the second possibility with the help of recent developments in continuous-variable quantum information. Concretely, we employ the solution to the Gaussian quantum marginal problem to show that the thermality of all individual Hawking modes is consistent with a global pure state of the radiation. Surprisingly, we find out that the mods of radiation of an astrophysical black hole are thermal until the very last burst. In contrast, the single-mode thermality of Hawking radiation originating from microscopic black holes, expected to evaporate through several quanta, is not excluded, though there are constraints on modes' frequencies. Our result paves the way towards a systematic study of multi-mode correlations in Hawking radiation.

Almost all of the entropy in the universe is in the form of Bekenstein-Hawking (BH) entropy of super-massive black holes. This entropy, if it satisfies Boltzmann's equation S=log N, hence represents almost all the accessible phase space of the Universe, somehow associated to objects which themselves fill out a very small fraction of ordinary three-dimensional space. Although time scales are very long, it is believed that black holes will eventually evaporate by emitting Hawking radiation, which is thermal when counted mode by mode. A pure quantum state collapsing to a black hole will hence eventually re-emerge as a state with strictly positive entropy, which constitutes the famous black hole information paradox. Expanding on a remark by Hawking we posit that BH entropy is a thermodynamic entropy, which must be distinguished from information-theoretic entropy. The paradox can then be explained by information return in Hawking radiation. The novel perspective advanced here is that if BH entropy counts the number of accessible physical states in a quantum black hole, then the paradox can be seen as an instance of the fundamental problem of statistical mechanics. We suggest a specific analogy to the increase of the entropy in a solvation process. We further show that the huge phase volume (N), which must be made available to the universe in a gravitational collapse, cannot originate from the entanglement between ordinary matter and/or radiation inside and outside the black hole. We argue that, instead, the quantum degrees of freedom of the gravitational field must get activated near the singularity, resulting in a final state of the 'entangled entanglement' form involving both matter and gravity.

A black hole evaporates by Hawking radiation. Each mode of that radiation is thermal. If the total state is nevertheless to be pure, modes must be entangled. Estimating the minimum size of this entanglement has been an important outstanding issue. We develop a new theory of constrained random symplectic transformations, based on the assumptions that the total state is pure and Gaussian with given marginals. In the random constrained symplectic model we then compute the distribution of mode-mode correlations, from which we bound mode-mode entanglement. Modes of frequency much larger than [k(B)T(H)(t)/h] are not populated at time t and drop out of the analysis. Among other relatively thinly populated modes (earlytime high-frequency modes and/or late modes of any frequency), we find correlations and hence entanglement to be strongly suppressed. Relatively highly populated modes (early-time low-frequency modes) can, on the other hand, be strongly correlated, but a detailed analysis reveals that they are nevertheless very unlikely to be entangled. Our analysis hence establishes that restoring unitarity after a complete evaporation of a black hole does not require any significant quantum entanglement between any pair of Hawking modes. Our analysis further gives exact general expressions for the distribution of modemode correlations in random, pure, Gaussian states with given marginals, which may have applications beyond black hole physics.

We present a derivation of the Feynman-Vernon approach to open quantum systems in the language of super-operators. We show that this gives a new and more direct derivation of the generating function of energy changes in a bath, or baths. As found previously, this generating function is given by a Feynman-Vernon-like influence functional, with only time shifts in the kernels coupling the forward and backward paths. We further show that the new approach extends to an-harmonic and possible non-equilibrium baths, provided that the interactions are bi-linear, and that the baths do not interact between themselves. Such baths are characterized by non-trivial cumulants. Every non-zero cumulant of certain environment correlation functions is thus a kernel in a higher-order term in the Feynman-Vernon action.

We consider classical spin systems evolving in continuous time with interactions given by a locally tree-like graph. Several approximate analysis methods have earlier been reported based on the idea of Belief Propagation / cavity method. We introduce a new such method which can be derived in a more systematic manner using the theory of Random Point Processes. Within this approach, the master equation governing the system's dynamics is closed via a set of differential equations for the auxiliary cavity probabilities. The numerical results improve on the earlier versions of the closure on several important classes of problems. We re-visit here the cases of the Ising ferromagnet and the Viana-Bray spin-glass model.

In this paper we consider the thermal power of a heat flow through a qubit between two baths. The baths are modeled as a set of harmonic oscillators initially at equilibrium, at two temperatures. Heat is defined as the change of energy of the cold bath, and thermal power is defined as expected heat per unit time, in the long-time limit. The qubit and the baths interact as in the spin-boson model, i.e., through qubit operator sigma(z). We compute thermal power in an approximation analogous to a "noninteracting blip" (NIBA) and express it in the polaron picture as products of correlation functions of the two baths, and a time derivative of a correlation function of the cold bath. In the limit of weak interaction we recover known results in terms of a sum of correlation functions of the two baths, a correlation functions of the cold bath only, and the energy split.