Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “execution time”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 127 records · Page 7

Acceleration of Graph Neural Network-Based Prediction Models in Chemistry via Co-Design Optimization on Intelligence Processing Units

Atomic structure prediction and associated property calculations are the bedrock of chemical physics. Since high-fidelity ab initio modeling techniques for computing the structure and properties can be prohibitively expensive, this motivates the development of machine-learning (ML) models that make these predictions more efficiently. Training graph neural networks over large atomistic databases introduces unique computational challenges such as the need to process millions of small graphs with variable size and support communication patterns that are distinct from learning over large graphs such as social networks. We demonstrate a novel hardware-software co-design approach to scale up the training of atomistic graph neural networks (GNN) for structure and property prediction. First, to eliminate redundant computation and memory associated with alternative padding techniques and to improve throughput via minimizing communication, we formulate the effective coalescing of the batches of variable-size atomistic graphs as the bin packing problem and introduce a hardware-agnostic algorithm to pack these batches. In addition, we propose hardware-specific optimizations including a planner and vectorization for the gather-scatter operations targeted for Graphcore’s Intelligence Processing Unit (IPU), as well as model-specific optimizations such as merged communication collectives and optimized softplus. Putting these all together, we demonstrate the effectiveness of the proposed co-design approach by providing an implementation of a well-established atomistic GNN on the Graphcore IPUs. We evaluate the training performance on multiple atomistic graph databases with varying degrees of graph counts, sizes and sparsity. Here, we demonstrate that such a co-design approach can reduce the training time of atomistic GNNs and can improve the performance by up to 1.5× compared to the baseline implementation of the model on the IPUs. Additionally, we compare our IPU implementation with a Nvidia GPU-based implementation and show that our atomistic GNN implementation on the IPUs can run 1.8× faster on average compared to the execution time on the GPUs.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

GPU-accelerated denoising of 3D magnetic resonance images

The raw computational power of GPU accelerators enables fast denoising of 3D MR images using bilateral filtering, anisotropic diffusion, and non-local means. In practice, applying these filtering operations requires setting multiple parameters. This study was designed to provide better guidance to practitioners for choosing the most appropriate parameters by answering two questions: what parameters yield the best denoising results in practice? And what tuning is necessary to achieve optimal performance on a modern GPU? To answer the first question, we use two different metrics, mean squared error (MSE) and mean structural similarity (MSSIM), to compare denoising quality against a reference image. Surprisingly, the best improvement in structural similarity with the bilateral filter is achieved with a small stencil size that lies within the range of real-time execution on an NVIDIA Tesla M2050 GPU. Moreover, inappropriate choices for parameters, especially scaling parameters, can yield very poor denoising performance. To answer the second question, we perform an autotuning study to empirically determine optimal memory tiling on the GPU. The variation in these results suggests that such tuning is an essential step in achieving real-time performance. These results have important implications for the real-time application of denoising to MR images in clinical settings that require fast turn-around times.

97 MATHEMATICS AND COMPUTING↗

Field programmable gate array-assigned complex-valued computation and its limits

We discuss how leveraging Field Programmable Gate Array (FPGA) technology as part of a high performance computing platform reduces latency to meet the demanding real time constraints of a quantum optics simulation. Implementations of complex-valued operations using fixed point numeric on a Virtex-5 FPGA compare favorably to more conventional solutions on a central processing unit. Our investigation explores the performance of multiple fixed point options along with a traditional 64 bits floating point version. With this information, the lowest execution times can be estimated. Relative error is examined to ensure simulation accuracy is maintained.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Level-set segmentation of pulmonary nodules in megavolt electronic portal images using a CT prior

Purpose: Pulmonary nodules present unique problems during radiation treatment due to nodule position uncertainty that is caused by respiration. The radiation field has to be enlarged to account for nodule motion during treatment. The purpose of this work is to provide a method of locating a pulmonary nodule in a megavolt portal image that can be used to reduce the internal target volume (ITV) during radiation therapy. A reduction in the ITV would result in a decrease in radiation toxicity to healthy tissue. Methods: Eight patients with nonsmall cell lung cancer were used in this study. CT scans that include the pulmonary nodule were captured with a GE Healthcare LightSpeed RT 16 scanner. Megavolt portal images were acquired with a Varian Trilogy unit equipped with an AS1000 electronic portal imaging device. The nodule localization method uses grayscale morphological filtering and level-set segmentation with a prior. The treatment-time portion of the algorithm is implemented on a graphical processing unit. Results: The method was retrospectively tested on eight cases that include a total of 151 megavolt portal image frames. The method reduced the nodule position uncertainty by an average of 40% for seven out of the eight cases. The treatment phase portion of the method has a subsecond execution time that makes it suitable for near-real-time nodule localization. Conclusions: A method was developed to localize a pulmonary nodule in a megavolt portal image. The method uses the characteristics of the nodule in a prior CT scan to enhance the nodule in the portal image and to identify the nodule region by level-set segmentation. In a retrospective study, the method reduced the nodule position uncertainty by an average of 40% for seven out of the eight cases studied.

60 APPLIED LIFE SCIENCES↗

Just-in-Time Compilation and Link-Time Optimization for OpenMP Target Offloading

Following the mass adoption of external accelerators for high performance computing, the overall performance of many applications has become increasingly dependent on relatively small accelerated kernels. As static analysis is fundamentally limited by dynamic values and external definitions, standard ahead-of-time compilation is not always sufficient to achieve the best performance. Furthermore, many users looking to port an existing application to run on an external accelerator will not want to fundamentally restructure their programs. These and other problems can be addressed through both link-time optimization (LTO) and just-in-time (JIT) compilation, but until now had sparse and inconsistent support from the compiler. In this work, we present a new compilation method that enables device-side LTO as well as a transparent JIT compilation tool-chain for OpenMP target offloading. Our contributions include an entirely new device linking and embedding scheme to enable LTO as well as a novel JIT engine to efficiently optimize OpenMP offloading regions at run-time. We also introduce a persistent caching system to improve end-to-end runtime using the JIT engine and minimize kernel launching overheads. We measure the performance of our LTO and JIT implementation via several real-world scientific applications. With our optimizations we observe significant improvements through LTO on large applications as well as significant end-to-end execution time improvement using JIT.

Tian, Shilei↗

Advanced Shuttle Strategies for Parallel QCCD Architectures

Trapped ions (TIs) are at the forefront of quantum computing implementation, offering unparalleled coherence, fidelity, and connectivity. However, the scalability of TI systems is hampered by the limited capacity of individual ion traps, necessitating intricate ion shuttling for advanced computational tasks. The quantum charge-coupled device (QCCD) framework has emerged as a promising solution, facilitating ion mobility for universal quantum computation. Current QCCD architectures predominantly feature a linear topology, which is increasingly recognized as inefficient for complex quantum operations. Anticipating the shift toward more efficacious designs, this article introduces an innovative quantum scheduling strategy optimized for parallel QCCD topologies. Our strategy proposes a probabilistic formula for ion movement, alongside ingenious methods for local layer generation and layer compression, yielding a significant reduction in ion shuttle times. Through simulations, we demonstrate that our strategy not only substantially outstrips the linear model but also exhibits better performance over other parallel strategies that employ greedy algorithms. This is achieved through our nuanced resolution of complexities, such as traffic blocks and trap capacity limitations. The consequent reduction in shuttle operations leads to lower energy consumption and an enhancement in the quantum computer's fidelity, ultimately accelerating program execution times.

43 PARTICLE ACCELERATORS↗

MCNP and GADRAS Comparisons

To facilitate the timely execution of System Threat Reviews (STRs) for DNDO, and also to develop a methodology for performing STRs, LANL performed comparisons of several radiation transport codes (MCNP, GADRAS, and Gamma-Designer) that have been previously utilized to compute radiation signatures. While each of these codes has strengths, it is of paramount interest to determine the limitations of each of the respective codes and also to identify the most time efficient means by which to produce computational results, given the large number of parametric cases that are anticipated in performing STR's. These comparisons serve to identify regions of applicability for each code and provide estimates of uncertainty that may be anticipated. Furthermore, while performing these comparisons, examination of the sensitivity of the results to modeling assumptions was also examined. These investigations serve to enable the creation of the LANL methodology for performing STRs. Given the wide variety of radiation test sources, scenarios, and detectors, LANL calculated comparisons of the following parameters: decay data, multiplicity, device (n,γ) leakages, and radiation transport through representative scenes and shielding. This investigation was performed to understand potential limitations utilizing specific codes for different aspects of the STR challenges.

07 ISOTOPE AND RADIATION SOURCES↗

The HARNESS Workbench: Unified and Adaptive Access to Diverse HPC Platforms

The primary goal of the Harness WorkBench (HWB) project is to investigate innovative software environments that will help enhance the overall productivity of applications science on diverse HPC platforms. Two complementary frameworks were designed: one, a virtualized command toolkit for application building, deployment, and execution, that provides a common view across diverse HPC systems, in particular the DOE leadership computing platforms (Cray, IBM, SGI, and clusters); and two, a unified runtime environment that consolidates access to runtime services via an adaptive framework for execution-time and post processing activities. A prototype of the first was developed based on the concept of a 'system-call virtual machine' (SCVM), to enhance portability of the HPC application deployment process across heterogeneous high-end machines. The SCVM approach to portable builds is based on the insertion of toolkit-interpretable directives into original application build scripts. Modifications resulting from these directives preserve the semantics of the original build instruction flow. The execution of the build script is controlled by our toolkit that intercepts build script commands in a manner transparent to the end-user. We have applied this approach to a scientific production code (Gamess-US) on the Cray-XT5 machine. The second facet, termed Unibus, aims to facilitate provisioning and aggregation of multifaceted resources from resource providers and end-users perspectives. To achieve that, Unibus proposes a Capability Model and mediators (resource drivers) to virtualize access to diverse resources, and soft and successive conditioning to enable automatic and user-transparent resource provisioning. A proof of concept implementation has demonstrated the viability of this approach on high end machines, grid systems and computing clouds.

97 MATHEMATICS AND COMPUTING↗

A beamforming algorithm for bistatic SAR image formation.

Beamforming is a methodology for collection-mode-independent SAR image formation. It is essentially equivalent to backprojection. The authors have in previous papers developed this idea and discussed the advantages and disadvantages of the approach to monostatic SAR image formation vis--vis the more standard and time-tested polar formatting algorithm (PFA). In this paper we show that beamforming for bistatic SAR imaging leads again to a very simple image formation algorithm that requires a minimal number of lines of code and that allows the image to be directly formed onto a three-dimensional surface model, thus automatically creating an orthorectified image. The same disadvantage of beamforming applied to monostatic SAR imaging applies to the bistatic case, however, in that the execution time for the beamforming algorithm is quite long compared to that of PFA. Fast versions of beamforming do exist to help alleviate this issue. Results of image reconstructions from phase history data are presented.

97 MATHEMATICS AND COMPUTING↗

On The Theory of Speculative Checkpointing: Time and Energy Considerations

Collective checkpoint-rollback is the most popular approach to dealing with fail-stop errors on high-performance computing (HPC) platforms. Prior work has focused on choosing checkpoint intervals that minimize the total cost of checkpoint-rollback. In this paper, we introduce the notion of speculative checkpointing, where we probabilistically skip some checkpoints. Careful choice of check- points to be taken or skipped has the potential to further reduce the total checkpoint-rollback overhead. We mathematically formulate the overall checkpoint-rollback cost in the presence of speculation. We consider the choice of speculation being a fixed probability or a probability distribution. We formulate two criteria to be minimized: total execution time and approximate total energy. We derive the criteria for beneficial speculative checkpointing for exponential and arbitrary failure distributions. Further, we analyze the joint optimization of energy and time to mathematically express the trade-offs.

Subasi, Omer↗

Optimization and Parallelization of the Thermal-Hydraulic Sub-channel Code CTF for High-Fidelity Multi-physics Applications

This paper describes major improvements to the computational infrastructure of the CTF sub-channel code so that full-core sub-channel-resolved simulations can now be performed in much shorter run-times, either in stand-alone mode or as part of coupled-code multi-physics calculations. These improvements support the goals of the Department Of Energy (DOE) Consortium for Advanced Simulations of Light Water (CASL) Energy Innovation Hub to develop high fidelity multi-physics simulation tools for nuclear energy design and analysis. A set of serial code optimizations--including fixing computational inefficiencies, optimizing the numerical approach, and making smarter data storage choices--are first described and shown to reduce both execution time and memory usage by about a factor of ten. Next, a Single Program Multiple Data (SPMD) parallelization strategy targeting distributed memory Multiple Instruction Multiple Data (MIMD) platforms and utilizing domain-decomposition is presented. In this approach, data communication between processors is accomplished by inserting standard MPI calls at strategic points in the code. The domain decomposition approach implemented assigns one MPI process to each fuel assembly, with each domain being represented by its own CTF input file. The creation of CTF input files, both for serial and parallel runs, is also fully automated through use of a pre-processor utility that takes a greatly reduced set of user input over the traditional CTF input file. To run CTF in parallel, two additional libraries are currently needed; MPI, for inter-processor message passing, and the Parallel Extensible Toolkit for Scientific Computation (PETSc), which is leveraged to solve the global pressure matrix in parallel. Results presented include a set of testing and verification calculations and performance tests assessing parallel scaling characteristics up to a full core, sub-channel-resolved model of Watts Bar Unit 1 under hot full-power conditions (193 17x17 assemblies, ~56,000 pins, ~59,000 sub-channels, ~2.8 million thermal-hydraulic (TH) control volumes). Results demonstrate that CTF can now perform full core analysis (not previously possible due to excessively long runtimes and memory requirements) on the order of 20 minutes. This new capability is not only useful to standalone CTF users, but is also being leveraged in support of coupled code multi-physics calculations being done in the CASL program.

COBRA-TF↗

Peer-to-peer architectures for exascale computing : LDRD final report.

The goal of this research was to investigate the potential for employing dynamic, decentralized software architectures to achieve reliability in future high-performance computing platforms. These architectures, inspired by peer-to-peer networks such as botnets that already scale to millions of unreliable nodes, hold promise for enabling scientific applications to run usefully on next-generation exascale platforms ({approx} 10{sup 18} operations per second). Traditional parallel programming techniques suffer rapid deterioration of performance scaling with growing platform size, as the work of coping with increasingly frequent failures dominates over useful computation. Our studies suggest that new architectures, in which failures are treated as ubiquitous and their effects are considered as simply another controllable source of error in a scientific computation, can remove such obstacles to exascale computing for certain applications. We have developed a simulation framework, as well as a preliminary implementation in a large-scale emulation environment, for exploration of these 'fault-oblivious computing' approaches. High-performance computing (HPC) faces a fundamental problem of increasing total component failure rates due to increasing system sizes, which threaten to degrade system reliability to an unusable level by the time the exascale range is reached ({approx} 10{sup 18} operations per second, requiring of order millions of processors). As computer scientists seek a way to scale system software for next-generation exascale machines, it is worth considering peer-to-peer (P2P) architectures that are already capable of supporting 10{sup 6}-10{sup 7} unreliable nodes. Exascale platforms will require a different way of looking at systems and software because the machine will likely not be available in its entirety for a meaningful execution time. Realistic estimates of failure rates range from a few times per day to more than once per hour for these platforms. P2P architectures give us a starting point for crafting applications and system software for exascale. In the context of the Internet, P2P applications (e.g., file sharing, botnets) have already solved this problem for 10{sup 6}-10{sup 7} nodes. Usually based on a fractal distributed hash table structure, these systems have proven robust in practice to constant and unpredictable outages, failures, and even subversion. For example, a recent estimate of botnet turnover (i.e., the number of machines leaving and joining) is about 11% per week. Nonetheless, P2P networks remain effective despite these failures: The Conficker botnet has grown to {approx} 5 x 10{sup 6} peers. Unlike today's system software and applications, those for next-generation exascale machines cannot assume a static structure and, to be scalable over millions of nodes, must be decentralized. P2P architectures achieve both, and provide a promising model for 'fault-oblivious computing'. This project aimed to study the dynamics of P2P networks in the context of a design for exascale systems and applications. Having no single point of failure, the most successful P2P architectures are adaptive and self-organizing. While there has been some previous work applying P2P to message passing, little attention has been previously paid to the tightly coupled exascale domain. Typically, the per-node footprint of P2P systems is small, making them ideal for HPC use. The implementation on each peer node cooperates en masse to 'heal' disruptions rather than relying on a controlling 'master' node. Understanding this cooperative behavior from a complex systems viewpoint is essential to predicting useful environments for the inextricably unreliable exascale platforms of the future. We sought to obtain theoretical insight into the stability and large-scale behavior of candidate architectures, and to work toward leveraging Sandia's Emulytics platform to test promising candidates in a realistic (ultimately {ge} 10{sup 7} nodes) setting. Our primary example applications are drawn from linear algebra: a Jacobi relaxation solver for the heat equation, and the closely related technique of value iteration in optimization. We aimed to apply P2P concepts in designing implementations capable of surviving an unreliable machine of 10{sup 6} nodes.

99 GENERAL AND MISCELLANEOUS↗

Load Balancing of Dynamical Nucleation Theory Monte Carlo Simulations through Resource Sharing Barriers

The dynamical nucleation theory Monte Carlo (DNTMC) application from the NWChem computational chemistry suite utilizes a Markov chain Monte Carlo, two-level parallel structure, with periodic synchronization points that assemble the results of independent finer-grained calculations. Like many such applications, the existing code employs a static partitioning of processes into groups and assigns each group a piece of the finer-grained parallel calculation. A significant cause of performance degradation is load imbalance among groups since the time requirements of the inner-parallel calculation varies widely with the input problem and as a result of the Monte Carlo simulation. We present a novel approach to load balancing such calculations with minimal changes to the application. We introduce the concept of a resource sharing barrier (RSB) – a barrier that allows process groups waiting on other processes’ work to actively contribute to their completion. The RSB load balancing technique is applied to the production DNTMC application code, resulting in a small code change of 200 lines and a reduction in execution time of up to 37%.

Arafat, Md H.↗

Persistent memory as an effective alternative to random access memory in metagenome assembly

Abstract Background The assembly of metagenomes decomposes members of complex microbe communities and allows the characterization of these genomes without laborious cultivation or single-cell metagenomics. Metagenome assembly is a process that is memory intensive and time consuming. Multi-terabyte sequences can become too large to be assembled on a single computer node, and there is no reliable method to predict the memory requirement due to data-specific memory consumption pattern. Currently, out-of-memory (OOM) is one of the most prevalent factors that causes metagenome assembly failures. Results In this study, we explored the possibility of using Persistent Memory (PMem) as a less expensive substitute for dynamic random access memory (DRAM) to reduce OOM and increase the scalability of metagenome assemblers. We evaluated the execution time and memory usage of three popular metagenome assemblers (MetaSPAdes, MEGAHIT, and MetaHipMer2) in datasets up to one terabase. We found that PMem can enable metagenome assemblers on terabyte-sized datasets by partially or fully substituting DRAM. Depending on the configured DRAM/PMEM ratio, running metagenome assemblies with PMem can achieve a similar speed as DRAM, while in the worst case it showed a roughly two-fold slowdown. In addition, different assemblers displayed distinct memory/speed trade-offs in the same hardware/software environment. Conclusions We demonstrated that PMem is capable of expanding the capacity of DRAM to allow larger metagenome assembly with a potential tradeoff in speed. Because PMem can be used directly without any application-specific code modification, these findings are likely to be generalized to other memory-intensive bioinformatics applications.

59 BASIC BIOLOGICAL SCIENCES↗

SU-D-206-02: Evaluation of Partial Storage of the System Matrix for Cone Beam Computed Tomography Using a GPU Platform

Purpose: Iterative reconstruction algorithms in computed tomography (CT) require a fast method for computing the intersections between the photons’ trajectories and the object, also called ray-tracing or system matrix computation. This work evaluates different ways to store the system matrix, aiming to reconstruct dense image grids in reasonable time. Methods: We propose an optimized implementation of the Siddon’s algorithm using graphics processing units (GPUs) with a novel data storage scheme. The algorithm computes a part of the system matrix on demand, typically, for one projection angle. The proposed method was enhanced with accelerating options: storage of larger subsets of the system matrix, systematic reuse of data via geometric symmetries, an arithmetic-rich parallel code and code configuration via machine learning. It was tested on geometries mimicking a cone beam CT acquisition of a human head. To realistically assess the execution time, the ray-tracing routines were integrated into a regularized Poisson-based reconstruction algorithm. The proposed scheme was also compared to a different approach, where the system matrix is fully pre-computed and loaded at reconstruction time. Results: Fast ray-tracing of realistic acquisition geometries, which often lack spatial symmetry properties, was enabled via the proposed method. Ray-tracing interleaved with projection and backprojection operations required significant additional time. In most cases, ray-tracing was shown to use about 66 % of the total reconstruction time. In absolute terms, tracing times varied from 3.6 s to 7.5 min, depending on the problem size. The presence of geometrical symmetries allowed for non-negligible ray-tracing and reconstruction time reduction. Arithmetic-rich parallel code and machine learning permitted a modest reconstruction time reduction, in the order of 1 %. Conclusion: Partial system matrix storage permitted the reconstruction of higher 3D image grid sizes and larger projection datasets at the cost of additional time, when compared to the fully pre-computed approach. This work was supported in part by the Fonds de recherche du Quebec - Nature et technologies (FRQ-NT). The authors acknowledge partial support by the CREATE Medical Physics Research Training Network grant of the Natural Sciences and Engineering Research Council of Canada (Grant No. 432290).

60 APPLIED LIFE SCIENCES↗

Deploy threading in Nalu solver stack

The goal of the ExaWind project is to enable predictive simulations of wind farms composed of many MW-scale turbines situated in complex terrain. Predictive simulations will require computational fluid dynamics (CFD) simulations for which the mesh resolves the geometry of the turbines, and captures the rotation and large deflections of blades. Whereas such simulations for a single turbine are arguably petascale class, multi-turbine wind farm simulations will require exascale-class resources. The primary code in the ExaWind project is Nalu, which is an unstructured-grid solver for the acousticallyincompressible Navier-Stokes equations, and mass continuity is maintained through pressure projection. The model consists of the mass-continuity Poisson-type equation for pressure and a momentum equation for the velocity. For such modeling approaches, simulation times are dominated by linear-system setup and solution for the continuity and momentum systems. For the ExaWind challenge problem, the moving meshes greatly affect overall solver costs as re-initialization of matrices and re-computation of preconditioners is required at every time step In this Milestone, we examine the effect of threading on the solver stack performance against flat-MPI results obtained from previous milestones using Haswell performance data full-turbine simulations. Whereas the momentum equations are solved only with the Trilinos solvers, we investigate two algebraic-multigrid preconditioners for the continuity equations: Trilinos/Muelu and HYPRE/BoomerAMG. These two packages embody smoothed-aggregation and classical Ruge-Stiiben AMG methods, respectively. In our FY18 Q2 report, we described our efforts to improve setup and solve of the continuity equations under flat-MPI parallelism. While significant improvement was demonstrated in the solve phase, setup times remained larger than expected. Starting with the optimized settings described in the Q2 report, we explore here simulation performance where OpenMP threading is employed in the solver stack. For Trilinos, threading is acheived through the Kokkos abstraction where, whereas HYPRE/BoomerAMG employs straight OpenMP. We examined results for our mid-resolution baseline turbine simulation configuration (229M DOF). Simulations on 2048 Haswell cores explored the effect of decreasing the number of MPI ranks while increasing the number of threads. Both HYPRE and Trilinos exhibited similar overal solution times, and both showed dramatic increases in simulation time in the shift from MPI ranks to OpenMP threads. This increase is attributed to the large amount of work per MPI rank starting at the single-thread configuration. Decreasing MPI ranks, while increasing threads, may be increasing simulation time due to thread synchronization and start-up overhead contributing to the latency and serial time in the model. These result showed that an MPI+OpenMP parallel decomposition will be more effective as the amount per MPI rank computation per MPI rank decreases and the communication latency increases. This idea was demonstrated in a strong scaling study of our low-resolution baseline model (29M DOF) with the Trilinos-HYPRE configuration. While MPI-only results showed scaling improvement out to about 1536 cores, engaging threading carried scaling improvements out to 4128 cores — roughly 7000 DOF per core. This is an important result as improved strong scaling is needed for simulations to be executed over sufficiently long simulated durations (i.e., for many timesteps). In addition to threading work described above, the team examined solver-performance improvements by exploring communication-overhead in the HYPRE-GMRES implementation through a communicationoptimal- GMRE algorithm (CO-GMRES), and offloading compute-intensive solver actions to GPUs. To those ends, a HYPRE mini-app was allow us to easily test different solver approaches and HYPRE parameter settings without running the entire Nalu code. With GPU acceleration on the Summitdev supercomputer, a 20x speedup was achieved for the overall preconditioner and solver execution time for the mini-app. A study on Haswell processors showed that CO-GMRES provides benefits as one increases MPI ranks.

17 WIND ENERGY↗

PARIS: Predicting application resilience using machine learning

The traditional method to study application resilience to errors in HPC applications uses fault injection (FI), a time-consuming approach. Furthermore, while analytical models have been built to overcome the inefficiencies of FI, they lack accuracy. In this paper, we present PARIS, a machine-learning method to predict application resilience that avoids the time-consuming process of random FI and provides higher prediction accuracy than analytical models. PARIS captures the implicit relationship between application characteristics and application resilience, which is difficult to capture using most analytical models. We overcome many technical challenges for feature construction, extraction, and selection to use machine learning in our prediction approach. Our evaluation on 16 HPC benchmarks shows that PARIS achieves high prediction accuracy. PARIS is up to 450x faster than random FI (49x on average). Compared to the state-of-the-art analytical model, PARIS is at least 63% better in terms of accuracy and has comparable execution time on average.

97 MATHEMATICS AND COMPUTING↗

GridOPTICS/GridPACK

GridPACK is a software framework consisting of a set of modules designed to simplify the development of programs that model the power grid and run on parallel, high performance computing platforms. It also contains several fully developed applications, including powerflow, dynamic simulation, state estimation, Kalman filter analysis (dynamic state estimation), contingency analysis and real time path rating. These applications can be used either standalone or as components in more complicated workflows that combine several different types of application together. The framework modules are available as a combination of libraries and software templates and consist of components for setting up and distributing power grid networks, support for modeling the behavior of individual buses and branches in the network, converting the network models to the corresponding algebraic equations, and parallel routines for manipulating and solving large algebraic systems. The framework also contains a module for distributing tasks evenly amongst computing resources, even if individual tasks vary widely in their execution times. Additional modules support input and output, basic statistical analysis of contingency based calculations, distributed data structures, as well as basic profiling and error management.

Palmer, Bruce↗