Search NASA⌕ Search

SEARCH · Search NASA

Results for “compute intensive problems”

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 235 records · Page 13

Proceedings of the 1998 International Computational Accelerator Physics Conference

The 1998 International Computational Accelerator Physics Conference, ICAP’98, was held September 14-18, 1998, at the Monterey Convention Center in Monterey, California. The meeting continued the tradition of four earlier Computational Accelerator Physics conferences, held in La Jolla, California (1988), Los Alamos, New Mexico (1990), Pleasanton, California (1993), and Williamsburg, Virginia (1996). ICAP’98 was organized by the Stanford Linear Accelerator Center (SLAC), Los Alamos National Laboratory (LANL), and the National Energy Research Scientific Computing Center (NERSC), and received support from the U.S. Department of Energy. The aim of this conference series is to discuss recent advances in Computational Accelerator Physics and their applications to solving important problems related to present and future accelerators. The ICAP’98 program included plenary sessions on modeling challenges in next-generation accelerators, state-of-the art methods in computational accelerator physics, and a view of what high-end computing would be like in the 21st century. There were also focused sessions on accelerator design and control, modeling high intensity beam phenomena, electromagnetic modeling, advanced algorithms, software environments and libraries, visualization and graphic interfaces, modeling advanced accelerator concepts, modeling the beam-beam interaction, computing wakefields, and combined particle/field simulations. In addition to these oral sessions there was a poster session and a vendor display. Prior to the start of the technical program, a half-day tutorial on High Performance Computing was hosted by NERSC. The ICAP series recognizes that computing and, in particular, advanced computing, is playing an increasingly important role in the successful design and development of major accelerator facilities, and in research and development leading to advances in accelerator technology. We expect that this new emphasis on high-end computing will continue in future ICAP conferences.

Ko, Kwok↗

Extreme problem for partially nonoverlapping domains on a Riemann sphere

The results of this work are referred to the well-known trend of the geometric theory of functions of complex variable, namely, to the extreme problems on the classes of nonoverlapping domains. It was started by Lavrent’ev’s classical work [1], where, in particular, the problem of the product of conformal radii of two nonoverlapping domains was first solved. Now, this trend is intensively developed. The main results can be found in [2–8] and [9–13]. Our results present a generalization of some results in [7].

97 MATHEMATICS AND COMPUTING↗

Celeritas: GPU-accelerated particle transport for detector simulation in High Energy Physics experiments

Within the next decade, experimental High Energy Physics (HEP) will enter a new era of scientific discovery through a set of targeted programs recommended by the Particle Physics Project Prioritization Panel (P5), including the upcoming High Luminosity Large Hadron Collider (LHC) HL-LHC upgrade and the Deep Underground Neutrino Experiment (DUNE). These efforts in the Energy and Intensity Frontiers will require an unprecedented amount of computational capacity on many fronts including Monte Carlo (MC) detector simulation. In order to alleviate this impending computational bottleneck, the Celeritas MC particle transport code is designed to leverage the new generation of heterogeneous computer architectures, including the exascale computing power of U.S. Department of Energy (DOE) Leadership Computing Facilities (LCFs), to model targeted HEP detector problems at the full fidelity of Geant4. This paper presents the planned roadmap for Celeritas, including its proposed code architecture, physics capabilities, and strategies for integrating it with existing and future experimental HEP computing workflows.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Adaptive Angular Quadrature Scheme for a backwards-in-time Method of Characteristics Solution to the Radiative Transfer Equation [Slides]

Radiative transfer/radiation transport are important problems to solve in astrophysics and high energy density physics. Various methods exist to solve radiation transport, such as Monte Carlo (MC), Discrete Ordinates (S N ), Method of Characteristics (MOC), and the spherical harmonics (P N ) method. Method of Characteristics requires “launching” of rays in discrete directions. Unresolved details of angular mesh create ray effects and can miss sources in the domain. Ray effects can lead to unphysical “stepping” in solution and incorrect energy deposition. Adaptive quadrature schemes can be used to detect and mitigate these effects. The Method of Characteristics (MOC) is a common method for solving hyperbolic PDEs in radiation transport and supersonic flow problems. Generally in MOC for radiation transport, virtual particles are tracked from birth to the end of a timestep. This requires interpolation to go from final location to cell averaged or corner values of angular intensity. Backwards-in-Time (BIT) particle tracking avoids this by prescribing the final position of the virtual particle at the cell nodes/corners. Angular intensities are computed at time k + 1 by launching ray back to previous timestep(s), or t = 0. Scheme allows solution to be computed as the characteristic ray is traced backwards in time.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Efficient co-solution of time step size and independent state in simulations of fluid-driven fracture propagation with embedded meshes

Here we present an efficient time-continuation scheme for fluid-driven fracture propagation problems in the extended finite element method framework. The approach applies a monolithic solution strategy to a fully coupled and implicit approximation of hydro-mechanical systems in conjunction with simultaneous linear elastic propagation of multiple fractures. At the end of each time step, the process ensures that the weakest fracture tip is in an equilibrium propagation regime. Furthermore, the solution process provides an initialization procedure for the newly created fracture spaces and an a priori estimate of the stress intensity factor growth rate, improving simulation robustness, and efficiency. The solution process is validated using the Kristianovich-Geertsma-de Klerk analytical solution under the toughness- and viscosity-dominated regimes. It is also extended to and demonstrated on problems with multiple fractures undergoing simultaneous propagation with stress shadow interactions. Numerical examples demonstrate that the solution process can reduce the required computational cost by one order of magnitude compared to other existing methods.

42 ENGINEERING↗

Integrated State Estimation and Contingency Analysis Software Implementation using High Performance Computing Techniques

Power system simulation tools are traditionally developed in sequential mode and codes are optimized for single core computing only. However, the increasing complexity in the power grid models requires more intensive computation. The traditional simulation tools will soon not be able to meet the grid operation requirements. Therefore, power system simulation tools need to evolve accordingly to provide faster and better results for grid operations. This paper presents an integrated state estimation and contingency analysis software implementation using high performance computing techniques. The software is able to solve large size state estimation problems within one second and achieve a near-linear speedup of 9,800 with 10,000 cores for contingency analysis application. The performance evaluation is presented to show its effectiveness.

Chen, Yousu↗

Accelerating DCA++ (Dynamical Cluster Approximation) Scientific Application on the Summit supercomputer

Optimizing scientific applications on today’s accelerator-based high performance computing systems can be challenging, especially when multiple GPUs and CPUs with heterogeneous memories and persistent non-volatile memories are present. An example is Summit, an accelerator-based system at the Oak Ridge Leadership Computing Facility (OLCF) that is rated as the world’s fastest supercomputer to-date. New strategies are thus needed to expose the parallelism in legacy applications, while being amenable to efficient mapping to the underlying architecture.In this paper we discuss our experiences and strategies to port a scientific application, DCA++, to Summit. DCA++ is a highperformance research application that solves quantum manybody problems with a cutting edge quantum cluster algorithm, the dynamical cluster approximation.Our strategies aim to synergize the strengths of the different programming models in the code. These include: (a) streamlining the interactions between the CPU threads and the GPUs, (b) implementing computing kernels on the GPUs and decreasing CPU-GPU memory transfers, (c) allowing asynchronous GPU communications, and (d) increasing compute intensity by combining linear algebraic operations.Full-scale production runs using all 4600 Summit nodes attained a peak performance of 73.5 PFLOPS with a mixed precision implementation.We observed a perfect strong and weak scaling for the quantum Monte Carlo solver in DCA++, while encountering about 2× input/output (I/O) and MPI communication overhead on the time-to-solution for the full machine run. Our hardware agnostic optimizations are designed to alleviate the communication and I/O challenges observed, while improving the compute intensity and obtaining optimal performance on a complex, hybrid architecture like Summit.

Balduzzi, Giovanni↗

A robust approach to Gaussian process implementation

Abstract. Gaussian process (GP) regression is a flexible modeling technique used to predict outputs and to capture uncertainty in the predictions. However, the GP regression process becomes computationally intensive when the training spatial dataset has a large number of observations. To address this challenge, we introduce a scalable GP algorithm, termed MuyGPs, which incorporates nearest-neighbor and leave-one-out cross-validation during training. This approach enables the evaluation of large spatial datasets with state-of-the-art accuracy and speed in certain spatial problems. Despite these advantages, conventional quadratic loss functions used in the MuyGPs optimization, such as root mean squared error (RMSE), are highly influenced by outliers. We explore the behavior of MuyGPs in cases involving outlying observations and, subsequently, develop a robust approach to handle and mitigate their impact. Specifically, we introduce a novel leave-one-out loss function based on the pseudo-Huber function (LOOPH) that effectively accounts for outliers in large spatial datasets within the MuyGPs framework. Our simulation study shows that the LOOPH loss method maintains accuracy despite outlying observations, establishing MuyGPs as a powerful tool for mitigating unusual observation impacts in the large data regime. In the analysis of US ozone data, MuyGPs provides accurate predictions and uncertainty quantification, demonstrating its utility in managing data anomalies. Through these efforts, we advance the understanding of GP regression in spatial contexts.

Mukangango, Juliette↗

EqualChance: Addressing Intra-set Write Variation to Increase Lifetime of Non-volatile Caches

To address the limitations of SRAM such as high-leakage and low-density, researchers have explored use of non-volatile memory (NVM) devices, such as ReRAM (resistive RAM) and STT-RAM (spin transfer torque RAM) for designing on-chip caches. A crucial limitation of NVMs, however, is that their write endurance is low and the large intra-set write variation introduced by existing cache management policies may further exacerbate this problem, thereby reducing the cache lifetime significantly. We present EqualChance, a technique to increase cache lifetime by reducing intra-set write variation. EqualChance works by periodically changing the physical cache-block location of a write-intensive data item within a set to achieve wear-leveling. Simulations using workloads from SPEC CPU2006 suite and HPC (high-performance computing) field show that EqualChance improves the cache lifetime by 4.29X. Also, its implementation overhead is small, and it incurs very small performance and energy loss.

cache lifetime↗

New approaches to Bayesian uncertainty quantification for Nuclear Science (Final Technical Report)

Inverse problems play a central role in experimentation and theory/data comparisons for many areas of modern Nuclear Physics (NP) and High-Energy Physics (HEP). Bayes’s Theorem is a powerful tool for solving Inverse Problems, providing conceptually transparent and unbiased constraints on theoretical parameters and their uncertainties (“Bayesian Inference”) and enabling the quantification of agreement or tension between models and data. However, analyses based on Bayesian Inference are often challenging for NP and HEP applications, either because of the large number of parameters in the problem, the high computational cost, or both. We propose a multi-institutional collaboration to develop and deploy novel Bayesian analysis tools that advance the scientific scope of a broad range of current and future NP experiments. This project brings together NP domain scientists working on several high-profile NP projects for which new, high-performance Bayesian Uncertainty Quantification (“Bayesian UQ”) methods are essential to carry out the science, and data scientists who are developing state-of-the-art methods applicable to these problems. The NP projects in this proposal comprise measurements of the mass and fundamental nature of the neutrino; study of the Quark-Gluon Plasma that filled the early universe; and mapping of natural and anthropogenic radiation environments. While these NP projects have very different scientific goals, with datasets and analysis approaches that differ significantly, they share common requirements for improving computationally intensive Bayesian analyses using advanced Machine Learning algorithms and will benefit strongly from a coherent effort to develop general solutions. This proposal brings together these projects and forefront ML-based data science algorithms to develop such general solutions. The methods developed in this project will also be more widely applicable, thereby advancing science in the larger Nuclear Physics portfolio.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Diffraction of Rayleigh Waves on a Compliant Inclusion in the Elastic Half Space

We consider a three-dimensional dynamical problem of diffraction of Rayleigh plane waves on a circular compliant inclusion in the elastic half space. To solve the problem, we use the method of boundary integral equations. The dynamical stress intensity factors in the vicinities of points of the contour of the inclusion are analyzed.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

The Data Mine model for accessible partnerships in data science

Abstract The Data Mine at Purdue University is a pioneering experiential learning community for undergraduate and graduate students of any background to learn data science. The first data‐intensive experience embedded in a large learning community, The Data Mine had nearly 1300 students in academic year (AY) 2022–2023 and nearly 1700 students for AY 2023–2024. The Data Mine embodies data‐infused education, research, and collaboration. Students learn Python, R, SQL, and shell‐scripting, while working on weekly projects within a high‐performance computing (HPC) cluster. In the Corporate Partners cohort, students work on teams of 5–15 students, led by a paid student team leader. Each cohort follows an Agile approach, working on data‐intensive projects provided by industry partners and mentored by company employees. Students develop professional and data skills throughout the academic year, from August through April. Many students return in subsequent years to the program, increasing their tenure with a Corporate Partner. Student teams are inherently interdisciplinary; students from 133 different majors are involved in the program, ranging from new incoming students through PhD level students. These interdisciplinary teams of students bring new perspectives to challenging problems in which data science is a key part of the solution. The interdisciplinary teams foster an environment of synthesis with ideas and solutions. Students come together with different life experiences, different levels of technical skill, but also varying ways they navigate paths to solutions because of the variety of majors represented, resulting in a more creative and robust solution than a traditional data science program. This article is categorized under: Applications of Computational Statistics > Education in Computational Statistics

Betz, Margaret A.↗

A Compute Capable SSD Architecture for Next-Generation Non-volatile Memories

Existing storage technologies (e.g., disks and ash) are failing to cope with the processor and main memory speed and are limiting the overall perfor- mance of many large scale I/O or data-intensive applications. Emerging fast byte-addressable non-volatile memory (NVM) technologies, such as phase-change memory (PCM), spin-transfer torque memory (STTM) and memristor are very promising and are approaching DRAM-like performance with lower power con- sumption and higher density as process technology scales. These new memories are narrowing down the performance gap between the storage and the main mem- ory and are putting forward challenging problems on existing SSD architecture, I/O interface (e.g, SATA, PCIe) and software. This dissertation addresses those challenges and presents a novel SSD architecture called XSSD. XSSD o oads com- putation in storage to exploit fast NVMs and reduce the redundant data tra c across the I/O bus. XSSD o ers a exible RPC-based programming framework that developers can use for application development on SSD without dealing with the complication of the underlying architecture and communication management. We have built a prototype of XSSD on the BEE3 FPGA prototyping system. We implement various data-intensive applications and achieve speedup and energy ef- ciency of 1.5-8.9 and 1.7-10.27 respectively. This dissertation also compares XSSD with previous work on intelligent storage and intelligent memory. The existing ecosystem and these new enabling technologies make this system more viable than earlier ones.

97 MATHEMATICS AND COMPUTING↗

Accelerating the numerical simulation of magnetic field lines in tokamaks using the GPU

TRIP3D is a field line simulation code that numerically integrates a set of nonlinear magnetic field line differential equations. The code is used to study properties of magnetic islands and stochastic or chaotic field line topologies that are important for designing non-axisymmetric magnetic perturbation coils for controlling plasma instabilities in future machines. The code is very computationally intensive and for large runs can take on the order of days to complete on a traditional single CPU. This work describes how the code was converted from Fortran to C and then restructured to take advantage of GPU computing using NVIDIA's CUDA. The reduction in computing time has been dramatic where runs that previously took clays now take hours allowing a scale of problem to be examined that would previously not have been attempted. These gains have been accomplished without significant hardware expense. Performance, correctness, code flexibility, and implementation time have been analyzed to gauge the success and applicability of these methods when compared to the traditional multi-CPU approach.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

WE-G-207-04: Non-Local Total-Variation (NLTV) Combined with Reweighted L1-Norm for Compressed Sensing Based CT Reconstruction

Purpose: Compressed sensing (CS) has been used for CT (4DCT/CBCT) reconstruction with few projections to reduce dose of radiation. Total-variation (TV) in L1-minimization (min.) with local information is the prevalent technique in CS, while it can be prone to noise. To address the problem, this work proposes to apply a new image processing technique, called non-local TV (NLTV), to CS based CT reconstruction, and incorporate reweighted L1-norm into it for more precise reconstruction. Methods: TV minimizes intensity variations by considering two local neighboring voxels, which can be prone to noise, possibly damaging the reconstructed CT image. NLTV, contrarily, utilizes more global information by computing a weight function of current voxel relative to surrounding search area. In fact, it might be challenging to obtain an optimal solution due to difficulty in defining the weight function with appropriate parameters. Introducing reweighted L1-min., designed for approximation to ideal L0-min., can reduce the dependence on defining the weight function, therefore improving accuracy of the solution. This work implemented the NLTV combined with reweighted L1-min. by Split Bregman Iterative method. For evaluation, a noisy digital phantom and a pelvic CT images are employed to compare the quality of images reconstructed by TV, NLTV and reweighted NLTV. Results: In both cases, conventional and reweighted NLTV outperform TV min. in signal-to-noise ratio (SNR) and root-mean squared errors of the reconstructed images. Relative to conventional NLTV, NLTV with reweighted L1-norm was able to slightly improve SNR, while greatly increasing the contrast between tissues due to additional iterative reweighting process. Conclusion: NLTV min. can provide more precise compressed sensing based CT image reconstruction by incorporating the reweighted L1-norm, while maintaining greater robustness to the noise effect than TV min.

60 APPLIED LIFE SCIENCES↗

Review of multi-fidelity models

Multi-fidelity models provide a framework for integrating computational models of varying complexity, allowing for accurate predictions while optimizing computational resources. These models are especially beneficial when acquiring high-accuracy data is costly or computationally intensive. This review offers a comprehensive analysis of multi-fidelity models, focusing on their applications in scientific and engineering fields, particularly in optimization and uncertainty quantification. It classifies publications on multi-fidelity modeling according to several criteria, including application area, surrogate model selection, types of fidelity, combination methods and year of publication. The study investigates techniques for combining different fidelity levels, with an emphasis on multi-fidelity surrogate models. Here this work discusses reproducibility, open-sourcing methodologies and benchmarking procedures to promote transparency. The manuscript also includes educational toy problems to enhance understanding. Additionally, this paper outlines best practices for presenting multi-fidelity-related savings in a standardized, succinct and yet thorough manner. The review concludes by examining current trends in multi-fidelity modeling, including emerging techniques, recent advancements, and promising research directions.

42 ENGINEERING↗