Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “algorithms scalability”

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 469 records · Page 26

Analysis of a Computational Framework for Bayesian Inverse Problems: Ensemble Kalman Updates and MAP Estimators under Mesh Refinement

This paper analyzes a popular computational framework to solve infinite-dimensional Bayesian inverse problems, discretizing the prior and the forward model in a finite-dimensional weighted inner product space. We demonstrate the benefit of working on a weighted space by establishing operator-norm bounds for finite element and graph-based discretizations of Matérn-type priors and deconvolution forward models. For linear-Gaussian inverse problems, we develop a general theory to characterize the error in the approximation to the posterior. We also embed the computational framework into ensemble Kalman methods and MAP estimators for nonlinear inverse problems. Furthermore, our operator-norm bounds for prior discretizations guarantee the scalability and accuracy of these algorithms under mesh refinement.

Bayesian inverse problem↗

Toward Large-Scale Image Segmentation on Summit

Semantic segmentation of images is an important computer vision task that emerges in a variety of application domains such as medical imaging, robotic vision and autonomous vehicles to name a few. While these domain-specific image analysis tasks involve relatively small image sizes (~ 10 2 × 10 2 ), there are many applications that need to train machine learning models on image data with extents that are orders of magnitude larger (~10 4 × 10 4 ). Training deep neural network (DNN) models on large extent images is extremely memory-intensive and often exceeds the memory limitations of a single graphical processing unit, a hardware accelerator of choice for computer vision workloads. Here, an efficient, sample parallel approach to train U-Net models on large extent image data sets is presented. Its advantages and limitations are analyzed and near-linear strong-scaling speedup demonstrated on 256 nodes (1536 GPUs) of the Summit supercomputer. Using a single node of the Summit supercomputer, an early evaluation of a recently released model parallel framework called GPipe is demonstrated to deliver ~ 2X speedup in executing a U-Net model with an order of magnitude larger number of trainable parameters than reported before. Performance bottlenecks for pipelined training of U-Net models are identified and mitigation strategies to improve the speedups are discussed. Together, these results open up the possibility of combining both approaches into a unified scalable pipelined and data parallel algorithm to efficiently train U-Net models with very large receptive fields on data sets of ultra-large extent images.

Seal, Sudip↗

BD-CATS-IO

Modern cosmology and plasma physics codes are now capable of simulating trillions of particles on petascale systems. Each timestep output from such simulations is on the order of 10s of TBs. Summarizing and visualizing raw particle data is challenging, and scientists often focus on density structures, whether in the real 3D space, or a high-dimensional phase space. In this work, code was developed for reading 10s of terabytes of data and for writing the resultant cluster IDs for each particle. This code will be part of a highly scalable version of the clustering algorithm DBSCAN, and apply it to the largest datasets produced by state-of-the-art codes. The overall function of BD-CATS (Big Data Clustering at Trillion Particles Scale) is to perform end-to-end analysis: loading the data, dynamic partitioning, computing kdtrees, performing clustering analysis, and storing the results. BD-CATS-IO provides the functionality of loading the data and storing the results.

Byna, Surendra↗

DECA: scalable XHMM exome copy-number variant calling with ADAM and Apache Spark

Abstract Background XHMM is a widely used tool for copy-number variant (CNV) discovery from whole exome sequencing data but can require hours to days to run for large cohorts. A more scalable implementation would reduce the need for specialized computational resources and enable increased exploration of the configuration parameter space to obtain the best possible results. Results DECA is a horizontally scalable implementation of the XHMM algorithm using the ADAM framework and Apache Spark that incorporates novel algorithmic optimizations to eliminate unneeded computation. DECA parallelizes XHMM on both multi-core shared memory computers and large shared-nothing Spark clusters. We performed CNV discovery from the read-depth matrix in 2535 exomes in 9.3 min on a 16-core workstation (35.3× speedup vs. XHMM), 12.7 min using 10 executor cores on a Spark cluster (18.8× speedup vs. XHMM), and 9.8 min using 32 executor cores on Amazon AWS’ Elastic MapReduce. We performed CNV discovery from the original BAM files in 292 min using 640 executor cores on a Spark cluster. Conclusions We describe DECA’s performance, our algorithmic and implementation enhancements to XHMM to obtain that performance, and our lessons learned porting a complex genome analysis application to ADAM and Spark. ADAM and Apache Spark are a performant and productive platform for implementing large-scale genome analyses, but efficiently utilizing large clusters can require algorithmic optimizations and careful attention to Spark’s configuration parameters.

Linderman, Michael D. (ORCID:0000000296437148)↗

Decoupling Superconducting Qubits from the Quantum Bus/Readout Resonator to Enable Scaling

Superconducting qubits have made great strides in coherence time, gating, and algorithms. However, to achieve real scalability, more is required. We propose to study the problem of coupling and decoupling a transmon, a popular type of superconducting qubit, from its host resonator, which serves the dual role of a bus connecting qubits together and a readout channel. The transmon couples to its host resonator via its electric-dipole moment. We plan to use a trick of quantum mechanics to null the dipole moment and decouple the transmon. In doing so, we hope to study a variety of physics associated with multi-qubit operation, control, and readout.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Revolutionizing Neuromorphic Computing for Science (Brochure on the 2024 ASCR Workshop on Neuromorphic Computing for Science)

The ASCR basic research needs for Neuromorphic Computing for Science workshop was held in September 2024. The workshop brochure and report aim to inform and draft a set of grand challenges for advancing the field of neuromorphic computing and developing proof of principle neuromorphic circuits applicable for High Performance Computer (HPC) acceleration for scientific discovery, and brainstorm ideas needed for a successful, robust, and world leading basic research program. The resulting priority research directions are: (1) Neuromorphic computing circuit primitives; (2) Connectivity, communication, and hardware integration; (3) Neuroscience-derived dynamics and algorithms; and (4) Ecosystem for scalable neuromorphic co-design. Breakthroughs in understanding, designing, and prototyping the circuitry and simulation capabilities for a truly neuromorphic computer are essential to enable progress in the field.

97 MATHEMATICS AND COMPUTING↗

Revolutionizing Neuromorphic Computing for Science (Report for the 2024 ASCR Workshop on Neuromorphic Computing for Science)

The ASCR basic research needs for Neuromorphic Computing for Science workshop was held in September 2024. The workshop brochure and report aim to inform and draft a set of grand challenges for advancing the field of neuromorphic computing and developing proof of principle neuromorphic circuits applicable for High Performance Computer (HPC) acceleration for scientific discovery, and brainstorm ideas needed for a successful, robust, and world leading basic research program. The resulting priority research directions are: (1) Neuromorphic computing circuit primitives; (2) Connectivity, communication, and hardware integration; (3) Neuroscience-derived dynamics and algorithms; and (4) Ecosystem for scalable neuromorphic co-design. Breakthroughs in understanding, designing, and prototyping the circuitry and simulation capabilities for a truly neuromorphic computer are essential to enable progress in the field.

97 MATHEMATICS AND COMPUTING↗

Extending substructure based iterative solvers to multiple load and repeated analyses

Direct solvers currently dominate commercial finite element structural software, but do not scale well in the fine granularity regime targeted by emerging parallel processors. Substructure based iterative solvers--often called also domain decomposition algorithms--lend themselves better to parallel processing, but must overcome several obstacles before earning their place in general purpose structural analysis programs. One such obstacle is the solution of systems with many or repeated right hand sides. Such systems arise, for example, in multiple load static analyses and in implicit linear dynamics computations. Direct solvers are well-suited for these problems because after the system matrix has been factored, the multiple or repeated solutions can be obtained through relatively inexpensive forward and backward substitutions. On the other hand, iterative solvers in general are ill-suited for these problems because they often must restart from scratch for every different right hand side. In this paper, we present a methodology for extending the range of applications of domain decomposition methods to problems with multiple or repeated right hand sides. Basically, we formulate the overall problem as a series of minimization problems over K-orthogonal and supplementary subspaces, and tailor the preconditioned conjugate gradient algorithm to solve them efficiently. The resulting solution method is scalable, whereas direct factorization schemes and forward and backward substitution algorithms are not. We illustrate the proposed methodology with the solution of static and dynamic structural problems, and highlight its potential to outperform forward and backward substitutions on parallel computers. As an example, we show that for a linear structural dynamics problem with 11640 degrees of freedom, every time-step beyond time-step 15 is solved in a single iteration and consumes 1.0 second on a 32 processor iPSC-860 system; for the same problem and the same parallel processor, a pair of forward/backward substitutions at each step consumes 15.0 seconds.

Farhat, Charbel↗

Nanoscale Cluster Detection in Massive Atom Probe Tomography Data

Recent technological advances in atom probe tomography (APT) have led to unprecedented data acquisition capabilities that routinely generate data sets containing hundreds of millions of atoms. Detecting nanoscale clusters of different atom types present in these enormous amounts of data and analyzing their spatial correlations with one another are fundamental to understanding the structural properties of the material from which the data is derived. Extant algorithms for nanoscale cluster detection do not scale to large data sets. Here, a scalable, CUDA-based implementation of an autocorrelation algorithm is presented. It isolates spatial correlations amongst atomic clusters present in massive APT data sets in linear time using a linear amount of storage. Correctness of the algorithm is demonstrated using large synthetically generated data with known spatial distributions. Benefits and limitations of using GPU-acceleration for autocorrelation-based APT data analyses are presented with supporting performance results on data sets with up to billions of atoms. To our knowledge, this is the first nanoscale cluster detection algorithm that scales to massive APT data sets and executes on commodity hardware.

Seal, Sudip K↗

Cold Plasma Measurements

We have continued the simulation campaign in support of our ongoing magnetospheric cold plasma research project. This project aims to develop the next-generation particle instruments to measure the properties of the cold particle populations in the Earth’s magnetosphere. For this purpose, simulations have been performed with a Particle-In-Cell (PIC) code called the Curvilinear PIC (CPIC). The code is formulated in curvilinear geometry and couples the standard PIC algorithm with algorithms for the generation and adaptation of the underlaying computational mesh. It conforms to complex objects like spacecraft and it can place more grid points in regions where higher resolution is needed. The code also features a scalable solver based on the multigrid algorithm and it is fully parallelized via domain decomposition and MPI.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Parallel adaptive mesh refinement within the PUMAA3D Project

To enable the solution of large-scale applications on distributed memory architectures, we are designing and implementing parallel algorithms for the fundamental tasks of unstructured mesh computation. In this paper, we discuss efficient algorithms developed for two of these tasks: parallel adaptive mesh refinement and mesh partitioning. The algorithms are discussed in the context of two-dimensional finite element solution on triangular meshes, but are suitable for use with a variety of element types and with h- or p-refinement. Results demonstrating the scalability and efficiency of the refinement algorithm and the quality of the mesh partitioning are presented for several test problems on the Intel DELTA.

Freitag, Lori↗

Error-Correcting Codes for Code-Division Multiplexed TES Detectors

Transition edge sensors (TESs) have proven to be highly sensitive and versatile X-ray spectrometers. Upcoming missions, including Athena X-IFU, will rely on highly multiplexed focal planes where more than 32 TES pixels are read out using a single SQUID amplifier channel. We have implemented an error correction algorithm for code-division multiplexed TES signals that can compensate for potential failures of individual SQUID readout channels and that is both scalable and easily implemented in hardware. We present this algorithm for error correction and show results of laboratory tests to assess algorithm performance at recovering TES channels after a SQUID failure.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Computational toolkit for predicting thickness of 2D materials using machine learning and autogenerated dataset by large language model

The thickness of 2D materials not only plays a crucial role in determining the performance of nanoelectronic and optoelectronic devices but also introduces complexities in predicting volume-dependent properties, such as energy storage capacity, due to the intrinsic vacuum within these materials. Although a plethora of experimental techniques, including but not limited to optical contrast, Raman spectroscopy, nonlinear optical spectroscopy, near-field optical imaging, and hyperspectral imaging, facilitate the measurement of 2D material thickness, comprehensive data for many materials remain elusive. Over the past decade, the exponential proliferation of 2D materials and their heterostructures has outstripped the capabilities of conventional experimental and computational approaches. In this evolving landscape, machine learning (ML) has emerged as an indispensable tool, offering a scalable approach to augment these traditional methodologies. Addressing the critical gap, we introduce THICK2D—Thickness Hierarchy Inference and Calculation Kit for 2D Materials. This Python-based computational framework harnesses an autogenerated thickness database, developed using large language models, and advanced ML algorithms to facilitate the rapid and scalable estimation of material thickness, relying solely on crystallographic data. To demonstrate the utility and robustness of THICK2D, we successfully used the toolkit to predict the thickness of more than 8000 2D-based materials, sourced from two extensive 2D materials databases. THICK2D is disseminated as an open-source utility, accessible on GitHub at https://github.com/gmp007/THICK2D, and archived on Zenodo at https://10.5281/zenodo.11216648.

Ekuma, Chinedu E. (ORCID:0000000258527556)↗

Is disaggregation the holy grail of energy efficiency? The case of electricity

This paper aims to address two timely energy problems. First, significant low-cost energy reductions can be made in the residential and commercial sectors, but these savings have not been achievable to date. Second, billions of dollars are being spent to install smart meters, yet the energy saving and financial benefits of this infrastructure - without careful consideration of the human element - will not reach its full potential. We believe that we can address these problems by strategically marrying them, using disaggregation. Disaggregation refers to a set of statistical approaches for extracting end-use and/or appliance level data from an aggregate, or whole-building, energy signal. In this paper, we explain how appliance level data affords numerous benefits, and why using the algorithms in conjunction with smart meters is the most cost-effective and scalable solution for getting this data. We review disaggregation algorithms and their requirements, and evaluate the extent to which smart meters can meet those requirements. Research, technology, and policy recommendations are also outlined. (C) 2012 Elsevier Ltd. All rights reserved.

Armel, KC↗

Program Instrumentation and Trace Analysis

Several attempts have been made recently to apply techniques such as model checking and theorem proving to the analysis of programs. This shall be seen as a current trend to analyze real software systems instead of just their designs. This includes our own effort to develop a model checker for Java, the Java PathFinder 1, one of the very first of its kind in 1998. However, model checking cannot handle very large programs without some kind of abstraction of the program. This paper describes a complementary scalable technique to handle such large programs. Our interest is turned on the observation part of the equation: How much information can be extracted about a program from observing a single execution trace? It is our intention to develop a technology that can be applied automatically and to large full-size applications, with minimal modification to the code. We present a tool, Java PathExplorer (JPaX), for exploring execution traces of Java programs. The tool prioritizes scalability for completeness, and is directed towards detecting errors in programs, not to prove correctness. One core element in JPaX is an instrumentation package that allows to instrument Java byte code files to log various events when executed. The instrumentation is driven by a user provided script that specifies what information to log. Examples of instructions that such a script can contain are: 'report name and arguments of all called methods defined in class C, together with a timestamp'; 'report all updates to all variables'; and 'report all acquisitions and releases of locks'. In more complex instructions one can specify that certain expressions should be evaluated and even that certain code should be executed under various conditions. The instrumentation package can hence be seen as implementing Aspect Oriented Programming for Java in the sense that one can add functionality to a Java program without explicitly changing the code of the original program, but one rather writes an aspect and compiles it into the original program using the instrumentation. Another core element of JPaX is an observation package that supports the analysis of the generated event stream. Two kinds of analysis are currently supported. In temporal analysis the execution trace is evaluated against formulae written in temporal logic. We have implemented a temporal logic evaluator on finite traces using the Maude rewriting system from SRI International, USA. Temporal logic is defined in Maude by giving its syntax as a signature and its semantics as rewrite equations. The resulting semantics is extremely efficient and can handle event streams of hundreds of millions events in few minutes. Furthermore, the implementation is very succinct. The second form of even stream analysis supported is error pattern analysis where an execution trace is analyzed using various error detection algorithms that can identify error-prone programming practices that may potentially lead to errors in some different executions. Two such algorithms focusing on concurrency errors have been implemented in JPaX, one for deadlocks and the other for data races. It is important to note, that a deadlock or data race potential does not need to occur in order for its potential to be detected with these algorithms. This is what makes them very scalable in practice. The data race algorithm implemented is the Eraser algorithm from Compaq, however adopted to Java. The tool is currently being applied to a code base for controlling a spacecraft by the developers of that software in order to evaluate its applicability.

Havelund, Klaus↗

MPI-hybrid Parallelism for Volume Rendering on Large, Multi-core Systems

This work studies the performance and scalability characteristics of"hybrid'" parallel programming and execution as applied to raycasting volume rendering -- a staple visualization algorithm -- on a large, multi-core platform. Historically, the Message Passing Interface (MPI) has become the de-facto standard for parallel programming and execution on modern parallel systems. As the computing industry trends towards multi-core processors, with four- and six-core chips common today and 128-core chips coming soon, we wish to better understand how algorithmic and parallel programming choices impact performance and scalability on large, distributed-memory multi-core systems. Our findings indicate that the hybrid-parallel implementation, at levels of concurrency ranging from 1,728 to 216,000, performs better, uses a smaller absolute memory footprint, and consumes less communication bandwidth than the traditional, MPI-only implementation.

97 MATHEMATICS AND COMPUTING↗

Hybrid Parallelism for Volume Rendering on Large, Multi-core Systems

This work studies the performance and scalability characteristics of"hybrid'"parallel programming and execution as applied to raycasting volume rendering -- a staple visualization algorithm -- on a large, multi-core platform. Historically, the Message Passing Interface (MPI) has become the de-facto standard for parallel programming and execution on modern parallel systems. As the computing industry trends towards multi-core processors, with four- and six-core chips common today and 128-core chips coming soon, we wish to better understand how algorithmic and parallel programming choices impact performance and scalability on large, distributed-memory multi-core systems. Our findings indicate that the hybrid-parallel implementation, at levels of concurrency ranging from 1,728 to 216,000, performs better, uses a smaller absolute memory footprint, and consumes less communication bandwidth than the traditional, MPI-only implementation.

97 MATHEMATICS AND COMPUTING↗