Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “algorithmic efficiency”

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 1,765 records · Page 98

Optimizing blocking and nonblocking reduction operations for multicore systems: Hierarchical design and implementation

Many scientific simulations, using the Message Passing Interface (MPI) programming model, are sensitive to the performance and scalability of reduction collective operations such as MPI_Allreduce and MPI_Reduce. These operations are the most widely used abstractions to perform mathematical operations over all processes that are part of the simulation. In this work, we propose a hierarchical design to implement the reduction operations on multicore systems. This design aims to improve the efficiency of reductions by 1) tailoring the algorithms and customizing the implementations for various communication mechanisms in the system 2) providing the ability to configure the depth of hierarchy to match the system architecture, and 3) providing the ability to independently progress each of this hierarchy. Using this design, we implement MPI_Allreduce and MPI_Reduce operations (and its nonblocking variants MPI_Iallreduce and MPI_Ireduce) for all message sizes, and evaluate on multiple architectures including InfiniBand and Cray XT5. We leverage and enhance our existing infrastructure, Cheetah, which is a framework for implementing hierarchical collective operations to implement these reductions. The experimental results show that the Cheetah reduction operations outperform the production-grade MPI implementations such as Open MPI default, Cray MPI, and MVAPICH2, demonstrating its efficiency, flexibility and portability. On Infini-Band systems, with a microbenchmark, a 512-process Cheetah nonblocking Allreduce and Reduce achieves a speedup of 23x and 10x, respectively, compared to the default Open MPI reductions. The blocking variants of the reduction operations also show similar performance benefits. A 512-process nonblocking Cheetah Allreduce achieves a speedup of 3x, compared to the default MVAPICH2 Allreduce implementation. On a Cray XT5 system, a 6144-process Cheetah Allreduce outperforms the Cray MPI by 145%. The evaluation with an application kernel, Conjugate Gradient solver, shows that the Cheetah reductions speeds up total time to solution by 195%, demonstrating the potential benefits for scientific simulations.

Venkata, Manjunath Gorentla↗

Optimizing Blocking and Nonblocking Reduction Operations for Multicore Systems: Hierarchical Design and Implementation

Many scientific simulations, using the Message Passing Interface (MPI) programming model, are sensitive to the performance and scalability of reduction collective operations such as MPI Allreduce and MPI Reduce. These operations are the most widely used abstractions to perform mathematical operations over all processes that are part of the simulation. In this work, we propose a hierarchical design to implement the reduction operations on multicore systems. This design aims to improve the efficiency of reductions by 1) tailoring the algorithms and customizing the implementations for various communication mechanisms in the system 2) providing the ability to configure the depth of hierarchy to match the system architecture, and 3) providing the ability to independently progress each of this hierarchy. Using this design, we implement MPI Allreduce and MPI Reduce operations (and its nonblocking variants MPI Iallreduce and MPI Ireduce) for all message sizes, and evaluate on multiple architectures including InfiniBand and Cray XT5. We leverage and enhance our existing infrastructure, Cheetah, which is a framework for implementing hierarchical collective operations to implement these reductions. The experimental results show that the Cheetah reduction operations outperform the production-grade MPI implementations such as Open MPI default, Cray MPI, and MVAPICH2, demonstrating its efficiency, flexibility and portability. On Infini- Band systems, with a microbenchmark, a 512-process Cheetah nonblocking Allreduce and Reduce achieves a speedup of 23x and 10x, respectively, compared to the default Open MPI reductions. The blocking variants of the reduction operations also show similar performance benefits. A 512-process nonblocking Cheetah Allreduce achieves a speedup of 3x, compared to the default MVAPICH2 Allreduce implementation. On a Cray XT5 system, a 6144-process Cheetah Allreduce outperforms the Cray MPI by 145%. The evaluation with an application kernel, Conjugate Gradient solver, shows that the Cheetah reductions speeds up total time to solution by 195%, demonstrating the potential benefits for scientific simulations.

Gorentla Venkata, Manjunath↗

Trigger Rate of High Energy Electrons in the Mu2e Experiment

aThe Mu2e Collaboration will look for Charged Lepton Flavor Violation (CLFV) decay processes, with a sensitivity goal of 10E-17. The experiment is searching for a 105 MeV electron, which will be detected using a straw tracker solenoid and a crystal calorimeter. With a goal processing time of less than 5 milliseconds per event and a target rejection greater than 100, it is critical that the trigger software is optimized for speed and performance. The trigger is responsible for parsing data from the detector and determining which events to keep. By examining the properties CLFV event helices are expected to have, a specific set of parameters, or cuts, were used to create Mu2e triggering algorithms. The following research is an investigation into the efficiency of the trigger and a calculation of the rate at which these events will be triggered. With the combined contribution of two helix reconstruction algorithms, the trigger will be expected to accept approximately on e CLFV candidate electron event per second, with a triggering rate of .86 Hz.

43 PARTICLE ACCELERATORS↗

Fast Bayesian optimization of Needle-in-a-Haystack problems using zooming memory-based initialization (ZoMBI)

Abstract Needle-in-a-Haystack problems exist across a wide range of applications including rare disease prediction, ecological resource management, fraud detection, and material property optimization. A Needle-in-a-Haystack problem arises when there is an extreme imbalance of optimum conditions relative to the size of the dataset. However, current state-of-the-art optimization algorithms are not designed with the capabilities to find solutions to these challenging multidimensional Needle-in-a-Haystack problems, resulting in slow convergence or pigeonholing into a local minimum. In this paper, we present a Zooming Memory-Based Initialization algorithm, entitled ZoMBI, that builds on conventional Bayesian optimization principles to quickly and efficiently optimize Needle-in-a-Haystack problems in both less time and fewer experiments. The ZoMBI algorithm demonstrates compute time speed-ups of 400× compared to traditional Bayesian optimization as well as efficiently discovering optima in under 100 experiments that are up to 3× more highly optimized than those discovered by similar methods.

36 MATERIALS SCIENCE↗

A Transfer Learning Strategy for Improving the Data Efficiency of Deep Reinforcement Learning Control in Smart Buildings

Reinforcement learning (RL) is a powerful tool that has shown promising results in many domains such as robotics and game-playing. Because RL algorithms learn optimal control policies by continuously interacting with their environments, these algorithms require a lot of data to learn, which limits their application to a wide range of domains. For this reason, there is an immense need for improving the training and data efficiency of RL. Towards addressing this research gap, this paper proposes a transfer learning (TL) approach to improve the efficiency of the RL algorithms by reducing data need and, thus, reducing training time. To demonstrate the proposed approach, a knowledge transfer from a set of buildings to another building was conducted. The results show that the proposed TL approach is a promising method that can efficiently harness the information from similar RL tasks and reduce the data needs of RL algorithms.

Amasyali, Kadir↗

Application of convergence acceleration to the reactor kinetic equations: A comparative study

This presentation provides a comparison of two methodologies for the solution of reactor kinetic equations, namely for a standard finite difference and a semi-analytical approach. The above-mentioned methods are implemented in a convergence acceleration framework to enhance their efficiency and a comparative study is reported to verify whether it is more convenient to use a rudimentary but fast algorithm (finite difference) with respect to the more refined but computationally intense approach of the semi-analytical method. Performance on several test cases from the literature are compared. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Perspectives on Research in Computational Plasma Physics With Applications to Experiments

Computational plasma physics was born some 50 years ago with the development of high-speed computers. Computer simulation is well suited to the challenging nature of fundamental and applied plasma science, which exhibits enormous ranges of time and space scales, and whose underlying mathematical framework is comprised of nonlinear partial differential equations and nonlinear kinetic equations. Computational plasma physics has advanced from relatively simple calculations with limited dimensionality and scope to sophisticated and comprehensive simulation models. Moreover, computational plasma physics has matured as a significant scientific discipline with a rigorous mathematical foundation and voluminous literature. While computational plasma physics has greatly benefitted from the growth in computing capability by many orders of magnitude, the contributions from innovation in methods and algorithms have been no less important. This talk presents a personal perspective on the development and application of computational plasma physics to plasmas in nature and in the laboratory. The examples described will be based on experience over the course of a career in computational plasma physics and illustrate both the fundamental plasma phenomena and the behavior of laboratory plasmas. Some of the specific examples include simulations of microinstabilities in the magnetic mirror, tokamak and spheromak plasmas, laser-plasma interactions, and the Knudsen-layer phenomena, as it affects fusion performance. The examples also illustrate the development of appropriate models and algorithms that are well suited to simulating the phenomena of interest efficiently. Finally, a particular interest of the author has been the development of multiple time-scale algorithms.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

THE HIGH A{sub V} Quasar Survey: Reddened Quasi-Stellar Objects selected from optical/near-infrared photometry. II

Quasi-stellar objects (QSOs) whose spectral energy distributions (SEDs) are reddened by dust either in their host galaxies or in intervening absorber galaxies are to a large degree missed by optical color selection criteria like the ones used by the Sloan Digital Sky Survey (SDSS). To overcome this bias against red QSOs, we employ a combined optical and near-infrared (near-IR) color selection. In this paper, we present a spectroscopic follow-up campaign of a sample of red candidate QSOs which were selected from the SDSS and the UKIRT Infrared Deep Sky Survey (UKIDSS). The spectroscopic data and SDSS/UKIDSS photometry are supplemented by mid-infrared photometry from the Wide-field Infrared Survey Explorer. In our sample of 159 candidates, 154 (97%) are confirmed to be QSOs. We use a statistical algorithm to identify sightlines with plausible intervening absorption systems and identify nine such cases assuming dust in the absorber similar to Large Magellanic Cloud sightlines. We find absorption systems toward 30 QSOs, 2 of which are consistent with the best-fit absorber redshift from the statistical modeling. Furthermore, we observe a broad range in SED properties of the QSOs as probed by the rest-frame 2 μm flux. We find QSOs with a strong excess as well as QSOs with a large deficit at rest-frame 2 μm relative to a QSO template. Potential solutions to these discrepancies are discussed. Overall, our study demonstrates the high efficiency of the optical/near-IR selection of red QSOs.

79 ASTRONOMY AND ASTROPHYSICS↗

The CXSFIT spectral fitting code: Past, present and future

Magnetically confined plasma experiments generate a wealth of spectroscopic data. The first step toward extracting physical parameters is to fit a spectral model to the often complex spectra. The CXSFIT (Charge eXchange Spectroscopy FITting) spectral fitting code was originally developed for fitting charge exchange spectra on JET from the late 1980s onward and has been further developed over decades to keep up with the needs of the users. The primary use is to efficiently fit a large number of spectra with many constrained Gaussian spectral lines of which the physical parameters can be coupled in a user-friendly manner. More recent additions to the code include time-dependent couplings between parameters, flexible background subtraction, and a non-linear coupling scheme between fit parameters. The latter was a pre-requisite for implementing Zeeman and motional Stark effect multiplets in the library of spectral features. The ability to save and replay “fit recipes,” even when multiple iterations are required, has ensured the traceability of the results and is one of the keys to the longevity and success of the code. The code is also in use on other tokamaks (AUG, ST-40) and to fit data from other spectroscopic diagnostics on JET. In this paper, we document the current capabilities and philosophy behind the structure of the code, including some of the algorithms used to calculate spectral features numerically efficiently. We also provide an outline of how CXSFIT could be transferred into a framework that would be able to meet the spectral fitting requirements of future devices, such as ITER.

Delabie, Ephrem G.↗

Sage: parallel semi-asymmtric graph algorithms for NVRAMs

Non-volatile main memory (NVRAM) technologies provide an attractive set of features for large-scale graph analytics, including byte-addressability, low idle power, and improved memory-density. NVRAM systems today have an order of magnitude more NVRAM than traditional memory (DRAM). NVRAM systems could therefore potentially allow very large graph problems to be solved on a single machine, at a modest cost. However, a significant challenge in achieving high performance is in accounting for the fact that NVRAM writes can be much more expensive than NVRAM reads. In this paper, we propose an approach to parallel graph analytics using the Parallel Semi-Asymmetric Model (PSAM), in which the graph is stored as a read-only data structure (in NVRAM), and the amount of mutable memory is kept proportional to the number of vertices. Similar to the popular semi-external and semi-streaming models for graph analytics, the PSAM approach assumes that the vertices of the graph fit in a fast read-write memory (DRAM), but the edges do not. In NVRAM systems, our approach eliminates writes to the NVRAM, among other benefits. To experimentally study this new setting, we develop Sage, a parallel semi-asymmetric graph engine with which we implement provably-efficient (and often work-optimal) PSAM algorithms for over a dozen fundamental graph problems. We experimentally study Sage using a 48--core machine on the largest publicly-available real-world graph (the Hyperlink Web graph with over 3.5 billion vertices and 128 billion edges) equipped with Optane DC Persistent Memory, and show that Sage outperforms the fastest prior systems designed for NVRAM. Importantly, we also show that Sage nearly matches the fastest prior systems running solely in DRAM, by effectively hiding the costs of repeatedly accessing NVRAM versus DRAM.

97 MATHEMATICS AND COMPUTING↗

Lossy Compression: An Online Multi-Stage Technology for High-Fidelity Synchro- Waveform Measurements

Effective real-time monitoring and analysis of distributed grids necessitate the use of synchro-waveform measurements, which capture almost all high-frequency disturbances and transient phenomena. However, due to limitations in high-speed measurements and network bandwidth, it is challenging to transfer all high-fidelity synchro-waveforms losslessly and successfully. To cope with these challenges, a hybrid-based online multi-stage compression algorithm is proposed to significantly improve the compression efficiency for synchro-waveform measurements. Initially, the multiple discrete Wavelet transformation is deployed to deconstruct the waveform components. The delta encoding is further developed to decrease the magnitude. In conjunction with the Lempel-Ziv-Markov chain, the hybrid compression algorithm is implemented to achieve real-time compression for the synchro-waveform measurements. Moreover, an innovative error index that synergizes the time and frequency domain error and correlation is formulated to evaluate the waveform distortion. By integrating compression ratio, suitable parameters can be optimally selected. Finally, the simulation, laboratory experiments, as well as field tests across a spectrum of sampling frequencies and time intervals are conducted to substantiate the efficacy of the proposed method. Here, the outcomes demonstrated that a compression ratio of approximately 15.5 and 17.83 can be reached for 0.5 s and 1 s data under both offline and online scenarios, which equates to a substantial 93.5% to 94.39% reduction in data storage requirements.

High-fidelity synchro-waveform measurements↗

Efficient graph representation framework for chemical molecule similarity tasks

Graph data has emerged in numerous scientific domains and machine learning techniques have been widely used for analysis and learning of diverse data for prediction and decision. Machine learning techniques can readily address complex problems by leveraging their structural information. But graphs cannot be directly used for existing machine learning algorithms unless encoded as vectors. The problem of efficient representation of graphs is a substantial challenge in graph machine learning. In this paper, we propose a novel two-stage framework for the representation of chemical molecule graphs based on the strengths of Graph Isomorphism Networks (GINs) and Siamese autoencoders. In the first stage, the GIN model is constructed and trained using the structural information of chemical molecule graphs. Node attributes, edge attributes, and edge indices are used as input data, while graph attributes are used as labels. The GIN model effectively captures the structural characteristics of graphs and can accurately predict graph attributes, i.e., molecular properties. It also generates Graph Embeddings, represented as vectors that encode the structural information of graphs. In the second stage, Graph Embedding vectors are further optimized for downstream similarity tasks while preserving the graph structural information. The Siamese autoencoder is constructed and trained, which reduces the dimensionality of the Graph Embedding vectors, while maximizing the preservation of structural information in the original high-dimensional vectors. The resulting low-dimensional Graph Embeddings can be effectively utilized for tasks such as approximate nearest neighbor search. The experimental results demonstrate the effectiveness of our proposed framework in accurately predicting graph similarity.

Ma, Jiaji↗

A Hybrid-Learning Algorithm for Online Dynamic State Estimation in Multimachine Power Systems

With the increasing penetration of distributed generators in the smart grids, having knowledge of rapid real-time electromechanical dynamic states has become crucial to system stability control. Conventional Supervisory Control and Data Acquisition (SCADA)-based dynamic state estimation (DSE) techniques are limited by the slow sampling rates, while the emerging phasor measurement units (PMUs) technology enables rapid real-time measurements at network nodes. Using generator bus terminal voltages, we propose a hybrid-learning DSE (HL-DSE) algorithm to estimate the synchronous machine rotor angle and speed in real time. The HL-DSE takes the power system model into account and trains neuroestimators with real-time data in an online manner. Compared with traditional DSE methods, the HL-DSE overcomes limitations by using a data-driven approach in conjunction with the physical power system model. The time efficiency, accuracy, convergence, and robustness of the proposed algorithm are tested under noises and fault conditions in both small- and large-scale test systems. Simulation results show that the proposed HL-DSE is much more computationally efficient than widely used Kalman filter (KF)-based methods while maintaining comparable accuracy and robustness. In particular, HL-DSE is over 100 times faster than square-root unscented KF (SR-UKF) and 80 times faster than extended KF (EKF). The advantages and challenges of the HL-DSE are also discussed.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Efficient Machine Learning Approach for Optimizing Scientific Computing Applications on Emerging HPC Architectures

Efficient parallel implementations of scientific applications on multi-core CPUs with accelerators such as GPUs and Xeon Phis is challenging. This requires - exploiting the data parallel architecture of the accelerator along with the vector pipelines of modern x86 CPU architectures, load balancing, and efficient memory transfer between different devices. It is relatively easy to meet these requirements for highly structured scientific applications. In contrast, a number of scientific and engineering applications are unstructured. Getting performance on accelerators for these applications is extremely challenging because many of these applications employ irregular algorithms which exhibit data-dependent control-ow and irregular memory accesses. Furthermore, these applications are often iterative with dependency between steps, and thus making it hard to parallelize across steps. As a result, parallelism in these applications is often limited to a single step. Numerical simulation of charged particles beam dynamics is one such application where the distribution of work and memory access pattern at each time step is irregular. Applications with these properties tend to present significant branch and memory divergence, load imbalance between different processor cores, and poor compute and memory utilization. Prior research on parallelizing such irregular applications have been focused around optimizing the irregular, data-dependent memory accesses and control-ow during a single step of the application independent of the other steps, with the assumption that these patterns are completely unpredictable. We observed that the structure of computation leading to control-ow divergence and irregular memory accesses in one step is similar to that in the next step. It is possible to predict this structure in the current step by observing the computation structure of previous steps. In this dissertation, we present novel machine learning based optimization techniques to address the parallel implementation challenges of such irregular applications on different HPC architectures. In particular, we use supervised learning to predict the computation structure and use it to address the control-ow and memory access irregularities in the parallel implementation of such applications on GPUs, Xeon Phis, and heterogeneous architectures composed of multi-core CPUs with GPUs or Xeon Phis. We use numerical simulation of charged particles beam dynamics simulation as a motivating example throughout the dissertation to present our new approach, though they should be equally applicable to a wide range of irregular applications. The machine learning approach presented here use predictive analytics and forecasting techniques to adaptively model and track the irregular memory access pattern at each time step of the simulation to anticipate the future memory access pattern. Access pattern forecasts can then be used to formulate optimization decisions during application execution which improves the performance of the application at a future time step based on the observations from earlier time steps. In heterogeneous architectures, forecasts can also be used to improve the memory performance and resource utilization of all the processing units to deliver a good aggregate performance. We used these optimization techniques and anticipation strategy to design a cache-aware, memory efficient parallel algorithm to address the irregularities in the parallel implementation of charged particles beam dynamics simulation on different HPC architectures. Experimental result using a diverse mix of HPC architectures shows that our approach in using anticipation strategy is effective in maximizing data reuse, ensuring workload balance, minimizing branch and memory divergence, and in improving resource utilization.

Arumugam, Kamesh↗

Synchronization of coupled large-scale Boolean networks

This paper investigates the complete synchronization and partial synchronization of two large-scale Boolean networks. First, the aggregation algorithm towards large-scale Boolean network is reviewed. Second, the aggregation algorithm is applied to study the complete synchronization and partial synchronization of large-scale Boolean networks. Finally, an illustrative example is presented to show the efficiency of the proposed results.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

A multi-dimensional, energy- and charge-conserving, nonlinearly implicit, electromagnetic Vlasov–Darwin particle-in-cell algorithm

For decades, the Vlasov–Darwin model has been recognized to be attractive for particle-in-cell (PIC) kinetic plasma simulations in non-radiative electromagnetic regimes, to avoid radiative noise issues and gain computational efficiency. However, the Darwin model results in an elliptic set of field equations that renders conventional explicit time integration unconditionally unstable. We explore a fully implicit PIC algorithm for the Vlasov–Darwin model in multiple dimensions, which overcomes many difficulties of traditional semi-implicit Darwin PIC algorithms. The finite-difference scheme for Darwin field equations and particle equations of motion is space–time-centered, employing particle sub-cycling and orbit-averaging. This algorithm conserves total energy, local charge, canonical-momentum in the ignorable direction, and preserves the Coulomb gauge exactly. An asymptotically well-posed fluid preconditioner allows efficient use of large cell sizes, which are determined by accuracy considerations, not stability, and can be orders of magnitude larger than required in a standard explicit electromagnetic PIC simulation. Finally, we demonstrate the accuracy and efficiency properties of the algorithm with various numerical experiments in 2D–3V.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Solving reaction dynamics with quantum computing algorithms

The description of quantum many-body dynamics is extremely challenging on classical computers, as it can involve many degrees of freedom. However, the time evolution of quantum states is a natural application for quantum computers that are designed to efficiently perform unitary transformations. Here, in this paper, we study quantum algorithms for response functions, relevant for describing different reactions governed by linear response. We focus on nuclear-physics applications and consider a qubit-efficient mapping on the lattice, which can efficiently represent the large volumes required for realistic scattering simulations. For the case of a contact interaction, we develop an algorithm for time evolution based on the Trotter approximation that scales logarithmically with the lattice size and is combined with quantum phase estimation. We eventually focus on the nuclear two-body system and a typical response function relevant for electron scattering as an example. We also investigate ground-state preparation and examine the total circuit depth required for a realistic calculation and the hardware noise level required to interpret the signal.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗