Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallel algorithms”

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 361 records · Page 20

Machine Learning for Big Data: A Study to Understand Limits at Scale

This report aims to empirically understand the limits of machine learning when applied to Big Data. We observe that recent innovations in being able to collect, access, organize, integrate, and query massive amounts of data from a wide variety of data sources have brought statistical data mining and machine learning under more scrutiny, evaluation and application for gleaning insights from the data than ever before. Much is expected from algorithms without understanding their limitations at scale while dealing with massive datasets. In that context, we pose and address the following questions How does a machine learning algorithm perform on measures such as accuracy and execution time with increasing sample size and feature dimensionality? Does training with more samples guarantee better accuracy? How many features to compute for a given problem? Do more features guarantee better accuracy? Do efforts to derive and calculate more features and train on larger samples worth the effort? As problems become more complex and traditional binary classification algorithms are replaced with multi-task, multi-class categorization algorithms do parallel learners perform better? What happens to the accuracy of the learning algorithm when trained to categorize multiple classes within the same feature space? Towards finding answers to these questions, we describe the design of an empirical study and present the results. We conclude with the following observations (i) accuracy of the learning algorithm increases with increasing sample size but saturates at a point, beyond which more samples do not contribute to better accuracy/learning, (ii) the richness of the feature space dictates performance - both accuracy and training time, (iii) increased dimensionality often reflected in better performance (higher accuracy in spite of longer training times) but the improvements are not commensurate the efforts for feature computation and training and (iv) accuracy of the learning algorithms drop significantly with multi-class learners training on the same feature matrix and (v) learning algorithms perform well when categories in labeled data are independent (i.e., no relationship or hierarchy exists among categories).

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

Nonconforming Mesh Refinement for High-Order Finite Elements

In this study, we propose a general algorithm for nonconforming adaptive mesh refinement (AMR) of unstructured meshes in high-order finite element codes. Our focus is on $h$-refinement with a fixed polynomial order. The algorithm handles triangular, quadrilateral, hexahedral, and prismatic meshes of arbitrarily high-order curvature, for any order finite element space in the de Rham sequence. We present a flexible data structure for meshes with hanging nodes and a general procedure to construct the conforming interpolation operator, both in serial and in parallel. The algorithm and data structure allow anisotropic refinement of tensor product elements in two and three dimensions and support unlimited refinement ratios of adjacent elements. We report numerical experiments verifying the correctness of the algorithms and perform a parallel scaling study to show that we can adapt meshes containing billions of elements and run efficiently on 393,000 parallel tasks. Finally, we illustrate the integration of dynamic AMR into a high-order Lagrangian hydrodynamics solver.

97 MATHEMATICS AND COMPUTING↗

Update on Development of Mesh Generation Algorithms in MeshKit

MeshKit uses a graph-based design for coding all its meshing algorithms, which includes the Reactor Geometry (and mesh) Generation (RGG) algorithms. This report highlights the developmental updates of all the algorithms, results and future work. Parallel versions of algorithms, documentation and performance results are reported. RGG GUI design was updated to incorporate new features requested by the users; boundary layer generation and parallel RGG support were added to the GUI. Key contributions to the release, upgrade and maintenance of other SIGMA1 libraries (CGM and MOAB) were made. Several fundamental meshing algorithms for creating a robust parallel meshing pipeline in MeshKit are under development. Results and current status of automated, open-source and high quality nuclear reactor assembly mesh generation algorithms such as trimesher, quadmesher, interval matching and multi-sweeper are reported.

Jain, Rajeev↗

Integrated Network Decompositions and Dynamic Programming for Graph Optimization (INDDGO)

The INDDGO software package offers a set of tools for finding exact solutions to graph optimization problems via tree decompositions and dynamic programming algorithms. Currently the framework offers serial and parallel (distributed memory) algorithms for finding tree decompositions and solving the maximum weighted independent set problem. The parallel dynamic programming algorithm is implemented on top of the MADNESS task-based runtime.

Source record↗

Integrated Network Decompositions and Dynamic Programming for Graph Optimization (INDDGO)

The INDDGO software package offers a set of tools for finding exact solutions to graph optimization problems via tree decompositions and dynamic programming algorithms. Currently the framework offers serial and parallel (distributed memory) algorithms for finding tree decompositions and solving the maximum weighted independent set problem. The parallel dynamic programming algorithm is implemented on top of the MADNESS task-based runtime.

Source record↗

Asynchronously parallel optimization solver for finding multiple minima

This paper proposes and analyzes an asynchronously parallel optimization algorithm for finding multiple, high-quality minima of nonlinear optimization problems. Our multistart algorithm considers all previously evaluated points when determining where to start or continue a local optimization run. Theoretical results show that, under certain assumptions, the algorithm almost surely starts a finite number of local optimization runs and identifies, or has a single local optimization run converging to, every minimum. The algorithm is applicable to general optimization settings, but our numerical results focus on the case when derivatives are unavailable. In numerical tests, a PYTHON implementation of the algorithm is shown to yield good approximations of many minima (including a global minimum), and this ability scales well with additional resources. Our implementation’s time to solution is shown also to scale well even when the time to evaluate the function evaluation is highly variable.

97 MATHEMATICS AND COMPUTING↗

PV output variability modeling using satellite imagery.

High frequency irradiance variability measured on the ground is caused by the formation, dissipation, and passage of clouds in the sky. If we can identify and associate different cloud types/patterns from satellite imagery, we may be able to predict irradiance variability in areas lacking sensors. With satellite imagery covering the entire U.S., this allows for more accurate integration planning and power flow modeling over wide areas. Satellite imagery from southern Nevada was analyzed at 15 minute intervals over a year. Methods for image stabilization, cloud detection, and textural classification of clouds were developed and tested. High Performance Computing parallel processing algorithms were also investigated and tested. Artificial Neural Networks using imagery as inputs were trained on ground-based measurements of irradiance to model the variability and were tested to show some promise as a means for predicting irradiance variability.

99 GENERAL AND MISCELLANEOUS↗

Implementation of Parallel Dynamic Simulation on Shared-Memory vs. Distributed-Memory Environments

Power system dynamic simulation computes the system response to a sequence of large disturbance, such as sudden changes in generation or load, or a network short circuit followed by protective branch switching operation. It consists of a large set of differential and algebraic equations, which is computational intensive and challenging to solve using single-processor based dynamic simulation solution. High-performance computing (HPC) based parallel computing is a very promising technology to speed up the computation and facilitate the simulation process. This paper presents two different parallel implementations of power grid dynamic simulation using Open Multi-processing (OpenMP) on shared-memory platform, and Message Passing Interface (MPI) on distributed-memory clusters, respectively. The difference of the parallel simulation algorithms and architectures of the two HPC technologies are illustrated, and their performances for running parallel dynamic simulation are compared and demonstrated.

dynamic simulation↗

A Scalable Hybrid Scheme for Ray-Casting of Unstructured Volume Data

We present an algorithm for parallel volume rendering that is a hybrid between classical object order and image order techniques. The algorithm operates on unstructured grids (and structured ones), and thus can deal with block boundaries interleaving in complex ways. It also deals effectively with cases that are prone to load imbalance, i.e., cases where cell sizes differ dramatically, either because of the nature of the input data, or because of the effects of the camera transformation. The algorithm divides work over resources such that each phase of its processing is bounded in the amount of computation it can perform. We demonstrate its efficacy through a series of studies, varying over camera position, data set size, transfer function, image size, and processor count. At its biggest, our experiments scaled up to 8,192 processors and operated on data sets with more than one billion cells. In total, we find that our hybrid algorithm performs well in all cases. This is because our algorithm naturally adapts its computation based on workload, and can operate like either an object order technique or an image order technique in scenarios where those techniques are efficient.

97 MATHEMATICS AND COMPUTING↗

Parallelism in System Tools

The Cray XT, when employed in conjunction with the Lustre filesystem, has provided the ability to generate huge amounts of data in the form of many files. Typically, this is accommodated by satisfying the requests of large numbers of Lustre clients in parallel. In contrast, a single service node (Lustre client) cannot adequately service such datasets. This means that the use of traditional UNIX tools like cp, tar, et alli (with have no parallel capability) can result in substantial impact to user productivity. For example, to copy a 10 TB dataset from the service node using cp would take about 24 hours, under more or less ideal conditions. During production operation, this could easily extend to 36 hours. In this paper, we introduce the Lustre User Toolkit for Cray XT, developed at the Oak Ridge Leadership Computing Facility (OLCF). We will show that Linux commands, implementing highly parallel I/O algorithms, provide orders of magnitude greater performance, greatly reducing impact to productivity.

97 MATHEMATICS AND COMPUTING↗

Improvement of parallelization efficiency of batch pattern BP training algorithm using Open MPI

The use of tuned collective's module of Open MPI to improve a parallelization efficiency of parallel batch pattern back propagation training algorithm of a multilayer perceptron is considered in this paper. The multilayer perceptron model and the usual sequential batch pattern training algorithm are theoretically described. An algorithmic description of a parallel version of the batch pattern training method is introduced. The obtained parallelization efficiency results using Open MPI tuned collective's module and MPICH2 are compared. Our results show that (i) Open MPI tuned collective's module outperforms MPICH2 implementation both on SMP computer and computational cluster and (ii) different internal algorithms of MPI_Allreduce() collective operation give better results on different scenarios and different parallel systems. Therefore the properties of the communication network and user application should be taken into account when a specific collective algorithm is used.

97 MATHEMATICS AND COMPUTING↗

Computing Maximum Cardinality Matchings in Parallel on Bipartite Graphs via Tree-Grafting

It is difficult to obtain high performance when computing matchings on parallel processors because matching algorithms explicitly or implicitly search for paths in the graph, and when these paths become long, there is little concurrency. In spite of this limitation, we present a new algorithm and its shared-memory parallelization that achieves good performance and scalability in computing maximum cardinality matchings in bipartite graphs. This algorithm searches for augmenting paths via specialized breadth-first searches (BFS) from multiple source vertices, hence creating more parallelism than single source algorithms. Algorithms that employ multiple-source searches cannot discard a search tree once no augmenting path is discovered from the tree, unlike algorithms that rely on single-source searches. We describe a novel tree-grafting method that eliminates most of the redundant edge traversals resulting from this property of multiple-source searches. We also employ the recent direction-optimizing BFS algorithm as a subroutine to discover augmenting paths faster. Our algorithm compares favorably with the current best algorithms in terms of the number of edges traversed, the average augmenting path length, and the number of iterations. Here, we provide a proof of correctness for our algorithm. Our NUMA-aware implementation is scalable to 80 threads of an Intel multiprocessor and to 240 threads on an Intel Knights Corner coprocessor. On average, our parallel algorithm runs an order of magnitude faster than the fastest algorithms available. The performance improvement is more significant on graphs with small matching number.

97 MATHEMATICS AND COMPUTING↗

Optimization of multigrid based elliptic solver for large scale simulations in the FLASH code: OPTIMIZATION OF MULTIGRID BASED ELLIPTIC SOLVER IN THE FLASH CODE

FLASH is a multiphysics multiscale adaptive mesh refinement (AMR) code originally designed for simulation of reactive flows often found in Astrophysics. With its wide user base and flexible applications configuration capability, FLASH has a dual task of maintaining scalability and portability in all its solvers. The scalability of fully explicit solvers in the code is tied very closely to that of the underlying mesh. Others such as the Poisson solver based on a multigrid method have more complex scaling behavior. Multigrid methods suffer from processor starvation and dominating communication costs at coarser grids with increase in the number of processors. In this paper, we propose a combination of uniform grid mesh with AMR mesh, and the merger of two different sets of solvers to overcome the scalability limitation of the Poisson solver in FLASH. The principal challenge in the proposed merger is the efficiency of the communication algorithm to map the mesh back and forth between uniform grid and AMR. We present two different parallel mapping algorithms and also discuss results from performance studies of the two implementations. Copyright © 2012 John Wiley & Sons, Ltd.

Computer Science↗

Machine Committee Framework for Power Grid Disturbances Analysis Using Synchrophasors Data

Events detection is a key challenge in power grid frequency disturbances analysis. Accurate detection of events is crucial for situational awareness of the power system. In this paper, we study the problem of events detection in power grid frequency disturbance analysis using synchrophasors data streams. Current events detection approaches for power grid rely on individual detection algorithm. This study integrates some of the existing detection algorithms using the concept of machine committee to develop improved detection approaches for grid disturbance analysis. Specifically, we propose two algorithms—an Event Detection Machine Committee (EDMC) algorithm and a Change-Point Detection Machine Committee (CPDMC) algorithm. Both algorithms use parallel architecture to fuse detection knowledge of its individual methods to arrive at an overall output. The EDMC algorithm combines five individual event detection methods, while the CPDMC algorithm combines two change-point detection methods. Each method performs the detection task separately. The overall output of each algorithm is then computed using a voting strategy. The proposed algorithms are evaluated using three case studies of actual power grid disturbances. Compared with the individual results of the various detection methods, we found that the EDMC algorithm is a better fit for analyzing synchrophasors data; it improves the detection accuracy; and it is suitable for practical scenarios.

24 POWER TRANSMISSION AND DISTRIBUTION↗

CEAZ: Accelerating Parallel I/O Via Hardware-Algorithm Co-Designed Adaptive Lossy Compression

As supercomputers continue to grow to exa-scale, the amount of data that needs to be saved or transmitted is exploding. To this end, many previous works have studied using error-bounded lossy compressors to reduce the data size and improve the I/O performance. However, little work has been done for effectively offloading lossy compression onto FPGA-based SmartNICs to reduce the compression overhead. In this paper, we propose a hardware-algorithm co-design of efficient and adaptive lossy compressor for scientific data on FPGAs (called CEAZ) to accelerate parallel I/O. Our contribution is fourfold: (1) We propose an efficient Huffman coding approach that can adaptively update Huffman codewords online based on codewords generated offline (from a variety of representative scientific datasets). (2) We derive a theoretical analysis to support a precise control of compression ratio under an error-bounded compression mode, enabling accurate offline Huffman codewords generation. This also help us create a fixed-ratio compression mode for consistent throughput. (3) We develop an efficient compression pipeline by adopting cuSZ’s dual-quantization algorithm to our hardware use case. (4) We evaluate CEAC on five real-world datasets with both a single FPGA board and 256 nodes from Bridges2 supercomputer. Experiments show that CEAZ outperforms the second-best FPGA-based lossy compressor by 2× of throughput and 9.6× of compression ratio. It also improves MPI_File_write and MPI_Gather throughputs by up to 32.7× and 31.4×, respectively.

Zhang, Chengming↗

TorchBraid: High-Performance Layer-Parallel Training of Deep Neural Networks with MPI and GPU Acceleration

TorchBraid is a high-performance implementation of layer-parallel training for deep neural networks (DNNs) supporting MPI-based parallelism and GPU acceleration. Layer-parallel training has been developed to overcome the serialization inherent in forward and backward propagation of DNNs that limits utilization of computational resources in the strong scaling limit. To achieve this, TorchBraid integrates the PyTorch neural network framework with the state-of-the-art XBraid time-parallel library. Furthermore, this article presents the use and performance of TorchBraid, in addition to solutions for overcoming the algorithmic challenges inherent in combining automatic differentiation with layer-parallel. Results are presented with and without GPU acceleration for the Tiny ImageNet and MNIST image classification data sets, as well as recurrent neural networks. Overall, TorchBraid enables fast training of DNNs, both in a strong and weak scaling context. In addition to the TorchBraid software, several new advances in applying layer-parallel algorithms are detailed. Integration of layer-parallel with data-parallel algorithms is presented for the first time, showing the computational advantages of the combination. Standard deep learning techniques, like batch-normalization, are developed for layer-parallel training. Finally, a new approach combining layer-parallel with spatial coarsening in order to accelerate training for 3D image classification shows roughly a 10× speedup over serial execution.

Layer-parallel↗

A parallel additive Schwarz preconditioned Jacobi-Davidson algorithm for polynomial eigenvalue problems in quantum dot simulation

We develop a parallel Jacobi-Davidson approach for finding a partial set of eigenpairs of large sparse polynomial eigenvalue problems with application in quantum dot simulation. A Jacobi-Davidson eigenvalue solver is implemented based on the Portable, Extensible Toolkit for Scientific Computation (PETSc). The eigensolver thus inherits PETSc's efficient and various parallel operations, linear solvers, preconditioning schemes, and easy usages. The parallel eigenvalue solver is then used to solve higher degree polynomial eigenvalue problems arising in numerical simulations of three dimensional quantum dots governed by Schroedinger's equations. We find that the parallel restricted additive Schwarz preconditioner in conjunction with a parallel Krylov subspace method (e.g. GMRES) can solve the correction equations, the most costly step in the Jacobi-Davidson algorithm, very efficiently in parallel. Besides, the overall performance is quite satisfactory. We have observed near perfect superlinear speedup by using up to 320 processors. The parallel eigensolver can find all target interior eigenpairs of a quintic polynomial eigenvalue problem with more than 32 million variables within 12 minutes by using 272 Intel 3.0 GHz processors.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Reinforcement Learning for Load-balanced Parallel Particle Tracing

We explore an online reinforcement learning (RL) paradigm to dynamically optimize parallel particle tracing performance in distributed-memory systems. Our method combines three novel components: (1) a work donation algorithm, (2) a high-order workload estimation model, and (3) a communication cost model. First, we design an RL-based work donation algorithm. Our algorithm monitors workloads of processes and creates RL agents to donate data blocks and particles from high-workload processes to low-workload processes to minimize program execution time. The agents learn the donation strategy on the fly based on reward and cost functions designed to consider processes' workload changes and data transfer costs of donation actions. Second, we propose a workload estimation model, helping RL agents estimate the workload distribution of processes in future computations. Third, we design a communication cost model that considers both block and particle data exchange costs, helping RL agents make effective decisions with minimized communication costs. We demonstrate that our algorithm adapts to different flow behaviors in large-scale fluid dynamics, ocean, and weather simulation data. Our algorithm improves parallel particle tracing performance in terms of parallel efficiency, load balance, and costs of I/O and communication for evaluations with up to 16,384 processors.

Distributed and parallel particle tracing↗