Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Multiple processors”

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 415 records · Page 23

Multiple Coarse Grid Multigrid Methods for Solving Elliptic Problems

In this paper we describe some classes of multigrid methods for solving large linear systems arising in the solution by finite difference methods of certain boundary value problems involving Poisson's equation on rectangular regions. If parallel computing systems are used, then with standard multigrid methods many of the processors will be idle when one is working at the coarsest grid levels. We describe the use of Multiple Coarse Grid MultiGrid (MCGMG) methods. Here one first constructs a periodic set of equations corresponding to the given system. One then constructs a set of coarse grids such that for each grid corresponding to the grid size h there are four grids corresponding to the grid size 2*h. Multigrid operations such as restriction of residuals and interpolation of corrections are done in parallel at each grid level. For suitable choices of the multigrid operators the MCGMG method is equivalent to the Parallel Superconvergent MultiGrid (PSMG) method of Frederickson and McBryan. The convergence properties of MCGMG methods can be accurately analyzed using spectral methods.

Xiao, Shengyou↗

The Global Trigger Processor: A VXS Switch Module for Triggering Large Scale Data Acquisition Systems

The 12 GeV upgrade for Jefferson Lab's Continuous Electron Beam Accelerator Facility requires the development of a new data acquisition system to accommodate the proposed 200 kHz Level 1 trigger rates expected for fixed target experiments at 12 GeV. As part of a suite of trigger electronics comprised of VXS switch and payload modules, the Global Trigger Processor (GTP) will handle up to 32,768 channels of preprocessed trigger information data from the multiple detector systems that surround the beam target at a system clock rate of 250 MHz. The GTP is configured with user programmable Physics trigger equations and when trigger conditions are satisfied, the GTP will activate the storage of data for subsequent analysis. The GTP features an Altera Stratix IV GX FPGA allowing interface to 16 Sub-System Processor modules via 32 5-Gbps links, DDR2 and flash memory devices, two gigabit Ethernet interfaces using Nios II embedded processors, fiber optic transceivers, and trigger output signals. The GTP's high-bandwidth interconnect with the payload modules in the VXS crate, the Ethernet interface for parameter control, status monitoring, and remote update, and the inherent nature of its FPGA give it the flexibility to be used large variety of tasks and adapt to future needs. This paper details the responsibilities of the GTP, the hardware's role in meeting those requirements, and elements of the VXS architecture that facilitated the design of the trigger system. Also presented will be the current status of development including significant milestones and challenges.

Kaneta, Scott↗

A Software Framework for Comparing Training Approaches for Spiking Neuromorphic Systems

There are a wide variety of training approaches for spiking neural networks for neuromorphic deployment. However, it is often not clear how these training algorithms perform or compare when applied across multiple neuromorphic hardware platforms and multiple datasets. In this work, we present a software framework for comparing performance across four neuromorphic training algorithms across three neuromorphic simulators and four simple classification tasks. We introduce an approach for training a spiking neural network using a decision tree, and we compare this approach to training algorithms based on evolutionary algorithms, back-propagation, and reservoir computing. We present a hyperparameter optimization approach to tune the hyperparameters of the algorithm, and show that these optimized hyperparameters depend on the processor, algorithm, and classification task. Finally, we compare the performance of the optimized algorithms across multiple metrics, including accuracy, training time, and resulting network size, and we show that there is not one best training algorithm across all datasets and performance metrics.

Schuman, Catherine↗

Automated target recognition and tracking using an optical pattern recognition neural network

The on-going development of an automatic target recognition and tracking system at the Jet Propulsion Laboratory is presented. This system is an optical pattern recognition neural network (OPRNN) that is an integration of an innovative optical parallel processor and a feature extraction based neural net training algorithm. The parallel optical processor provides high speed and vast parallelism as well as full shift invariance. The neural network algorithm enables simultaneous discrimination of multiple noisy targets in spite of their scales, rotations, perspectives, and various deformations. This fully developed OPRNN system can be effectively utilized for the automated spacecraft recognition and tracking that will lead to success in the Automated Rendezvous and Capture (AR&C) of the unmanned Cargo Transfer Vehicle (CTV). One of the most powerful optical parallel processors for automatic target recognition is the multichannel correlator. With the inherent advantages of parallel processing capability and shift invariance, multiple objects can be simultaneously recognized and tracked using this multichannel correlator. This target tracking capability can be greatly enhanced by utilizing a powerful feature extraction based neural network training algorithm such as the neocognitron. The OPRNN, currently under investigation at JPL, is constructed with an optical multichannel correlator where holographic filters have been prepared using the neocognitron training algorithm. The computation speed of the neocognitron-type OPRNN is up to 10(exp 14) analog connections/sec that enabling the OPRNN to outperform its state-of-the-art electronics counterpart by at least two orders of magnitude.

Chao, Tien-Hsin↗

High-Performance Sparse Matrix-Matrix Products on Intel KNL and Multicore Architectures

Sparse matrix-matrix multiplication (SpGEMM) is a computational primitive that is vastly used in areas ranging from traditional numerical applications to recent big data analysis and machine learning. While many SpGEMM algorithms have been proposed, hardware specific optimizations for multi- and many-core processors are lacking and a detailed analysis of their performance under various use cases and matrices is not available. We firstly identify and mitigate multiple bottlenecks with memory management and thread scheduling on Intel Xeon Phi (Knights Landing or KNL). Specifically targeting multi- and many-core processors, we develop a hash-table-based algorithm and optimize a heap-based shared-memory SpGEMM algorithm. We investigate their performance together with other publicly available codes. Different from the literature, our evaluation also includes use cases that are representative of real graph algorithms, such as multi-source breadth-first search or triangle counting. Our hash-table and heap-based algorithms are showing significant speedups from libraries in the majority of the cases while different algorithms dominate the other scenarios with different matrix size, sparsity, compression factor and operation type. We wrap up in-depth evaluation results and make a recipe to give the best SpGEMM algorithm for target scenario. In conclusion, a critical finding is that hash-table-based SpGEMM gets a significant performance boost if the nonzeros are not required to be sorted within each row of the output matrix.

97 MATHEMATICS AND COMPUTING↗

Portable Data-Parallel Surface Reconstruction on a Uniform Rectilinear Grid

With the increasing heterogeneity and on-node parallelism of high-performance computing hardware, a major challenge is to develop portable and efficient algorithms and software. In this work, we present our implementation of a portable code to perform surface reconstruction using NVIDIA's Thrust library. Surface reconstruction is a technique commonly used in volume tracking methods for simulations of multimaterial flow with interfaces. Here, we have designed a 3D mesh data structure that is easily mapped to the 1D vectors used by Thrust and at the same time is simple to use and uses familiar data structure terminology (such as cells, faces, vertices, and edges). With this new data structure in place, we have implemented a piecewise linear interface reconstruction algorithm in 3 dimensions that effectively exploits the symmetry present in a uniform rectilinear computational cell. Finally, we report performance results, which show that a single implementation of these algorithms can be compiled to multiple backends (specifically, multi-core CPUs, NVIDIA GPUs, and Intel Xeon Phi processors), making efficient use of the available parallelism on each. We also compare performance of our implementation to a legacy FORTRAN implementation in Message Passing Interface (MPI) and show performance parity on single and multi-core CPU and achieved good parallel speed-ups on GPU. In conclusion, our research demonstrates the advantage of performance portability of the underlying data-parallel programming model.

97 MATHEMATICS AND COMPUTING↗

PARVMEC: An Efficient, Scalable Implementation of the Variational Moments Equilibrium Code

The ability to sustain magnetically confined plasma in a state of stable equilibrium is crucial for optimal and cost-effective operations of fusion devices like tokamaks and stellarators. The Variational Moments Equilibrium Code (VMEC) is the de-facto serial application used by fusion scientists to compute magnetohydrodynamics (MHD) equilibria and study the physics of three dimensional plasmas in confined configurations. Modern fusion energy experiments have larger system scales with more interactive experimental workflows, both demanding faster analysis turnaround times on computational workloads that are stressing the capabilities of sequential VMEC. In this paper, we present PARVMEC, an efficient, parallel version of its sequential counterpart, capable of scaling to thousands of processors on distributed memory machines. PARVMEC is a non-linear code, with multiple numerical physics modules, each with its own computational complexity. A detailed speedup analysis supported by scaling results on 1,024 cores of a Cray XC30 supercomputer is presented. Depending on the mode of PARVMEC execution, speedup improvements of one to two orders of magnitude are reported. PARVMEC equips fusion scientists for the first time with a state-of-theart capability for rapid, high fidelity analyses of magnetically confined plasmas at unprecedented scales.

Seal, Sudip K↗

Distributed merge trees

Improved simulations and sensors are producing datasets whose increasing complexity exhausts our ability to visualize and comprehend them directly. To cope with this problem, we can detect and extract significant features in the data and use them as the basis for subsequent analysis. Topological methods are valuable in this context because they provide robust and general feature definitions. As the growth of serial computational power has stalled, data analysis is becoming increasingly dependent on massively parallel machines. To satisfy the computational demand created by complex datasets, algorithms need to effectively utilize these computer architectures. The main strength of topological methods, their emphasis on global information, turns into an obstacle during parallelization. We present two approaches to alleviate this problem. We develop a distributed representation of the merge tree that avoids computing the global tree on a single processor and lets us parallelize subsequent queries. To account for the increasing number of cores per processor, we develop a new data structure that lets us take advantage of multiple shared-memory cores to parallelize the work on a single node. Finally, we present experiments that illustrate the strengths of our approach as well as help identify future challenges. Copyright © 2013 ACM 978-1-4503-1922-5/13/ 02...$15.00.

Morozov, D↗

Ex-Core Radiation Transport Modeling with VERA User Manual

It is important to be able to accurately predict the neutron flux outside the immediate reactor core for a variety of safety and material analyses. Monte Carlo radiation transport calculations are required to produce these high–fidelity ex-core responses. The Virtual Environment for Reactor Applications (VERA) provides the automated capability to launch independent Shift fixed-source and eigenvalue Monte Carlo (MC) calculations for user-specified state points during a standard Virtual Environment for Reactor Applications (VERA) calculation. VERA couples MPACT with COBRA-TF (CTF) to Shift to perform ex-core tallies for multiple state points concurrently, with each component capable of parallel execution on independent processor domains. In these ex-core calculations, MPACT is coupled to CTF and performs the in-core depletion and heat transfer calculation, followed by a fixed-source Shift transport calculation including ex-core regions to produce ex-core responses. The fission source, fuel pin temperatures, moderator temperature and density, boron concentration, and fuel pin depleted isotopic compositions can be transferred to Shift from the MPACT calculation. Specifically, VERA performs fluence calculations in the core barrel outward to the end of the pressure vessel and detector response calculations in ex-core detectors. It also performs the requested tallies in any user-defined ex-core regions. VERA takes advantage of the General Geometry (GG) package in Shift. This gives VERA the flexibility to explicitly model features outside the core barrel, including detailed vessel models, detectors, and power plant details. A very limited set of experimental and numerical benchmarks is available for ex-core simulation comparison. The Consortium for Advanced Simulation of Light Water Reactors has developed a set of ex-core benchmark problems to include as part of the VERA verification and validation set of problems. The ex-core capability in VERA has been tested on small representative assembly problems, multi-assembly problems, as well as quarter-core and full-core problems. VERA View has also been extended to visualize these vessel fluence results from VERA. This manual serves to present a guide to VERA users about the methodology behind ex-core calculations and the details of input, output, and analysis of results from these calculations. Details in this version of the manual are based on features in VERA 4.0.1.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Implicit schemes and parallel computing in unstructured grid CFD

The development of implicit schemes for obtaining steady state solutions to the Euler and Navier-Stokes equations on unstructured grids is outlined. Applications are presented that compare the convergence characteristics of various implicit methods. Next, the development of explicit and implicit schemes to compute unsteady flows on unstructured grids is discussed. Next, the issues involved in parallelizing finite volume schemes on unstructured meshes in an MIMD (multiple instruction/multiple data stream) fashion are outlined. Techniques for partitioning unstructured grids among processors and for extracting parallelism in explicit and implicit solvers are discussed. Finally, some dynamic load balancing ideas, which are useful in adaptive transient computations, are presented.

Venkatakrishnam, V.↗

An MPA-IO interface to HPSS

This paper describes an implementation of the proposed MPI-IO (Message Passing Interface - Input/Output) standard for parallel I/O. Our system uses third-party transfer to move data over an external network between the processors where it is used and the I/O devices where it resides. Data travels directly from source to destination, without the need for shuffling it among processors or funneling it through a central node. Our distributed server model lets multiple compute nodes share the burden of coordinating data transfers. The system is built on the High Performance Storage System (HPSS), and a prototype version runs on a Meiko CS-2 parallel computer.

Jones, Terry↗

An MPI-IO Interface to HPSS

This paper describes an implementation of the proposed MPI-IO standard for parallel I/O. Our system uses third-party transfer to move data over an external network between the processors where it is used and the I/O devices where it resides. Data travels directly from source to destination, without the need for shuffling it among processors or funneling it through a central node. Our distributed server model lets multiple compute nodes share the burden of coordinating data transfers. The system is built on the High Performance Storage System (HPSS), and a prototype version runs on a Meiko CS-2 parallel computer.

Parallel Processing↗

Hardware/Software Issues for Video Guidance Systems: The Coreco Frame Grabber

The F64 frame grabber is a high performance video image acquisition and processing board utilizing the TMS320C40 and TMS34020 processors. The hardware is designed for the ISA 16 bit bus and supports multiple digital or analog cameras. It has an acquisition rate of 40 million pixels per second, with a variable sampling frequency of 510 kHz to MO MHz. The board has a 4MB frame buffer memory expandable to 32 MB, and has a simultaneous acquisition and processing capability. It supports both VGA and RGB displays, and accepts all analog and digital video input standards.

Bales, John W.↗

Tools for Analysis and Visualization of Large Time-Varying CFD Data Sets

In the second year, we continued to built upon and improve our scanline-based direct volume renderer that we developed in the first year of this grant. This extremely general rendering approach can handle regular or irregular grids, including overlapping multiple grids, and polygon mesh surfaces. It runs in parallel on multi-processors. It can also be used in conjunction with a k-d tree hierarchy, where approximate models and error terms are stored in the nodes of the tree, and approximate fast renderings can be created. We have extended our software to handle time-varying data where the data changes but the grid does not. We are now working on extending it to handle more general time-varying data. We have also developed a new extension of our direct volume renderer that uses automatic decimation of the 3D grid, as opposed to an explicit hierarchy. We explored this alternative approach as being more appropriate for very large data sets, where the extra expense of a tree may be unacceptable. We also describe a new approach to direct volume rendering using hardware 3D textures and incorporates lighting effects. Volume rendering using hardware 3D textures is extremely fast, and machines capable of using this technique are becoming more moderately priced. While this technique, at present, is limited to use with regular grids, we are pursuing possible algorithms extending the approach to more general grid types. We have also begun to explore a new method for determining the accuracy of approximate models based on the light field method described at ACM SIGGRAPH '96. In our initial implementation, we automatically image the volume from 32 equi-distant positions on the surface of an enclosing tessellated sphere. We then calculate differences between these images under different conditions of volume approximation or decimation. We are studying whether this will give a quantitative measure of the effects of approximation. We have created new tools for exploring the differences between images produced by various rendering methods. Images created by our software can be stored in the SGI RGB format. Our idtools software reads in pair of images and compares them using various metrics. The differences of the images using the RGB, HSV, and HSL color models can be calculated and shown. We can also calculate the auto-correlation function and the Fourier transform of the image and image differences. We will explore how these image differences compare in order to find useful metrics for quantifying the success of various visualization approaches. In general, progress was consistent with our research plan for the second year of the grant.

Wilhelms, Jane↗

Applications of Digital Image Processing 11

A new technique, digital image velocimetry, is proposed for the measurement of instantaneous velocity fields of time dependent flows. A time sequence of single-exposure images of seed particles are captured with a high-speed camera, and a finite number of the single-exposure images are sampled within a prescribed period in time. The sampled images are then digitized on an image processor, enhanced, and superimposed to construct an image which is equivalent to a multiple exposure image used in both laser speckle velocimetry and particle image velocimetry. The superimposed image and a single-exposure Image are digitally Fourier transformed for extraction of information on the velocity field. A great enhancement of the dynamic range of the velocity measurement is accomplished through the new technique by manipulating the Fourier transform of both the single-exposure image and the superimposed image. Also the direction of the velocity vector is unequivocally determined. With the use of a high-speed video camera, the whole process from image acquisition to velocity determination can be carried out electronically; thus this technique can be developed into a real-time capability.

Cho, Y. -C.↗

Design and Implementation of a Mechanical Control System for the Scanning Microwave Limb Sounder

The Scanning Microwave Limb Sounder (SMLS) will use technological improvements in low noise mixers to provide precise data on the Earth's atmospheric composition with high spatial resolution. This project focuses on the design and implementation of a real time control system needed for airborne engineering tests of the SMLS. The system must coordinate the actuation of optical components using four motors with encoder readback, while collecting synchronized telemetric data from a GPS receiver and 3-axis gyrometric system. A graphical user interface for testing the control system was also designed using Python. Although the system could have been implemented with a FPGA-based setup, we chose to use a low cost processor development kit manufactured by XMOS. The XMOS architecture allows parallel execution of multiple tasks on separate threads-making it ideal for this application and is easily programmed using XC (a subset of C). The necessary communication interfaces were implemented in software, including Ethernet, with significant cost and time reduction compared to an FPGA-based approach. For these reasons, the XMOS technology is an attractive, cost effective, alternative to FPGA-based technologies for this design and similar rapid prototyping projects.

Mars Science Laboratory Robotics↗

Distributed Merge Trees

Improved simulations and sensors are producing datasets whose increasing complexity exhausts our ability to visualize and comprehend them directly. To cope with this problem, we can detect and extract significant features in the data and use them as the basis for subsequent analysis. Topological methods are valuable in this context because they provide robust and general feature definitions. As the growth of serial computational power has stalled, data analysis is becoming increasingly dependent on massively parallel machines. To satisfy the computational demand created by complex datasets, algorithms need to effectively utilize these computer architectures. The main strength of topological methods, their emphasis on global information, turns into an obstacle during parallelization. We present two approaches to alleviate this problem. We develop a distributed representation of the merge tree that avoids computing the global tree on a single processor and lets us parallelize subsequent queries. To account for the increasing number of cores per processor, we develop a new data structure that lets us take advantage of multiple shared-memory cores to parallelize the work on a single node. Finally, we present experiments that illustrate the strengths of our approach as well as help identify future challenges.

97 MATHEMATICS AND COMPUTING↗