Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “COMPUTER ARCHITECTURE”

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 1,657 records · Page 92

SCinet Architecture: Featured at the International Conference for High Performance Computing,Networking, Storage and Analysis 2016

SCinet is the purpose-built network that operates during the International Conference for High Performance Computing,Networking, Storage and Analysis (Super Computing or SC). Created each year for the conference, SCinet brings to life a high-capacity network that supports applications and experiments that are a hallmark of the SC conference. The network links the convention center to research and commercial networks around the world. This resource serves as a platform for exhibitors to demonstrate the advanced computing resources of their home institutions and elsewhere by supporting a wide variety of applications. Volunteers from academia, government and industry work together to design and deliver the SCinet infrastructure. Industry vendors and carriers donate millions of dollars in equipment and services needed to build and support the local and wide area networks. Planning begins more than a year in advance of each SC conference and culminates in a high intensity installation in the days leading up to the conference. The SCinet architecture for SC16 illustrates a dramatic increase in participation from the vendor community, particularly those that focus on network equipment. Software-Defined Networking (SDN) and Data Center Networking (DCN) are present in nearly all aspects of the design.

Lyonnais, Marc↗

Optimal processor assignment for pipeline computations

The availability of large scale multitasked parallel architectures introduces the following processor assignment problem for pipelined computations. Given a set of tasks and their precedence constraints, along with their experimentally determined individual responses times for different processor sizes, find an assignment of processor to tasks. Two objectives are of interest: minimal response given a throughput requirement, and maximal throughput given a response time requirement. These assignment problems differ considerably from the classical mapping problem in which several tasks share a processor; instead, it is assumed that a large number of processors are to be assigned to a relatively small number of tasks. Efficient assignment algorithms were developed for different classes of task structures. For a p processor system and a series parallel precedence graph with n constituent tasks, an O(np2) algorithm is provided that finds the optimal assignment for the response time optimization problem; it was found that the assignment optimizing the constrained throughput in O(np2log p) time. Special cases of linear, independent, and tree graphs are also considered.

Nicol, David M.↗

Scalable Contour Tree Computation by Data Parallel Peak Pruning

As data sets grow to exascale, automated data analysis and visualization are increasingly important, to intermediate human understanding and to reduce demands on disk storage via in situ analysis. Trends in architecture of high performance computing systems necessitate analysis algorithms to make effective use of combinations of massively multicore and distributed systems. One of the principal analytic tools is the contour tree, which analyses relationships between contours to identify features of more than local importance. Unfortunately, the predominant algorithms for computing the contour tree are explicitly serial, and founded on serial metaphors, which has limited the scalability of this form of analysis. While there is some work on distributed contour tree computation, and separately on hybrid GPU-CPU computation, there is no efficient algorithm with strong formal guarantees on performance allied with fast practical performance. Furthermore, we report the first shared SMP algorithm for fully parallel contour tree computation, with formal guarantees of O(lg V lg t) parallel steps and O(V lg V) work for data with V samples and t contour tree supernodes, and implementations with more than 30× parallel speed up on both CPU using TBB and GPU using Thrust and up 70× speed up compared to the serial sweep and merge algorithm.

97 MATHEMATICS AND COMPUTING↗

VecPAC: A Vectorizable and Precision-Aware CGRA

This paper proposes VecPAC -- a vectorizable and precision-aware coarse-grained reconfigurable array (CGRA) design. VecPAC integrates CGRA tiles with scalar functional units and specialized tiles with vector functional units that can trade off the number of vector lanes for the accuracy of the computation. We discuss the architecture design and present the related compilation framework. The experimental evaluation on a set of applications from three different domains (embedded, machine learning, and high-performance computing) shows that the hybrid design of VecPAC outperforms CGRAs with only scalar functional units by 1.48x, while providing higher scalability.

Tan, Cheng↗

From biological neural networks to thinking machines: Transitioning biological organizational principles to computer technology

The three-dimensional organization of the vestibular macula is under study by computer assisted reconstruction and simulation methods as a model for more complex neural systems. One goal of this research is to transition knowledge of biological neural network architecture and functioning to computer technology, to contribute to the development of thinking computers. Maculas are organized as weighted neural networks for parallel distributed processing of information. The network is characterized by non-linearity of its terminal/receptive fields. Wiring appears to develop through constrained randomness. A further property is the presence of two main circuits, highly channeled and distributed modifying, that are connected through feedforward-feedback collaterals and biasing subcircuit. Computer simulations demonstrate that differences in geometry of the feedback (afferent) collaterals affects the timing and the magnitude of voltage changes delivered to the spike initiation zone. Feedforward (efferent) collaterals act as voltage followers and likely inhibit neurons of the distributed modifying circuit. These results illustrate the importance of feedforward-feedback loops, of timing, and of inhibition in refining neural network output. They also suggest that it is the distributed modifying network that is most involved in adaptation, memory, and learning. Tests of macular adaptation, through hyper- and microgravitational studies, support this hypothesis since synapses in the distributed modifying circuit, but not the channeled circuit, are altered. Transitioning knowledge of biological systems to computer technology, however, remains problematical.

Ross, Muriel D.↗

Accelerated Constrained Sparse Tensor Factorization on Massively Parallel Architectures

This study presents the first constrained sparse tensor factorization (cSTF) framework that optimizes and fully offloads computation to massively parallel GPU architectures, and the first performance characterization of cSTF on GPU architectures. In contrast to prior work on tensor factorization, where the matricized tensor times Khatri-Rao product (MTTKRP) is the primary performance bottleneck, our systematic analysis of the cSTF algorithm on GPUs reveals that adding constraints creates an additional bottleneck in the update operation for many real-world sparse tensors. While executing the update operation on the GPU brings significant speedup over its CPU counterpart, it remains a significant bottleneck. To further accelerate the update operation, we propose cuADMM, a new update algorithm that leverages algorithmic and code optimization strategies to minimize both computation and data movement on GPUs. As a result, our framework delivers significantly improved performance compared to prior state-of-the-art. On 10 real-world sparse tensors, our framework achieves geometric mean speedup of 5.1 × (max 41.59 ×) and 7.01 × (max 58.05 ×) on the NIVIDA A100 and H100 GPUs, respectively, over the state-of-the-art SPLATT library running on a 26-core Intel Ice Lake Xeon CPU.

Soh, Yongseok↗

COMPILE: a GWAS computational pipeline for gene discovery in complex genomes

Abstract Background Genome-Wide Association Studies (GWAS) are used to identify genes and alleles that contribute to quantitative traits in large and genetically diverse populations. However, traits with complex genetic architectures create an enormous computational load for discovery of candidate genes with acceptable statistical certainty. We developed a streamlined computational pipeline for GWAS (COMPILE) to accelerate identification and annotation of candidate maize genes associated with a quantitative trait, and then matches maize genes to their closest rice and Arabidopsis homologs by sequence similarity. Results COMPILE executed GWAS using a Mixed Linear Model that incorporated, without compression, recent advancements in population structure control, then linked significant Quantitative Trait Loci (QTL) to candidate genes and RNA regulatory elements contained in any genome. COMPILE was validated using published data to identify QTL associated with the traits of α-tocopherol biosynthesis and flowering time, and identified published candidate genes as well as additional genes and non-coding RNAs. We then applied COMPILE to 274 genotypes of the maize Goodman Association Panel to identify candidate loci contributing to resistance of maize stems to penetration by larvae of the European Corn Borer ( Ostrinia nubilalis ). Candidate genes included those that encode a gene of unknown function, WRKY and MYB-like transcriptional factors, receptor-kinase signaling, riboflavin synthesis, nucleotide-sugar interconversion, and prolyl hydroxylation. Expression of the gene of unknown function has been associated with pathogen stress in maize and in rice homologs closest in sequence identity. Conclusions The relative speed of data analysis using COMPILE allowed comparison of population size and compression. Limitations in population size and diversity are major constraints for a trait and are not overcome by increasing marker density. COMPILE is customizable and is readily adaptable for application to species with robust genomic and proteome databases.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An S N Algorithm for Modern Architectures

LANL discrete ordinates transport packages are required to perform large, computationally intensive time-dependent calculations on massively parallel architectures, where even a single such calculation may need many months to complete. While KBA methods scale out well to very large numbers of compute nodes, we are limited by practical constraints on the number of such nodes we can actually apply to any given calculation. Instead, we describe a modified KBA algorithm that allows realization of the reductions in solution time offered by both the current, and future, architectural changes within a compute node.

97 MATHEMATICS AND COMPUTING↗

Parallelizable adjoint stencil computations using transposed forward-mode algorithmic differentiation

Algorithmic differentiation (AD) is a tool for generating discrete adjoint solvers, which efficiently compute gradients of functions with many inputs, for example for use in gradient-based optimization. AD is often applied to large computations such as stencil operators, which are an important part of most structured-mesh PDE solvers. Stencil computations are often parallelized, for example by using OpenMP, and optimized by using techniques such as cache-blocking and tiling to fully utilize multicore CPUs and many-core accelerators and GPUs. Differentiating these codes with conventional reverse-mode AD results in adjoint codes that cannot be expressed as stencil operations and may not be easily parallelizable. They thus leave most of the compute power of modern architectures unused. We present a method that combines forward-mode AD and loop transformation to generate adjoint solvers that use the same memory access pattern as the original computation that they are derived from and can benefit from the same optimization techniques. The effectiveness of this method is demonstrated by generating a scalable adjoint CFD solver for multicore CPUs and Xeon Phi accelerators.

algorithmic differentiation↗

Domain decomposition: A bridge between nature and parallel computers

Domain decomposition is an intuitive organizing principle for a partial differential equation (PDE) computation, both physically and architecturally. However, its significance extends beyond the readily apparent issues of geometry and discretization, on one hand, and of modular software and distributed hardware, on the other. Engineering and computer science aspects are bridged by an old but recently enriched mathematical theory that offers the subject not only unity, but also tools for analysis and generalization. Domain decomposition induces function-space and operator decompositions with valuable properties. Function-space bases and operator splittings that are not derived from domain decompositions generally lack one or more of these properties. The evolution of domain decomposition methods for elliptically dominated problems has linked two major algorithmic developments of the last 15 years: multilevel and Krylov methods. Domain decomposition methods may be considered descendants of both classes with an inheritance from each: they are nearly optimal and at the same time efficiently parallelizable. Many computationally driven application areas are ripe for these developments. A progression is made from a mathematically informal motivation for domain decomposition methods to a specific focus on fluid dynamics applications. To be introductory rather than comprehensive, simple examples are provided while convergence proofs and algorithmic details are left to the original references; however, an attempt is made to convey their most salient features, especially where this leads to algorithmic insight.

Keyes, David E.↗

Parallel fast gauss transform

We present fast adaptive parallel algorithms to compute the sum of N Gaussians at N points. Direct sequential computation of this sum would take O(N{sup 2}) time. The parallel time complexity estimates for our algorithms are O(N/n{sub p}) for uniform point distributions and O( (N/n{sub p}) log (N/n{sub p}) + n{sub p}log n{sub p}) for non-uniform distributions using n{sub p} CPUs. We incorporate a plane-wave representation of the Gaussian kernel which permits 'diagonal translation'. We use parallel octrees and a new scheme for translating the plane-waves to efficiently handle non-uniform distributions. Computing the transform to six-digit accuracy at 120 billion points took approximately 140 seconds using 4096 cores on the Jaguar supercomputer. Our implementation is 'kernel-independent' and can handle other 'Gaussian-type' kernels even when explicit analytic expression for the kernel is not known. These algorithms form a new class of core computational machinery for solving parabolic PDEs on massively parallel architectures.

99 GENERAL AND MISCELLANEOUS↗

Design and Implementation of a High-Performance CCA Event Service

Event services based on publish-subscribe architectures are well established components of distributed computing applications. Recently, an event service has been proposed as part of the Common Component Architecture (CCA) for high-performance computing applications. In this paper we describe our implementation, experimental evaluation, and initial experience with a high performance CCA event service that exploits efficient communications mechanisms commonly used on high-performance computing (HPC) platforms. We describe the CCA event service model, and discuss briefly the possible implementation strategies of that model. We then present the design and implementation of the event service using the Aggregate Remote Memory Copy Interface (ARMCI) as an underlying communication layer for this mechanism. Two alternative implementations are presented and evaluated on a Cray XD-1 platform. The performance results demonstrate that event delivery latencies are low and that the event service is able to achieve high throughput levels. Finally, we describe the use of the event service in an application for high-speed processing of data from a mass spectrometer, and conclude by discussing some possible extensions to the event service for other HPC applications.

97 MATHEMATICS AND COMPUTING↗

Access control and privacy in large distributed systems

Large scale distributed systems consists of workstations, mainframe computers, supercomputers and other types of servers, all connected by a computer network. These systems are being used in a variety of applications including the support of collaborative scientific research. In such an environment, issues of access control and privacy arise. Access control is required for several reasons, including the protection of sensitive resources and cost control. Privacy is also required for similar reasons, including the protection of a researcher's proprietary results. A possible architecture for integrating available computer and communications security technologies into a system that meet these requirements is described. This architecture is meant as a starting point for discussion, rather that the final answer.

Leiner, B. M.↗

Access control and privacy in large distributed systems

Large scale distributed systems consists of workstations, mainframe computers, supercomputers and other types of servers, all connected by a computer network. These systems are being used in a variety of applications including the support of collaborative scientific research. In such an environment, issues of access control and privacy arise. Access control is required for several reasons, including the protection of sensitive resources and cost control. Privacy is also required for similar reasons, including the protection of a researcher's proprietary results. A possible architecture for integrating available computer and communications security technologies into a system that meet these requirements is described. This architecture is meant as a starting point for discussion, rather that the final answer.

Leiner, B. M.↗

Generic Divide and Conquer Internet-Based Computing

The growth of Internet-based applications and the proliferation of networking technologies have been transforming traditional commercial application areas as well as computer and computational sciences and engineering. This growth stimulates the exploration of Peer to Peer (P2P) software technologies that can open new research and application opportunities not only for the commercial world, but also for the scientific and high-performance computing applications community. The general goal of this project is to achieve better understanding of the transition to Internet-based high-performance computing and to develop solutions for some of the technical challenges of this transition. In particular, we are interested in creating long-term motivation for end users to provide their idle processor time to support computationally intensive tasks. We believe that a practical P2P architecture should provide useful service to both clients with high-performance computing needs and contributors of lower-end computing resources. To achieve this, we are designing dual -service architecture for P2P high-performance divide-and conquer computing; we are also experimenting with a prototype implementation. Our proposed architecture incorporates a master server, utilizes dual satellite servers, and operates on the Internet in a dynamically changing large configuration of lower-end nodes provided by volunteer contributors. A dual satellite server comprises a high-performance computing engine and a lower-end contributor service engine. The computing engine provides generic support for divide and conquer computations. The service engine is intended to provide free useful HTTP-based services to contributors of lower-end computing resources. Our proposed architecture is complementary to and accessible from computational grids, such as Globus, Legion, and Condor. Grids provide remote access to existing higher-end computing resources; in contrast, our goal is to utilize idle processor time of lower-end Internet nodes. Our project is focused on a generic divide and conquer paradigm and on mobile applications of this paradigm that can operate on a loose and ever changing pool of lower-end Internet nodes.

Follen, Gregory J.↗

Transferring ecosystem simulation codes to supercomputers

Many ecosystem simulation computer codes have been developed in the last twenty-five years. This development took place initially on main-frame computers, then mini-computers, and more recently, on micro-computers and workstations. Supercomputing platforms (both parallel and distributed systems) have been largely unused, however, because of the perceived difficulty in accessing and using the machines. Also, significant differences in the system architectures of sequential, scalar computers and parallel and/or vector supercomputers must be considered. We have transferred a grassland simulation model (developed on a VAX) to a Cray Y-MP/C90. We describe porting the model to the Cray and the changes we made to exploit the parallelism in the application and improve code execution. The Cray executed the model 30 times faster than the VAX and 10 times faster than a Unix workstation. We achieved an additional speedup of 30 percent by using the compiler's vectoring and 'in-line' capabilities. The code runs at only about 5 percent of the Cray's peak speed because it ineffectively uses the vector and parallel processing capabilities of the Cray. We expect that by restructuring the code, it could execute an additional six to ten times faster.

Skiles, J. W.↗