Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “distributed memory parallel systems”

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 235 records · Page 13

Center for Computing Research Highlights

Sandia has a legacy of leadership in the advancement of high performance computing (HPC) at extreme scales. First-of-a-kind scalable distributed-memory parallel platforms such as the Intel Paragon, ASCI Red (the world’s first teraflops computer), and Red Storm (co-developed with Cray) helped form the basis for one of the most successful supercomputer product lines ever: the Cray XT series. Sandia also has pioneered system software elements—including lightweight operating systems, the Portals network programming interface, advanced interconnection network designs, and scalable I/O— that are critical to achieving scalability on large computing systems.

97 MATHEMATICS AND COMPUTING↗

PCLIPS

CLIPS is an expert system, created specifically to allow rapid implementation of an expert system. CLIPS is written in C, and thus needs a very small amount of memory to run. Parallel CLIPS (PCLIPS) is an extension to CLIPS which is intended to be used in situations where a group of expert systems are expected to run simultaneously and occasionally communicate with each other on an integrated network. PCLIPS is a coarse-grained data distribution system. Its main goal is to take information in one knowledge base and distribute it to other knowledge bases so that all the executing expert systems are able to use that knowledge to solve their disparate problems.

Krolak, Patrick D.↗

Forward-Weighted CADIS Method for Variance Reduction of Monte Carlo Reactor Analyses

Current state-of-the-art tools and methods used to perform 'real' commercial reactor analyses use high-fidelity transport codes to produce few-group parameters at the assembly level for use in low-order methods applied at the core level. Monte Carlo (MC) methods, which allow detailed and accurate modeling of the full geometry and energy details and are considered the 'gold standard' for radiation transport solutions, are playing an ever-increasing role in correcting and/or verifying the several-decade-old methodology used in current practice. However, the prohibitive computational requirements associated with obtaining fully converged system-wide solutions restrict the role of MC to benchmarking deterministic results at a limited number of state-points for a limited number of relevant quantities. A goal of current research at Oak Ridge National Laboratory (ORNL) is to change this paradigm by enabling the direct use of MC for full-core reactor analyses. The most significant of the many technical challenges that must be overcome is the slow non-uniform convergence of system-wide MC estimates and the memory requirements associated with detailed solutions throughout a reactor (problems involving hundreds of millions of different material and tally regions due to fuel irradiation, temperature distributions, and the needs associated with multi-physics code coupling). To address these challenges, research has focused on development in the following two areas: (1) a hybrid deterministic/MC method for determining high-precision fluxes throughout the problem space in k-eigenvalue problems and (2) an efficient MC domain-decomposition algorithm that partitions the problem phase space onto multiple processors for massively parallel systems, with statistical uncertainty estimation. The focus of this paper is limited to the first area mentioned above. It describes the FW-CADIS method applied to variance reduction of MC reactor analyses and provides initial results for calculating group-wise fluxes throughout a generic 2-D pressurized water reactor (PWR) quarter core model.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Scalable Knowledge Graph Analytics at 136 Petaflop/s

We are motivated by newly proposed methods for data mining large-scale corpora of scholarly publications, such as the full biomedical literature, which may consist of tens of millions of papers spanning decades of research. In this setting, analysts seek to discover how concepts relate to one another. They construct graph representations from annotated text databases and then formulate the relationship-mining problem as one of computing all-pairs shortest paths (APSP), which becomes a significant bottleneck. In this context, we present a new high-performance algorithm and implementation of the Floyd-Warshall algorithm for distributed-memory parallel computers accelerated by GPUs, which we call DSNAPSHOT (Distributed Accelerated Semiring All-Pairs Shortest Path). For our largest experiments, we ran DSNAPSHOT on a connected input graph with millions of vertices using 4, 096nodes (24,576GPUs) of the Oak Ridge National Laboratory's Summit supercomputer system. We find DSNAPSHOT achieves a sustained performance of 136×1015 floating-point operations per second (136petaflop/s) at a parallel efficiency of 90% under weak scaling and, in absolute speed, 70% of the best possible performance given our computation (in the single-precision tropical semiring or “min-plus” algebra). Looking forward, we believe this novel capability will enable the mining of scholarly knowledge corpora when embedded and integrated into artificial intelligence-driven natural language processing workflows at scale.

Kannan, Ramakrishnan {ramki}↗

Concurrent Finite-Element Analysis On Hypercube Computers

Improved approach to execution of finite-element codes on hypercube and similar concurrent data processors increases efficiency of computation for many different types of problems. Based on flexible general model of computation on, and communication among, large-node parallel processors. Hybrid combining direct methods within subdomains and preconditioned-conjugate-gradient (PCG) iteration on remaining boundary system to obtain method of solution both robust and efficient. Preserves general structure and function of conventional sequential finite-element-method software. Generalized to both distributed- and shared-memory multicomputers, eliminating degree of machine specificity restricting general usefulness.

Lyzenga, G. A.↗

A Distributed-Memory Package for Dense Hierarchically Semi-Separable Matrix Computations Using Randomization

In this paper, we present a distributed-memory library for computations with dense structured matrices. A matrix is considered structured if its off-diagonal blocks can be approximated by a rank-deficient matrix with low numerical rank. Here, we use Hierarchically Semi-Separable (HSS) representations. Such matrices appear in many applications, for example, finite-element methods, boundary element methods, and so on. Exploiting this structure allows for fast solution of linear systems and/or fast computation of matrix-vector products, which are the two main building blocks of matrix computations. The compression algorithm that we use, that computes the HSS form of an input dense matrix, relies on randomized sampling with a novel adaptive sampling mechanism. We discuss the parallelization of this algorithm and also present the parallelization of structured matrix-vector product, structured factorization, and solution routines. The efficiency of the approach is demonstrated on large problems from different academic and industrial applications, on up to 8,000 cores. Finally, this work is part of a more global effort, the STRUctured Matrices PACKage (STRUMPACK) software package for computations with sparse and dense structured matrices. Hence, although useful on their own right, the routines also represent a step in the direction of a distributed-memory sparse solver.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗

A communication-avoiding 3D sparse triangular solver

We present a novel distributed memory algorithm to improve the strong scalability of the solution of a sparse triangular system. This operation appears in the solve phase of direct methods for solving general sparse linear systems, Ax = b. Our 3D sparse triangular solver employs several techniques, including a 3D MPI process grid, elimination tree parallelism, and data replication, all of which reduce the per-process communication when combined. We present analytical models to understand the communication cost of our algorithm and show that our 3D sparse triangular solver can reduce the per-process communication volume asymptotically by a factor of O(n1/4) and O(n1/6) for problems arising from the finite element discretizations of 2D "planar" and 3D "non-planar" PDEs, respectively. We implement our algorithm for use in SuperLU_DIST3D, using a hybrid MPI+OpenMP programming model. Our 3D triangular solve algorithm, when run on 12k cores of Cray XC30, outperforms the current state-of-the-art 2D algorithm by 7.2x for planar and 2.7x for the non-planar sparse matrices, respectively.

Sao, Piyush↗

Asynchronous interactive control systems

A class of interactive control systems is derived by generalizing interactive manipulator control systems. The general structural properties of such systems are discussed and an appropriate general software implementation is proposed. This is based on the fact that tasks of interactive control systems can be represented as a network of a finite set of actions which have specific operational characteristics and specific resource requirements, and which are of limited duration. This has enabled the decomposition of the overall control algorithm into a set of subalgorithms, called subcontrollers, which can operate simultaneously and asynchronously. Coordinate transformations of sensor feedback data and actuator set-points have enabled the further simplification of the subcontrollers and have reduced their conflicting resource requirements. The modules of the decomposed control system are implemented as parallel processes with disjoint memory space communicating only by I/O. The synchronization mechanisms for dynamic resource allocation among subcontrollers and other synchronization mechanisms are also discussed in this paper. Such a software organization is suitable for the general form of multiprocessing using computer networks with distributed storage.

Vuskovic, M. I.↗

Parallel discrete event simulation: A shared memory approach

With traditional event list techniques, evaluating a detailed discrete event simulation model can often require hours or even days of computation time. Parallel simulation mimics the interacting servers and queues of a real system by assigning each simulated entity to a processor. By eliminating the event list and maintaining only sufficient synchronization to insure causality, parallel simulation can potentially provide speedups that are linear in the number of processors. A set of shared memory experiments is presented using the Chandy-Misra distributed simulation algorithm to simulate networks of queues. Parameters include queueing network topology and routing probabilities, number of processors, and assignment of network nodes to processors. These experiments show that Chandy-Misra distributed simulation is a questionable alternative to sequential simulation of most queueing network models.

Reed, Daniel A.↗

A Massively Parallel Adaptive Fast-Multipole Method on Heterogeneous Architectures

We present new scalable algorithms and an implementation of the kernel-independent fast multiple method (KIFMM), employing hybrid distributed memory message passing (via MPI) and shared memory/streaming using graphics processing unit (GPU) acceleration to rapidly evaluate two-body non-oscillatory potentials. On traditional CPU-only systems, our implementation scales well up to 30 billion unknowns on 65k cores (AMD/CRAY-based Kraken system at NSF/NICS) on tree data structures with 25 levels between leaves. On GPU-enabled systems, we achieve 30 x speedup for problems of up to 256 million points on 256 GPUs (Lincoln at NSF/NCSA) over a comparable CPU-only implementation. Both of these demonstrations represent the largest and fastest of their kind of which we are aware. We achieve scalability at extreme core counts by extending the initial work of Ying et al. (ACM/IEEE SC 03) with a new approach to scalable MPI-based tree construction and partitioning. For the sub-components of KIFMM, which direct- and approximate-interactions, target evaluation, and source-to-multipole translations, we use CUDA-based GPU-acceleration to achieve excellent performance. To do so requires carefully constructed data structure transformations, which we describe, and whose cost we show is minor. Taken together, these components show promise for ultrascalable FMM in the petascale era and beyond.

99 GENERAL AND MISCELLANEOUS↗

Summer Proceedings 2016: The Center for Computing Research at Sandia National Laboratories

Solving sparse linear systems from the discretization of elliptic partial differential equations (PDEs) is an important building block in many engineering applications. Sparse direct solvers can solve general linear systems, but are usually slower and use much more memory than effective iterative solvers. To overcome these two disadvantages, a hierarchical solver (LoRaSp) based on H2-matrices was introduced in [22]. Here, we have developed a parallel version of the algorithm in LoRaSp to solve large sparse matrices on distributed memory machines. On a single processor, the factorization time of our parallel solver scales almost linearly with the problem size for three-dimensional problems, as opposed to the quadratic scalability of many existing sparse direct solvers. Moreover, our solver leads to almost constant numbers of iterations, when used as a preconditioner for Poisson problems. On more than one processor, our algorithm has significant speedups compared to sequential runs. With this parallel algorithm, we are able to solve large problems much faster than many existing packages as demonstrated by the numerical experiments.

97 MATHEMATICS AND COMPUTING↗

Portable programming on parallel/networked computers using the Application Portable Parallel Library (APPL)

The Application Portable Parallel Library (APPL) is a subroutine-based library of communication primitives that is callable from applications written in FORTRAN or C. APPL provides a consistent programmer interface to a variety of distributed and shared-memory multiprocessor MIMD machines. The objective of APPL is to minimize the effort required to move parallel applications from one machine to another, or to a network of homogeneous machines. APPL encompasses many of the message-passing primitives that are currently available on commercial multiprocessor systems. This paper describes APPL (version 2.3.1) and its usage, reports the status of the APPL project, and indicates possible directions for the future. Several applications using APPL are discussed, as well as performance and overhead results.

Quealy, Angela↗

Modeling the asynchronous Jacobi method without communication delays

Asynchronous iterative methods for solving linear systems are gaining renewed interest due to the high cost of synchronization points in massively parallel codes. Historically, theory on asynchronous iterative methods has focused on asymptotic behavior, while the transient behavior remains poorly understood. In this paper, we study a model of the asynchronous Jacobi method without communication delays, which we call simplified asynchronous Jacobi. Simplified asynchronous Jacobi can be used to model asynchronous Jacobi implemented in shared memory or distributed memory with fast communication networks. Our analysis uses the idea of a propagation matrix, which is similar in concept to an iteration matrix. We show that simplified asynchronous Jacobi can continue to reduce the residual when some processes are slower than other processes. We also show that simplified asynchronous Jacobi can converge when synchronous Jacobi does not. We verify our analysis of simplified asynchronous Jacobi using results from asynchronous Jacobi implemented in shared and distributed memory.

97 MATHEMATICS AND COMPUTING↗

Parallel Optical Flow Processing of 4D Cardiac CT Data on Multicore Clusters, In: 2014 IEEE 17th International Conference on Computational Science and Engineering

Optical flow is the distribution of apparent velocities of movement of brightness patterns in a sequence of images. For large 3D image sequences, optical flow applications are time consuming and memory-bound. To cope with these problems, in this paper, we present parallel optical flow processing of 4D cardiac CT data on multicore cluster systems to significantly shorten the time for computing velocity fields of the heart in order to aid cardiologists in diagnosing heart disease such as myocardial infarction and cardiac dysrhythmia in time. First, we modify and extend two traditional 2D optical flow methods Horn-Schunck and Lucas-Kanade to three-dimensional cases to process the 4D cardiac CT data. Second, we extend Mat lab MPI to support parallel computing with Mat lab and Octave on these cluster systems. Then we develop the parallel Mat lab/Octave optical flow applications for the 4D cardiac CT data in detail. Our experimental results show that these parallel optical flow applications have good scalability with close to linear speedup, and are able to shorten the image processing time significantly from more than 5 hours on 4 cores to 1.5 minutes on 1024 cores.

Engineering↗

User-Defined Data Distributions in High-Level Programming Languages

One of the characteristic features of today s high performance computing systems is a physically distributed memory. Efficient management of locality is essential for meeting key performance requirements for these architectures. The standard technique for dealing with this issue has involved the extension of traditional sequential programming languages with explicit message passing, in the context of a processor-centric view of parallel computation. This has resulted in complex and error-prone assembly-style codes in which algorithms and communication are inextricably interwoven. This paper presents a high-level approach to the design and implementation of data distributions. Our work is motivated by the need to improve the current parallel programming methodology by introducing a paradigm supporting the development of efficient and reusable parallel code. This approach is currently being implemented in the context of a new programming language called Chapel, which is designed in the HPCS project Cascade.

physically distributed memory↗

Tuning collective communication for Partitioned Global Address Space programming models

Partitioned Global Address Space (PGAS) languages offer programmers the convenience of a shared memory programming style combined with locality control necessary to run on large-scale distributed memory systems. Even within a PGAS language programmers often need to perform global communication operations such as broadcasts or reductions, which are best performed as collective operations in which a group of threads work together to perform the operation. In this study we consider the problem of implementing collective communication within PGAS languages and explore some of the design trade-offs in both the interface and implementation. In particular, PGAS collectives have semantic issues that are different than in send–receive style message passing programs, and different implementation approaches that take advantage of the one-sided communication style in these languages. We present an implementation framework for PGAS collectives as part of the GASNet communication layer, which supports shared memory, distributed memory and hybrids. The framework supports a broad set of algorithms for each collective, over which the implementation may be automatically tuned. In conclusion, we demonstrate the benefit of optimized GASNet collectives using application benchmarks written in UPC, and demonstrate that the GASNet collectives can deliver scalable performance on a variety of state-of-the-art parallel machines including a Cray XT4, an IBM BlueGene/P, and a Sun Constellation system with InfiniBand interconnect.

97 MATHEMATICS AND COMPUTING↗

Petascale direct numerical simulation of blood flow on 200K cores and heterogeneous architectures

We present a fast, petaflop-scalable algorithm for Stokesian particulate flows. Our goal is the direct simulation of blood, which we model as a mixture of a Stokesian fluid (plasma) and red blood cells (RBCs). Directly simulating blood is a challenging multiscale, multiphysics problem. We report simulations with up to 260 million deformable RBCs. The largest simulation amounts to 90 billion unknowns in space. In terms of the number of cells, we improve the state-of-the art by several orders of magnitude: the previous largest simulation, at the same physical fidelity as ours, resolved the flow of O(1,000-10,000) RBCs. Our approach has three distinct characteristics: (1) we faithfully represent the physics of RBCs by using nonlinear solid mechanics to capture the deformations of each cell; (2) we accurately resolve the long-range, N-body, hydrodynamic interactions between RBCs (which are caused by the surrounding plasma); and (3) we allow for highly non-uniform spatial distributions of RBCs. The new method has been implemented in the software library MOBO (for 'Moving Boundaries'). We designed MOBO to support parallelism at all levels, including inter-node distributed memory parallelism, intra-node shared memory parallelism, data parallelism (vectorization), and fine-grained multithreading for GPUs. We have implemented and optimized the majority of the computation kernels on both Intel/AMD x86 and NVidia's Tesla/Fermi platforms for single and double floating point precision. Overall, the code has scaled on 256 CPU-GPUs on the Teragrid's Lincoln cluster and on 200,000 AMD cores of the Oak Ridge National Laboratory's Jaguar PF system. In our largest simulation, we have achieved 0.7 Petaflops/s of sustained performance on Jaguar.

59 BASIC BIOLOGICAL SCIENCES↗

FitCache: A Transparent Drop-In Framework for Multi-Tier Caching to Accelerate Distributed Deep Learning Workloads

Training in Deep learning (DL) remains highly compute- and data-intensive, with I/O becoming a critical bottleneck as models and datasets scale. Recent studies report that data loading can dominate training time, especially on large-scale HPC systems with shared parallel file systems (PFS). Existing caching approaches either rely on single-tier designs or require intrusive modifications to training pipelines, limiting their portability and effectiveness. In this work, we present FitCache, a transparent drop-in framework for multi-tier caching to accelerate distributed DL training by coordinating fast local memory (e.g., DRAM, Persistent Memory (PMem)) and NVMe as hierarchical caches atop PFS. Our design adapts to hardware diversity, i.e., if NVMe is missing, memory transparently acts as a caching tier, ensuring stable performance. FitCache transparently intercepts I/O requests and issues concurrent fetches across all tiers, returning data from the fastest responder without centralized metadata or static redirection paths. FitCache adapts to dynamic workloads and heterogeneous clusters while maintaining POSIX compatibility. Experiments on Frontier (2048 GPUs) and smaller research clusters show that FitCache reduces training time by up to 40% and per-batch I/O latency by up to 71.6% compared to Lustre Orion PFS, offering a drop-in solution for scalable DL training.

Hu, Guangxing [ORNL] (ORCID:0009000283203614)↗