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 145 records · Page 8

Center for Programming Models for Scalable Parallel Computing - Towards Enhancing OpenMP for Manycore and Heterogeneous Nodes (Final Report)

OpenMP was not well recognized at the beginning of the project, around year 2003, because of its limited use in DoE production applications and the inmature hardware support for an efficient implementation. Yet in the recent years, it has been graduately adopted both in HPC applications, mostly in the form of MPI+OpenMP hybrid code, and in mid-scale desktop applications for scientific and experimental studies. We have observed this trend and worked deligiently to improve our OpenMP compiler and runtimes, as well as to work with the OpenMP standard organization to make sure OpenMP are evolved in the direction close to DoE missions. In the Center for Programming Models for Scalable Parallel Computing project, the HPCTools team at the University of Houston (UH), directed by Dr. Barbara Chapman, has been working with project partners, external collaborators and hardware vendors to increase the scalability and applicability of OpenMP for multi-core (and future manycore) platforms and for distributed memory systems by exploring different programming models, language extensions, compiler optimizations, as well as runtime library support.

97 MATHEMATICS AND COMPUTING↗

An OpenACC-Based Unified Programming Model for Multi-accelerator Systems

This paper proposes a novel SPMD programming model of OpenACC. Our model integrates the different granularities of parallelism from vector-level parallelism to node-level parallelism into a single, unified model based on OpenACC. It allows programmers to write programs for multiple accelerators using a uniform programming model whether they are in shared or distributed memory systems. We implement a prototype of our model and evaluate its performance with a GPU-based supercomputer using three benchmark applications.

Accelerators↗

An integrated runtime and compile-time approach for parallelizing structured and block structured applications

Scientific and engineering applications often involve structured meshes. These meshes may be nested (for multigrid codes) and/or irregularly coupled (called multiblock or irregularly coupled regular mesh problems). A combined runtime and compile-time approach for parallelizing these applications on distributed memory parallel machines in an efficient and machine-independent fashion was described. A runtime library which can be used to port these applications on distributed memory machines was designed and implemented. The library is currently implemented on several different systems. To further ease the task of application programmers, methods were developed for integrating this runtime library with compilers for HPK-like parallel programming languages. How this runtime library was integrated with the Fortran 90D compiler being developed at Syracuse University is discussed. Experimental results to demonstrate the efficacy of our approach are presented. A multiblock Navier-Stokes solver template and a multigrid code were experimented with. Our experimental results show that our primitives have low runtime communication overheads. Further, the compiler parallelized codes perform within 20 percent of the code parallelized by manually inserting calls to the runtime library.

Agrawal, Gagan↗

Parallel Programming in MCNP6

Monte Carlo N-Particle (MCNP)1 is a general-purpose Monte Carlo particle transport code developed by Los Alamos National Laboratory (LANL). To efficiently handle long simulations, MCNP version 6 (MCNP6) supports parallel execution using two primary programming models: • Shared-memory task-based threading using OpenMP (Open Multi-Processing), and • Distributed-memory calculations using MPI (Message Passing Interface). The OpenMP and MPI programming models enable MCNP6 to scale from desktop systems to high-performance computing (HPC) clusters, allowing users to run MCNP in one of three parallel modes: • OpenMP-only, • MPI-only, and • Hybrid (MPI + OpenMP). The choice of parallelization mode depends on the underlying computer architecture and the characteristics of the simulation problem.

97 MATHEMATICS AND COMPUTING↗

Parallel-vector algorithms for particle simulations on shared-memory multiprocessors

Over the last few decades, the computational demands of massive particle-based simulations for both scientific and industrial purposes have been continuously increasing. Hence, considerable efforts are being made to develop parallel computing techniques on various platforms. In such simulations, particles freely move within a given space, and so on a distributed-memory system, load balancing, i.e., assigning an equal number of particles to each processor, is not guaranteed. However, shared-memory systems achieve better load balancing for particle models, but suffer from the intrinsic drawback of memory access competition, particularly during (1) paring of contact candidates from among neighboring particles and (2) force summation for each particle. Here, novel algorithms are proposed to overcome these two problems. For the first problem, the key is a pre-conditioning process during which particle labels are sorted by a cell label in the domain to which the particles belong. Then, a list of contact candidates is constructed by pairing the sorted particle labels. For the latter problem, a table comprising the list indexes of the contact candidate pairs is created and used to sum the contact forces acting on each particle for all contacts according to Newton's third law. With just these methods, memory access competition is avoided without additional redundant procedures. The parallel efficiency and compatibility of these two algorithms were evaluated in discrete element method (DEM) simulations on four types of shared-memory parallel computers: a multicore multiprocessor computer, scalar supercomputer, vector supercomputer, and graphics processing unit. The computational efficiency of a DEM code was found to be drastically improved with our algorithms on all but the scalar supercomputer. Thus, the developed parallel algorithms are useful on shared-memory parallel computers with sufficient memory bandwidth.

97 MATHEMATICS AND COMPUTING↗

Dynamic object management for distributed data structures

In distributed-memory multiprocessors, remote memory accesses incur larger delays than local accesses. Hence, insightful allocation and access of distributed data can yield substantial performance gains. The authors argue for the use of dynamic data management policies encapsulated within individual distributed data structures. Distributed data structures offer performance, flexibility, abstraction, and system independence. This approach is supported by data from a trace-driven simulation study of parallel scientific benchmarks. Experimental data on memory locality, message count, message volume, and communication delay suggest that data-structure-specific data management is superior to a single, system-imposed policy.

Totty, Brian K.↗

HydraGNN v3.0

New or improved capabilities included in v3.0 release are as follows: 1. Enhancement in message passing layers through generalization of the class inheritance to enable the inclusion of a broader set of message passing policies Inclusion of equivariant message passing layers from the original implementations of: SchNet (https://pubs.aip.org/aip/jcp/article/148/24/241722/962591/SchNet-A-deep-learning-architecture-for-molecules); DimeNet++ (https://arxiv.org/abs/2011.14115); EGNN models (https://arxiv.org/pdf/2102.09844.pdf) 2. Restructuring of class inheritance for data management 3. Support of DDStore https://github.com/ORNL/DDStore capabilities for improved distributed data parallelism on large volumes of data that cannot fit on intra-node memory capacities 4. Large-scale system support for OLCF-Crusher and OLCF-Frontier

Lupo Pasini, Massimiliano [Oak Ridge National Labo↗

Efficient Support for Matrix Computations on Heterogeneous Multi-core and Multi-GPU Architectures

We present a new methodology for utilizing all CPU cores and all GPUs on a heterogeneous multicore and multi-GPU system to support matrix computations e ciently. Our approach is able to achieve the objectives of a high degree of parallelism, minimized synchronization, minimized communication, and load balancing. Our main idea is to treat the heterogeneous system as a distributed-memory machine, and to use a heterogeneous 1-D block cyclic distribution to allocate data to the host system and GPUs to minimize communication. We have designed heterogeneous algorithms with two di erent tile sizes (one for CPU cores and the other for GPUs) to cope with processor heterogeneity. We propose an auto-tuning method to determine the best tile sizes to attain both high performance and load balancing. We have also implemented a new runtime system and applied it to the Cholesky and QR factorizations. Our experiments on a compute node with two Intel Westmere hexa-core CPUs and three Nvidia Fermi GPUs demonstrate good weak scalability, strong scalability, load balance, and e ciency of our approach.

97 MATHEMATICS AND COMPUTING↗

Experiences using OpenMP based on Computer Directed Software DSM on a PC Cluster

In this work we report on our experiences running OpenMP programs on a commodity cluster of PCs running a software distributed shared memory (DSM) system. We describe our test environment and report on the performance of a subset of the NAS Parallel Benchmarks that have been automaticaly parallelized for OpenMP. We compare the performance of the OpenMP implementations with that of their message passing counterparts and discuss performance differences.

Hess, Matthias↗

A hybrid parallel framework for the cellular Potts model simulations

The Cellular Potts Model (CPM) has been widely used for biological simulations. However, most current implementations are either sequential or approximated, which can't be used for large scale complex 3D simulation. In this paper we present a hybrid parallel framework for CPM simulations. The time-consuming POE solving, cell division, and cell reaction operation are distributed to clusters using the Message Passing Interface (MPI). The Monte Carlo lattice update is parallelized on shared-memory SMP system using OpenMP. Because the Monte Carlo lattice update is much faster than the POE solving and SMP systems are more and more common, this hybrid approach achieves good performance and high accuracy at the same time. Based on the parallel Cellular Potts Model, we studied the avascular tumor growth using a multiscale model. The application and performance analysis show that the hybrid parallel framework is quite efficient. The hybrid parallel CPM can be used for the large scale simulation ({approx}10{sup 8} sites) of complex collective behavior of numerous cells ({approx}10{sup 6}).

97 MATHEMATICS AND COMPUTING↗

Locality-Aware Parallel Process Mapping for Multi-Core HPC Systems

High Performance Computing (HPC) systems are composed of servers containing an ever-increasing number of cores. With such high processor core counts, non-uniform memory access (NUMA) architectures are almost universally used to reduce inter-processor and memory communication bottlenecks by distributing processors and memory throughout a server-internal networking topology. Application studies have shown that the tuning of processes placement in a server s NUMA networking topology to the application can have a dramatic impact on performance. The performance implications are magnified when running a parallel job across multiple server nodes, especially with large scale HPC applications. This paper presents the Locality-Aware Mapping Algorithm (LAMA) for distributing the individual processes of a parallel application across processing resources in an HPC system, paying particular attention to the internal server NUMA topologies. The algorithm is able to support both homogeneous and heterogeneous hardware systems, and dynamically adapts to the available hardware and user-specified process layout at run-time. As implemented in Open MPI, the LAMA provides 362,880 mapping permutations and is able to naturally scale out to additional hardware resources as they become available in future architectures.

99 GENERAL AND MISCELLANEOUS↗

Implementing Connected Component Labeling as a User Defined Operator for SciDB

We have implemented a flexible User Defined Operator (UDO) for labeling connected components of a binary mask expressed as an array in SciDB, a parallel distributed database management system based on the array data model. This UDO is able to process very large multidimensional arrays by exploiting SciDB's memory management mechanism that efficiently manipulates arrays whose memory requirements far exceed available physical memory. The UDO takes as primary inputs a binary mask array and a binary stencil array that specifies the connectivity of a given cell to its neighbors. The UDO returns an array of the same shape as the input mask array with each foreground cell containing the label of the component it belongs to. By default, dimensions are treated as non-periodic, but the UDO also accepts optional input parameters to specify periodicity in any of the array dimensions. The UDO requires four stages to completely label connected components. In the first stage, labels are computed for each subarray or chunk of the mask array in parallel across SciDB instances using the weighted quick union (WQU) with half-path compression algorithm. In the second stage, labels around chunk boundaries from the first stage are stored in a temporary SciDB array that is then replicated across all SciDB instances. Equivalences are resolved by again applying the WQU algorithm to these boundary labels. In the third stage, relabeling is done for each chunk using the resolved equivalences. In the fourth stage, the resolved labels, which so far are "flattened" coordinates of the original binary mask array, are renamed with sequential integers for legibility. The UDO is demonstrated on a 3-D mask of O(1011) elements, with O(108) foreground cells and O(106) connected components. The operator completes in 19 minutes using 84 SciDB instances.

UDO↗

Electronic Neural Networks

Memory based on neural network models content-addressable and fault-tolerant. System includes electronic equivalent of synaptic network; particular, matrix of programmable binary switching elements over which data distributed. Switches programmed in parallel by outputs of serial-input/parallel-output shift registers. Input and output terminals of bank of high-gain nonlinear amplifiers connected in nonlinear-feedback configuration by switches and by memory-prompting shift registers.

Lambe, John↗

Sparse Linear Algebra Toolkit for Computational Aerodynamics

Finding solutions to sparse linear systems of equations is an essential step in Computational Engineering applications of interest to NASA. Linear systems of equations are composed and solved in almost every computational engineering application. The characteristics of linear systems vary greatly from one application to another. Accordingly, there are a wide variety of methods for the solution of linear systems of equations. The operations and methods prepared by the authors are focused on linear systems of interest to NASA, primarily those associated with Computational Fluid Dynamics (CFD), Aeroelasticity, and Aeroacoustics. The Sparse Linear Algebra Toolkit (SLAT) is a coordinated collection of software featuring operations, methods, and data structures that are useful when solving sparse linear systems of equations on modern computer architectures. The implemented operations and methods are designed and tuned for parallelism in shared memory, in distributed memory, and across the hybrid combination of distributed-shared memory. The toolkit includes novel methods and implementations for modern architectures and facilitates development of new approaches for meeting NASA’s evolving computational engineering challenges using evolving computer architectures that are not available in vendor libraries. In this paper, significant features and interfaces within SLAT are presented and verified for simulations performed with NASA’s CFD solver, FUN3D. The runtime and scaling performance of the Generalized Minimum Residual (GMRES) method implemented in SLAT is analyzed for the linear subproblems within the solution of turbulent Navier-Stokes equations employed in the simulation of high-lift configurations. Prior to this work, the SPARSKIT GMRES implementation was the only Krylov subspace method available within FUN3D. A strong scaling study shows the SLAT GMRES implementation facilitates accurate Reynolds-averaged Navier-Stokes CFD solutions between 15% and 56% faster than the SPARSKIT GMRES implementation.

Stephen L Wood↗

symPACK: A GPU-Capable Fan-Out Sparse Cholesky Solver

Sparse symmetric positive definite systems of equations are ubiquitous in scientific workloads and applications. Parallel sparse Cholesky factorization is the method of choice for solving such linear systems. Therefore, the development of parallel sparse Cholesky codes that can efficiently run on today’s large-scale heterogeneous distributed-memory platforms is of vital importance. Modern supercomputers offer nodes that contain a mix of CPUs and GPUs. To fully utilize the computing power of these nodes, scientific codes must be adapted to offload expensive computations to GPUs. We present symPACK, a GPU-capable parallel sparse Cholesky solver that uses one-sided communication primitives and remote procedure calls provided by the UPC++ library. We also utilize the UPC++ "memory kinds" feature to enable efficient communication of GPU-resident data. We show that on a number of large problems, symPACK outperforms comparable state-of-the-art GPU-capable Cholesky factorization codes by up to 14x on the NERSC Perlmutter supercomputer.

Bellavita, Julian↗

Efficacy of Code Optimization on Cache-based Processors

The current common wisdom in the U.S. is that the powerful, cost-effective supercomputers of tomorrow will be based on commodity (RISC) micro-processors with cache memories. Already, most distributed systems in the world use such hardware as building blocks. This shift away from vector supercomputers and towards cache-based systems has brought about a change in programming paradigm, even when ignoring issues of parallelism. Vector machines require inner-loop independence and regular, non-pathological memory strides (usually this means: non-power-of-two strides) to allow efficient vectorization of array operations. Cache-based systems require spatial and temporal locality of data, so that data once read from main memory and stored in high-speed cache memory is used optimally before being written back to main memory. This means that the most cache-friendly array operations are those that feature zero or unit stride, so that each unit of data read from main memory (a cache line) contains information for the next iteration in the loop. Moreover, loops ought to be 'fat', meaning that as many operations as possible are performed on cache data-provided instruction caches do not overflow and enough registers are available. If unit stride is not possible, for example because of some data dependency, then care must be taken to avoid pathological strides, just ads on vector computers. For cache-based systems the issues are more complex, due to the effects of associativity and of non-unit block (cache line) size. But there is more to the story. Most modern micro-processors are superscalar, which means that they can issue several (arithmetic) instructions per clock cycle, provided that there are enough independent instructions in the loop body. This is another argument for providing fat loop bodies. With these restrictions, it appears fairly straightforward to produce code that will run efficiently on any cache-based system. It can be argued that although some of the important computational algorithms employed at NASA Ames require different programming styles on vector machines and cache-based machines, respectively, neither architecture class appeared to be favored by particular algorithms in principle. Practice tells us that the situation is more complicated. This report presents observations and some analysis of performance tuning for cache-based systems. We point out several counterintuitive results that serve as a cautionary reminder that memory accesses are not the only factors that determine performance, and that within the class of cache-based systems, significant differences exist.

VanderWijngaart, Rob F.↗

Modular Spiking Neural Circuits for Mapping Long Short-Term Memory on a Neurosynaptic Processor

Due to the distributed and asynchronous nature of neural computation through low-energy spikes, brain-inspired hardware systems offer high energy efficiency and massive parallelism. One such platform is the IBM TrueNorth neurosynaptic system. Recently, TrueNorth compatible representation learning algorithms have emerged, achieving close to the state-of-the-art performance in various data sets. However, its application in temporal sequence processing models, such as recurrent neural networks (RNNs), is still only at the proof of concept level. There is an inherent difficulty in capturing temporal dynamics of an RNN using spiking neurons, which is only exasperated by the hardware constraints in connectivity and synaptic weight resolution. This paper presents a design flow that overcomes these difficulties and maps a special case of recurrent networks called long short-term memory (LSTM) onto a spike-based platform. The framework utilizes various approximation techniques, such as activation discretization, weight quantization, and scaling and rounding, spiking neural circuits that implement the complex gating mechanisms, and a store-and-release technique to enable neuron synchronization and faithful storage. While the presented techniques can be applied to map LSTM to any spiking neural network (SNN) simulator/emulator, here we choose the TrueNorth chip as the target platform by adhering to its hardware constraints. Three LSTM applications, parity check, extended Reber grammar, and question classification, are evaluated. The tradeoffs among accuracy, performance, and energy tradeoffs achieved on TrueNorth are demonstrated. Finally, this is compared with the performance on an SNN platform without hardware constraints, which represents the upper bound of the achievable accuracy.

97 MATHEMATICS AND COMPUTING↗

Performance Measurement, Visualization and Modeling of Parallel and Distributed Programs

This paper presents a methodology for debugging the performance of message-passing programs on both tightly coupled and loosely coupled distributed-memory machines. The AIMS (Automated Instrumentation and Monitoring System) toolkit, a suite of software tools for measurement and analysis of performance, is introduced and its application illustrated using several benchmark programs drawn from the field of computational fluid dynamics. AIMS includes (i) Xinstrument, a powerful source-code instrumentor, which supports both Fortran77 and C as well as a number of different message-passing libraries including Intel's NX Thinking Machines' CMMD, and PVM; (ii) Monitor, a library of timestamping and trace -collection routines that run on supercomputers (such as Intel's iPSC/860, Delta, and Paragon and Thinking Machines' CM5) as well as on networks of workstations (including Convex Cluster and SparcStations connected by a LAN); (iii) Visualization Kernel, a trace-animation facility that supports source-code clickback, simultaneous visualization of computation and communication patterns, as well as analysis of data movements; (iv) Statistics Kernel, an advanced profiling facility, that associates a variety of performance data with various syntactic components of a parallel program; (v) Index Kernel, a diagnostic tool that helps pinpoint performance bottlenecks through the use of abstract indices; (vi) Modeling Kernel, a facility for automated modeling of message-passing programs that supports both simulation -based and analytical approaches to performance prediction and scalability analysis; (vii) Intrusion Compensator, a utility for recovering true performance from observed performance by removing the overheads of monitoring and their effects on the communication pattern of the program; and (viii) Compatibility Tools, that convert AIMS-generated traces into formats used by other performance-visualization tools, such as ParaGraph, Pablo, and certain AVS/Explorer modules.

Yan, Jerry C.↗