Search NASA⌕ Search

SEARCH · Search NASA

Results for “Distributed memory”

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 811 records · Page 45

Effects of Ordering Strategies and Programming Paradigms on Sparse Matrix Computations

The Conjugate Gradient (CG) algorithm is perhaps the best-known iterative technique to solve sparse linear systems that are symmetric and positive definite. For systems that are ill-conditioned, it is often necessary to use a preconditioning technique. In this paper, we investigate the effects of various ordering and partitioning strategies on the performance of parallel CG and ILU(O) preconditioned CG (PCG) using different programming paradigms and architectures. Results show that for this class of applications: ordering significantly improves overall performance on both distributed and distributed shared-memory systems, that cache reuse may be more important than reducing communication, that it is possible to achieve message-passing performance using shared-memory constructs through careful data ordering and distribution, and that a hybrid MPI+OpenMP paradigm increases programming complexity with little performance gains. A implementation of CG on the Cray MTA does not require special ordering or partitioning to obtain high efficiency and scalability, giving it a distinct advantage for adaptive applications; however, it shows limited scalability for PCG due to a lack of thread level parallelism.

Oliker, Leonid↗

Holographic optical interconnects for VLSI

This paper introduces new applications and design trade-offs anticipated for free-space optical interconnections of VLSI chips. New implementation s of VLSI functions are described that use the capability of making optical inputs at any point on a chip and take advantage of greater flexibility in on-chip signal routing. These include n-port addressable memories, CPU clock phase distribution, hardware multipliers, and dynamic memory refresh, as well as enhanced testability. Fault tolerance and production yields may be improved by reprogramming the optical imaging system to circumvent defective elements. These attributes, as well as those related to performance alone, will affect the design methodology of future VLSI ICs. This paper focuses on identifying the design issues, their possible solutions, and their impact on VLSI design techniques and, finally, presents some preliminary measurements on various system components.

Bergman, L. A.↗

Performance Assessment of OVERFLOW on Distributed Computing Environment

The aerodynamic computer code, OVERFLOW, with a multi-zone overset grid feature, has been parallelized to enhance its performance on distributed and shared memory paradigms. Practical application benchmarks have been set to assess the efficiency of code's parallelism on high-performance architectures. The code's performance has also been experimented with in the context of the distributed computing paradigm on distant computer resources using the Information Power Grid (IPG) toolkit, Globus. Two parallel versions of the code, namely OVERFLOW-MPI and -MLP, have developed around the natural coarse grained parallelism inherent in a multi-zonal domain decomposition paradigm. The algorithm invokes a strategy that forms a number of groups, each consisting of a zone, a cluster of zones and/or a partition of a large zone. Each group can be thought of as a process with one or multithreads assigned to it and that all groups run in parallel. The -MPI version of the code uses explicit message-passing based on the standard MPI library for sending and receiving interzonal boundary data across processors. The -MLP version employs no message-passing paradigm; the boundary data is transferred through the shared memory. The -MPI code is suited for both distributed and shared memory architectures, while the -MLP code can only be used on shared memory platforms. The IPG applications are implemented by the -MPI code using the Globus toolkit. While a computational task is distributed across multiple computer resources, the parallelism can be explored on each resource alone. Performance studies are achieved with some practical aerodynamic problems with complex geometries, consisting of 2.5 up to 33 million grid points and a large number of zonal blocks. The computations were executed primarily on SGI Origin 2000 multiprocessors and on the Cray T3E. OVERFLOW's IPG applications are carried out on NASA homogeneous metacomputing machines located at three sites, Ames, Langley and Glenn. Plans for the future will exploit the distributed parallel computing capability on various homogeneous and heterogeneous resources and large scale benchmarks. Alternative IPG toolkits will be used along with sophisticated zonal grouping strategies to minimize the communication time across the computer resources.

Djomehri, M. Jahed↗

NavP: Structured and Multithreaded Distributed Parallel Programming

We present Navigational Programming (NavP) -- a distributed parallel programming methodology based on the principles of migrating computations and multithreading. The four major steps of NavP are: (1) Distribute the data using the data communication pattern in a given algorithm; (2) Insert navigational commands for the computation to migrate and follow large-sized distributed data; (3) Cut the sequential migrating thread and construct a mobile pipeline; and (4) Loop back for refinement. NavP is significantly different from the current prevailing Message Passing (MP) approach. The advantages of NavP include: (1) NavP is structured distributed programming and it does not change the code structure of an original algorithm. This is in sharp contrast to MP as MP implementations in general do not resemble the original sequential code; (2) NavP implementations are always competitive with the best MPI implementations in terms of performance. Approaches such as DSM or HPF have failed to deliver satisfying performance as of today in contrast, even if they are relatively easy to use compared to MP; (3) NavP provides incremental parallelization, which is beyond the reach of MP; and (4) NavP is a unifying approach that allows us to exploit both fine- (multithreading on shared memory) and coarse- (pipelined tasks on distributed memory) grained parallelism. This is in contrast to the currently popular hybrid use of MP+OpenMP, which is known to be complex to use. We present experimental results that demonstrate the effectiveness of NavP.

navigational programming (NavP)↗

Distributed trace using central performance counter memory

A plurality of processing cores, are central storage unit having at least memory connected in a daisy chain manner, forming a daisy chain ring layout on an integrated chip. At least one of the plurality of processing cores places trace data on the daisy chain connection for transmitting the trace data to the central storage unit, and the central storage unit detects the trace data and stores the trace data in the memory co-located in with the central storage unit.

97 MATHEMATICS AND COMPUTING↗

Distributed trace using central performance counter memory

A plurality of processing cores, are central storage unit having at least memory connected in a daisy chain manner, forming a daisy chain ring layout on an integrated chip. At least one of the plurality of processing cores places trace data on the daisy chain connection for transmitting the trace data to the central storage unit, and the central storage unit detects the trace data and stores the trace data in the memory co-located in with the central storage unit.

97 MATHEMATICS AND COMPUTING↗

Distributed state-space generation of discrete-state stochastic models

High-level formalisms such as stochastic Petri nets can be used to model complex systems. Analysis of logical and numerical properties of these models of ten requires the generation and storage of the entire underlying state space. This imposes practical limitations on the types of systems which can be modeled. Because of the vast amount of memory consumed, we investigate distributed algorithms for the generation of state space graphs. The distributed construction allows us to take advantage of the combined memory readily available on a network of workstations. The key technical problem is to find effective methods for on-the-fly partitioning, so that the state space is evenly distributed among processors. In this paper we report on the implementation of a distributed state-space generator that may be linked to a number of existing system modeling tools. We discuss partitioning strategies in the context of Petri net models, and report on performance observed on a network of workstations, as well as on a distributed memory multi-computer.

Ciardo, Gianfranco↗

Distributed simulation using a real-time shared memory network

The Advanced Control Technology Branch of the NASA Lewis Research Center performs research in the area of advanced digital controls for aeronautic and space propulsion systems. This work requires the real-time implementation of both control software and complex dynamical models of the propulsion system. We are implementing these systems in a distributed, multi-vendor computer environment. Therefore, a need exists for real-time communication and synchronization between the distributed multi-vendor computers. A shared memory network is a potential solution which offers several advantages over other real-time communication approaches. A candidate shared memory network was tested for basic performance. The shared memory network was then used to implement a distributed simulation of a ramjet engine. The accuracy and execution time of the distributed simulation was measured and compared to the performance of the non-partitioned simulation. The ease of partitioning the simulation, the minimal time required to develop for communication between the processors and the resulting execution time all indicate that the shared memory network is a real-time communication technique worthy of serious consideration.

Simon, Donald L.↗

SEU/SRAM as a Process Monitor

The SEU/SRAM is a 4-kbit Static Random Access Memory (SRAM) designed to detect Single-Event Upsets (SEUs) produced by high energy particles. This device was used to determine the distribution in the memory cell spontaneous flip potential.

distribution residual plots↗

Distributed Quantum Computing with Photons and Atomic Memories

The promise of universal quantum computing requires scalable single- and inter-qubit control interactions. Currently, three of the leading candidate platforms for quantum computing are based on superconducting circuits, trapped ions, and neutral atom arrays. However, these systems have strong interaction with environmental and control noises that introduce decoherence of qubit states and gate operations. Alternatively, photons are well decoupled from the environment and have advantages of speed and timing for quantum computing. Photonic systems have already demonstrated capability for solving specific intractable problems like Boson sampling, but face challenges for practically scalable universal quantum computing solutions because it is extremely difficult for a single photon to “talk” to another deterministically. Here, a universal distributed quantum computing scheme based on photons and atomic-ensemble-based quantum memories is proposed. Taking the established photonic advantages, two-qubit nonlinear interaction is mediated by converting photonic qubits into quantum memory states and employing Rydberg blockade for the controlled gate operation. Spatial and temporal scalability of this scheme is demonstrated further. Furthermore, these results show photon-atom network hybrid approach can be a potential solution to universal distributed quantum computing.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Distributed Halide

Many image processing tasks are naturally expressed as a pipeline of small computational kernels known as stencils. Halide is a popular domain-specific language and compiler designed to implement image processing algorithms. Halide uses simple language constructs to express what to compute and a separate scheduling co-language for expressing when and where to perform the computation. This approach has demonstrated performance comparable to or better than hand-optimized code. Until now, however, Halide has been restricted to parallel shared memory execution, limiting its performance for memory-bandwidth-bound pipelines or large-scale image processing tasks. We present an extension to Halide to support distributed-memory parallel execution of complex stencil pipelines. These extensions compose with the existing scheduling constructs in Halide, allowing expression of complex computation and communication strategies. Existing Halide applications can be distributed with minimal changes, allowing programmers to explore the tradeoff between recomputation and communication with little effort. Approximately 10 new of lines code are needed even for a 200 line, 99 stage application. On nine image processing benchmarks, our extensions give up to a 1.4× speedup on a single node over regular multithreaded execution with the same number of cores, by mitigating the effects of non-uniform memory access. The distributed benchmarks achieve up to 18× speedup on a 16 node testing machine and up to 57× speedup on 64 nodes of the NERSC Cori supercomputer.

97 MATHEMATICS AND COMPUTING↗

Predictive Materials Design of Magnetic Random-Access Memory Based on Nanoscale Atomic Structure and Element Distribution

Magnetic tunnel junctions (MTJs) capable of electrical read and write operations have emerged as a canonical building block for non-volatile memory and logic. However, the cause of the wide-spread device properties found experimentally in various MTJ stacks, including tunneling magnetoresistance (TMR), perpendicular magnetic anisotropy (PMA), and voltage-controlled magnetic anisotropy (VCMA), remains elusive. Here, using high-resolution transmission electron microscopy and energy-dispersive X-ray spectroscopy, we found that the MTJ crystallization quality, boron diffusion out of the CoFeB fixed layer, and minimal CoFe oxidation correlate with the TMR. As with the CoFeB free layer, seed layer diffusion into the free layer/MgO interface is negatively correlated with the interfacial PMA, while the metal-oxides concentrations in the free layer correlate with the VCMA. Combined with formation enthalpy and thermal diffusion analysis, we further established a predictive materials design framework to guide the complex design space explorations for high-performance MTJs. In this paper, we demonstrate experimentally high PMA and VCMA values of 1.74 mJ/m2 and 115 fJ/V-m with annealing stability above 400 °C.

36 MATERIALS SCIENCE↗

A Parallel Graph Environment for Real-World Data Analytics Workflows

Economic competitiveness and national security depend increasingly on the insightful analysis of large data sets. The diversity of real-world data sources and analytic workflows impose challenging hardware and software requirements for parallel graph platforms. The irregular nature of graph methods is not supported well by the deep memory hierarchies of conventional distributed systems, requiring new processor and runtime system designs to tolerate memory and synchronization latencies. Moreover, the efficiency of relational table operations and matrix computations are not attainable when data is stored in common graph data structures. In this paper, we present HAGGLE, a high-performance, scalable data analytics platform. The platform’s hybrid data model supports a variety of distributed, thread-safe data structures, parallel programming constructs, and persistent and streaming data. An abstract runtime layer enables us to map the stack to conventional, distributed computer systems with accelerators. The runtime uses multithreading, active messages, and data aggregation to hide memory and synchronization latencies on large-scale systems.

Graph Analytics↗

A Scalable Graph Analytics Framework for Programming with Big Data in R (pbdR)

Many disciplines such as biology, economics, engineering, physics, and the social sciences represent their data as graphs to capture patterns, trends, and associations. There are are many commercially available graph libraries in different programming languages to analyze these complex graphs. But there is no distributed graph library package in R - the popular statistical programming language to analyze graphs that bigger than a single machine's memory. Many domain experts prefer R over the numerous other alternatives. Towards this, we present a distributed graph analytics framework for R called programming with big graph using R (pBGR.) Our proposed framework leverages the Programming with Big Data in R (pbdR) ecosystem that provides scalable R packages for distributed computing in data science. We present an early prototype implementation of this framework using the distributed-memory parallel graph library CombBLAS and evaluate the framework's performance on leadership class computing platforms. Our experimental results demonstrate that the proposed framework is capable of performing large-scale parallel graph mining through the easyto-use R language. This enhanced graph processing capability coupled with other statistical tools already available in R, should be valuable to many domain experts.

Hasan, S M Shamimul↗

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.↗

Scalable Work Stealing

Irregular and dynamic parallel applications pose significant challenges to achieving scalable performance on large-scale multicore clusters. These applications often require ongoing, dynamic load balancing in order to maintain efficiency. While effective at small scale, centralized load balancing schemes quickly become a bottleneck on large-scale clusters. Work stealing is a popular approach to distributed dynamic load balancing; however its performance on large-scale clusters is not well understood. Prior work on work stealing has largely focused on shared memory machines. In this work we investigate the design and scalability of work stealing on modern distributed memory systems. We demonstrate high efficiency and low overhead when scaling to 8,192 processors for three benchmark codes: a producer-consumer benchmark, the unbalanced tree search benchmark, and a multiresolution analysis kernel.

99 GENERAL AND MISCELLANEOUS↗

High Performance Data Analytics on a Commodity Private Cloud

Sandia is researching novel methods of analysis for large datasets. The term “Big Data” has been used frequently in recent years to describe a discipline of computer science that attempts to distil an increasing quantity of data into a result or meaningful subset that can be easily visualized and acted upon by human consumers. These data are frequently characterized by one or more of the following: high volume, velocity, variety, and veracity. Sandia scientists and researchers are no strangers to large datasets, which often typify the result of high-performance computing workloads. The challenge for Sandia data scientists, however, comes at the arrival of exascale computing, which has become mandated by executive order for national computing centers. Because of power and networking limitations, exascale computing precludes the movement of datasets at exascale from data production to data analysis. Such analysis must be done in-situ, to mitigate the time and power costs of network transport. Currently, many of Sandia’s data analytics environments exist as independent entities, often specialized for individual use cases, interconnected to other data analytics, high-performance computing, and storage systems. Sandia is procuring, installing, and researching future platforms for such data analytics. A memory footprint of 256 GB per node, fourteen data rate (FDR) InfiniBand networking, PCIe connected non-volatile memory (NVMe) data storage, and an object data store will characterize the coming institutional data analytics platforms. The combined systems will contain 300 compute nodes. A cloud-inspired software stack resulting from collaboration between the Sandia centers will allow the systems to support a flexible mix of user and system-facing workloads and use cases including code development and testing in R and Python, Apache Hadoop and Spark, distributed in-memory databases, graph analytics, emulytics, system log and predictive analytics, and data visualization for institutional and mission customers. These systems will arrive at the end of FY 2015 and production is planned by the spring of FY 2016.

97 MATHEMATICS AND COMPUTING↗