Scheduling for Locality in Shared-Memory Multiprocessors

DTIC Science & Technology

1993-05-01

Submitted in Partial Fulfillment of the Requirements for the Degree ’)iIC Q(JALfryT INSPECTED 5 DOCTOR OF PHILOSOPHY I Accesion For Supervised by NTIS CRAM... architecture on parallel program performance, explain the implications of this trend on popular parallel programming models, and propose system software to 0...decomoosition and scheduling algorithms. I. SUIUECT TERMS IS. NUMBER OF PAGES shared-memory multiprocessors; architecture trends; loop 110 scheduling

Kokkos: Enabling manycore performance portability through polymorphic memory access patterns

DOE PAGES

Carter Edwards, H.; Trott, Christian R.; Sunderland, Daniel

2014-07-22

The manycore revolution can be characterized by increasing thread counts, decreasing memory per thread, and diversity of continually evolving manycore architectures. High performance computing (HPC) applications and libraries must exploit increasingly finer levels of parallelism within their codes to sustain scalability on these devices. We found that a major obstacle to performance portability is the diverse and conflicting set of constraints on memory access patterns across devices. Contemporary portable programming models address manycore parallelism (e.g., OpenMP, OpenACC, OpenCL) but fail to address memory access patterns. The Kokkos C++ library enables applications and domain libraries to achieve performance portability on diversemore » manycore architectures by unifying abstractions for both fine-grain data parallelism and memory access patterns. In this paper we describe Kokkos’ abstractions, summarize its application programmer interface (API), present performance results for unit-test kernels and mini-applications, and outline an incremental strategy for migrating legacy C++ codes to Kokkos. Furthermore, the Kokkos library is under active research and development to incorporate capabilities from new generations of manycore architectures, and to address a growing list of applications and domain libraries.« less

Real-time implementations of image segmentation algorithms on shared memory multicore architecture: a survey (Conference Presentation)

NASA Astrophysics Data System (ADS)

Akil, Mohamed

2017-05-01

The real-time processing is getting more and more important in many image processing applications. Image segmentation is one of the most fundamental tasks image analysis. As a consequence, many different approaches for image segmentation have been proposed. The watershed transform is a well-known image segmentation tool. The watershed transform is a very data intensive task. To achieve acceleration and obtain real-time processing of watershed algorithms, parallel architectures and programming models for multicore computing have been developed. This paper focuses on the survey of the approaches for parallel implementation of sequential watershed algorithms on multicore general purpose CPUs: homogeneous multicore processor with shared memory. To achieve an efficient parallel implementation, it's necessary to explore different strategies (parallelization/distribution/distributed scheduling) combined with different acceleration and optimization techniques to enhance parallelism. In this paper, we give a comparison of various parallelization of sequential watershed algorithms on shared memory multicore architecture. We analyze the performance measurements of each parallel implementation and the impact of the different sources of overhead on the performance of the parallel implementations. In this comparison study, we also discuss the advantages and disadvantages of the parallel programming models. Thus, we compare the OpenMP (an application programming interface for multi-Processing) with Ptheads (POSIX Threads) to illustrate the impact of each parallel programming model on the performance of the parallel implementations.

An Evaluation of Architectural Platforms for Parallel Navier-Stokes Computations

NASA Technical Reports Server (NTRS)

Jayasimha, D. N.; Hayder, M. E.; Pillay, S. K.

1996-01-01

We study the computational, communication, and scalability characteristics of a computational fluid dynamics application, which solves the time accurate flow field of a jet using the compressible Navier-Stokes equations, on a variety of parallel architecture platforms. The platforms chosen for this study are a cluster of workstations (the LACE experimental testbed at NASA Lewis), a shared memory multiprocessor (the Cray YMP), and distributed memory multiprocessors with different topologies - the IBM SP and the Cray T3D. We investigate the impact of various networks connecting the cluster of workstations on the performance of the application and the overheads induced by popular message passing libraries used for parallelization. The work also highlights the importance of matching the memory bandwidth to the processor speed for good single processor performance. By studying the performance of an application on a variety of architectures, we are able to point out the strengths and weaknesses of each of the example computing platforms.

Parallelizing Navier-Stokes Computations on a Variety of Architectural Platforms

NASA Technical Reports Server (NTRS)

Jayasimha, D. N.; Hayder, M. E.; Pillay, S. K.

1997-01-01

We study the computational, communication, and scalability characteristics of a Computational Fluid Dynamics application, which solves the time accurate flow field of a jet using the compressible Navier-Stokes equations, on a variety of parallel architectural platforms. The platforms chosen for this study are a cluster of workstations (the LACE experimental testbed at NASA Lewis), a shared memory multiprocessor (the Cray YMP), distributed memory multiprocessors with different topologies-the IBM SP and the Cray T3D. We investigate the impact of various networks, connecting the cluster of workstations, on the performance of the application and the overheads induced by popular message passing libraries used for parallelization. The work also highlights the importance of matching the memory bandwidth to the processor speed for good single processor performance. By studying the performance of an application on a variety of architectures, we are able to point out the strengths and weaknesses of each of the example computing platforms.

Cache write generate for parallel image processing on shared memory architectures.

PubMed

Wittenbrink, C M; Somani, A K; Chen, C H

1996-01-01

We investigate cache write generate, our cache mode invention. We demonstrate that for parallel image processing applications, the new mode improves main memory bandwidth, CPU efficiency, cache hits, and cache latency. We use register level simulations validated by the UW-Proteus system. Many memory, cache, and processor configurations are evaluated.

Programming parallel architectures: The BLAZE family of languages

NASA Technical Reports Server (NTRS)

Mehrotra, Piyush

1988-01-01

Programming multiprocessor architectures is a critical research issue. An overview is given of the various approaches to programming these architectures that are currently being explored. It is argued that two of these approaches, interactive programming environments and functional parallel languages, are particularly attractive since they remove much of the burden of exploiting parallel architectures from the user. Also described is recent work by the author in the design of parallel languages. Research on languages for both shared and nonshared memory multiprocessors is described, as well as the relations of this work to other current language research projects.

Efficient Parallelization of a Dynamic Unstructured Application on the Tera MTA

NASA Technical Reports Server (NTRS)

Oliker, Leonid; Biswas, Rupak

1999-01-01

The success of parallel computing in solving real-life computationally-intensive problems relies on their efficient mapping and execution on large-scale multiprocessor architectures. Many important applications are both unstructured and dynamic in nature, making their efficient parallel implementation a daunting task. This paper presents the parallelization of a dynamic unstructured mesh adaptation algorithm using three popular programming paradigms on three leading supercomputers. We examine an MPI message-passing implementation on the Cray T3E and the SGI Origin2OOO, a shared-memory implementation using cache coherent nonuniform memory access (CC-NUMA) of the Origin2OOO, and a multi-threaded version on the newly-released Tera Multi-threaded Architecture (MTA). We compare several critical factors of this parallel code development, including runtime, scalability, programmability, and memory overhead. Our overall results demonstrate that multi-threaded systems offer tremendous potential for quickly and efficiently solving some of the most challenging real-life problems on parallel computers.

A GaAs vector processor based on parallel RISC microprocessors

NASA Astrophysics Data System (ADS)

Misko, Tim A.; Rasset, Terry L.

A vector processor architecture based on the development of a 32-bit microprocessor using gallium arsenide (GaAs) technology has been developed. The McDonnell Douglas vector processor (MVP) will be fabricated completely from GaAs digital integrated circuits. The MVP architecture includes a vector memory of 1 megabyte, a parallel bus architecture with eight processing elements connected in parallel, and a control processor. The processing elements consist of a reduced instruction set CPU (RISC) with four floating-point coprocessor units and necessary memory interface functions. This architecture has been simulated for several benchmark programs including complex fast Fourier transform (FFT), complex inner product, trigonometric functions, and sort-merge routine. The results of this study indicate that the MVP can process a 1024-point complex FFT at a speed of 112 microsec (389 megaflops) while consuming approximately 618 W of power in a volume of approximately 0.1 ft-cubed.

The Tera Multithreaded Architecture and Unstructured Meshes

NASA Technical Reports Server (NTRS)

Bokhari, Shahid H.; Mavriplis, Dimitri J.

1998-01-01

The Tera Multithreaded Architecture (MTA) is a new parallel supercomputer currently being installed at San Diego Supercomputing Center (SDSC). This machine has an architecture quite different from contemporary parallel machines. The computational processor is a custom design and the machine uses hardware to support very fine grained multithreading. The main memory is shared, hardware randomized and flat. These features make the machine highly suited to the execution of unstructured mesh problems, which are difficult to parallelize on other architectures. We report the results of a study carried out during July-August 1998 to evaluate the execution of EUL3D, a code that solves the Euler equations on an unstructured mesh, on the 2 processor Tera MTA at SDSC. Our investigation shows that parallelization of an unstructured code is extremely easy on the Tera. We were able to get an existing parallel code (designed for a shared memory machine), running on the Tera by changing only the compiler directives. Furthermore, a serial version of this code was compiled to run in parallel on the Tera by judicious use of directives to invoke the "full/empty" tag bits of the machine to obtain synchronization. This version achieves 212 and 406 Mflop/s on one and two processors respectively, and requires no attention to partitioning or placement of data issues that would be of paramount importance in other parallel architectures.

A Parallel Saturation Algorithm on Shared Memory Architectures

NASA Technical Reports Server (NTRS)

Ezekiel, Jonathan; Siminiceanu

2007-01-01

Symbolic state-space generators are notoriously hard to parallelize. However, the Saturation algorithm implemented in the SMART verification tool differs from other sequential symbolic state-space generators in that it exploits the locality of ring events in asynchronous system models. This paper explores whether event locality can be utilized to efficiently parallelize Saturation on shared-memory architectures. Conceptually, we propose to parallelize the ring of events within a decision diagram node, which is technically realized via a thread pool. We discuss the challenges involved in our parallel design and conduct experimental studies on its prototypical implementation. On a dual-processor dual core PC, our studies show speed-ups for several example models, e.g., of up to 50% for a Kanban model, when compared to running our algorithm only on a single core.

The OpenMP Implementation of NAS Parallel Benchmarks and its Performance

NASA Technical Reports Server (NTRS)

Jin, Hao-Qiang; Frumkin, Michael; Yan, Jerry

1999-01-01

As the new ccNUMA architecture became popular in recent years, parallel programming with compiler directives on these machines has evolved to accommodate new needs. In this study, we examine the effectiveness of OpenMP directives for parallelizing the NAS Parallel Benchmarks. Implementation details will be discussed and performance will be compared with the MPI implementation. We have demonstrated that OpenMP can achieve very good results for parallelization on a shared memory system, but effective use of memory and cache is very important.

Parallel reduced-instruction-set-computer architecture for real-time symbolic pattern matching

NASA Astrophysics Data System (ADS)

Parson, Dale E.

1991-03-01

This report discusses ongoing work on a parallel reduced-instruction- set-computer (RISC) architecture for automatic production matching. The PRIOPS compiler takes advantage of the memoryless character of automatic processing by translating a program's collection of automatic production tests into an equivalent combinational circuit-a digital circuit without memory, whose outputs are immediate functions of its inputs. The circuit provides a highly parallel, fine-grain model of automatic matching. The compiler then maps the combinational circuit onto RISC hardware. The heart of the processor is an array of comparators capable of testing production conditions in parallel, Each comparator attaches to private memory that contains virtual circuit nodes-records of the current state of nodes and busses in the combinational circuit. All comparator memories hold identical information, allowing simultaneous update for a single changing circuit node and simultaneous retrieval of different circuit nodes by different comparators. Along with the comparator-based logic unit is a sequencer that determines the current combination of production-derived comparisons to try, based on the combined success and failure of previous combinations of comparisons. The memoryless nature of automatic matching allows the compiler to designate invariant memory addresses for virtual circuit nodes, and to generate the most effective sequences of comparison test combinations. The result is maximal utilization of parallel hardware, indicating speed increases and scalability beyond that found for course-grain, multiprocessor approaches to concurrent Rete matching. Future work will consider application of this RISC architecture to the standard (controlled) Rete algorithm, where search through memory dominates portions of matching.

Content-addressable read/write memories for image analysis

NASA Technical Reports Server (NTRS)

Snyder, W. E.; Savage, C. D.

1982-01-01

The commonly encountered image analysis problems of region labeling and clustering are found to be cases of search-and-rename problem which can be solved in parallel by a system architecture that is inherently suitable for VLSI implementation. This architecture is a novel form of content-addressable memory (CAM) which provides parallel search and update functions, allowing speed reductions down to constant time per operation. It has been proposed in related investigations by Hall (1981) that, with VLSI, CAM-based structures with enhanced instruction sets for general purpose processing will be feasible.

Computer architecture evaluation for structural dynamics computations: Project summary

NASA Technical Reports Server (NTRS)

Standley, Hilda M.

1989-01-01

The intent of the proposed effort is the examination of the impact of the elements of parallel architectures on the performance realized in a parallel computation. To this end, three major projects are developed: a language for the expression of high level parallelism, a statistical technique for the synthesis of multicomputer interconnection networks based upon performance prediction, and a queueing model for the analysis of shared memory hierarchies.

Parallel k-means++ for Multiple Shared-Memory Architectures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Mackey, Patrick S.; Lewis, Robert R.

2016-09-22

In recent years k-means++ has become a popular initialization technique for improved k-means clustering. To date, most of the work done to improve its performance has involved parallelizing algorithms that are only approximations of k-means++. In this paper we present a parallelization of the exact k-means++ algorithm, with a proof of its correctness. We develop implementations for three distinct shared-memory architectures: multicore CPU, high performance GPU, and the massively multithreaded Cray XMT platform. We demonstrate the scalability of the algorithm on each platform. In addition we present a visual approach for showing which platform performed k-means++ the fastest for varyingmore » data sizes.« less

Array processor architecture

NASA Technical Reports Server (NTRS)

Barnes, George H. (Inventor); Lundstrom, Stephen F. (Inventor); Shafer, Philip E. (Inventor)

1983-01-01

A high speed parallel array data processing architecture fashioned under a computational envelope approach includes a data base memory for secondary storage of programs and data, and a plurality of memory modules interconnected to a plurality of processing modules by a connection network of the Omega gender. Programs and data are fed from the data base memory to the plurality of memory modules and from hence the programs are fed through the connection network to the array of processors (one copy of each program for each processor). Execution of the programs occur with the processors operating normally quite independently of each other in a multiprocessing fashion. For data dependent operations and other suitable operations, all processors are instructed to finish one given task or program branch before all are instructed to proceed in parallel processing fashion on the next instruction. Even when functioning in the parallel processing mode however, the processors are not locked-step but execute their own copy of the program individually unless or until another overall processor array synchronization instruction is issued.

Optical memories in digital computing

NASA Technical Reports Server (NTRS)

Alford, C. O.; Gaylord, T. K.

1979-01-01

High capacity optical memories with relatively-high data-transfer rate and multiport simultaneous access capability may serve as basis for new computer architectures. Several computer structures that might profitably use memories are: a) simultaneous record-access system, b) simultaneously-shared memory computer system, and c) parallel digital processing structure.

Fast adaptive composite grid methods on distributed parallel architectures

NASA Technical Reports Server (NTRS)

Lemke, Max; Quinlan, Daniel

1992-01-01

The fast adaptive composite (FAC) grid method is compared with the adaptive composite method (AFAC) under variety of conditions including vectorization and parallelization. Results are given for distributed memory multiprocessor architectures (SUPRENUM, Intel iPSC/2 and iPSC/860). It is shown that the good performance of AFAC and its superiority over FAC in a parallel environment is a property of the algorithm and not dependent on peculiarities of any machine.

Optoelectronic-cache memory system architecture.

PubMed

Chiarulli, D M; Levitan, S P

1996-05-10

We present an investigation of the architecture of an optoelectronic cache that can integrate terabit optical memories with the electronic caches associated with high-performance uniprocessors and multiprocessors. The use of optoelectronic-cache memories enables these terabit technologies to provide transparently low-latency secondary memory with frame sizes comparable with disk pages but with latencies that approach those of electronic secondary-cache memories. This enables the implementation of terabit memories with effective access times comparable with the cycle times of current microprocessors. The cache design is based on the use of a smart-pixel array and combines parallel free-space optical input-output to-and-from optical memory with conventional electronic communication to the processor caches. This cache and the optical memory system to which it will interface provide a large random-access memory space that has a lower overall latency than that of magnetic disks and disk arrays. In addition, as a consequence of the high-bandwidth parallel input-output capabilities of optical memories, fault service times for the optoelectronic cache are substantially less than those currently achievable with any rotational media.

Architecture Adaptive Computing Environment

NASA Technical Reports Server (NTRS)

Dorband, John E.

2006-01-01

Architecture Adaptive Computing Environment (aCe) is a software system that includes a language, compiler, and run-time library for parallel computing. aCe was developed to enable programmers to write programs, more easily than was previously possible, for a variety of parallel computing architectures. Heretofore, it has been perceived to be difficult to write parallel programs for parallel computers and more difficult to port the programs to different parallel computing architectures. In contrast, aCe is supportable on all high-performance computing architectures. Currently, it is supported on LINUX clusters. aCe uses parallel programming constructs that facilitate writing of parallel programs. Such constructs were used in single-instruction/multiple-data (SIMD) programming languages of the 1980s, including Parallel Pascal, Parallel Forth, C*, *LISP, and MasPar MPL. In aCe, these constructs are extended and implemented for both SIMD and multiple- instruction/multiple-data (MIMD) architectures. Two new constructs incorporated in aCe are those of (1) scalar and virtual variables and (2) pre-computed paths. The scalar-and-virtual-variables construct increases flexibility in optimizing memory utilization in various architectures. The pre-computed-paths construct enables the compiler to pre-compute part of a communication operation once, rather than computing it every time the communication operation is performed.

An Adaptive Memory Interface Controller for Improving Bandwidth Utilization of Hybrid and Reconfigurable Systems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Castellana, Vito G.; Tumeo, Antonino; Ferrandi, Fabrizio

Emerging applications such as data mining, bioinformatics, knowledge discovery, social network analysis are irregular. They use data structures based on pointers or linked lists, such as graphs, unbalanced trees or unstructures grids, which generates unpredictable memory accesses. These data structures usually are large, but difficult to partition. These applications mostly are memory bandwidth bounded and have high synchronization intensity. However, they also have large amounts of inherent dynamic parallelism, because they potentially perform a task for each one of the element they are exploring. Several efforts are looking at accelerating these applications on hybrid architectures, which integrate general purpose processorsmore » with reconfigurable devices. Some solutions, which demonstrated significant speedups, include custom-hand tuned accelerators or even full processor architectures on the reconfigurable logic. In this paper we present an approach for the automatic synthesis of accelerators from C, targeted at irregular applications. In contrast to typical High Level Synthesis paradigms, which construct a centralized Finite State Machine, our approach generates dynamically scheduled hardware components. While parallelism exploitation in typical HLS-generated accelerators is usually bound within a single execution flow, our solution allows concurrently running multiple execution flow, thus also exploiting the coarser grain task parallelism of irregular applications. Our approach supports multiple, multi-ported and distributed memories, and atomic memory operations. Its main objective is parallelizing as many memory operations as possible, independently from their execution time, to maximize the memory bandwidth utilization. This significantly differs from current HLS flows, which usually consider a single memory port and require precise scheduling of memory operations. A key innovation of our approach is the generation of a memory interface controller, which dynamically maps concurrent memory accesses to multiple ports. We present a case study on a typical irregular kernel, Graph Breadth First search (BFS), exploring different tradeoffs in terms of parallelism and number of memories.« less

RRAM-based parallel computing architecture using k-nearest neighbor classification for pattern recognition

NASA Astrophysics Data System (ADS)

Jiang, Yuning; Kang, Jinfeng; Wang, Xinan

2017-03-01

Resistive switching memory (RRAM) is considered as one of the most promising devices for parallel computing solutions that may overcome the von Neumann bottleneck of today’s electronic systems. However, the existing RRAM-based parallel computing architectures suffer from practical problems such as device variations and extra computing circuits. In this work, we propose a novel parallel computing architecture for pattern recognition by implementing k-nearest neighbor classification on metal-oxide RRAM crossbar arrays. Metal-oxide RRAM with gradual RESET behaviors is chosen as both the storage and computing components. The proposed architecture is tested by the MNIST database. High speed (~100 ns per example) and high recognition accuracy (97.05%) are obtained. The influence of several non-ideal device properties is also discussed, and it turns out that the proposed architecture shows great tolerance to device variations. This work paves a new way to achieve RRAM-based parallel computing hardware systems with high performance.

Parallelization of Program to Optimize Simulated Trajectories (POST3D)

NASA Technical Reports Server (NTRS)

Hammond, Dana P.; Korte, John J. (Technical Monitor)

2001-01-01

This paper describes the parallelization of the Program to Optimize Simulated Trajectories (POST3D). POST3D uses a gradient-based optimization algorithm that reaches an optimum design point by moving from one design point to the next. The gradient calculations required to complete the optimization process, dominate the computational time and have been parallelized using a Single Program Multiple Data (SPMD) on a distributed memory NUMA (non-uniform memory access) architecture. The Origin2000 was used for the tests presented.

Parallel processing approach to transform-based image coding

NASA Astrophysics Data System (ADS)

Normile, James O.; Wright, Dan; Chu, Ken; Yeh, Chia L.

1991-06-01

This paper describes a flexible parallel processing architecture designed for use in real time video processing. The system consists of floating point DSP processors connected to each other via fast serial links, each processor has access to a globally shared memory. A multiple bus architecture in combination with a dual ported memory allows communication with a host control processor. The system has been applied to prototyping of video compression and decompression algorithms. The decomposition of transform based algorithms for decompression into a form suitable for parallel processing is described. A technique for automatic load balancing among the processors is developed and discussed, results ar presented with image statistics and data rates. Finally techniques for accelerating the system throughput are analyzed and results from the application of one such modification described.

A message passing kernel for the hypercluster parallel processing test bed

NASA Technical Reports Server (NTRS)

Blech, Richard A.; Quealy, Angela; Cole, Gary L.

1989-01-01

A Message-Passing Kernel (MPK) for the Hypercluster parallel-processing test bed is described. The Hypercluster is being developed at the NASA Lewis Research Center to support investigations of parallel algorithms and architectures for computational fluid and structural mechanics applications. The Hypercluster resembles the hypercube architecture except that each node consists of multiple processors communicating through shared memory. The MPK efficiently routes information through the Hypercluster, using a message-passing protocol when necessary and faster shared-memory communication whenever possible. The MPK also interfaces all of the processors with the Hypercluster operating system (HYCLOPS), which runs on a Front-End Processor (FEP). This approach distributes many of the I/O tasks to the Hypercluster processors and eliminates the need for a separate I/O support program on the FEP.

Sparse distributed memory overview

NASA Technical Reports Server (NTRS)

Raugh, Mike

1990-01-01

The Sparse Distributed Memory (SDM) project is investigating the theory and applications of massively parallel computing architecture, called sparse distributed memory, that will support the storage and retrieval of sensory and motor patterns characteristic of autonomous systems. The immediate objectives of the project are centered in studies of the memory itself and in the use of the memory to solve problems in speech, vision, and robotics. Investigation of methods for encoding sensory data is an important part of the research. Examples of NASA missions that may benefit from this work are Space Station, planetary rovers, and solar exploration. Sparse distributed memory offers promising technology for systems that must learn through experience and be capable of adapting to new circumstances, and for operating any large complex system requiring automatic monitoring and control. Sparse distributed memory is a massively parallel architecture motivated by efforts to understand how the human brain works. Sparse distributed memory is an associative memory, able to retrieve information from cues that only partially match patterns stored in the memory. It is able to store long temporal sequences derived from the behavior of a complex system, such as progressive records of the system's sensory data and correlated records of the system's motor controls.

CUDA Optimization Strategies for Compute- and Memory-Bound Neuroimaging Algorithms

PubMed Central

Lee, Daren; Dinov, Ivo; Dong, Bin; Gutman, Boris; Yanovsky, Igor; Toga, Arthur W.

2011-01-01

As neuroimaging algorithms and technology continue to grow faster than CPU performance in complexity and image resolution, data-parallel computing methods will be increasingly important. The high performance, data-parallel architecture of modern graphical processing units (GPUs) can reduce computational times by orders of magnitude. However, its massively threaded architecture introduces challenges when GPU resources are exceeded. This paper presents optimization strategies for compute- and memory-bound algorithms for the CUDA architecture. For compute-bound algorithms, the registers are reduced through variable reuse via shared memory and the data throughput is increased through heavier thread workloads and maximizing the thread configuration for a single thread block per multiprocessor. For memory-bound algorithms, fitting the data into the fast but limited GPU resources is achieved through reorganizing the data into self-contained structures and employing a multi-pass approach. Memory latencies are reduced by selecting memory resources whose cache performance are optimized for the algorithm's access patterns. We demonstrate the strategies on two computationally expensive algorithms and achieve optimized GPU implementations that perform up to 6× faster than unoptimized ones. Compared to CPU implementations, we achieve peak GPU speedups of 129× for the 3D unbiased nonlinear image registration technique and 93× for the non-local means surface denoising algorithm. PMID:21159404

CUDA optimization strategies for compute- and memory-bound neuroimaging algorithms.

PubMed

Lee, Daren; Dinov, Ivo; Dong, Bin; Gutman, Boris; Yanovsky, Igor; Toga, Arthur W

2012-06-01

As neuroimaging algorithms and technology continue to grow faster than CPU performance in complexity and image resolution, data-parallel computing methods will be increasingly important. The high performance, data-parallel architecture of modern graphical processing units (GPUs) can reduce computational times by orders of magnitude. However, its massively threaded architecture introduces challenges when GPU resources are exceeded. This paper presents optimization strategies for compute- and memory-bound algorithms for the CUDA architecture. For compute-bound algorithms, the registers are reduced through variable reuse via shared memory and the data throughput is increased through heavier thread workloads and maximizing the thread configuration for a single thread block per multiprocessor. For memory-bound algorithms, fitting the data into the fast but limited GPU resources is achieved through reorganizing the data into self-contained structures and employing a multi-pass approach. Memory latencies are reduced by selecting memory resources whose cache performance are optimized for the algorithm's access patterns. We demonstrate the strategies on two computationally expensive algorithms and achieve optimized GPU implementations that perform up to 6× faster than unoptimized ones. Compared to CPU implementations, we achieve peak GPU speedups of 129× for the 3D unbiased nonlinear image registration technique and 93× for the non-local means surface denoising algorithm. Copyright © 2010 Elsevier Ireland Ltd. All rights reserved.

Maximal clique enumeration with data-parallel primitives

DOE Office of Scientific and Technical Information (OSTI.GOV)

Lessley, Brenton; Perciano, Talita; Mathai, Manish

The enumeration of all maximal cliques in an undirected graph is a fundamental problem arising in several research areas. We consider maximal clique enumeration on shared-memory, multi-core architectures and introduce an approach consisting entirely of data-parallel operations, in an effort to achieve efficient and portable performance across different architectures. We study the performance of the algorithm via experiments varying over benchmark graphs and architectures. Overall, we observe that our algorithm achieves up to a 33-time speedup and 9-time speedup over state-of-the-art distributed and serial algorithms, respectively, for graphs with higher ratios of maximal cliques to total cliques. Further, we attainmore » additional speedups on a GPU architecture, demonstrating the portable performance of our data-parallel design.« less

A parallel Monte Carlo code for planar and SPECT imaging: implementation, verification and applications in (131)I SPECT.

PubMed

Dewaraja, Yuni K; Ljungberg, Michael; Majumdar, Amitava; Bose, Abhijit; Koral, Kenneth F

2002-02-01

This paper reports the implementation of the SIMIND Monte Carlo code on an IBM SP2 distributed memory parallel computer. Basic aspects of running Monte Carlo particle transport calculations on parallel architectures are described. Our parallelization is based on equally partitioning photons among the processors and uses the Message Passing Interface (MPI) library for interprocessor communication and the Scalable Parallel Random Number Generator (SPRNG) to generate uncorrelated random number streams. These parallelization techniques are also applicable to other distributed memory architectures. A linear increase in computing speed with the number of processors is demonstrated for up to 32 processors. This speed-up is especially significant in Single Photon Emission Computed Tomography (SPECT) simulations involving higher energy photon emitters, where explicit modeling of the phantom and collimator is required. For (131)I, the accuracy of the parallel code is demonstrated by comparing simulated and experimental SPECT images from a heart/thorax phantom. Clinically realistic SPECT simulations using the voxel-man phantom are carried out to assess scatter and attenuation correction.

Implementation of Parallel Dynamic Simulation on Shared-Memory vs. Distributed-Memory Environments

DOE Office of Scientific and Technical Information (OSTI.GOV)

Jin, Shuangshuang; Chen, Yousu; Wu, Di

2015-12-09

Power system dynamic simulation computes the system response to a sequence of large disturbance, such as sudden changes in generation or load, or a network short circuit followed by protective branch switching operation. It consists of a large set of differential and algebraic equations, which is computational intensive and challenging to solve using single-processor based dynamic simulation solution. High-performance computing (HPC) based parallel computing is a very promising technology to speed up the computation and facilitate the simulation process. This paper presents two different parallel implementations of power grid dynamic simulation using Open Multi-processing (OpenMP) on shared-memory platform, and Messagemore » Passing Interface (MPI) on distributed-memory clusters, respectively. The difference of the parallel simulation algorithms and architectures of the two HPC technologies are illustrated, and their performances for running parallel dynamic simulation are compared and demonstrated.« less

Some fast elliptic solvers on parallel architectures and their complexities

NASA Technical Reports Server (NTRS)

Gallopoulos, E.; Saad, Y.

1989-01-01

The discretization of separable elliptic partial differential equations leads to linear systems with special block tridiagonal matrices. Several methods are known to solve these systems, the most general of which is the Block Cyclic Reduction (BCR) algorithm which handles equations with nonconstant coefficients. A method was recently proposed to parallelize and vectorize BCR. In this paper, the mapping of BCR on distributed memory architectures is discussed, and its complexity is compared with that of other approaches including the Alternating-Direction method. A fast parallel solver is also described, based on an explicit formula for the solution, which has parallel computational compelxity lower than that of parallel BCR.

Some fast elliptic solvers on parallel architectures and their complexities

NASA Technical Reports Server (NTRS)

Gallopoulos, E.; Saad, Youcef

1989-01-01

The discretization of separable elliptic partial differential equations leads to linear systems with special block triangular matrices. Several methods are known to solve these systems, the most general of which is the Block Cyclic Reduction (BCR) algorithm which handles equations with nonconsistant coefficients. A method was recently proposed to parallelize and vectorize BCR. Here, the mapping of BCR on distributed memory architectures is discussed, and its complexity is compared with that of other approaches, including the Alternating-Direction method. A fast parallel solver is also described, based on an explicit formula for the solution, which has parallel computational complexity lower than that of parallel BCR.

Scalable Triadic Analysis of Large-Scale Graphs: Multi-Core vs. Multi-Processor vs. Multi-Threaded Shared Memory Architectures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chin, George; Marquez, Andres; Choudhury, Sutanay

2012-09-01

Triadic analysis encompasses a useful set of graph mining methods that is centered on the concept of a triad, which is a subgraph of three nodes and the configuration of directed edges across the nodes. Such methods are often applied in the social sciences as well as many other diverse fields. Triadic methods commonly operate on a triad census that counts the number of triads of every possible edge configuration in a graph. Like other graph algorithms, triadic census algorithms do not scale well when graphs reach tens of millions to billions of nodes. To enable the triadic analysis ofmore » large-scale graphs, we developed and optimized a triad census algorithm to efficiently execute on shared memory architectures. We will retrace the development and evolution of a parallel triad census algorithm. Over the course of several versions, we continually adapted the code’s data structures and program logic to expose more opportunities to exploit parallelism on shared memory that would translate into improved computational performance. We will recall the critical steps and modifications that occurred during code development and optimization. Furthermore, we will compare the performances of triad census algorithm versions on three specific systems: Cray XMT, HP Superdome, and AMD multi-core NUMA machine. These three systems have shared memory architectures but with markedly different hardware capabilities to manage parallelism.« less

Computational performance of a smoothed particle hydrodynamics simulation for shared-memory parallel computing

NASA Astrophysics Data System (ADS)

Nishiura, Daisuke; Furuichi, Mikito; Sakaguchi, Hide

2015-09-01

The computational performance of a smoothed particle hydrodynamics (SPH) simulation is investigated for three types of current shared-memory parallel computer devices: many integrated core (MIC) processors, graphics processing units (GPUs), and multi-core CPUs. We are especially interested in efficient shared-memory allocation methods for each chipset, because the efficient data access patterns differ between compute unified device architecture (CUDA) programming for GPUs and OpenMP programming for MIC processors and multi-core CPUs. We first introduce several parallel implementation techniques for the SPH code, and then examine these on our target computer architectures to determine the most effective algorithms for each processor unit. In addition, we evaluate the effective computing performance and power efficiency of the SPH simulation on each architecture, as these are critical metrics for overall performance in a multi-device environment. In our benchmark test, the GPU is found to produce the best arithmetic performance as a standalone device unit, and gives the most efficient power consumption. The multi-core CPU obtains the most effective computing performance. The computational speed of the MIC processor on Xeon Phi approached that of two Xeon CPUs. This indicates that using MICs is an attractive choice for existing SPH codes on multi-core CPUs parallelized by OpenMP, as it gains computational acceleration without the need for significant changes to the source code.

Implementations of BLAST for parallel computers.

PubMed

Jülich, A

1995-02-01

The BLAST sequence comparison programs have been ported to a variety of parallel computers-the shared memory machine Cray Y-MP 8/864 and the distributed memory architectures Intel iPSC/860 and nCUBE. Additionally, the programs were ported to run on workstation clusters. We explain the parallelization techniques and consider the pros and cons of these methods. The BLAST programs are very well suited for parallelization for a moderate number of processors. We illustrate our results using the program blastp as an example. As input data for blastp, a 799 residue protein query sequence and the protein database PIR were used.

Implementation of digital equality comparator circuit on memristive memory crossbar array using material implication logic

NASA Astrophysics Data System (ADS)

Haron, Adib; Mahdzair, Fazren; Luqman, Anas; Osman, Nazmie; Junid, Syed Abdul Mutalib Al

2018-03-01

One of the most significant constraints of Von Neumann architecture is the limited bandwidth between memory and processor. The cost to move data back and forth between memory and processor is considerably higher than the computation in the processor itself. This architecture significantly impacts the Big Data and data-intensive application such as DNA analysis comparison which spend most of the processing time to move data. Recently, the in-memory processing concept was proposed, which is based on the capability to perform the logic operation on the physical memory structure using a crossbar topology and non-volatile resistive-switching memristor technology. This paper proposes a scheme to map digital equality comparator circuit on memristive memory crossbar array. The 2-bit, 4-bit, 8-bit, 16-bit, 32-bit, and 64-bit of equality comparator circuit are mapped on memristive memory crossbar array by using material implication logic in a sequential and parallel method. The simulation results show that, for the 64-bit word size, the parallel mapping exhibits 2.8× better performance in total execution time than sequential mapping but has a trade-off in terms of energy consumption and area utilization. Meanwhile, the total crossbar area can be reduced by 1.2× for sequential mapping and 1.5× for parallel mapping both by using the overlapping technique.

Architectures for reasoning in parallel

NASA Technical Reports Server (NTRS)

Hall, Lawrence O.

1989-01-01

The research conducted has dealt with rule-based expert systems. The algorithms that may lead to effective parallelization of them were investigated. Both the forward and backward chained control paradigms were investigated in the course of this work. The best computer architecture for the developed and investigated algorithms has been researched. Two experimental vehicles were developed to facilitate this research. They are Backpac, a parallel backward chained rule-based reasoning system and Datapac, a parallel forward chained rule-based reasoning system. Both systems have been written in Multilisp, a version of Lisp which contains the parallel construct, future. Applying the future function to a function causes the function to become a task parallel to the spawning task. Additionally, Backpac and Datapac have been run on several disparate parallel processors. The machines are an Encore Multimax with 10 processors, the Concert Multiprocessor with 64 processors, and a 32 processor BBN GP1000. Both the Concert and the GP1000 are switch-based machines. The Multimax has all its processors hung off a common bus. All are shared memory machines, but have different schemes for sharing the memory and different locales for the shared memory. The main results of the investigations come from experiments on the 10 processor Encore and the Concert with partitions of 32 or less processors. Additionally, experiments have been run with a stripped down version of EMYCIN.

A new scheduling algorithm for parallel sparse LU factorization with static pivoting

DOE Office of Scientific and Technical Information (OSTI.GOV)

Grigori, Laura; Li, Xiaoye S.

2002-08-20

In this paper we present a static scheduling algorithm for parallel sparse LU factorization with static pivoting. The algorithm is divided into mapping and scheduling phases, using the symmetric pruned graphs of L' and U to represent dependencies. The scheduling algorithm is designed for driving the parallel execution of the factorization on a distributed-memory architecture. Experimental results and comparisons with SuperLU{_}DIST are reported after applying this algorithm on real world application matrices on an IBM SP RS/6000 distributed memory machine.

A Multi-Level Parallelization Concept for High-Fidelity Multi-Block Solvers

NASA Technical Reports Server (NTRS)

Hatay, Ferhat F.; Jespersen, Dennis C.; Guruswamy, Guru P.; Rizk, Yehia M.; Byun, Chansup; Gee, Ken; VanDalsem, William R. (Technical Monitor)

1997-01-01

The integration of high-fidelity Computational Fluid Dynamics (CFD) analysis tools with the industrial design process benefits greatly from the robust implementations that are transportable across a wide range of computer architectures. In the present work, a hybrid domain-decomposition and parallelization concept was developed and implemented into the widely-used NASA multi-block Computational Fluid Dynamics (CFD) packages implemented in ENSAERO and OVERFLOW. The new parallel solver concept, PENS (Parallel Euler Navier-Stokes Solver), employs both fine and coarse granularity in data partitioning as well as data coalescing to obtain the desired load-balance characteristics on the available computer platforms. This multi-level parallelism implementation itself introduces no changes to the numerical results, hence the original fidelity of the packages are identically preserved. The present implementation uses the Message Passing Interface (MPI) library for interprocessor message passing and memory accessing. By choosing an appropriate combination of the available partitioning and coalescing capabilities only during the execution stage, the PENS solver becomes adaptable to different computer architectures from shared-memory to distributed-memory platforms with varying degrees of parallelism. The PENS implementation on the IBM SP2 distributed memory environment at the NASA Ames Research Center obtains 85 percent scalable parallel performance using fine-grain partitioning of single-block CFD domains using up to 128 wide computational nodes. Multi-block CFD simulations of complete aircraft simulations achieve 75 percent perfect load-balanced executions using data coalescing and the two levels of parallelism. SGI PowerChallenge, SGI Origin 2000, and a cluster of workstations are the other platforms where the robustness of the implementation is tested. The performance behavior on the other computer platforms with a variety of realistic problems will be included as this on-going study progresses.

High-performance computing — an overview

NASA Astrophysics Data System (ADS)

Marksteiner, Peter

1996-08-01

An overview of high-performance computing (HPC) is given. Different types of computer architectures used in HPC are discussed: vector supercomputers, high-performance RISC processors, various parallel computers like symmetric multiprocessors, workstation clusters, massively parallel processors. Software tools and programming techniques used in HPC are reviewed: vectorizing compilers, optimization and vector tuning, optimization for RISC processors; parallel programming techniques like shared-memory parallelism, message passing and data parallelism; and numerical libraries.

A FAST ITERATIVE METHOD FOR SOLVING THE EIKONAL EQUATION ON TETRAHEDRAL DOMAINS

PubMed Central

Fu, Zhisong; Kirby, Robert M.; Whitaker, Ross T.

2014-01-01

Generating numerical solutions to the eikonal equation and its many variations has a broad range of applications in both the natural and computational sciences. Efficient solvers on cutting-edge, parallel architectures require new algorithms that may not be theoretically optimal, but that are designed to allow asynchronous solution updates and have limited memory access patterns. This paper presents a parallel algorithm for solving the eikonal equation on fully unstructured tetrahedral meshes. The method is appropriate for the type of fine-grained parallelism found on modern massively-SIMD architectures such as graphics processors and takes into account the particular constraints and capabilities of these computing platforms. This work builds on previous work for solving these equations on triangle meshes; in this paper we adapt and extend previous two-dimensional strategies to accommodate three-dimensional, unstructured, tetrahedralized domains. These new developments include a local update strategy with data compaction for tetrahedral meshes that provides solutions on both serial and parallel architectures, with a generalization to inhomogeneous, anisotropic speed functions. We also propose two new update schemes, specialized to mitigate the natural data increase observed when moving to three dimensions, and the data structures necessary for efficiently mapping data to parallel SIMD processors in a way that maintains computational density. Finally, we present descriptions of the implementations for a single CPU, as well as multicore CPUs with shared memory and SIMD architectures, with comparative results against state-of-the-art eikonal solvers. PMID:25221418

Array processor architecture connection network

NASA Technical Reports Server (NTRS)

Barnes, George H. (Inventor); Lundstrom, Stephen F. (Inventor); Shafer, Philip E. (Inventor)

1982-01-01

A connection network is disclosed for use between a parallel array of processors and a parallel array of memory modules for establishing non-conflicting data communications paths between requested memory modules and requesting processors. The connection network includes a plurality of switching elements interposed between the processor array and the memory modules array in an Omega networking architecture. Each switching element includes a first and a second processor side port, a first and a second memory module side port, and control logic circuitry for providing data connections between the first and second processor ports and the first and second memory module ports. The control logic circuitry includes strobe logic for examining data arriving at the first and the second processor ports to indicate when the data arriving is requesting data from a requesting processor to a requested memory module. Further, connection circuitry is associated with the strobe logic for examining requesting data arriving at the first and the second processor ports for providing a data connection therefrom to the first and the second memory module ports in response thereto when the data connection so provided does not conflict with a pre-established data connection currently in use.

Vascular system modeling in parallel environment - distributed and shared memory approaches

PubMed Central

Jurczuk, Krzysztof; Kretowski, Marek; Bezy-Wendling, Johanne

2011-01-01

The paper presents two approaches in parallel modeling of vascular system development in internal organs. In the first approach, new parts of tissue are distributed among processors and each processor is responsible for perfusing its assigned parts of tissue to all vascular trees. Communication between processors is accomplished by passing messages and therefore this algorithm is perfectly suited for distributed memory architectures. The second approach is designed for shared memory machines. It parallelizes the perfusion process during which individual processing units perform calculations concerning different vascular trees. The experimental results, performed on a computing cluster and multi-core machines, show that both algorithms provide a significant speedup. PMID:21550891

Scalable Parallel Density-based Clustering and Applications

NASA Astrophysics Data System (ADS)

Patwary, Mostofa Ali

2014-04-01

Recently, density-based clustering algorithms (DBSCAN and OPTICS) have gotten significant attention of the scientific community due to their unique capability of discovering arbitrary shaped clusters and eliminating noise data. These algorithms have several applications, which require high performance computing, including finding halos and subhalos (clusters) from massive cosmology data in astrophysics, analyzing satellite images, X-ray crystallography, and anomaly detection. However, parallelization of these algorithms are extremely challenging as they exhibit inherent sequential data access order, unbalanced workload resulting in low parallel efficiency. To break the data access sequentiality and to achieve high parallelism, we develop new parallel algorithms, both for DBSCAN and OPTICS, designed using graph algorithmic techniques. For example, our parallel DBSCAN algorithm exploits the similarities between DBSCAN and computing connected components. Using datasets containing up to a billion floating point numbers, we show that our parallel density-based clustering algorithms significantly outperform the existing algorithms, achieving speedups up to 27.5 on 40 cores on shared memory architecture and speedups up to 5,765 using 8,192 cores on distributed memory architecture. In our experiments, we found that while achieving the scalability, our algorithms produce clustering results with comparable quality to the classical algorithms.

Nonvolatile “AND,” “OR,” and “NOT” Boolean logic gates based on phase-change memory

DOE Office of Scientific and Technical Information (OSTI.GOV)

Li, Y.; Zhong, Y. P.; Deng, Y. F.

2013-12-21

Electronic devices or circuits that can implement both logic and memory functions are regarded as the building blocks for future massive parallel computing beyond von Neumann architecture. Here we proposed phase-change memory (PCM)-based nonvolatile logic gates capable of AND, OR, and NOT Boolean logic operations verified in SPICE simulations and circuit experiments. The logic operations are parallel computing and results can be stored directly in the states of the logic gates, facilitating the combination of computing and memory in the same circuit. These results are encouraging for ultralow-power and high-speed nonvolatile logic circuit design based on novel memory devices.

Gilgamesh: A Multithreaded Processor-In-Memory Architecture for Petaflops Computing

NASA Technical Reports Server (NTRS)

Sterling, T. L.; Zima, H. P.

2002-01-01

Processor-in-Memory (PIM) architectures avoid the von Neumann bottleneck in conventional machines by integrating high-density DRAM and CMOS logic on the same chip. Parallel systems based on this new technology are expected to provide higher scalability, adaptability, robustness, fault tolerance and lower power consumption than current MPPs or commodity clusters. In this paper we describe the design of Gilgamesh, a PIM-based massively parallel architecture, and elements of its execution model. Gilgamesh extends existing PIM capabilities by incorporating advanced mechanisms for virtualizing tasks and data and providing adaptive resource management for load balancing and latency tolerance. The Gilgamesh execution model is based on macroservers, a middleware layer which supports object-based runtime management of data and threads allowing explicit and dynamic control of locality and load balancing. The paper concludes with a discussion of related research activities and an outlook to future work.

Enabling the High Level Synthesis of Data Analytics Accelerators

DOE Office of Scientific and Technical Information (OSTI.GOV)

Minutoli, Marco; Castellana, Vito G.; Tumeo, Antonino

Conventional High Level Synthesis (HLS) tools mainly tar- get compute intensive kernels typical of digital signal pro- cessing applications. We are developing techniques and ar- chitectural templates to enable HLS of data analytics appli- cations. These applications are memory intensive, present fine-grained, unpredictable data accesses, and irregular, dy- namic task parallelism. We discuss an architectural tem- plate based around a distributed controller to efficiently ex- ploit thread level parallelism. We present a memory in- terface that supports parallel memory subsystems and en- ables implementing atomic memory operations. We intro- duce a dynamic task scheduling approach to efficiently ex- ecute heavilymore » unbalanced workload. The templates are val- idated by synthesizing queries from the Lehigh University Benchmark (LUBM), a well know SPARQL benchmark.« less

An architecture for real-time vision processing

NASA Technical Reports Server (NTRS)

Chien, Chiun-Hong

1994-01-01

To study the feasibility of developing an architecture for real time vision processing, a task queue server and parallel algorithms for two vision operations were designed and implemented on an i860-based Mercury Computing System 860VS array processor. The proposed architecture treats each vision function as a task or set of tasks which may be recursively divided into subtasks and processed by multiple processors coordinated by a task queue server accessible by all processors. Each idle processor subsequently fetches a task and associated data from the task queue server for processing and posts the result to shared memory for later use. Load balancing can be carried out within the processing system without the requirement for a centralized controller. The author concludes that real time vision processing cannot be achieved without both sequential and parallel vision algorithms and a good parallel vision architecture.

The potential of multi-port optical memories in digital computing

NASA Technical Reports Server (NTRS)

Alford, C. O.; Gaylord, T. K.

1975-01-01

A high-capacity memory with a relatively high data transfer rate and multi-port simultaneous access capability may serve as the basis for new computer architectures. The implementation of a multi-port optical memory is discussed. Several computer structures are presented that might profitably use such a memory. These structures include (1) a simultaneous record access system, (2) a simultaneously shared memory computer system, and (3) a parallel digital processing structure.

Multiprocessor architecture: Synthesis and evaluation

NASA Technical Reports Server (NTRS)

Standley, Hilda M.

1990-01-01

Multiprocessor computed architecture evaluation for structural computations is the focus of the research effort described. Results obtained are expected to lead to more efficient use of existing architectures and to suggest designs for new, application specific, architectures. The brief descriptions given outline a number of related efforts directed toward this purpose. The difficulty is analyzing an existing architecture or in designing a new computer architecture lies in the fact that the performance of a particular architecture, within the context of a given application, is determined by a number of factors. These include, but are not limited to, the efficiency of the computation algorithm, the programming language and support environment, the quality of the program written in the programming language, the multiplicity of the processing elements, the characteristics of the individual processing elements, the interconnection network connecting processors and non-local memories, and the shared memory organization covering the spectrum from no shared memory (all local memory) to one global access memory. These performance determiners may be loosely classified as being software or hardware related. This distinction is not clear or even appropriate in many cases. The effect of the choice of algorithm is ignored by assuming that the algorithm is specified as given. Effort directed toward the removal of the effect of the programming language and program resulted in the design of a high-level parallel programming language. Two characteristics of the fundamental structure of the architecture (memory organization and interconnection network) are examined.

Simulating Hydrologic Flow and Reactive Transport with PFLOTRAN and PETSc on Emerging Fine-Grained Parallel Computer Architectures

NASA Astrophysics Data System (ADS)

Mills, R. T.; Rupp, K.; Smith, B. F.; Brown, J.; Knepley, M.; Zhang, H.; Adams, M.; Hammond, G. E.

2017-12-01

As the high-performance computing community pushes towards the exascale horizon, power and heat considerations have driven the increasing importance and prevalence of fine-grained parallelism in new computer architectures. High-performance computing centers have become increasingly reliant on GPGPU accelerators and "manycore" processors such as the Intel Xeon Phi line, and 512-bit SIMD registers have even been introduced in the latest generation of Intel's mainstream Xeon server processors. The high degree of fine-grained parallelism and more complicated memory hierarchy considerations of such "manycore" processors present several challenges to existing scientific software. Here, we consider how the massively parallel, open-source hydrologic flow and reactive transport code PFLOTRAN - and the underlying Portable, Extensible Toolkit for Scientific Computation (PETSc) library on which it is built - can best take advantage of such architectures. We will discuss some key features of these novel architectures and our code optimizations and algorithmic developments targeted at them, and present experiences drawn from working with a wide range of PFLOTRAN benchmark problems on these architectures.

Programming model for distributed intelligent systems

NASA Technical Reports Server (NTRS)

Sztipanovits, J.; Biegl, C.; Karsai, G.; Bogunovic, N.; Purves, B.; Williams, R.; Christiansen, T.

1988-01-01

A programming model and architecture which was developed for the design and implementation of complex, heterogeneous measurement and control systems is described. The Multigraph Architecture integrates artificial intelligence techniques with conventional software technologies, offers a unified framework for distributed and shared memory based parallel computational models and supports multiple programming paradigms. The system can be implemented on different hardware architectures and can be adapted to strongly different applications.

SKIRT: Hybrid parallelization of radiative transfer simulations

NASA Astrophysics Data System (ADS)

Verstocken, S.; Van De Putte, D.; Camps, P.; Baes, M.

2017-07-01

We describe the design, implementation and performance of the new hybrid parallelization scheme in our Monte Carlo radiative transfer code SKIRT, which has been used extensively for modelling the continuum radiation of dusty astrophysical systems including late-type galaxies and dusty tori. The hybrid scheme combines distributed memory parallelization, using the standard Message Passing Interface (MPI) to communicate between processes, and shared memory parallelization, providing multiple execution threads within each process to avoid duplication of data structures. The synchronization between multiple threads is accomplished through atomic operations without high-level locking (also called lock-free programming). This improves the scaling behaviour of the code and substantially simplifies the implementation of the hybrid scheme. The result is an extremely flexible solution that adjusts to the number of available nodes, processors and memory, and consequently performs well on a wide variety of computing architectures.

GaAs Supercomputing: Architecture, Language, And Algorithms For Image Processing

NASA Astrophysics Data System (ADS)

Johl, John T.; Baker, Nick C.

1988-10-01

The application of high-speed GaAs processors in a parallel system matches the demanding computational requirements of image processing. The architecture of the McDonnell Douglas Astronautics Company (MDAC) vector processor is described along with the algorithms and language translator. Most image and signal processing algorithms can utilize parallel processing and show a significant performance improvement over sequential versions. The parallelization performed by this system is within each vector instruction. Since each vector has many elements, each requiring some computation, useful concurrent arithmetic operations can easily be performed. Balancing the memory bandwidth with the computation rate of the processors is an important design consideration for high efficiency and utilization. The architecture features a bus-based execution unit consisting of four to eight 32-bit GaAs RISC microprocessors running at a 200 MHz clock rate for a peak performance of 1.6 BOPS. The execution unit is connected to a vector memory with three buses capable of transferring two input words and one output word every 10 nsec. The address generators inside the vector memory perform different vector addressing modes and feed the data to the execution unit. The functions discussed in this paper include basic MATRIX OPERATIONS, 2-D SPATIAL CONVOLUTION, HISTOGRAM, and FFT. For each of these algorithms, assembly language programs were run on a behavioral model of the system to obtain performance figures.

Parallelizing ATLAS Reconstruction and Simulation: Issues and Optimization Solutions for Scaling on Multi- and Many-CPU Platforms

NASA Astrophysics Data System (ADS)

Leggett, C.; Binet, S.; Jackson, K.; Levinthal, D.; Tatarkhanov, M.; Yao, Y.

2011-12-01

Thermal limitations have forced CPU manufacturers to shift from simply increasing clock speeds to improve processor performance, to producing chip designs with multi- and many-core architectures. Further the cores themselves can run multiple threads as a zero overhead context switch allowing low level resource sharing (Intel Hyperthreading). To maximize bandwidth and minimize memory latency, memory access has become non uniform (NUMA). As manufacturers add more cores to each chip, a careful understanding of the underlying architecture is required in order to fully utilize the available resources. We present AthenaMP and the Atlas event loop manager, the driver of the simulation and reconstruction engines, which have been rewritten to make use of multiple cores, by means of event based parallelism, and final stage I/O synchronization. However, initial studies on 8 andl6 core Intel architectures have shown marked non-linearities as parallel process counts increase, with as much as 30% reductions in event throughput in some scenarios. Since the Intel Nehalem architecture (both Gainestown and Westmere) will be the most common choice for the next round of hardware procurements, an understanding of these scaling issues is essential. Using hardware based event counters and Intel's Performance Tuning Utility, we have studied the performance bottlenecks at the hardware level, and discovered optimization schemes to maximize processor throughput. We have also produced optimization mechanisms, common to all large experiments, that address the extreme nature of today's HEP code, which due to it's size, places huge burdens on the memory infrastructure of today's processors.

OS friendly microprocessor architecture: Hardware level computer security

NASA Astrophysics Data System (ADS)

Jungwirth, Patrick; La Fratta, Patrick

2016-05-01

We present an introduction to the patented OS Friendly Microprocessor Architecture (OSFA) and hardware level computer security. Conventional microprocessors have not tried to balance hardware performance and OS performance at the same time. Conventional microprocessors have depended on the Operating System for computer security and information assurance. The goal of the OS Friendly Architecture is to provide a high performance and secure microprocessor and OS system. We are interested in cyber security, information technology (IT), and SCADA control professionals reviewing the hardware level security features. The OS Friendly Architecture is a switched set of cache memory banks in a pipeline configuration. For light-weight threads, the memory pipeline configuration provides near instantaneous context switching times. The pipelining and parallelism provided by the cache memory pipeline provides for background cache read and write operations while the microprocessor's execution pipeline is running instructions. The cache bank selection controllers provide arbitration to prevent the memory pipeline and microprocessor's execution pipeline from accessing the same cache bank at the same time. This separation allows the cache memory pages to transfer to and from level 1 (L1) caching while the microprocessor pipeline is executing instructions. Computer security operations are implemented in hardware. By extending Unix file permissions bits to each cache memory bank and memory address, the OSFA provides hardware level computer security.

Parallel performance investigations of an unstructured mesh Navier-Stokes solver

NASA Technical Reports Server (NTRS)

Mavriplis, Dimitri J.

2000-01-01

A Reynolds-averaged Navier-Stokes solver based on unstructured mesh techniques for analysis of high-lift configurations is described. The method makes use of an agglomeration multigrid solver for convergence acceleration. Implicit line-smoothing is employed to relieve the stiffness associated with highly stretched meshes. A GMRES technique is also implemented to speed convergence at the expense of additional memory usage. The solver is cache efficient and fully vectorizable, and is parallelized using a two-level hybrid MPI-OpenMP implementation suitable for shared and/or distributed memory architectures, as well as clusters of shared memory machines. Convergence and scalability results are illustrated for various high-lift cases.

High Performance Programming Using Explicit Shared Memory Model on the Cray T3D

NASA Technical Reports Server (NTRS)

Saini, Subhash; Simon, Horst D.; Lasinski, T. A. (Technical Monitor)

1994-01-01

The Cray T3D is the first-phase system in Cray Research Inc.'s (CRI) three-phase massively parallel processing program. In this report we describe the architecture of the T3D, as well as the CRAFT (Cray Research Adaptive Fortran) programming model, and contrast it with PVM, which is also supported on the T3D We present some performance data based on the NAS Parallel Benchmarks to illustrate both architectural and software features of the T3D.

A parallel 3-D discrete wavelet transform architecture using pipelined lifting scheme approach for video coding

NASA Astrophysics Data System (ADS)

Hegde, Ganapathi; Vaya, Pukhraj

2013-10-01

This article presents a parallel architecture for 3-D discrete wavelet transform (3-DDWT). The proposed design is based on the 1-D pipelined lifting scheme. The architecture is fully scalable beyond the present coherent Daubechies filter bank (9, 7). This 3-DDWT architecture has advantages such as no group of pictures restriction and reduced memory referencing. It offers low power consumption, low latency and high throughput. The computing technique is based on the concept that lifting scheme minimises the storage requirement. The application specific integrated circuit implementation of the proposed architecture is done by synthesising it using 65 nm Taiwan Semiconductor Manufacturing Company standard cell library. It offers a speed of 486 MHz with a power consumption of 2.56 mW. This architecture is suitable for real-time video compression even with large frame dimensions.

Parallel algorithms for mapping pipelined and parallel computations

NASA Technical Reports Server (NTRS)

Nicol, David M.

1988-01-01

Many computational problems in image processing, signal processing, and scientific computing are naturally structured for either pipelined or parallel computation. When mapping such problems onto a parallel architecture it is often necessary to aggregate an obvious problem decomposition. Even in this context the general mapping problem is known to be computationally intractable, but recent advances have been made in identifying classes of problems and architectures for which optimal solutions can be found in polynomial time. Among these, the mapping of pipelined or parallel computations onto linear array, shared memory, and host-satellite systems figures prominently. This paper extends that work first by showing how to improve existing serial mapping algorithms. These improvements have significantly lower time and space complexities: in one case a published O(nm sup 3) time algorithm for mapping m modules onto n processors is reduced to an O(nm log m) time complexity, and its space requirements reduced from O(nm sup 2) to O(m). Run time complexity is further reduced with parallel mapping algorithms based on these improvements, which run on the architecture for which they create the mappings.

Learning, memory, and the role of neural network architecture.

PubMed

Hermundstad, Ann M; Brown, Kevin S; Bassett, Danielle S; Carlson, Jean M

2011-06-01

The performance of information processing systems, from artificial neural networks to natural neuronal ensembles, depends heavily on the underlying system architecture. In this study, we compare the performance of parallel and layered network architectures during sequential tasks that require both acquisition and retention of information, thereby identifying tradeoffs between learning and memory processes. During the task of supervised, sequential function approximation, networks produce and adapt representations of external information. Performance is evaluated by statistically analyzing the error in these representations while varying the initial network state, the structure of the external information, and the time given to learn the information. We link performance to complexity in network architecture by characterizing local error landscape curvature. We find that variations in error landscape structure give rise to tradeoffs in performance; these include the ability of the network to maximize accuracy versus minimize inaccuracy and produce specific versus generalizable representations of information. Parallel networks generate smooth error landscapes with deep, narrow minima, enabling them to find highly specific representations given sufficient time. While accurate, however, these representations are difficult to generalize. In contrast, layered networks generate rough error landscapes with a variety of local minima, allowing them to quickly find coarse representations. Although less accurate, these representations are easily adaptable. The presence of measurable performance tradeoffs in both layered and parallel networks has implications for understanding the behavior of a wide variety of natural and artificial learning systems.

HTMT-class Latency Tolerant Parallel Architecture for Petaflops Scale Computation

NASA Technical Reports Server (NTRS)

Sterling, Thomas; Bergman, Larry

2000-01-01

Computational Aero Sciences and other numeric intensive computation disciplines demand computing throughputs substantially greater than the Teraflops scale systems only now becoming available. The related fields of fluids, structures, thermal, combustion, and dynamic controls are among the interdisciplinary areas that in combination with sufficient resolution and advanced adaptive techniques may force performance requirements towards Petaflops. This will be especially true for compute intensive models such as Navier-Stokes are or when such system models are only part of a larger design optimization computation involving many design points. Yet recent experience with conventional MPP configurations comprising commodity processing and memory components has shown that larger scale frequently results in higher programming difficulty and lower system efficiency. While important advances in system software and algorithms techniques have had some impact on efficiency and programmability for certain classes of problems, in general it is unlikely that software alone will resolve the challenges to higher scalability. As in the past, future generations of high-end computers may require a combination of hardware architecture and system software advances to enable efficient operation at a Petaflops level. The NASA led HTMT project has engaged the talents of a broad interdisciplinary team to develop a new strategy in high-end system architecture to deliver petaflops scale computing in the 2004/5 timeframe. The Hybrid-Technology, MultiThreaded parallel computer architecture incorporates several advanced technologies in combination with an innovative dynamic adaptive scheduling mechanism to provide unprecedented performance and efficiency within practical constraints of cost, complexity, and power consumption. The emerging superconductor Rapid Single Flux Quantum electronics can operate at 100 GHz (the record is 770 GHz) and one percent of the power required by convention semiconductor logic. Wave Division Multiplexing optical communications can approach a peak per fiber bandwidth of 1 Tbps and the new Data Vortex network topology employing this technology can connect tens of thousands of ports providing a bi-section bandwidth on the order of a Petabyte per second with latencies well below 100 nanoseconds, even under heavy loads. Processor-in-Memory (PIM) technology combines logic and memory on the same chip exposing the internal bandwidth of the memory row buffers at low latency. And holographic storage photorefractive storage technologies provide high-density memory with access a thousand times faster than conventional disk technologies. Together these technologies enable a new class of shared memory system architecture with a peak performance in the range of a Petaflops but size and power requirements comparable to today's largest Teraflops scale systems. To achieve high-sustained performance, HTMT combines an advanced multithreading processor architecture with a memory-driven coarse-grained latency management strategy called "percolation", yielding high efficiency while reducing the much of the parallel programming burden. This paper will present the basic system architecture characteristics made possible through this series of advanced technologies and then give a detailed description of the new percolation approach to runtime latency management.

Performance Analysis of Multilevel Parallel Applications on Shared Memory Architectures

NASA Technical Reports Server (NTRS)

Jost, Gabriele; Jin, Haoqiang; Labarta, Jesus; Gimenez, Judit; Caubet, Jordi; Biegel, Bryan A. (Technical Monitor)

2002-01-01

In this paper we describe how to apply powerful performance analysis techniques to understand the behavior of multilevel parallel applications. We use the Paraver/OMPItrace performance analysis system for our study. This system consists of two major components: The OMPItrace dynamic instrumentation mechanism, which allows the tracing of processes and threads and the Paraver graphical user interface for inspection and analyses of the generated traces. We describe how to use the system to conduct a detailed comparative study of a benchmark code implemented in five different programming paradigms applicable for shared memory

Evaluating the performance of the particle finite element method in parallel architectures

NASA Astrophysics Data System (ADS)

Gimenez, Juan M.; Nigro, Norberto M.; Idelsohn, Sergio R.

2014-05-01

This paper presents a high performance implementation for the particle-mesh based method called particle finite element method two (PFEM-2). It consists of a material derivative based formulation of the equations with a hybrid spatial discretization which uses an Eulerian mesh and Lagrangian particles. The main aim of PFEM-2 is to solve transport equations as fast as possible keeping some level of accuracy. The method was found to be competitive with classical Eulerian alternatives for these targets, even in their range of optimal application. To evaluate the goodness of the method with large simulations, it is imperative to use of parallel environments. Parallel strategies for Finite Element Method have been widely studied and many libraries can be used to solve Eulerian stages of PFEM-2. However, Lagrangian stages, such as streamline integration, must be developed considering the parallel strategy selected. The main drawback of PFEM-2 is the large amount of memory needed, which limits its application to large problems with only one computer. Therefore, a distributed-memory implementation is urgently needed. Unlike a shared-memory approach, using domain decomposition the memory is automatically isolated, thus avoiding race conditions; however new issues appear due to data distribution over the processes. Thus, a domain decomposition strategy for both particle and mesh is adopted, which minimizes the communication between processes. Finally, performance analysis running over multicore and multinode architectures are presented. The Courant-Friedrichs-Lewy number used influences the efficiency of the parallelization and, in some cases, a weighted partitioning can be used to improve the speed-up. However the total cputime for cases presented is lower than that obtained when using classical Eulerian strategies.

Simplified Parallel Domain Traversal

DOE Office of Scientific and Technical Information (OSTI.GOV)

Erickson III, David J

2011-01-01

Many data-intensive scientific analysis techniques require global domain traversal, which over the years has been a bottleneck for efficient parallelization across distributed-memory architectures. Inspired by MapReduce and other simplified parallel programming approaches, we have designed DStep, a flexible system that greatly simplifies efficient parallelization of domain traversal techniques at scale. In order to deliver both simplicity to users as well as scalability on HPC platforms, we introduce a novel two-tiered communication architecture for managing and exploiting asynchronous communication loads. We also integrate our design with advanced parallel I/O techniques that operate directly on native simulation output. We demonstrate DStep bymore » performing teleconnection analysis across ensemble runs of terascale atmospheric CO{sub 2} and climate data, and we show scalability results on up to 65,536 IBM BlueGene/P cores.« less

Parallel Clustering Algorithm for Large-Scale Biological Data Sets

PubMed Central

Wang, Minchao; Zhang, Wu; Ding, Wang; Dai, Dongbo; Zhang, Huiran; Xie, Hao; Chen, Luonan; Guo, Yike; Xie, Jiang

2014-01-01

Backgrounds Recent explosion of biological data brings a great challenge for the traditional clustering algorithms. With increasing scale of data sets, much larger memory and longer runtime are required for the cluster identification problems. The affinity propagation algorithm outperforms many other classical clustering algorithms and is widely applied into the biological researches. However, the time and space complexity become a great bottleneck when handling the large-scale data sets. Moreover, the similarity matrix, whose constructing procedure takes long runtime, is required before running the affinity propagation algorithm, since the algorithm clusters data sets based on the similarities between data pairs. Methods Two types of parallel architectures are proposed in this paper to accelerate the similarity matrix constructing procedure and the affinity propagation algorithm. The memory-shared architecture is used to construct the similarity matrix, and the distributed system is taken for the affinity propagation algorithm, because of its large memory size and great computing capacity. An appropriate way of data partition and reduction is designed in our method, in order to minimize the global communication cost among processes. Result A speedup of 100 is gained with 128 cores. The runtime is reduced from serval hours to a few seconds, which indicates that parallel algorithm is capable of handling large-scale data sets effectively. The parallel affinity propagation also achieves a good performance when clustering large-scale gene data (microarray) and detecting families in large protein superfamilies. PMID:24705246

Multi-petascale highly efficient parallel supercomputer

DOE Office of Scientific and Technical Information (OSTI.GOV)

Asaad, Sameh; Bellofatto, Ralph E.; Blocksome, Michael A.

A Multi-Petascale Highly Efficient Parallel Supercomputer of 100 petaflop-scale includes node architectures based upon System-On-a-Chip technology, where each processing node comprises a single Application Specific Integrated Circuit (ASIC). The ASIC nodes are interconnected by a five dimensional torus network that optimally maximize the throughput of packet communications between nodes and minimize latency. The network implements collective network and a global asynchronous network that provides global barrier and notification functions. Integrated in the node design include a list-based prefetcher. The memory system implements transaction memory, thread level speculation, and multiversioning cache that improves soft error rate at the same time andmore » supports DMA functionality allowing for parallel processing message-passing.« less

A multiarchitecture parallel-processing development environment

NASA Technical Reports Server (NTRS)

Townsend, Scott; Blech, Richard; Cole, Gary

1993-01-01

A description is given of the hardware and software of a multiprocessor test bed - the second generation Hypercluster system. The Hypercluster architecture consists of a standard hypercube distributed-memory topology, with multiprocessor shared-memory nodes. By using standard, off-the-shelf hardware, the system can be upgraded to use rapidly improving computer technology. The Hypercluster's multiarchitecture nature makes it suitable for researching parallel algorithms in computational field simulation applications (e.g., computational fluid dynamics). The dedicated test-bed environment of the Hypercluster and its custom-built software allows experiments with various parallel-processing concepts such as message passing algorithms, debugging tools, and computational 'steering'. Such research would be difficult, if not impossible, to achieve on shared, commercial systems.

Integrating Cache Performance Modeling and Tuning Support in Parallelization Tools

NASA Technical Reports Server (NTRS)

Waheed, Abdul; Yan, Jerry; Saini, Subhash (Technical Monitor)

1998-01-01

With the resurgence of distributed shared memory (DSM) systems based on cache-coherent Non Uniform Memory Access (ccNUMA) architectures and increasing disparity between memory and processors speeds, data locality overheads are becoming the greatest bottlenecks in the way of realizing potential high performance of these systems. While parallelization tools and compilers facilitate the users in porting their sequential applications to a DSM system, a lot of time and effort is needed to tune the memory performance of these applications to achieve reasonable speedup. In this paper, we show that integrating cache performance modeling and tuning support within a parallelization environment can alleviate this problem. The Cache Performance Modeling and Prediction Tool (CPMP), employs trace-driven simulation techniques without the overhead of generating and managing detailed address traces. CPMP predicts the cache performance impact of source code level "what-if" modifications in a program to assist a user in the tuning process. CPMP is built on top of a customized version of the Computer Aided Parallelization Tools (CAPTools) environment. Finally, we demonstrate how CPMP can be applied to tune a real Computational Fluid Dynamics (CFD) application.

Mobile Thread Task Manager

NASA Technical Reports Server (NTRS)

Clement, Bradley J.; Estlin, Tara A.; Bornstein, Benjamin J.

2013-01-01

The Mobile Thread Task Manager (MTTM) is being applied to parallelizing existing flight software to understand the benefits and to develop new techniques and architectural concepts for adapting software to multicore architectures. It allocates and load-balances tasks for a group of threads that migrate across processors to improve cache performance. In order to balance-load across threads, the MTTM augments a basic map-reduce strategy to draw jobs from a global queue. In a multicore processor, memory may be "homed" to the cache of a specific processor and must be accessed from that processor. The MTTB architecture wraps access to data with thread management to move threads to the home processor for that data so that the computation follows the data in an attempt to avoid L2 cache misses. Cache homing is also handled by a memory manager that translates identifiers to processor IDs where the data will be homed (according to rules defined by the user). The user can also specify the number of threads and processors separately, which is important for tuning performance for different patterns of computation and memory access. MTTM efficiently processes tasks in parallel on a multiprocessor computer. It also provides an interface to make it easier to adapt existing software to a multiprocessor environment.

Comparing the OpenMP, MPI, and Hybrid Programming Paradigm on an SMP Cluster

NASA Technical Reports Server (NTRS)

Jost, Gabriele; Jin, Haoqiang; anMey, Dieter; Hatay, Ferhat F.

2003-01-01

With the advent of parallel hardware and software technologies users are faced with the challenge to choose a programming paradigm best suited for the underlying computer architecture. With the current trend in parallel computer architectures towards clusters of shared memory symmetric multi-processors (SMP), parallel programming techniques have evolved to support parallelism beyond a single level. Which programming paradigm is the best will depend on the nature of the given problem, the hardware architecture, and the available software. In this study we will compare different programming paradigms for the parallelization of a selected benchmark application on a cluster of SMP nodes. We compare the timings of different implementations of the same CFD benchmark application employing the same numerical algorithm on a cluster of Sun Fire SMP nodes. The rest of the paper is structured as follows: In section 2 we briefly discuss the programming models under consideration. We describe our compute platform in section 3. The different implementations of our benchmark code are described in section 4 and the performance results are presented in section 5. We conclude our study in section 6.

The DANTE Boltzmann transport solver: An unstructured mesh, 3-D, spherical harmonics algorithm compatible with parallel computer architectures

DOE Office of Scientific and Technical Information (OSTI.GOV)

McGhee, J.M.; Roberts, R.M.; Morel, J.E.

1997-06-01

A spherical harmonics research code (DANTE) has been developed which is compatible with parallel computer architectures. DANTE provides 3-D, multi-material, deterministic, transport capabilities using an arbitrary finite element mesh. The linearized Boltzmann transport equation is solved in a second order self-adjoint form utilizing a Galerkin finite element spatial differencing scheme. The core solver utilizes a preconditioned conjugate gradient algorithm. Other distinguishing features of the code include options for discrete-ordinates and simplified spherical harmonics angular differencing, an exact Marshak boundary treatment for arbitrarily oriented boundary faces, in-line matrix construction techniques to minimize memory consumption, and an effective diffusion based preconditioner formore » scattering dominated problems. Algorithm efficiency is demonstrated for a massively parallel SIMD architecture (CM-5), and compatibility with MPP multiprocessor platforms or workstation clusters is anticipated.« less

Long-range interactions and parallel scalability in molecular simulations

NASA Astrophysics Data System (ADS)

Patra, Michael; Hyvönen, Marja T.; Falck, Emma; Sabouri-Ghomi, Mohsen; Vattulainen, Ilpo; Karttunen, Mikko

2007-01-01

Typical biomolecular systems such as cellular membranes, DNA, and protein complexes are highly charged. Thus, efficient and accurate treatment of electrostatic interactions is of great importance in computational modeling of such systems. We have employed the GROMACS simulation package to perform extensive benchmarking of different commonly used electrostatic schemes on a range of computer architectures (Pentium-4, IBM Power 4, and Apple/IBM G5) for single processor and parallel performance up to 8 nodes—we have also tested the scalability on four different networks, namely Infiniband, GigaBit Ethernet, Fast Ethernet, and nearly uniform memory architecture, i.e. communication between CPUs is possible by directly reading from or writing to other CPUs' local memory. It turns out that the particle-mesh Ewald method (PME) performs surprisingly well and offers competitive performance unless parallel runs on PC hardware with older network infrastructure are needed. Lipid bilayers of sizes 128, 512 and 2048 lipid molecules were used as the test systems representing typical cases encountered in biomolecular simulations. Our results enable an accurate prediction of computational speed on most current computing systems, both for serial and parallel runs. These results should be helpful in, for example, choosing the most suitable configuration for a small departmental computer cluster.

Novel memory architecture for video signal processor

NASA Astrophysics Data System (ADS)

Hung, Jen-Sheng; Lin, Chia-Hsing; Jen, Chein-Wei

1993-11-01

An on-chip memory architecture for video signal processor (VSP) is proposed. This memory structure is a two-level design for the different data locality in video applications. The upper level--Memory A provides enough storage capacity to reduce the impact on the limitation of chip I/O bandwidth, and the lower level--Memory B provides enough data parallelism and flexibility to meet the requirements of multiple reconfigurable pipeline function units in a single VSP chip. The needed memory size is decided by the memory usage analysis for video algorithms and the number of function units. Both levels of memory adopted a dual-port memory scheme to sustain the simultaneous read and write operations. Especially, Memory B uses multiple one-read-one-write memory banks to emulate the real multiport memory. Therefore, one can change the configuration of Memory B to several sets of memories with variable read/write ports by adjusting the bus switches. Then the numbers of read ports and write ports in proposed memory can meet requirement of data flow patterns in different video coding algorithms. We have finished the design of a prototype memory design using 1.2- micrometers SPDM SRAM technology and will fabricated it through TSMC, in Taiwan.

Implementation of collisions on GPU architecture in the Vorpal code

NASA Astrophysics Data System (ADS)

Leddy, Jarrod; Averkin, Sergey; Cowan, Ben; Sides, Scott; Werner, Greg; Cary, John

2017-10-01

The Vorpal code contains a variety of collision operators allowing for the simulation of plasmas containing multiple charge species interacting with neutrals, background gas, and EM fields. These existing algorithms have been improved and reimplemented to take advantage of the massive parallelization allowed by GPU architecture. The use of GPUs is most effective when algorithms are single-instruction multiple-data, so particle collisions are an ideal candidate for this parallelization technique due to their nature as a series of independent processes with the same underlying operation. This refactoring required data memory reorganization and careful consideration of device/host data allocation to minimize memory access and data communication per operation. Successful implementation has resulted in an order of magnitude increase in simulation speed for a test-case involving multiple binary collisions using the null collision method. Work supported by DARPA under contract W31P4Q-16-C-0009.

Second International Workshop on Software Engineering and Code Design in Parallel Meteorological and Oceanographic Applications

NASA Technical Reports Server (NTRS)

OKeefe, Matthew (Editor); Kerr, Christopher L. (Editor)

1998-01-01

This report contains the abstracts and technical papers from the Second International Workshop on Software Engineering and Code Design in Parallel Meteorological and Oceanographic Applications, held June 15-18, 1998, in Scottsdale, Arizona. The purpose of the workshop is to bring together software developers in meteorology and oceanography to discuss software engineering and code design issues for parallel architectures, including Massively Parallel Processors (MPP's), Parallel Vector Processors (PVP's), Symmetric Multi-Processors (SMP's), Distributed Shared Memory (DSM) multi-processors, and clusters. Issues to be discussed include: (1) code architectures for current parallel models, including basic data structures, storage allocation, variable naming conventions, coding rules and styles, i/o and pre/post-processing of data; (2) designing modular code; (3) load balancing and domain decomposition; (4) techniques that exploit parallelism efficiently yet hide the machine-related details from the programmer; (5) tools for making the programmer more productive; and (6) the proliferation of programming models (F--, OpenMP, MPI, and HPF).

A mixed parallel strategy for the solution of coupled multi-scale problems at finite strains

NASA Astrophysics Data System (ADS)

Lopes, I. A. Rodrigues; Pires, F. M. Andrade; Reis, F. J. P.

2018-02-01

A mixed parallel strategy for the solution of homogenization-based multi-scale constitutive problems undergoing finite strains is proposed. The approach aims to reduce the computational time and memory requirements of non-linear coupled simulations that use finite element discretization at both scales (FE^2). In the first level of the algorithm, a non-conforming domain decomposition technique, based on the FETI method combined with a mortar discretization at the interface of macroscopic subdomains, is employed. A master-slave scheme, which distributes tasks by macroscopic element and adopts dynamic scheduling, is then used for each macroscopic subdomain composing the second level of the algorithm. This strategy allows the parallelization of FE^2 simulations in computers with either shared memory or distributed memory architectures. The proposed strategy preserves the quadratic rates of asymptotic convergence that characterize the Newton-Raphson scheme. Several examples are presented to demonstrate the robustness and efficiency of the proposed parallel strategy.

Ultra-High Density Holographic Memory Module with Solid-State Architecture

NASA Technical Reports Server (NTRS)

Markov, Vladimir B.

2000-01-01

NASA's terrestrial. space, and deep-space missions require technology that allows storing. retrieving, and processing a large volume of information. Holographic memory offers high-density data storage with parallel access and high throughput. Several methods exist for data multiplexing based on the fundamental principles of volume hologram selectivity. We recently demonstrated that a spatial (amplitude-phase) encoding of the reference wave (SERW) looks promising as a way to increase the storage density. The SERW hologram offers a method other than traditional methods of selectivity, such as spatial de-correlation between recorded and reconstruction fields, In this report we present the experimental results of the SERW-hologram memory module with solid-state architecture, which is of particular interest for space operations.

Research in Parallel Algorithms and Software for Computational Aerosciences

NASA Technical Reports Server (NTRS)

Domel, Neal D.

1996-01-01

Phase I is complete for the development of a Computational Fluid Dynamics parallel code with automatic grid generation and adaptation for the Euler analysis of flow over complex geometries. SPLITFLOW, an unstructured Cartesian grid code developed at Lockheed Martin Tactical Aircraft Systems, has been modified for a distributed memory/massively parallel computing environment. The parallel code is operational on an SGI network, Cray J90 and C90 vector machines, SGI Power Challenge, and Cray T3D and IBM SP2 massively parallel machines. Parallel Virtual Machine (PVM) is the message passing protocol for portability to various architectures. A domain decomposition technique was developed which enforces dynamic load balancing to improve solution speed and memory requirements. A host/node algorithm distributes the tasks. The solver parallelizes very well, and scales with the number of processors. Partially parallelized and non-parallelized tasks consume most of the wall clock time in a very fine grain environment. Timing comparisons on a Cray C90 demonstrate that Parallel SPLITFLOW runs 2.4 times faster on 8 processors than its non-parallel counterpart autotasked over 8 processors.

Research in Parallel Algorithms and Software for Computational Aerosciences

NASA Technical Reports Server (NTRS)

Domel, Neal D.

1996-01-01

Phase 1 is complete for the development of a computational fluid dynamics CFD) parallel code with automatic grid generation and adaptation for the Euler analysis of flow over complex geometries. SPLITFLOW, an unstructured Cartesian grid code developed at Lockheed Martin Tactical Aircraft Systems, has been modified for a distributed memory/massively parallel computing environment. The parallel code is operational on an SGI network, Cray J90 and C90 vector machines, SGI Power Challenge, and Cray T3D and IBM SP2 massively parallel machines. Parallel Virtual Machine (PVM) is the message passing protocol for portability to various architectures. A domain decomposition technique was developed which enforces dynamic load balancing to improve solution speed and memory requirements. A host/node algorithm distributes the tasks. The solver parallelizes very well, and scales with the number of processors. Partially parallelized and non-parallelized tasks consume most of the wall clock time in a very fine grain environment. Timing comparisons on a Cray C90 demonstrate that Parallel SPLITFLOW runs 2.4 times faster on 8 processors than its non-parallel counterpart autotasked over 8 processors.

Streaming parallel GPU acceleration of large-scale filter-based spiking neural networks.

PubMed

Slażyński, Leszek; Bohte, Sander

2012-01-01

The arrival of graphics processing (GPU) cards suitable for massively parallel computing promises affordable large-scale neural network simulation previously only available at supercomputing facilities. While the raw numbers suggest that GPUs may outperform CPUs by at least an order of magnitude, the challenge is to develop fine-grained parallel algorithms to fully exploit the particulars of GPUs. Computation in a neural network is inherently parallel and thus a natural match for GPU architectures: given inputs, the internal state for each neuron can be updated in parallel. We show that for filter-based spiking neurons, like the Spike Response Model, the additive nature of membrane potential dynamics enables additional update parallelism. This also reduces the accumulation of numerical errors when using single precision computation, the native precision of GPUs. We further show that optimizing simulation algorithms and data structures to the GPU's architecture has a large pay-off: for example, matching iterative neural updating to the memory architecture of the GPU speeds up this simulation step by a factor of three to five. With such optimizations, we can simulate in better-than-realtime plausible spiking neural networks of up to 50 000 neurons, processing over 35 million spiking events per second.

Optimizing SIEM Throughput on the Cloud Using Parallelization.

PubMed

Alam, Masoom; Ihsan, Asif; Khan, Muazzam A; Javaid, Qaisar; Khan, Abid; Manzoor, Jawad; Akhundzada, Adnan; Khan, Muhammad Khurram; Farooq, Sajid

2016-01-01

Processing large amounts of data in real time for identifying security issues pose several performance challenges, especially when hardware infrastructure is limited. Managed Security Service Providers (MSSP), mostly hosting their applications on the Cloud, receive events at a very high rate that varies from a few hundred to a couple of thousand events per second (EPS). It is critical to process this data efficiently, so that attacks could be identified quickly and necessary response could be initiated. This paper evaluates the performance of a security framework OSTROM built on the Esper complex event processing (CEP) engine under a parallel and non-parallel computational framework. We explain three architectures under which Esper can be used to process events. We investigated the effect on throughput, memory and CPU usage in each configuration setting. The results indicate that the performance of the engine is limited by the number of events coming in rather than the queries being processed. The architecture where 1/4th of the total events are submitted to each instance and all the queries are processed by all the units shows best results in terms of throughput, memory and CPU usage.

Parallel machine architecture for production rule systems

DOEpatents

Allen, Jr., John D.; Butler, Philip L.

1989-01-01

A parallel processing system for production rule programs utilizes a host processor for storing production rule right hand sides (RHS) and a plurality of rule processors for storing left hand sides (LHS). The rule processors operate in parallel in the recognize phase of the system recognize -Act Cycle to match their respective LHS's against a stored list of working memory elements (WME) in order to find a self consistent set of WME's. The list of WME is dynamically varied during the Act phase of the system in which the host executes or fires rule RHS's for those rules for which a self-consistent set has been found by the rule processors. The host transmits instructions for creating or deleting working memory elements as dictated by the rule firings until the rule processors are unable to find any further self-consistent working memory element sets at which time the production rule system is halted.

Inflated speedups in parallel simulations via malloc()

NASA Technical Reports Server (NTRS)

Nicol, David M.

1990-01-01

Discrete-event simulation programs make heavy use of dynamic memory allocation in order to support simulation's very dynamic space requirements. When programming in C one is likely to use the malloc() routine. However, a parallel simulation which uses the standard Unix System V malloc() implementation may achieve an overly optimistic speedup, possibly superlinear. An alternate implementation provided on some (but not all systems) can avoid the speedup anomaly, but at the price of significantly reduced available free space. This is especially severe on most parallel architectures, which tend not to support virtual memory. It is shown how a simply implemented user-constructed interface to malloc() can both avoid artificially inflated speedups, and make efficient use of the dynamic memory space. The interface simply catches blocks on the basis of their size. The problem is demonstrated empirically, and the effectiveness of the solution is shown both empirically and analytically.

GPU-completeness: theory and implications

NASA Astrophysics Data System (ADS)

Lin, I.-Jong

2011-01-01

This paper formalizes a major insight into a class of algorithms that relate parallelism and performance. The purpose of this paper is to define a class of algorithms that trades off parallelism for quality of result (e.g. visual quality, compression rate), and we propose a similar method for algorithmic classification based on NP-Completeness techniques, applied toward parallel acceleration. We will define this class of algorithm as "GPU-Complete" and will postulate the necessary properties of the algorithms for admission into this class. We will also formally relate his algorithmic space and imaging algorithms space. This concept is based upon our experience in the print production area where GPUs (Graphic Processing Units) have shown a substantial cost/performance advantage within the context of HPdelivered enterprise services and commercial printing infrastructure. While CPUs and GPUs are converging in their underlying hardware and functional blocks, their system behaviors are clearly distinct in many ways: memory system design, programming paradigms, and massively parallel SIMD architecture. There are applications that are clearly suited to each architecture: for CPU: language compilation, word processing, operating systems, and other applications that are highly sequential in nature; for GPU: video rendering, particle simulation, pixel color conversion, and other problems clearly amenable to massive parallelization. While GPUs establishing themselves as a second, distinct computing architecture from CPUs, their end-to-end system cost/performance advantage in certain parts of computation inform the structure of algorithms and their efficient parallel implementations. While GPUs are merely one type of architecture for parallelization, we show that their introduction into the design space of printing systems demonstrate the trade-offs against competing multi-core, FPGA, and ASIC architectures. While each architecture has its own optimal application, we believe that the selection of architecture can be defined in terms of properties of GPU-Completeness. For a welldefined subset of algorithms, GPU-Completeness is intended to connect the parallelism, algorithms and efficient architectures into a unified framework to show that multiple layers of parallel implementation are guided by the same underlying trade-off.

A Framework for Parallel Unstructured Grid Generation for Complex Aerodynamic Simulations

NASA Technical Reports Server (NTRS)

Zagaris, George; Pirzadeh, Shahyar Z.; Chrisochoides, Nikos

2009-01-01

A framework for parallel unstructured grid generation targeting both shared memory multi-processors and distributed memory architectures is presented. The two fundamental building-blocks of the framework consist of: (1) the Advancing-Partition (AP) method used for domain decomposition and (2) the Advancing Front (AF) method used for mesh generation. Starting from the surface mesh of the computational domain, the AP method is applied recursively to generate a set of sub-domains. Next, the sub-domains are meshed in parallel using the AF method. The recursive nature of domain decomposition naturally maps to a divide-and-conquer algorithm which exhibits inherent parallelism. For the parallel implementation, the Master/Worker pattern is employed to dynamically balance the varying workloads of each task on the set of available CPUs. Performance results by this approach are presented and discussed in detail as well as future work and improvements.

Manyscale Computing for Sensor Processing in Support of Space Situational Awareness

NASA Astrophysics Data System (ADS)

Schmalz, M.; Chapman, W.; Hayden, E.; Sahni, S.; Ranka, S.

2014-09-01

Increasing image and signal data burden associated with sensor data processing in support of space situational awareness implies continuing computational throughput growth beyond the petascale regime. In addition to growing applications data burden and diversity, the breadth, diversity and scalability of high performance computing architectures and their various organizations challenge the development of a single, unifying, practicable model of parallel computation. Therefore, models for scalable parallel processing have exploited architectural and structural idiosyncrasies, yielding potential misapplications when legacy programs are ported among such architectures. In response to this challenge, we have developed a concise, efficient computational paradigm and software called Manyscale Computing to facilitate efficient mapping of annotated application codes to heterogeneous parallel architectures. Our theory, algorithms, software, and experimental results support partitioning and scheduling of application codes for envisioned parallel architectures, in terms of work atoms that are mapped (for example) to threads or thread blocks on computational hardware. Because of the rigor, completeness, conciseness, and layered design of our manyscale approach, application-to-architecture mapping is feasible and scalable for architectures at petascales, exascales, and above. Further, our methodology is simple, relying primarily on a small set of primitive mapping operations and support routines that are readily implemented on modern parallel processors such as graphics processing units (GPUs) and hybrid multi-processors (HMPs). In this paper, we overview the opportunities and challenges of manyscale computing for image and signal processing in support of space situational awareness applications. We discuss applications in terms of a layered hardware architecture (laboratory > supercomputer > rack > processor > component hierarchy). Demonstration applications include performance analysis and results in terms of execution time as well as storage, power, and energy consumption for bus-connected and/or networked architectures. The feasibility of the manyscale paradigm is demonstrated by addressing four principal challenges: (1) architectural/structural diversity, parallelism, and locality, (2) masking of I/O and memory latencies, (3) scalability of design as well as implementation, and (4) efficient representation/expression of parallel applications. Examples will demonstrate how manyscale computing helps solve these challenges efficiently on real-world computing systems.

Locality Aware Concurrent Start for Stencil Applications

DOE Office of Scientific and Technical Information (OSTI.GOV)

Shrestha, Sunil; Gao, Guang R.; Manzano Franco, Joseph B.

Stencil computations are at the heart of many physical simulations used in scientific codes. Thus, there exists a plethora of optimization efforts for this family of computations. Among these techniques, tiling techniques that allow concurrent start have proven to be very efficient in providing better performance for these critical kernels. Nevertheless, with many core designs being the norm, these optimization techniques might not be able to fully exploit locality (both spatial and temporal) on multiple levels of the memory hierarchy without compromising parallelism. It is no longer true that the machine can be seen as a homogeneous collection of nodesmore » with caches, main memory and an interconnect network. New architectural designs exhibit complex grouping of nodes, cores, threads, caches and memory connected by an ever evolving network-on-chip design. These new designs may benefit greatly from carefully crafted schedules and groupings that encourage parallel actors (i.e. threads, cores or nodes) to be aware of the computational history of other actors in close proximity. In this paper, we provide an efficient tiling technique that allows hierarchical concurrent start for memory hierarchy aware tile groups. Each execution schedule and tile shape exploit the available parallelism, load balance and locality present in the given applications. We demonstrate our technique on the Intel Xeon Phi architecture with selected and representative stencil kernels. We show improvement ranging from 5.58% to 31.17% over existing state-of-the-art techniques.« less

Efficient and flexible memory architecture to alleviate data and context bandwidth bottlenecks of coarse-grained reconfigurable arrays

NASA Astrophysics Data System (ADS)

Yang, Chen; Liu, LeiBo; Yin, ShouYi; Wei, ShaoJun

2014-12-01

The computational capability of a coarse-grained reconfigurable array (CGRA) can be significantly restrained due to data and context memory bandwidth bottlenecks. Traditionally, two methods have been used to resolve this problem. One method loads the context into the CGRA at run time. This method occupies very small on-chip memory but induces very large latency, which leads to low computational efficiency. The other method adopts a multi-context structure. This method loads the context into the on-chip context memory at the boot phase. Broadcasting the pointer of a set of contexts changes the hardware configuration on a cycle-by-cycle basis. The size of the context memory induces a large area overhead in multi-context structures, which results in major restrictions on application complexity. This paper proposes a Predictable Context Cache (PCC) architecture to address the above context issues by buffering the context inside a CGRA. In this architecture, context is dynamically transferred into the CGRA. Utilizing a PCC significantly reduces the on-chip context memory and the complexity of the applications running on the CGRA is no longer restricted by the size of the on-chip context memory. Data preloading is the most frequently used approach to hide input data latency and speed up the data transmission process for the data bandwidth issue. Rather than fundamentally reducing the amount of input data, the transferred data and computations are processed in parallel. However, the data preloading method cannot work efficiently because data transmission becomes the critical path as the reconfigurable array scale increases. This paper also presents a Hierarchical Data Memory (HDM) architecture as a solution to the efficiency problem. In this architecture, high internal bandwidth is provided to buffer both reused input data and intermediate data. The HDM architecture relieves the external memory from the data transfer burden so that the performance is significantly improved. As a result of using PCC and HDM, experiments running mainstream video decoding programs achieved performance improvements of 13.57%-19.48% when there was a reasonable memory size. Therefore, 1080p@35.7fps for H.264 high profile video decoding can be achieved on PCC and HDM architecture when utilizing a 200 MHz working frequency. Further, the size of the on-chip context memory no longer restricted complex applications, which were efficiently executed on the PCC and HDM architecture.

A GPU-Accelerated Approach for Feature Tracking in Time-Varying Imagery Datasets.

PubMed

Peng, Chao; Sahani, Sandip; Rushing, John

2017-10-01

We propose a novel parallel connected component labeling (CCL) algorithm along with efficient out-of-core data management to detect and track feature regions of large time-varying imagery datasets. Our approach contributes to the big data field with parallel algorithms tailored for GPU architectures. We remove the data dependency between frames and achieve pixel-level parallelism. Due to the large size, the entire dataset cannot fit into cached memory. Frames have to be streamed through the memory hierarchy (disk to CPU main memory and then to GPU memory), partitioned, and processed as batches, where each batch is small enough to fit into the GPU. To reconnect the feature regions that are separated due to data partitioning, we present a novel batch merging algorithm to extract the region connection information across multiple batches in a parallel fashion. The information is organized in a memory-efficient structure and supports fast indexing on the GPU. Our experiment uses a commodity workstation equipped with a single GPU. The results show that our approach can efficiently process a weather dataset composed of terabytes of time-varying radar images. The advantages of our approach are demonstrated by comparing to the performance of an efficient CPU cluster implementation which is being used by the weather scientists.

Real-time FPGA architectures for computer vision

NASA Astrophysics Data System (ADS)

Arias-Estrada, Miguel; Torres-Huitzil, Cesar

2000-03-01

This paper presents an architecture for real-time generic convolution of a mask and an image. The architecture is intended for fast low level image processing. The FPGA-based architecture takes advantage of the availability of registers in FPGAs to implement an efficient and compact module to process the convolutions. The architecture is designed to minimize the number of accesses to the image memory and is based on parallel modules with internal pipeline operation in order to improve its performance. The architecture is prototyped in a FPGA, but it can be implemented on a dedicated VLSI to reach higher clock frequencies. Complexity issues, FPGA resources utilization, FPGA limitations, and real time performance are discussed. Some results are presented and discussed.

Efficient Graph Based Assembly of Short-Read Sequences on Hybrid Core Architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Sczyrba, Alex; Pratap, Abhishek; Canon, Shane

2011-03-22

Advanced architectures can deliver dramatically increased throughput for genomics and proteomics applications, reducing time-to-completion in some cases from days to minutes. One such architecture, hybrid-core computing, marries a traditional x86 environment with a reconfigurable coprocessor, based on field programmable gate array (FPGA) technology. In addition to higher throughput, increased performance can fundamentally improve research quality by allowing more accurate, previously impractical approaches. We will discuss the approach used by Convey?s de Bruijn graph constructor for short-read, de-novo assembly. Bioinformatics applications that have random access patterns to large memory spaces, such as graph-based algorithms, experience memory performance limitations on cache-based x86more » servers. Convey?s highly parallel memory subsystem allows application-specific logic to simultaneously access 8192 individual words in memory, significantly increasing effective memory bandwidth over cache-based memory systems. Many algorithms, such as Velvet and other de Bruijn graph based, short-read, de-novo assemblers, can greatly benefit from this type of memory architecture. Furthermore, small data type operations (four nucleotides can be represented in two bits) make more efficient use of logic gates than the data types dictated by conventional programming models.JGI is comparing the performance of Convey?s graph constructor and Velvet on both synthetic and real data. We will present preliminary results on memory usage and run time metrics for various data sets with different sizes, from small microbial and fungal genomes to very large cow rumen metagenome. For genomes with references we will also present assembly quality comparisons between the two assemblers.« less

A flexible algorithm for calculating pair interactions on SIMD architectures

NASA Astrophysics Data System (ADS)

Páll, Szilárd; Hess, Berk

2013-12-01

Calculating interactions or correlations between pairs of particles is typically the most time-consuming task in particle simulation or correlation analysis. Straightforward implementations using a double loop over particle pairs have traditionally worked well, especially since compilers usually do a good job of unrolling the inner loop. In order to reach high performance on modern CPU and accelerator architectures, single-instruction multiple-data (SIMD) parallelization has become essential. Avoiding memory bottlenecks is also increasingly important and requires reducing the ratio of memory to arithmetic operations. Moreover, when pairs only interact within a certain cut-off distance, good SIMD utilization can only be achieved by reordering input and output data, which quickly becomes a limiting factor. Here we present an algorithm for SIMD parallelization based on grouping a fixed number of particles, e.g. 2, 4, or 8, into spatial clusters. Calculating all interactions between particles in a pair of such clusters improves data reuse compared to the traditional scheme and results in a more efficient SIMD parallelization. Adjusting the cluster size allows the algorithm to map to SIMD units of various widths. This flexibility not only enables fast and efficient implementation on current CPUs and accelerator architectures like GPUs or Intel MIC, but it also makes the algorithm future-proof. We present the algorithm with an application to molecular dynamics simulations, where we can also make use of the effective buffering the method introduces.

Parallel, Distributed Scripting with Python

DOE Office of Scientific and Technical Information (OSTI.GOV)

Miller, P J

2002-05-24

Parallel computers used to be, for the most part, one-of-a-kind systems which were extremely difficult to program portably. With SMP architectures, the advent of the POSIX thread API and OpenMP gave developers ways to portably exploit on-the-box shared memory parallelism. Since these architectures didn't scale cost-effectively, distributed memory clusters were developed. The associated MPI message passing libraries gave these systems a portable paradigm too. Having programmers effectively use this paradigm is a somewhat different question. Distributed data has to be explicitly transported via the messaging system in order for it to be useful. In high level languages, the MPI librarymore » gives access to data distribution routines in C, C++, and FORTRAN. But we need more than that. Many reasonable and common tasks are best done in (or as extensions to) scripting languages. Consider sysadm tools such as password crackers, file purgers, etc ... These are simple to write in a scripting language such as Python (an open source, portable, and freely available interpreter). But these tasks beg to be done in parallel. Consider the a password checker that checks an encrypted password against a 25,000 word dictionary. This can take around 10 seconds in Python (6 seconds in C). It is trivial to parallelize if you can distribute the information and co-ordinate the work.« less

The AIS-5000 parallel processor

DOE Office of Scientific and Technical Information (OSTI.GOV)

Schmitt, L.A.; Wilson, S.S.

1988-05-01

The AIS-5000 is a commercially available massively parallel processor which has been designed to operate in an industrial environment. It has fine-grained parallelism with up to 1024 processing elements arranged in a single-instruction multiple-data (SIMD) architecture. The processing elements are arranged in a one-dimensional chain that, for computer vision applications, can be as wide as the image itself. This architecture has superior cost/performance characteristics than two-dimensional mesh-connected systems. The design of the processing elements and their interconnections as well as the software used to program the system allow a wide variety of algorithms and applications to be implemented. In thismore » paper, the overall architecture of the system is described. Various components of the system are discussed, including details of the processing elements, data I/O pathways and parallel memory organization. A virtual two-dimensional model for programming image-based algorithms for the system is presented. This model is supported by the AIS-5000 hardware and software and allows the system to be treated as a full-image-size, two-dimensional, mesh-connected parallel processor. Performance bench marks are given for certain simple and complex functions.« less

Memory-Intensive Benchmarks: IRAM vs. Cache-Based Machines

NASA Technical Reports Server (NTRS)

Biswas, Rupak; Gaeke, Brian R.; Husbands, Parry; Li, Xiaoye S.; Oliker, Leonid; Yelick, Katherine A.; Biegel, Bryan (Technical Monitor)

2002-01-01

The increasing gap between processor and memory performance has lead to new architectural models for memory-intensive applications. In this paper, we explore the performance of a set of memory-intensive benchmarks and use them to compare the performance of conventional cache-based microprocessors to a mixed logic and DRAM processor called VIRAM. The benchmarks are based on problem statements, rather than specific implementations, and in each case we explore the fundamental hardware requirements of the problem, as well as alternative algorithms and data structures that can help expose fine-grained parallelism or simplify memory access patterns. The benchmarks are characterized by their memory access patterns, their basic control structures, and the ratio of computation to memory operation.

Parallel Hough Transform-Based Straight Line Detection and Its FPGA Implementation in Embedded Vision

PubMed Central

Lu, Xiaofeng; Song, Li; Shen, Sumin; He, Kang; Yu, Songyu; Ling, Nam

2013-01-01

Hough Transform has been widely used for straight line detection in low-definition and still images, but it suffers from execution time and resource requirements. Field Programmable Gate Arrays (FPGA) provide a competitive alternative for hardware acceleration to reap tremendous computing performance. In this paper, we propose a novel parallel Hough Transform (PHT) and FPGA architecture-associated framework for real-time straight line detection in high-definition videos. A resource-optimized Canny edge detection method with enhanced non-maximum suppression conditions is presented to suppress most possible false edges and obtain more accurate candidate edge pixels for subsequent accelerated computation. Then, a novel PHT algorithm exploiting spatial angle-level parallelism is proposed to upgrade computational accuracy by improving the minimum computational step. Moreover, the FPGA based multi-level pipelined PHT architecture optimized by spatial parallelism ensures real-time computation for 1,024 × 768 resolution videos without any off-chip memory consumption. This framework is evaluated on ALTERA DE2-115 FPGA evaluation platform at a maximum frequency of 200 MHz, and it can calculate straight line parameters in 15.59 ms on the average for one frame. Qualitative and quantitative evaluation results have validated the system performance regarding data throughput, memory bandwidth, resource, speed and robustness. PMID:23867746

Parallel Hough Transform-based straight line detection and its FPGA implementation in embedded vision.

PubMed

Lu, Xiaofeng; Song, Li; Shen, Sumin; He, Kang; Yu, Songyu; Ling, Nam

2013-07-17

Hough Transform has been widely used for straight line detection in low-definition and still images, but it suffers from execution time and resource requirements. Field Programmable Gate Arrays (FPGA) provide a competitive alternative for hardware acceleration to reap tremendous computing performance. In this paper, we propose a novel parallel Hough Transform (PHT) and FPGA architecture-associated framework for real-time straight line detection in high-definition videos. A resource-optimized Canny edge detection method with enhanced non-maximum suppression conditions is presented to suppress most possible false edges and obtain more accurate candidate edge pixels for subsequent accelerated computation. Then, a novel PHT algorithm exploiting spatial angle-level parallelism is proposed to upgrade computational accuracy by improving the minimum computational step. Moreover, the FPGA based multi-level pipelined PHT architecture optimized by spatial parallelism ensures real-time computation for 1,024 × 768 resolution videos without any off-chip memory consumption. This framework is evaluated on ALTERA DE2-115 FPGA evaluation platform at a maximum frequency of 200 MHz, and it can calculate straight line parameters in 15.59 ms on the average for one frame. Qualitative and quantitative evaluation results have validated the system performance regarding data throughput, memory bandwidth, resource, speed and robustness.

Using CLIPS in the domain of knowledge-based massively parallel programming

NASA Technical Reports Server (NTRS)

Dvorak, Jiri J.

1994-01-01

The Program Development Environment (PDE) is a tool for massively parallel programming of distributed-memory architectures. Adopting a knowledge-based approach, the PDE eliminates the complexity introduced by parallel hardware with distributed memory and offers complete transparency in respect of parallelism exploitation. The knowledge-based part of the PDE is realized in CLIPS. Its principal task is to find an efficient parallel realization of the application specified by the user in a comfortable, abstract, domain-oriented formalism. A large collection of fine-grain parallel algorithmic skeletons, represented as COOL objects in a tree hierarchy, contains the algorithmic knowledge. A hybrid knowledge base with rule modules and procedural parts, encoding expertise about application domain, parallel programming, software engineering, and parallel hardware, enables a high degree of automation in the software development process. In this paper, important aspects of the implementation of the PDE using CLIPS and COOL are shown, including the embedding of CLIPS with C++-based parts of the PDE. The appropriateness of the chosen approach and of the CLIPS language for knowledge-based software engineering are discussed.

Bayer image parallel decoding based on GPU

NASA Astrophysics Data System (ADS)

Hu, Rihui; Xu, Zhiyong; Wei, Yuxing; Sun, Shaohua

2012-11-01

In the photoelectrical tracking system, Bayer image is decompressed in traditional method, which is CPU-based. However, it is too slow when the images become large, for example, 2K×2K×16bit. In order to accelerate the Bayer image decoding, this paper introduces a parallel speedup method for NVIDA's Graphics Processor Unit (GPU) which supports CUDA architecture. The decoding procedure can be divided into three parts: the first is serial part, the second is task-parallelism part, and the last is data-parallelism part including inverse quantization, inverse discrete wavelet transform (IDWT) as well as image post-processing part. For reducing the execution time, the task-parallelism part is optimized by OpenMP techniques. The data-parallelism part could advance its efficiency through executing on the GPU as CUDA parallel program. The optimization techniques include instruction optimization, shared memory access optimization, the access memory coalesced optimization and texture memory optimization. In particular, it can significantly speed up the IDWT by rewriting the 2D (Tow-dimensional) serial IDWT into 1D parallel IDWT. Through experimenting with 1K×1K×16bit Bayer image, data-parallelism part is 10 more times faster than CPU-based implementation. Finally, a CPU+GPU heterogeneous decompression system was designed. The experimental result shows that it could achieve 3 to 5 times speed increase compared to the CPU serial method.

Power and Performance Trade-offs for Space Time Adaptive Processing

DOE Office of Scientific and Technical Information (OSTI.GOV)

Gawande, Nitin A.; Manzano Franco, Joseph B.; Tumeo, Antonino

Computational efficiency – performance relative to power or energy – is one of the most important concerns when designing RADAR processing systems. This paper analyzes power and performance trade-offs for a typical Space Time Adaptive Processing (STAP) application. We study STAP implementations for CUDA and OpenMP on two computationally efficient architectures, Intel Haswell Core I7-4770TE and NVIDIA Kayla with a GK208 GPU. We analyze the power and performance of STAP’s computationally intensive kernels across the two hardware testbeds. We also show the impact and trade-offs of GPU optimization techniques. We show that data parallelism can be exploited for efficient implementationmore » on the Haswell CPU architecture. The GPU architecture is able to process large size data sets without increase in power requirement. The use of shared memory has a significant impact on the power requirement for the GPU. A balance between the use of shared memory and main memory access leads to an improved performance in a typical STAP application.« less

Spaceborne Processor Array

NASA Technical Reports Server (NTRS)

Chow, Edward T.; Schatzel, Donald V.; Whitaker, William D.; Sterling, Thomas

2008-01-01

A Spaceborne Processor Array in Multifunctional Structure (SPAMS) can lower the total mass of the electronic and structural overhead of spacecraft, resulting in reduced launch costs, while increasing the science return through dynamic onboard computing. SPAMS integrates the multifunctional structure (MFS) and the Gilgamesh Memory, Intelligence, and Network Device (MIND) multi-core in-memory computer architecture into a single-system super-architecture. This transforms every inch of a spacecraft into a sharable, interconnected, smart computing element to increase computing performance while simultaneously reducing mass. The MIND in-memory architecture provides a foundation for high-performance, low-power, and fault-tolerant computing. The MIND chip has an internal structure that includes memory, processing, and communication functionality. The Gilgamesh is a scalable system comprising multiple MIND chips interconnected to operate as a single, tightly coupled, parallel computer. The array of MIND components shares a global, virtual name space for program variables and tasks that are allocated at run time to the distributed physical memory and processing resources. Individual processor- memory nodes can be activated or powered down at run time to provide active power management and to configure around faults. A SPAMS system is comprised of a distributed Gilgamesh array built into MFS, interfaces into instrument and communication subsystems, a mass storage interface, and a radiation-hardened flight computer.

Automated problem scheduling and reduction of synchronization delay effects

NASA Technical Reports Server (NTRS)

Saltz, Joel H.

1987-01-01

It is anticipated that in order to make effective use of many future high performance architectures, programs will have to exhibit at least a medium grained parallelism. A framework is presented for partitioning very sparse triangular systems of linear equations that is designed to produce favorable preformance results in a wide variety of parallel architectures. Efficient methods for solving these systems are of interest because: (1) they provide a useful model problem for use in exploring heuristics for the aggregation, mapping and scheduling of relatively fine grained computations whose data dependencies are specified by directed acrylic graphs, and (2) because such efficient methods can find direct application in the development of parallel algorithms for scientific computation. Simple expressions are derived that describe how to schedule computational work with varying degrees of granularity. The Encore Multimax was used as a hardware simulator to investigate the performance effects of using the partitioning techniques presented in shared memory architectures with varying relative synchronization costs.

Parallel algorithms for modeling flow in permeable media. Annual report, February 15, 1995 - February 14, 1996

DOE Office of Scientific and Technical Information (OSTI.GOV)

G.A. Pope; K. Sephernoori; D.C. McKinney

1996-03-15

This report describes the application of distributed-memory parallel programming techniques to a compositional simulator called UTCHEM. The University of Texas Chemical Flooding reservoir simulator (UTCHEM) is a general-purpose vectorized chemical flooding simulator that models the transport of chemical species in three-dimensional, multiphase flow through permeable media. The parallel version of UTCHEM addresses solving large-scale problems by reducing the amount of time that is required to obtain the solution as well as providing a flexible and portable programming environment. In this work, the original parallel version of UTCHEM was modified and ported to CRAY T3D and CRAY T3E, distributed-memory, multiprocessor computersmore » using CRAY-PVM as the interprocessor communication library. Also, the data communication routines were modified such that the portability of the original code across different computer architectures was mad possible.« less

Parallel computation with the force

NASA Technical Reports Server (NTRS)

Jordan, H. F.

1985-01-01

A methodology, called the force, supports the construction of programs to be executed in parallel by a force of processes. The number of processes in the force is unspecified, but potentially very large. The force idea is embodied in a set of macros which produce multiproceossor FORTRAN code and has been studied on two shared memory multiprocessors of fairly different character. The method has simplified the writing of highly parallel programs within a limited class of parallel algorithms and is being extended to cover a broader class. The individual parallel constructs which comprise the force methodology are discussed. Of central concern are their semantics, implementation on different architectures and performance implications.

Aerodynamic Shape Optimization of Supersonic Aircraft Configurations via an Adjoint Formulation on Parallel Computers

NASA Technical Reports Server (NTRS)

Reuther, James; Alonso, Juan Jose; Rimlinger, Mark J.; Jameson, Antony

1996-01-01

This work describes the application of a control theory-based aerodynamic shape optimization method to the problem of supersonic aircraft design. The design process is greatly accelerated through the use of both control theory and a parallel implementation on distributed memory computers. Control theory is employed to derive the adjoint differential equations whose solution allows for the evaluation of design gradient information at a fraction of the computational cost required by previous design methods. The resulting problem is then implemented on parallel distributed memory architectures using a domain decomposition approach, an optimized communication schedule, and the MPI (Message Passing Interface) Standard for portability and efficiency. The final result achieves very rapid aerodynamic design based on higher order computational fluid dynamics methods (CFD). In our earlier studies, the serial implementation of this design method was shown to be effective for the optimization of airfoils, wings, wing-bodies, and complex aircraft configurations using both the potential equation and the Euler equations. In our most recent paper, the Euler method was extended to treat complete aircraft configurations via a new multiblock implementation. Furthermore, during the same conference, we also presented preliminary results demonstrating that this basic methodology could be ported to distributed memory parallel computing architectures. In this paper, our concern will be to demonstrate that the combined power of these new technologies can be used routinely in an industrial design environment by applying it to the case study of the design of typical supersonic transport configurations. A particular difficulty of this test case is posed by the propulsion/airframe integration.

Parallel protein secondary structure prediction based on neural networks.

PubMed

Zhong, Wei; Altun, Gulsah; Tian, Xinmin; Harrison, Robert; Tai, Phang C; Pan, Yi

2004-01-01

Protein secondary structure prediction has a fundamental influence on today's bioinformatics research. In this work, binary and tertiary classifiers of protein secondary structure prediction are implemented on Denoeux belief neural network (DBNN) architecture. Hydrophobicity matrix, orthogonal matrix, BLOSUM62 and PSSM (position specific scoring matrix) are experimented separately as the encoding schemes for DBNN. The experimental results contribute to the design of new encoding schemes. New binary classifier for Helix versus not Helix ( approximately H) for DBNN produces prediction accuracy of 87% when PSSM is used for the input profile. The performance of DBNN binary classifier is comparable to other best prediction methods. The good test results for binary classifiers open a new approach for protein structure prediction with neural networks. Due to the time consuming task of training the neural networks, Pthread and OpenMP are employed to parallelize DBNN in the hyperthreading enabled Intel architecture. Speedup for 16 Pthreads is 4.9 and speedup for 16 OpenMP threads is 4 in the 4 processors shared memory architecture. Both speedup performance of OpenMP and Pthread is superior to that of other research. With the new parallel training algorithm, thousands of amino acids can be processed in reasonable amount of time. Our research also shows that hyperthreading technology for Intel architecture is efficient for parallel biological algorithms.

Optimizing SIEM Throughput on the Cloud Using Parallelization

PubMed Central

Alam, Masoom; Ihsan, Asif; Javaid, Qaisar; Khan, Abid; Manzoor, Jawad; Akhundzada, Adnan; Khan, M Khurram; Farooq, Sajid

2016-01-01

Processing large amounts of data in real time for identifying security issues pose several performance challenges, especially when hardware infrastructure is limited. Managed Security Service Providers (MSSP), mostly hosting their applications on the Cloud, receive events at a very high rate that varies from a few hundred to a couple of thousand events per second (EPS). It is critical to process this data efficiently, so that attacks could be identified quickly and necessary response could be initiated. This paper evaluates the performance of a security framework OSTROM built on the Esper complex event processing (CEP) engine under a parallel and non-parallel computational framework. We explain three architectures under which Esper can be used to process events. We investigated the effect on throughput, memory and CPU usage in each configuration setting. The results indicate that the performance of the engine is limited by the number of events coming in rather than the queries being processed. The architecture where 1/4th of the total events are submitted to each instance and all the queries are processed by all the units shows best results in terms of throughput, memory and CPU usage. PMID:27851762

Performance prediction: A case study using a multi-ring KSR-1 machine

NASA Technical Reports Server (NTRS)

Sun, Xian-He; Zhu, Jianping

1995-01-01

While computers with tens of thousands of processors have successfully delivered high performance power for solving some of the so-called 'grand-challenge' applications, the notion of scalability is becoming an important metric in the evaluation of parallel machine architectures and algorithms. In this study, the prediction of scalability and its application are carefully investigated. A simple formula is presented to show the relation between scalability, single processor computing power, and degradation of parallelism. A case study is conducted on a multi-ring KSR1 shared virtual memory machine. Experimental and theoretical results show that the influence of topology variation of an architecture is predictable. Therefore, the performance of an algorithm on a sophisticated, heirarchical architecture can be predicted and the best algorithm-machine combination can be selected for a given application.

Optical resonators and neural networks

NASA Astrophysics Data System (ADS)

Anderson, Dana Z.

1986-08-01

It may be possible to implement neural network models using continuous field optical architectures. These devices offer the inherent parallelism of propagating waves and an information density in principle dictated by the wavelength of light and the quality of the bulk optical elements. Few components are needed to construct a relatively large equivalent network. Various associative memories based on optical resonators have been demonstrated in the literature, a ring resonator design is discussed in detail here. Information is stored in a holographic medium and recalled through a competitive processes in the gain medium supplying energy to the ring rsonator. The resonator memory is the first realized example of a neural network function implemented with this kind of architecture.

Parallel Navier-Stokes computations on shared and distributed memory architectures

NASA Technical Reports Server (NTRS)

Hayder, M. Ehtesham; Jayasimha, D. N.; Pillay, Sasi Kumar

1995-01-01

We study a high order finite difference scheme to solve the time accurate flow field of a jet using the compressible Navier-Stokes equations. As part of our ongoing efforts, we have implemented our numerical model on three parallel computing platforms to study the computational, communication, and scalability characteristics. The platforms chosen for this study are a cluster of workstations connected through fast networks (the LACE experimental testbed at NASA Lewis), a shared memory multiprocessor (the Cray YMP), and a distributed memory multiprocessor (the IBM SPI). Our focus in this study is on the LACE testbed. We present some results for the Cray YMP and the IBM SP1 mainly for comparison purposes. On the LACE testbed, we study: (1) the communication characteristics of Ethernet, FDDI, and the ALLNODE networks and (2) the overheads induced by the PVM message passing library used for parallelizing the application. We demonstrate that clustering of workstations is effective and has the potential to be computationally competitive with supercomputers at a fraction of the cost.

Exploring Machine Learning Techniques For Dynamic Modeling on Future Exascale Systems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Song, Shuaiwen; Tallent, Nathan R.; Vishnu, Abhinav

2013-09-23

Future exascale systems must be optimized for both power and performance at scale in order to achieve DOE’s goal of a sustained petaflop within 20 Megawatts by 2022 [1]. Massive parallelism of the future systems combined with complex memory hierarchies will form a barrier to efficient application and architecture design. These challenges are exacerbated with emerging complex architectures such as GPGPUs and Intel Xeon Phi as parallelism increases orders of magnitude and system power consumption can easily triple or quadruple. Therefore, we need techniques that can reduce the search space for optimization, isolate power-performance bottlenecks, identify root causes for software/hardwaremore » inefficiency, and effectively direct runtime scheduling.« less

MILC Code Performance on High End CPU and GPU Supercomputer Clusters

NASA Astrophysics Data System (ADS)

DeTar, Carleton; Gottlieb, Steven; Li, Ruizi; Toussaint, Doug

2018-03-01

With recent developments in parallel supercomputing architecture, many core, multi-core, and GPU processors are now commonplace, resulting in more levels of parallelism, memory hierarchy, and programming complexity. It has been necessary to adapt the MILC code to these new processors starting with NVIDIA GPUs, and more recently, the Intel Xeon Phi processors. We report on our efforts to port and optimize our code for the Intel Knights Landing architecture. We consider performance of the MILC code with MPI and OpenMP, and optimizations with QOPQDP and QPhiX. For the latter approach, we concentrate on the staggered conjugate gradient and gauge force. We also consider performance on recent NVIDIA GPUs using the QUDA library.

A novel parallel architecture for local histogram equalization

NASA Astrophysics Data System (ADS)

Ohannessian, Mesrob I.; Choueiter, Ghinwa F.; Diab, Hassan

2005-07-01

Local histogram equalization is an image enhancement algorithm that has found wide application in the pre-processing stage of areas such as computer vision, pattern recognition and medical imaging. The computationally intensive nature of the procedure, however, is a main limitation when real time interactive applications are in question. This work explores the possibility of performing parallel local histogram equalization, using an array of special purpose elementary processors, through an HDL implementation that targets FPGA or ASIC platforms. A novel parallelization scheme is presented and the corresponding architecture is derived. The algorithm is reduced to pixel-level operations. Processing elements are assigned image blocks, to maintain a reasonable performance-cost ratio. To further simplify both processor and memory organizations, a bit-serial access scheme is used. A brief performance assessment is provided to illustrate and quantify the merit of the approach.

Parallelization and automatic data distribution for nuclear reactor simulations

DOE Office of Scientific and Technical Information (OSTI.GOV)

Liebrock, L.M.

1997-07-01

Detailed attempts at realistic nuclear reactor simulations currently take many times real time to execute on high performance workstations. Even the fastest sequential machine can not run these simulations fast enough to ensure that the best corrective measure is used during a nuclear accident to prevent a minor malfunction from becoming a major catastrophe. Since sequential computers have nearly reached the speed of light barrier, these simulations will have to be run in parallel to make significant improvements in speed. In physical reactor plants, parallelism abounds. Fluids flow, controls change, and reactions occur in parallel with only adjacent components directlymore » affecting each other. These do not occur in the sequentialized manner, with global instantaneous effects, that is often used in simulators. Development of parallel algorithms that more closely approximate the real-world operation of a reactor may, in addition to speeding up the simulations, actually improve the accuracy and reliability of the predictions generated. Three types of parallel architecture (shared memory machines, distributed memory multicomputers, and distributed networks) are briefly reviewed as targets for parallelization of nuclear reactor simulation. Various parallelization models (loop-based model, shared memory model, functional model, data parallel model, and a combined functional and data parallel model) are discussed along with their advantages and disadvantages for nuclear reactor simulation. A variety of tools are introduced for each of the models. Emphasis is placed on the data parallel model as the primary focus for two-phase flow simulation. Tools to support data parallel programming for multiple component applications and special parallelization considerations are also discussed.« less

Extending the BEAGLE library to a multi-FPGA platform.

PubMed

Jin, Zheming; Bakos, Jason D

2013-01-19

Maximum Likelihood (ML)-based phylogenetic inference using Felsenstein's pruning algorithm is a standard method for estimating the evolutionary relationships amongst a set of species based on DNA sequence data, and is used in popular applications such as RAxML, PHYLIP, GARLI, BEAST, and MrBayes. The Phylogenetic Likelihood Function (PLF) and its associated scaling and normalization steps comprise the computational kernel for these tools. These computations are data intensive but contain fine grain parallelism that can be exploited by coprocessor architectures such as FPGAs and GPUs. A general purpose API called BEAGLE has recently been developed that includes optimized implementations of Felsenstein's pruning algorithm for various data parallel architectures. In this paper, we extend the BEAGLE API to a multiple Field Programmable Gate Array (FPGA)-based platform called the Convey HC-1. The core calculation of our implementation, which includes both the phylogenetic likelihood function (PLF) and the tree likelihood calculation, has an arithmetic intensity of 130 floating-point operations per 64 bytes of I/O, or 2.03 ops/byte. Its performance can thus be calculated as a function of the host platform's peak memory bandwidth and the implementation's memory efficiency, as 2.03 × peak bandwidth × memory efficiency. Our FPGA-based platform has a peak bandwidth of 76.8 GB/s and our implementation achieves a memory efficiency of approximately 50%, which gives an average throughput of 78 Gflops. This represents a ~40X speedup when compared with BEAGLE's CPU implementation on a dual Xeon 5520 and 3X speedup versus BEAGLE's GPU implementation on a Tesla T10 GPU for very large data sizes. The power consumption is 92 W, yielding a power efficiency of 1.7 Gflops per Watt. The use of data parallel architectures to achieve high performance for likelihood-based phylogenetic inference requires high memory bandwidth and a design methodology that emphasizes high memory efficiency. To achieve this objective, we integrated 32 pipelined processing elements (PEs) across four FPGAs. For the design of each PE, we developed a specialized synthesis tool to generate a floating-point pipeline with resource and throughput constraints to match the target platform. We have found that using low-latency floating-point operators can significantly reduce FPGA area and still meet timing requirement on the target platform. We found that this design methodology can achieve performance that exceeds that of a GPU-based coprocessor.

Optimized Laplacian image sharpening algorithm based on graphic processing unit

NASA Astrophysics Data System (ADS)

Ma, Tinghuai; Li, Lu; Ji, Sai; Wang, Xin; Tian, Yuan; Al-Dhelaan, Abdullah; Al-Rodhaan, Mznah

2014-12-01

In classical Laplacian image sharpening, all pixels are processed one by one, which leads to large amount of computation. Traditional Laplacian sharpening processed on CPU is considerably time-consuming especially for those large pictures. In this paper, we propose a parallel implementation of Laplacian sharpening based on Compute Unified Device Architecture (CUDA), which is a computing platform of Graphic Processing Units (GPU), and analyze the impact of picture size on performance and the relationship between the processing time of between data transfer time and parallel computing time. Further, according to different features of different memory, an improved scheme of our method is developed, which exploits shared memory in GPU instead of global memory and further increases the efficiency. Experimental results prove that two novel algorithms outperform traditional consequentially method based on OpenCV in the aspect of computing speed.

A simple modern correctness condition for a space-based high-performance multiprocessor

NASA Technical Reports Server (NTRS)

Probst, David K.; Li, Hon F.

1992-01-01

A number of U.S. national programs, including space-based detection of ballistic missile launches, envisage putting significant computing power into space. Given sufficient progress in low-power VLSI, multichip-module packaging and liquid-cooling technologies, we will see design of high-performance multiprocessors for individual satellites. In very high speed implementations, performance depends critically on tolerating large latencies in interprocessor communication; without latency tolerance, performance is limited by the vastly differing time scales in processor and data-memory modules, including interconnect times. The modern approach to tolerating remote-communication cost in scalable, shared-memory multiprocessors is to use a multithreaded architecture, and alter the semantics of shared memory slightly, at the price of forcing the programmer either to reason about program correctness in a relaxed consistency model or to agree to program in a constrained style. The literature on multiprocessor correctness conditions has become increasingly complex, and sometimes confusing, which may hinder its practical application. We propose a simple modern correctness condition for a high-performance, shared-memory multiprocessor; the correctness condition is based on a simple interface between the multiprocessor architecture and a high-performance, shared-memory multiprocessor; the correctness condition is based on a simple interface between the multiprocessor architecture and the parallel programming system.

Novel Highly Parallel and Systolic Architectures Using Quantum Dot-Based Hardware

NASA Technical Reports Server (NTRS)

Fijany, Amir; Toomarian, Benny N.; Spotnitz, Matthew

1997-01-01

VLSI technology has made possible the integration of massive number of components (processors, memory, etc.) into a single chip. In VLSI design, memory and processing power are relatively cheap and the main emphasis of the design is on reducing the overall interconnection complexity since data routing costs dominate the power, time, and area required to implement a computation. Communication is costly because wires occupy the most space on a circuit and it can also degrade clock time. In fact, much of the complexity (and hence the cost) of VLSI design results from minimization of data routing. The main difficulty in VLSI routing is due to the fact that crossing of the lines carrying data, instruction, control, etc. is not possible in a plane. Thus, in order to meet this constraint, the VLSI design aims at keeping the architecture highly regular with local and short interconnection. As a result, while the high level of integration has opened the way for massively parallel computation, practical and full exploitation of such a capability in many applications of interest has been hindered by the constraints on interconnection pattern. More precisely. the use of only localized communication significantly simplifies the design of interconnection architecture but at the expense of somewhat restricted class of applications. For example, there are currently commercially available products integrating; hundreds of simple processor elements within a single chip. However, the lack of adequate interconnection pattern among these processing elements make them inefficient for exploiting a large degree of parallelism in many applications.

Binary synaptic connections based on memory switching in a-Si:H for artificial neural networks

NASA Technical Reports Server (NTRS)

Thakoor, A. P.; Lamb, J. L.; Moopenn, A.; Khanna, S. K.

1987-01-01

A scheme for nonvolatile associative electronic memory storage with high information storage density is proposed which is based on neural network models and which uses a matrix of two-terminal passive interconnections (synapses). It is noted that the massive parallelism in the architecture would require the ON state of a synaptic connection to be unusually weak (highly resistive). Memory switching using a-Si:H along with ballast resistors patterned from amorphous Ge-metal alloys is investigated for a binary programmable read only memory matrix. The fabrication of a 1600 synapse test array of uniform connection strengths and a-Si:H switching elements is discussed.

Slices: A Scalable Partitioner for Finite Element Meshes

NASA Technical Reports Server (NTRS)

Ding, H. Q.; Ferraro, R. D.

1995-01-01

A parallel partitioner for partitioning unstructured finite element meshes on distributed memory architectures is developed. The element based partitioner can handle mixtures of different element types. All algorithms adopted in the partitioner are scalable, including a communication template for unpredictable incoming messages, as shown in actual timing measurements.

High Performance Programming Using Explicit Shared Memory Model on Cray T3D1

NASA Technical Reports Server (NTRS)

Simon, Horst D.; Saini, Subhash; Grassi, Charles

1994-01-01

The Cray T3D system is the first-phase system in Cray Research, Inc.'s (CRI) three-phase massively parallel processing (MPP) program. This system features a heterogeneous architecture that closely couples DEC's Alpha microprocessors and CRI's parallel-vector technology, i.e., the Cray Y-MP and Cray C90. An overview of the Cray T3D hardware and available programming models is presented. Under Cray Research adaptive Fortran (CRAFT) model four programming methods (data parallel, work sharing, message-passing using PVM, and explicit shared memory model) are available to the users. However, at this time data parallel and work sharing programming models are not available to the user community. The differences between standard PVM and CRI's PVM are highlighted with performance measurements such as latencies and communication bandwidths. We have found that the performance of neither standard PVM nor CRI s PVM exploits the hardware capabilities of the T3D. The reasons for the bad performance of PVM as a native message-passing library are presented. This is illustrated by the performance of NAS Parallel Benchmarks (NPB) programmed in explicit shared memory model on Cray T3D. In general, the performance of standard PVM is about 4 to 5 times less than obtained by using explicit shared memory model. This degradation in performance is also seen on CM-5 where the performance of applications using native message-passing library CMMD on CM-5 is also about 4 to 5 times less than using data parallel methods. The issues involved (such as barriers, synchronization, invalidating data cache, aligning data cache etc.) while programming in explicit shared memory model are discussed. Comparative performance of NPB using explicit shared memory programming model on the Cray T3D and other highly parallel systems such as the TMC CM-5, Intel Paragon, Cray C90, IBM-SP1, etc. is presented.

Designing Next Generation Massively Multithreaded Architectures for Irregular Applications

DOE Office of Scientific and Technical Information (OSTI.GOV)

Tumeo, Antonino; Secchi, Simone; Villa, Oreste

Irregular applications, such as data mining or graph-based computations, show unpredictable memory/network access patterns and control structures. Massively multi-threaded architectures with large node count, like the Cray XMT, have been shown to address their requirements better than commodity clusters. In this paper we present the approaches that we are currently pursuing to design future generations of these architectures. First, we introduce the Cray XMT and compare it to other multithreaded architectures. We then propose an evolution of the architecture, integrating multiple cores per node and next generation network interconnect. We advocate the use of hardware support for remote memory referencemore » aggregation to optimize network utilization. For this evaluation we developed a highly parallel, custom simulation infrastructure for multi-threaded systems. Our simulator executes unmodified XMT binaries with very large datasets, capturing effects due to contention and hot-spotting, while predicting execution times with greater than 90% accuracy. We also discuss the FPGA prototyping approach that we are employing to study efficient support for irregular applications in next generation manycore processors.« less

The language parallel Pascal and other aspects of the massively parallel processor

NASA Technical Reports Server (NTRS)

Reeves, A. P.; Bruner, J. D.

1982-01-01

A high level language for the Massively Parallel Processor (MPP) was designed. This language, called Parallel Pascal, is described in detail. A description of the language design, a description of the intermediate language, Parallel P-Code, and details for the MPP implementation are included. Formal descriptions of Parallel Pascal and Parallel P-Code are given. A compiler was developed which converts programs in Parallel Pascal into the intermediate Parallel P-Code language. The code generator to complete the compiler for the MPP is being developed independently. A Parallel Pascal to Pascal translator was also developed. The architecture design for a VLSI version of the MPP was completed with a description of fault tolerant interconnection networks. The memory arrangement aspects of the MPP are discussed and a survey of other high level languages is given.

Highly parallel sparse Cholesky factorization

NASA Technical Reports Server (NTRS)

Gilbert, John R.; Schreiber, Robert

1990-01-01

Several fine grained parallel algorithms were developed and compared to compute the Cholesky factorization of a sparse matrix. The experimental implementations are on the Connection Machine, a distributed memory SIMD machine whose programming model conceptually supplies one processor per data element. In contrast to special purpose algorithms in which the matrix structure conforms to the connection structure of the machine, the focus is on matrices with arbitrary sparsity structure. The most promising algorithm is one whose inner loop performs several dense factorizations simultaneously on a 2-D grid of processors. Virtually any massively parallel dense factorization algorithm can be used as the key subroutine. The sparse code attains execution rates comparable to those of the dense subroutine. Although at present architectural limitations prevent the dense factorization from realizing its potential efficiency, it is concluded that a regular data parallel architecture can be used efficiently to solve arbitrarily structured sparse problems. A performance model is also presented and it is used to analyze the algorithms.

Aerodynamic Shape Optimization of Supersonic Aircraft Configurations via an Adjoint Formulation on Parallel Computers

NASA Technical Reports Server (NTRS)

Reuther, James; Alonso, Juan Jose; Rimlinger, Mark J.; Jameson, Antony

1996-01-01

This work describes the application of a control theory-based aerodynamic shape optimization method to the problem of supersonic aircraft design. The design process is greatly accelerated through the use of both control theory and a parallel implementation on distributed memory computers. Control theory is employed to derive the adjoint differential equations whose solution allows for the evaluation of design gradient information at a fraction of the computational cost required by previous design methods (13, 12, 44, 38). The resulting problem is then implemented on parallel distributed memory architectures using a domain decomposition approach, an optimized communication schedule, and the MPI (Message Passing Interface) Standard for portability and efficiency. The final result achieves very rapid aerodynamic design based on higher order computational fluid dynamics methods (CFD). In our earlier studies, the serial implementation of this design method (19, 20, 21, 23, 39, 25, 40, 41, 42, 43, 9) was shown to be effective for the optimization of airfoils, wings, wing-bodies, and complex aircraft configurations using both the potential equation and the Euler equations (39, 25). In our most recent paper, the Euler method was extended to treat complete aircraft configurations via a new multiblock implementation. Furthermore, during the same conference, we also presented preliminary results demonstrating that the basic methodology could be ported to distributed memory parallel computing architectures [241. In this paper, our concem will be to demonstrate that the combined power of these new technologies can be used routinely in an industrial design environment by applying it to the case study of the design of typical supersonic transport configurations. A particular difficulty of this test case is posed by the propulsion/airframe integration.

Real-time field programmable gate array architecture for computer vision

NASA Astrophysics Data System (ADS)

Arias-Estrada, Miguel; Torres-Huitzil, Cesar

2001-01-01

This paper presents an architecture for real-time generic convolution of a mask and an image. The architecture is intended for fast low-level image processing. The field programmable gate array (FPGA)-based architecture takes advantage of the availability of registers in FPGAs to implement an efficient and compact module to process the convolutions. The architecture is designed to minimize the number of accesses to the image memory and it is based on parallel modules with internal pipeline operation in order to improve its performance. The architecture is prototyped in a FPGA, but it can be implemented on dedicated very- large-scale-integrated devices to reach higher clock frequencies. Complexity issues, FPGA resources utilization, FPGA limitations, and real-time performance are discussed. Some results are presented and discussed.

Optimizing Excited-State Electronic-Structure Codes for Intel Knights Landing: A Case Study on the BerkeleyGW Software

DOE Office of Scientific and Technical Information (OSTI.GOV)

Deslippe, Jack; da Jornada, Felipe H.; Vigil-Fowler, Derek

2016-10-06

We profile and optimize calculations performed with the BerkeleyGW code on the Xeon-Phi architecture. BerkeleyGW depends both on hand-tuned critical kernels as well as on BLAS and FFT libraries. We describe the optimization process and performance improvements achieved. We discuss a layered parallelization strategy to take advantage of vector, thread and node-level parallelism. We discuss locality changes (including the consequence of the lack of L3 cache) and effective use of the on-package high-bandwidth memory. We show preliminary results on Knights-Landing including a roofline study of code performance before and after a number of optimizations. We find that the GW methodmore » is particularly well-suited for many-core architectures due to the ability to exploit a large amount of parallelism over plane-wave components, band-pairs, and frequencies.« less

Automation of Data Traffic Control on DSM Architecture

NASA Technical Reports Server (NTRS)

Frumkin, Michael; Jin, Hao-Qiang; Yan, Jerry

2001-01-01

The design of distributed shared memory (DSM) computers liberates users from the duty to distribute data across processors and allows for the incremental development of parallel programs using, for example, OpenMP or Java threads. DSM architecture greatly simplifies the development of parallel programs having good performance on a few processors. However, to achieve a good program scalability on DSM computers requires that the user understand data flow in the application and use various techniques to avoid data traffic congestions. In this paper we discuss a number of such techniques, including data blocking, data placement, data transposition and page size control and evaluate their efficiency on the NAS (NASA Advanced Supercomputing) Parallel Benchmarks. We also present a tool which automates the detection of constructs causing data congestions in Fortran array oriented codes and advises the user on code transformations for improving data traffic in the application.

Parallel Implementation of MAFFT on CUDA-Enabled Graphics Hardware.

PubMed

Zhu, Xiangyuan; Li, Kenli; Salah, Ahmad; Shi, Lin; Li, Keqin

2015-01-01

Multiple sequence alignment (MSA) constitutes an extremely powerful tool for many biological applications including phylogenetic tree estimation, secondary structure prediction, and critical residue identification. However, aligning large biological sequences with popular tools such as MAFFT requires long runtimes on sequential architectures. Due to the ever increasing sizes of sequence databases, there is increasing demand to accelerate this task. In this paper, we demonstrate how graphic processing units (GPUs), powered by the compute unified device architecture (CUDA), can be used as an efficient computational platform to accelerate the MAFFT algorithm. To fully exploit the GPU's capabilities for accelerating MAFFT, we have optimized the sequence data organization to eliminate the bandwidth bottleneck of memory access, designed a memory allocation and reuse strategy to make full use of limited memory of GPUs, proposed a new modified-run-length encoding (MRLE) scheme to reduce memory consumption, and used high-performance shared memory to speed up I/O operations. Our implementation tested in three NVIDIA GPUs achieves speedup up to 11.28 on a Tesla K20m GPU compared to the sequential MAFFT 7.015.

Data Parallel Bin-Based Indexing for Answering Queries on Multi-Core Architectures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Gosink, Luke; Wu, Kesheng; Bethel, E. Wes

2009-06-02

The multi-core trend in CPUs and general purpose graphics processing units (GPUs) offers new opportunities for the database community. The increase of cores at exponential rates is likely to affect virtually every server and client in the coming decade, and presents database management systems with a huge, compelling disruption that will radically change how processing is done. This paper presents a new parallel indexing data structure for answering queries that takes full advantage of the increasing thread-level parallelism emerging in multi-core architectures. In our approach, our Data Parallel Bin-based Index Strategy (DP-BIS) first bins the base data, and then partitionsmore » and stores the values in each bin as a separate, bin-based data cluster. In answering a query, the procedures for examining the bin numbers and the bin-based data clusters offer the maximum possible level of concurrency; each record is evaluated by a single thread and all threads are processed simultaneously in parallel. We implement and demonstrate the effectiveness of DP-BIS on two multi-core architectures: a multi-core CPU and a GPU. The concurrency afforded by DP-BIS allows us to fully utilize the thread-level parallelism provided by each architecture--for example, our GPU-based DP-BIS implementation simultaneously evaluates over 12,000 records with an equivalent number of concurrently executing threads. In comparing DP-BIS's performance across these architectures, we show that the GPU-based DP-BIS implementation requires significantly less computation time to answer a query than the CPU-based implementation. We also demonstrate in our analysis that DP-BIS provides better overall performance than the commonly utilized CPU and GPU-based projection index. Finally, due to data encoding, we show that DP-BIS accesses significantly smaller amounts of data than index strategies that operate solely on a column's base data; this smaller data footprint is critical for parallel processors that possess limited memory resources (e.g., GPUs).« less

Geopotential Error Analysis from Satellite Gradiometer and Global Positioning System Observables on Parallel Architecture

NASA Technical Reports Server (NTRS)

Schutz, Bob E.; Baker, Gregory A.

1997-01-01

The recovery of a high resolution geopotential from satellite gradiometer observations motivates the examination of high performance computational techniques. The primary subject matter addresses specifically the use of satellite gradiometer and GPS observations to form and invert the normal matrix associated with a large degree and order geopotential solution. Memory resident and out-of-core parallel linear algebra techniques along with data parallel batch algorithms form the foundation of the least squares application structure. A secondary topic includes the adoption of object oriented programming techniques to enhance modularity and reusability of code. Applications implementing the parallel and object oriented methods successfully calculate the degree variance for a degree and order 110 geopotential solution on 32 processors of the Cray T3E. The memory resident gradiometer application exhibits an overall application performance of 5.4 Gflops, and the out-of-core linear solver exhibits an overall performance of 2.4 Gflops. The combination solution derived from a sun synchronous gradiometer orbit produce average geoid height variances of 17 millimeters.

Geopotential error analysis from satellite gradiometer and global positioning system observables on parallel architectures

NASA Astrophysics Data System (ADS)

Baker, Gregory Allen

The recovery of a high resolution geopotential from satellite gradiometer observations motivates the examination of high performance computational techniques. The primary subject matter addresses specifically the use of satellite gradiometer and GPS observations to form and invert the normal matrix associated with a large degree and order geopotential solution. Memory resident and out-of-core parallel linear algebra techniques along with data parallel batch algorithms form the foundation of the least squares application structure. A secondary topic includes the adoption of object oriented programming techniques to enhance modularity and reusability of code. Applications implementing the parallel and object oriented methods successfully calculate the degree variance for a degree and order 110 geopotential solution on 32 processors of the Cray T3E. The memory resident gradiometer application exhibits an overall application performance of 5.4 Gflops, and the out-of-core linear solver exhibits an overall performance of 2.4 Gflops. The combination solution derived from a sun synchronous gradiometer orbit produce average geoid height variances of 17 millimeters.

Accelerating next generation sequencing data analysis with system level optimizations.

PubMed

Kathiresan, Nagarajan; Temanni, Ramzi; Almabrazi, Hakeem; Syed, Najeeb; Jithesh, Puthen V; Al-Ali, Rashid

2017-08-22

Next generation sequencing (NGS) data analysis is highly compute intensive. In-memory computing, vectorization, bulk data transfer, CPU frequency scaling are some of the hardware features in the modern computing architectures. To get the best execution time and utilize these hardware features, it is necessary to tune the system level parameters before running the application. We studied the GATK-HaplotypeCaller which is part of common NGS workflows, that consume more than 43% of the total execution time. Multiple GATK 3.x versions were benchmarked and the execution time of HaplotypeCaller was optimized by various system level parameters which included: (i) tuning the parallel garbage collection and kernel shared memory to simulate in-memory computing, (ii) architecture-specific tuning in the PairHMM library for vectorization, (iii) including Java 1.8 features through GATK source code compilation and building a runtime environment for parallel sorting and bulk data transfer (iv) the default 'on-demand' mode of CPU frequency is over-clocked by using 'performance-mode' to accelerate the Java multi-threads. As a result, the HaplotypeCaller execution time was reduced by 82.66% in GATK 3.3 and 42.61% in GATK 3.7. Overall, the execution time of NGS pipeline was reduced to 70.60% and 34.14% for GATK 3.3 and GATK 3.7 respectively.

Toward Millions of File System IOPS on Low-Cost, Commodity Hardware

PubMed Central

Zheng, Da; Burns, Randal; Szalay, Alexander S.

2013-01-01

We describe a storage system that removes I/O bottlenecks to achieve more than one million IOPS based on a user-space file abstraction for arrays of commodity SSDs. The file abstraction refactors I/O scheduling and placement for extreme parallelism and non-uniform memory and I/O. The system includes a set-associative, parallel page cache in the user space. We redesign page caching to eliminate CPU overhead and lock-contention in non-uniform memory architecture machines. We evaluate our design on a 32 core NUMA machine with four, eight-core processors. Experiments show that our design delivers 1.23 million 512-byte read IOPS. The page cache realizes the scalable IOPS of Linux asynchronous I/O (AIO) and increases user-perceived I/O performance linearly with cache hit rates. The parallel, set-associative cache matches the cache hit rates of the global Linux page cache under real workloads. PMID:24402052

Toward Millions of File System IOPS on Low-Cost, Commodity Hardware.

PubMed

Zheng, Da; Burns, Randal; Szalay, Alexander S

2013-01-01

We describe a storage system that removes I/O bottlenecks to achieve more than one million IOPS based on a user-space file abstraction for arrays of commodity SSDs. The file abstraction refactors I/O scheduling and placement for extreme parallelism and non-uniform memory and I/O. The system includes a set-associative, parallel page cache in the user space. We redesign page caching to eliminate CPU overhead and lock-contention in non-uniform memory architecture machines. We evaluate our design on a 32 core NUMA machine with four, eight-core processors. Experiments show that our design delivers 1.23 million 512-byte read IOPS. The page cache realizes the scalable IOPS of Linux asynchronous I/O (AIO) and increases user-perceived I/O performance linearly with cache hit rates. The parallel, set-associative cache matches the cache hit rates of the global Linux page cache under real workloads.

Design and evaluation of an architecture for a digital signal processor for instrumentation applications

NASA Astrophysics Data System (ADS)

Fellman, Ronald D.; Kaneshiro, Ronald T.; Konstantinides, Konstantinos

1990-03-01

The authors present the design and evaluation of an architecture for a monolithic, programmable, floating-point digital signal processor (DSP) for instrumentation applications. An investigation of the most commonly used algorithms in instrumentation led to a design that satisfies the requirements for high computational and I/O (input/output) throughput. In the arithmetic unit, a 16- x 16-bit multiplier and a 32-bit accumulator provide the capability for single-cycle multiply/accumulate operations, and three format adjusters automatically adjust the data format for increased accuracy and dynamic range. An on-chip I/O unit is capable of handling data block transfers through a direct memory access port and real-time data streams through a pair of parallel I/O ports. I/O operations and program execution are performed in parallel. In addition, the processor includes two data memories with independent addressing units, a microsequencer with instruction RAM, and multiplexers for internal data redirection. The authors also present the structure and implementation of a design environment suitable for the algorithmic, behavioral, and timing simulation of a complete DSP system. Various benchmarking results are reported.

Avoiding and tolerating latency in large-scale next-generation shared-memory multiprocessors

NASA Technical Reports Server (NTRS)

Probst, David K.

1993-01-01

A scalable solution to the memory-latency problem is necessary to prevent the large latencies of synchronization and memory operations inherent in large-scale shared-memory multiprocessors from reducing high performance. We distinguish latency avoidance and latency tolerance. Latency is avoided when data is brought to nearby locales for future reference. Latency is tolerated when references are overlapped with other computation. Latency-avoiding locales include: processor registers, data caches used temporally, and nearby memory modules. Tolerating communication latency requires parallelism, allowing the overlap of communication and computation. Latency-tolerating techniques include: vector pipelining, data caches used spatially, prefetching in various forms, and multithreading in various forms. Relaxing the consistency model permits increased use of avoidance and tolerance techniques. Each model is a mapping from the program text to sets of partial orders on program operations; it is a convention about which temporal precedences among program operations are necessary. Information about temporal locality and parallelism constrains the use of avoidance and tolerance techniques. Suitable architectural primitives and compiler technology are required to exploit the increased freedom to reorder and overlap operations in relaxed models.

A CFD Heterogeneous Parallel Solver Based on Collaborating CPU and GPU

NASA Astrophysics Data System (ADS)

Lai, Jianqi; Tian, Zhengyu; Li, Hua; Pan, Sha

2018-03-01

Since Graphic Processing Unit (GPU) has a strong ability of floating-point computation and memory bandwidth for data parallelism, it has been widely used in the areas of common computing such as molecular dynamics (MD), computational fluid dynamics (CFD) and so on. The emergence of compute unified device architecture (CUDA), which reduces the complexity of compiling program, brings the great opportunities to CFD. There are three different modes for parallel solution of NS equations: parallel solver based on CPU, parallel solver based on GPU and heterogeneous parallel solver based on collaborating CPU and GPU. As we can see, GPUs are relatively rich in compute capacity but poor in memory capacity and the CPUs do the opposite. We need to make full use of the GPUs and CPUs, so a CFD heterogeneous parallel solver based on collaborating CPU and GPU has been established. Three cases are presented to analyse the solver’s computational accuracy and heterogeneous parallel efficiency. The numerical results agree well with experiment results, which demonstrate that the heterogeneous parallel solver has high computational precision. The speedup on a single GPU is more than 40 for laminar flow, it decreases for turbulent flow, but it still can reach more than 20. What’s more, the speedup increases as the grid size becomes larger.

Implementation of the DPM Monte Carlo code on a parallel architecture for treatment planning applications.

PubMed

Tyagi, Neelam; Bose, Abhijit; Chetty, Indrin J

2004-09-01

We have parallelized the Dose Planning Method (DPM), a Monte Carlo code optimized for radiotherapy class problems, on distributed-memory processor architectures using the Message Passing Interface (MPI). Parallelization has been investigated on a variety of parallel computing architectures at the University of Michigan-Center for Advanced Computing, with respect to efficiency and speedup as a function of the number of processors. We have integrated the parallel pseudo random number generator from the Scalable Parallel Pseudo-Random Number Generator (SPRNG) library to run with the parallel DPM. The Intel cluster consisting of 800 MHz Intel Pentium III processor shows an almost linear speedup up to 32 processors for simulating 1 x 10(8) or more particles. The speedup results are nearly linear on an Athlon cluster (up to 24 processors based on availability) which consists of 1.8 GHz+ Advanced Micro Devices (AMD) Athlon processors on increasing the problem size up to 8 x 10(8) histories. For a smaller number of histories (1 x 10(8)) the reduction of efficiency with the Athlon cluster (down to 83.9% with 24 processors) occurs because the processing time required to simulate 1 x 10(8) histories is less than the time associated with interprocessor communication. A similar trend was seen with the Opteron Cluster (consisting of 1400 MHz, 64-bit AMD Opteron processors) on increasing the problem size. Because of the 64-bit architecture Opteron processors are capable of storing and processing instructions at a faster rate and hence are faster as compared to the 32-bit Athlon processors. We have validated our implementation with an in-phantom dose calculation study using a parallel pencil monoenergetic electron beam of 20 MeV energy. The phantom consists of layers of water, lung, bone, aluminum, and titanium. The agreement in the central axis depth dose curves and profiles at different depths shows that the serial and parallel codes are equivalent in accuracy.

Running ATLAS workloads within massively parallel distributed applications using Athena Multi-Process framework (AthenaMP)

NASA Astrophysics Data System (ADS)

Calafiura, Paolo; Leggett, Charles; Seuster, Rolf; Tsulaia, Vakhtang; Van Gemmeren, Peter

2015-12-01

AthenaMP is a multi-process version of the ATLAS reconstruction, simulation and data analysis framework Athena. By leveraging Linux fork and copy-on-write mechanisms, it allows for sharing of memory pages between event processors running on the same compute node with little to no change in the application code. Originally targeted to optimize the memory footprint of reconstruction jobs, AthenaMP has demonstrated that it can reduce the memory usage of certain configurations of ATLAS production jobs by a factor of 2. AthenaMP has also evolved to become the parallel event-processing core of the recently developed ATLAS infrastructure for fine-grained event processing (Event Service) which allows the running of AthenaMP inside massively parallel distributed applications on hundreds of compute nodes simultaneously. We present the architecture of AthenaMP, various strategies implemented by AthenaMP for scheduling workload to worker processes (for example: Shared Event Queue and Shared Distributor of Event Tokens) and the usage of AthenaMP in the diversity of ATLAS event processing workloads on various computing resources: Grid, opportunistic resources and HPC.

New NAS Parallel Benchmarks Results

NASA Technical Reports Server (NTRS)

Yarrow, Maurice; Saphir, William; VanderWijngaart, Rob; Woo, Alex; Kutler, Paul (Technical Monitor)

1997-01-01

NPB2 (NAS (NASA Advanced Supercomputing) Parallel Benchmarks 2) is an implementation, based on Fortran and the MPI (message passing interface) message passing standard, of the original NAS Parallel Benchmark specifications. NPB2 programs are run with little or no tuning, in contrast to NPB vendor implementations, which are highly optimized for specific architectures. NPB2 results complement, rather than replace, NPB results. Because they have not been optimized by vendors, NPB2 implementations approximate the performance a typical user can expect for a portable parallel program on distributed memory parallel computers. Together these results provide an insightful comparison of the real-world performance of high-performance computers. New NPB2 features: New implementation (CG), new workstation class problem sizes, new serial sample versions, more performance statistics.

Efficient Parallel Video Processing Techniques on GPU: From Framework to Implementation

PubMed Central

Su, Huayou; Wen, Mei; Wu, Nan; Ren, Ju; Zhang, Chunyuan

2014-01-01

Through reorganizing the execution order and optimizing the data structure, we proposed an efficient parallel framework for H.264/AVC encoder based on massively parallel architecture. We implemented the proposed framework by CUDA on NVIDIA's GPU. Not only the compute intensive components of the H.264 encoder are parallelized but also the control intensive components are realized effectively, such as CAVLC and deblocking filter. In addition, we proposed serial optimization methods, including the multiresolution multiwindow for motion estimation, multilevel parallel strategy to enhance the parallelism of intracoding as much as possible, component-based parallel CAVLC, and direction-priority deblocking filter. More than 96% of workload of H.264 encoder is offloaded to GPU. Experimental results show that the parallel implementation outperforms the serial program by 20 times of speedup ratio and satisfies the requirement of the real-time HD encoding of 30 fps. The loss of PSNR is from 0.14 dB to 0.77 dB, when keeping the same bitrate. Through the analysis to the kernels, we found that speedup ratios of the compute intensive algorithms are proportional with the computation power of the GPU. However, the performance of the control intensive parts (CAVLC) is much related to the memory bandwidth, which gives an insight for new architecture design. PMID:24757432

VINE-A NUMERICAL CODE FOR SIMULATING ASTROPHYSICAL SYSTEMS USING PARTICLES. II. IMPLEMENTATION AND PERFORMANCE CHARACTERISTICS

DOE Office of Scientific and Technical Information (OSTI.GOV)

Nelson, Andrew F.; Wetzstein, M.; Naab, T.

2009-10-01

We continue our presentation of VINE. In this paper, we begin with a description of relevant architectural properties of the serial and shared memory parallel computers on which VINE is intended to run, and describe their influences on the design of the code itself. We continue with a detailed description of a number of optimizations made to the layout of the particle data in memory and to our implementation of a binary tree used to access that data for use in gravitational force calculations and searches for smoothed particle hydrodynamics (SPH) neighbor particles. We describe the modifications to the codemore » necessary to obtain forces efficiently from special purpose 'GRAPE' hardware, the interfaces required to allow transparent substitution of those forces in the code instead of those obtained from the tree, and the modifications necessary to use both tree and GRAPE together as a fused GRAPE/tree combination. We conclude with an extensive series of performance tests, which demonstrate that the code can be run efficiently and without modification in serial on small workstations or in parallel using the OpenMP compiler directives on large-scale, shared memory parallel machines. We analyze the effects of the code optimizations and estimate that they improve its overall performance by more than an order of magnitude over that obtained by many other tree codes. Scaled parallel performance of the gravity and SPH calculations, together the most costly components of most simulations, is nearly linear up to at least 120 processors on moderate sized test problems using the Origin 3000 architecture, and to the maximum machine sizes available to us on several other architectures. At similar accuracy, performance of VINE, used in GRAPE-tree mode, is approximately a factor 2 slower than that of VINE, used in host-only mode. Further optimizations of the GRAPE/host communications could improve the speed by as much as a factor of 3, but have not yet been implemented in VINE. Finally, we find that although parallel performance on small problems may reach a plateau beyond which more processors bring no additional speedup, performance never decreases, a factor important for running large simulations on many processors with individual time steps, where only a small fraction of the total particles require updates at any given moment.« less

Kalman filter tracking on parallel architectures

NASA Astrophysics Data System (ADS)

Cerati, G.; Elmer, P.; Krutelyov, S.; Lantz, S.; Lefebvre, M.; McDermott, K.; Riley, D.; Tadel, M.; Wittich, P.; Wurthwein, F.; Yagil, A.

2017-10-01

We report on the progress of our studies towards a Kalman filter track reconstruction algorithm with optimal performance on manycore architectures. The combinatorial structure of these algorithms is not immediately compatible with an efficient SIMD (or SIMT) implementation; the challenge for us is to recast the existing software so it can readily generate hundreds of shared-memory threads that exploit the underlying instruction set of modern processors. We show how the data and associated tasks can be organized in a way that is conducive to both multithreading and vectorization. We demonstrate very good performance on Intel Xeon and Xeon Phi architectures, as well as promising first results on Nvidia GPUs.

A Cerebellar-model Associative Memory as a Generalized Random-access Memory

NASA Technical Reports Server (NTRS)

Kanerva, Pentti

1989-01-01

A versatile neural-net model is explained in terms familiar to computer scientists and engineers. It is called the sparse distributed memory, and it is a random-access memory for very long words (for patterns with thousands of bits). Its potential utility is the result of several factors: (1) a large pattern representing an object or a scene or a moment can encode a large amount of information about what it represents; (2) this information can serve as an address to the memory, and it can also serve as data; (3) the memory is noise tolerant--the information need not be exact; (4) the memory can be made arbitrarily large and hence an arbitrary amount of information can be stored in it; and (5) the architecture is inherently parallel, allowing large memories to be fast. Such memories can become important components of future computers.

Japanese project aims at supercomputer that executes 10 gflops

DOE Office of Scientific and Technical Information (OSTI.GOV)

Burskey, D.

1984-05-03

Dubbed supercom by its multicompany design team, the decade-long project's goal is an engineering supercomputer that can execute 10 billion floating-point operations/s-about 20 times faster than today's supercomputers. The project, guided by Japan's Ministry of International Trade and Industry (MITI) and the Agency of Industrial Science and Technology encompasses three parallel research programs, all aimed at some angle of the superconductor. One program should lead to superfast logic and memory circuits, another to a system architecture that will afford the best performance, and the last to the software that will ultimately control the computer. The work on logic and memorymore » chips is based on: GAAS circuit; Josephson junction devices; and high electron mobility transistor structures. The architecture will involve parallel processing.« less

Extending the BEAGLE library to a multi-FPGA platform

PubMed Central

2013-01-01

Background Maximum Likelihood (ML)-based phylogenetic inference using Felsenstein’s pruning algorithm is a standard method for estimating the evolutionary relationships amongst a set of species based on DNA sequence data, and is used in popular applications such as RAxML, PHYLIP, GARLI, BEAST, and MrBayes. The Phylogenetic Likelihood Function (PLF) and its associated scaling and normalization steps comprise the computational kernel for these tools. These computations are data intensive but contain fine grain parallelism that can be exploited by coprocessor architectures such as FPGAs and GPUs. A general purpose API called BEAGLE has recently been developed that includes optimized implementations of Felsenstein’s pruning algorithm for various data parallel architectures. In this paper, we extend the BEAGLE API to a multiple Field Programmable Gate Array (FPGA)-based platform called the Convey HC-1. Results The core calculation of our implementation, which includes both the phylogenetic likelihood function (PLF) and the tree likelihood calculation, has an arithmetic intensity of 130 floating-point operations per 64 bytes of I/O, or 2.03 ops/byte. Its performance can thus be calculated as a function of the host platform’s peak memory bandwidth and the implementation’s memory efficiency, as 2.03 × peak bandwidth × memory efficiency. Our FPGA-based platform has a peak bandwidth of 76.8 GB/s and our implementation achieves a memory efficiency of approximately 50%, which gives an average throughput of 78 Gflops. This represents a ~40X speedup when compared with BEAGLE’s CPU implementation on a dual Xeon 5520 and 3X speedup versus BEAGLE’s GPU implementation on a Tesla T10 GPU for very large data sizes. The power consumption is 92 W, yielding a power efficiency of 1.7 Gflops per Watt. Conclusions The use of data parallel architectures to achieve high performance for likelihood-based phylogenetic inference requires high memory bandwidth and a design methodology that emphasizes high memory efficiency. To achieve this objective, we integrated 32 pipelined processing elements (PEs) across four FPGAs. For the design of each PE, we developed a specialized synthesis tool to generate a floating-point pipeline with resource and throughput constraints to match the target platform. We have found that using low-latency floating-point operators can significantly reduce FPGA area and still meet timing requirement on the target platform. We found that this design methodology can achieve performance that exceeds that of a GPU-based coprocessor. PMID:23331707

Bioinspired architecture approach for a one-billion transistor smart CMOS camera chip

NASA Astrophysics Data System (ADS)

Fey, Dietmar; Komann, Marcus

2007-05-01

In the paper we present a massively parallel VLSI architecture for future smart CMOS camera chips with up to one billion transistors. To exploit efficiently the potential offered by future micro- or nanoelectronic devices traditional on central structures oriented parallel architectures based on MIMD or SIMD approaches will fail. They require too long and too many global interconnects for the distribution of code or the access to common memory. On the other hand nature developed self-organising and emergent principles to manage successfully complex structures based on lots of interacting simple elements. Therefore we developed a new as Marching Pixels denoted emergent computing paradigm based on a mixture of bio-inspired computing models like cellular automaton and artificial ants. In the paper we present different Marching Pixels algorithms and the corresponding VLSI array architecture. A detailed synthesis result for a 0.18 μm CMOS process shows that a 256×256 pixel image is processed in less than 10 ms assuming a moderate 100 MHz clock rate for the processor array. Future higher integration densities and a 3D chip stacking technology will allow the integration and processing of Mega pixels within the same time since our architecture is fully scalable.

Optimization of atmospheric transport models on HPC platforms

NASA Astrophysics Data System (ADS)

de la Cruz, Raúl; Folch, Arnau; Farré, Pau; Cabezas, Javier; Navarro, Nacho; Cela, José María

2016-12-01

The performance and scalability of atmospheric transport models on high performance computing environments is often far from optimal for multiple reasons including, for example, sequential input and output, synchronous communications, work unbalance, memory access latency or lack of task overlapping. We investigate how different software optimizations and porting to non general-purpose hardware architectures improve code scalability and execution times considering, as an example, the FALL3D volcanic ash transport model. To this purpose, we implement the FALL3D model equations in the WARIS framework, a software designed from scratch to solve in a parallel and efficient way different geoscience problems on a wide variety of architectures. In addition, we consider further improvements in WARIS such as hybrid MPI-OMP parallelization, spatial blocking, auto-tuning and thread affinity. Considering all these aspects together, the FALL3D execution times for a realistic test case running on general-purpose cluster architectures (Intel Sandy Bridge) decrease by a factor between 7 and 40 depending on the grid resolution. Finally, we port the application to Intel Xeon Phi (MIC) and NVIDIA GPUs (CUDA) accelerator-based architectures and compare performance, cost and power consumption on all the architectures. Implications on time-constrained operational model configurations are discussed.

Supercomputers Of The Future

NASA Technical Reports Server (NTRS)

Peterson, Victor L.; Kim, John; Holst, Terry L.; Deiwert, George S.; Cooper, David M.; Watson, Andrew B.; Bailey, F. Ron

1992-01-01

Report evaluates supercomputer needs of five key disciplines: turbulence physics, aerodynamics, aerothermodynamics, chemistry, and mathematical modeling of human vision. Predicts these fields will require computer speed greater than 10(Sup 18) floating-point operations per second (FLOP's) and memory capacity greater than 10(Sup 15) words. Also, new parallel computer architectures and new structured numerical methods will make necessary speed and capacity available.

A high performance linear equation solver on the VPP500 parallel supercomputer

DOE Office of Scientific and Technical Information (OSTI.GOV)

Nakanishi, Makoto; Ina, Hiroshi; Miura, Kenichi

1994-12-31

This paper describes the implementation of two high performance linear equation solvers developed for the Fujitsu VPP500, a distributed memory parallel supercomputer system. The solvers take advantage of the key architectural features of VPP500--(1) scalability for an arbitrary number of processors up to 222 processors, (2) flexible data transfer among processors provided by a crossbar interconnection network, (3) vector processing capability on each processor, and (4) overlapped computation and transfer. The general linear equation solver based on the blocked LU decomposition method achieves 120.0 GFLOPS performance with 100 processors in the LIN-PACK Highly Parallel Computing benchmark.

The force on the flex: Global parallelism and portability

NASA Technical Reports Server (NTRS)

Jordan, H. F.

1986-01-01

A parallel programming methodology, called the force, supports the construction of programs to be executed in parallel by an unspecified, but potentially large, number of processes. The methodology was originally developed on a pipelined, shared memory multiprocessor, the Denelcor HEP, and embodies the primitive operations of the force in a set of macros which expand into multiprocessor Fortran code. A small set of primitives is sufficient to write large parallel programs, and the system has been used to produce 10,000 line programs in computational fluid dynamics. The level of complexity of the force primitives is intermediate. It is high enough to mask detailed architectural differences between multiprocessors but low enough to give the user control over performance. The system is being ported to a medium scale multiprocessor, the Flex/32, which is a 20 processor system with a mixture of shared and local memory. Memory organization and the type of processor synchronization supported by the hardware on the two machines lead to some differences in efficient implementations of the force primitives, but the user interface remains the same. An initial implementation was done by retargeting the macros to Flexible Computer Corporation's ConCurrent C language. Subsequently, the macros were caused to directly produce the system calls which form the basis for ConCurrent C. The implementation of the Fortran based system is in step with Flexible Computer Corporations's implementation of a Fortran system in the parallel environment.

Phase space simulation of collisionless stellar systems on the massively parallel processor

NASA Technical Reports Server (NTRS)

White, Richard L.

1987-01-01

A numerical technique for solving the collisionless Boltzmann equation describing the time evolution of a self gravitating fluid in phase space was implemented on the Massively Parallel Processor (MPP). The code performs calculations for a two dimensional phase space grid (with one space and one velocity dimension). Some results from calculations are presented. The execution speed of the code is comparable to the speed of a single processor of a Cray-XMP. Advantages and disadvantages of the MPP architecture for this type of problem are discussed. The nearest neighbor connectivity of the MPP array does not pose a significant obstacle. Future MPP-like machines should have much more local memory and easier access to staging memory and disks in order to be effective for this type of problem.

Design of Belief Propagation Based on FPGA for the Multistereo CAFADIS Camera

PubMed Central

Magdaleno, Eduardo; Lüke, Jonás Philipp; Rodríguez, Manuel; Rodríguez-Ramos, José Manuel

2010-01-01

In this paper we describe a fast, specialized hardware implementation of the belief propagation algorithm for the CAFADIS camera, a new plenoptic sensor patented by the University of La Laguna. This camera captures the lightfield of the scene and can be used to find out at which depth each pixel is in focus. The algorithm has been designed for FPGA devices using VHDL. We propose a parallel and pipeline architecture to implement the algorithm without external memory. Although the BRAM resources of the device increase considerably, we can maintain real-time restrictions by using extremely high-performance signal processing capability through parallelism and by accessing several memories simultaneously. The quantifying results with 16 bit precision have shown that performances are really close to the original Matlab programmed algorithm. PMID:22163404

Design of belief propagation based on FPGA for the multistereo CAFADIS camera.

PubMed

Magdaleno, Eduardo; Lüke, Jonás Philipp; Rodríguez, Manuel; Rodríguez-Ramos, José Manuel

2010-01-01

In this paper we describe a fast, specialized hardware implementation of the belief propagation algorithm for the CAFADIS camera, a new plenoptic sensor patented by the University of La Laguna. This camera captures the lightfield of the scene and can be used to find out at which depth each pixel is in focus. The algorithm has been designed for FPGA devices using VHDL. We propose a parallel and pipeline architecture to implement the algorithm without external memory. Although the BRAM resources of the device increase considerably, we can maintain real-time restrictions by using extremely high-performance signal processing capability through parallelism and by accessing several memories simultaneously. The quantifying results with 16 bit precision have shown that performances are really close to the original Matlab programmed algorithm.

Improving the performance of heterogeneous multi-core processors by modifying the cache coherence protocol

NASA Astrophysics Data System (ADS)

Fang, Juan; Hao, Xiaoting; Fan, Qingwen; Chang, Zeqing; Song, Shuying

2017-05-01

In the Heterogeneous multi-core architecture, CPU and GPU processor are integrated on the same chip, which poses a new challenge to the last-level cache management. In this architecture, the CPU application and the GPU application execute concurrently, accessing the last-level cache. CPU and GPU have different memory access characteristics, so that they have differences in the sensitivity of last-level cache (LLC) capacity. For many CPU applications, a reduced share of the LLC could lead to significant performance degradation. On the contrary, GPU applications can tolerate increase in memory access latency when there is sufficient thread-level parallelism. Taking into account the GPU program memory latency tolerance characteristics, this paper presents a method that let GPU applications can access to memory directly, leaving lots of LLC space for CPU applications, in improving the performance of CPU applications and does not affect the performance of GPU applications. When the CPU application is cache sensitive, and the GPU application is insensitive to the cache, the overall performance of the system is improved significantly.

GPU accelerated cell-based adaptive mesh refinement on unstructured quadrilateral grid

NASA Astrophysics Data System (ADS)

Luo, Xisheng; Wang, Luying; Ran, Wei; Qin, Fenghua

2016-10-01

A GPU accelerated inviscid flow solver is developed on an unstructured quadrilateral grid in the present work. For the first time, the cell-based adaptive mesh refinement (AMR) is fully implemented on GPU for the unstructured quadrilateral grid, which greatly reduces the frequency of data exchange between GPU and CPU. Specifically, the AMR is processed with atomic operations to parallelize list operations, and null memory recycling is realized to improve the efficiency of memory utilization. It is found that results obtained by GPUs agree very well with the exact or experimental results in literature. An acceleration ratio of 4 is obtained between the parallel code running on the old GPU GT9800 and the serial code running on E3-1230 V2. With the optimization of configuring a larger L1 cache and adopting Shared Memory based atomic operations on the newer GPU C2050, an acceleration ratio of 20 is achieved. The parallelized cell-based AMR processes have achieved 2x speedup on GT9800 and 18x on Tesla C2050, which demonstrates that parallel running of the cell-based AMR method on GPU is feasible and efficient. Our results also indicate that the new development of GPU architecture benefits the fluid dynamics computing significantly.

SMT-Aware Instantaneous Footprint Optimization

DOE Office of Scientific and Technical Information (OSTI.GOV)

Roy, Probir; Liu, Xu; Song, Shuaiwen

Modern architectures employ simultaneous multithreading (SMT) to increase thread-level parallelism. SMT threads share many functional units and the whole memory hierarchy of a physical core. Without a careful code design, SMT threads can easily contend with each other for these shared resources, causing severe performance degradation. Minimizing SMT thread contention for HPC applications running on dedicated platforms is very challenging, because they usually spawn threads within Single Program Multiple Data (SPMD) models. To address this important issue, we introduce a simple scheme for SMT-aware code optimization, which aims to reduce the memory contention across SMT threads.

Large-Constraint-Length, Fast Viterbi Decoder

NASA Technical Reports Server (NTRS)

Collins, O.; Dolinar, S.; Hsu, In-Shek; Pollara, F.; Olson, E.; Statman, J.; Zimmerman, G.

1990-01-01

Scheme for efficient interconnection makes VLSI design feasible. Concept for fast Viterbi decoder provides for processing of convolutional codes of constraint length K up to 15 and rates of 1/2 to 1/6. Fully parallel (but bit-serial) architecture developed for decoder of K = 7 implemented in single dedicated VLSI circuit chip. Contains six major functional blocks. VLSI circuits perform branch metric computations, add-compare-select operations, and then store decisions in traceback memory. Traceback processor reads appropriate memory locations and puts out decoded bits. Used as building block for decoders of larger K.

Computing NLTE Opacities -- Node Level Parallel Calculation

DOE Office of Scientific and Technical Information (OSTI.GOV)

Holladay, Daniel

Presentation. The goal: to produce a robust library capable of computing reasonably accurate opacities inline with the assumption of LTE relaxed (non-LTE). Near term: demonstrate acceleration of non-LTE opacity computation. Far term (if funded): connect to application codes with in-line capability and compute opacities. Study science problems. Use efficient algorithms that expose many levels of parallelism and utilize good memory access patterns for use on advanced architectures. Portability to multiple types of hardware including multicore processors, manycore processors such as KNL, GPUs, etc. Easily coupled to radiation hydrodynamics and thermal radiative transfer codes.

A performance comparison of the IBM RS/6000 and the Astronautics ZS-1

DOE Office of Scientific and Technical Information (OSTI.GOV)

Smith, W.M.; Abraham, S.G.; Davidson, E.S.

1991-01-01

Concurrent uniprocessor architectures, of which vector and superscalar are two examples, are designed to capitalize on fine-grain parallelism. The authors have developed a performance evaluation method for comparing and improving these architectures, and in this article they present the methodology and a detailed case study of two machines. The runtime of many programs is dominated by time spent in loop constructs - for example, Fortran Do-loops. Loops generally comprise two logical processes: The access process generates addresses for memory operations while the execute process operates on floating-point data. Memory access patterns typically can be generated independently of the data inmore » the execute process. This independence allows the access process to slip ahead, thereby hiding memory latency. The IBM 360/91 was designed in 1967 to achieve slip dynamically, at runtime. One CPU unit executes integer operations while another handles floating-point operations. Other machines, including the VAX 9000 and the IBM RS/6000, use a similar approach.« less

The medial temporal lobe-conduit of parallel connectivity: a model for attention, memory, and perception.

PubMed

Mozaffari, Brian

2014-01-01

Based on the notion that the brain is equipped with a hierarchical organization, which embodies environmental contingencies across many time scales, this paper suggests that the medial temporal lobe (MTL)-located deep in the hierarchy-serves as a bridge connecting supra- to infra-MTL levels. Bridging the upper and lower regions of the hierarchy provides a parallel architecture that optimizes information flow between upper and lower regions to aid attention, encoding, and processing of quick complex visual phenomenon. Bypassing intermediate hierarchy levels, information conveyed through the MTL "bridge" allows upper levels to make educated predictions about the prevailing context and accordingly select lower representations to increase the efficiency of predictive coding throughout the hierarchy. This selection or activation/deactivation is associated with endogenous attention. In the event that these "bridge" predictions are inaccurate, this architecture enables the rapid encoding of novel contingencies. A review of hierarchical models in relation to memory is provided along with a new theory, Medial-temporal-lobe Conduit for Parallel Connectivity (MCPC). In this scheme, consolidation is considered as a secondary process, occurring after a MTL-bridged connection, which eventually allows upper and lower levels to access each other directly. With repeated reactivations, as contingencies become consolidated, less MTL activity is predicted. Finally, MTL bridging may aid processing transient but structured perceptual events, by allowing communication between upper and lower levels without calling on intermediate levels of representation.

A class Hierarchical, object-oriented approach to virtual memory management

NASA Technical Reports Server (NTRS)

Russo, Vincent F.; Campbell, Roy H.; Johnston, Gary M.

1989-01-01

The Choices family of operating systems exploits class hierarchies and object-oriented programming to facilitate the construction of customized operating systems for shared memory and networked multiprocessors. The software is being used in the Tapestry laboratory to study the performance of algorithms, mechanisms, and policies for parallel systems. Described here are the architectural design and class hierarchy of the Choices virtual memory management system. The software and hardware mechanisms and policies of a virtual memory system implement a memory hierarchy that exploits the trade-off between response times and storage capacities. In Choices, the notion of a memory hierarchy is captured by abstract classes. Concrete subclasses of those abstractions implement a virtual address space, segmentation, paging, physical memory management, secondary storage, and remote (that is, networked) storage. Captured in the notion of a memory hierarchy are classes that represent memory objects. These classes provide a storage mechanism that contains encapsulated data and have methods to read or write the memory object. Each of these classes provides specializations to represent the memory hierarchy.

Studying an Eulerian Computer Model on Different High-performance Computer Platforms and Some Applications

NASA Astrophysics Data System (ADS)

Georgiev, K.; Zlatev, Z.

2010-11-01

The Danish Eulerian Model (DEM) is an Eulerian model for studying the transport of air pollutants on large scale. Originally, the model was developed at the National Environmental Research Institute of Denmark. The model computational domain covers Europe and some neighbour parts belong to the Atlantic Ocean, Asia and Africa. If DEM model is to be applied by using fine grids, then its discretization leads to a huge computational problem. This implies that such a model as DEM must be run only on high-performance computer architectures. The implementation and tuning of such a complex large-scale model on each different computer is a non-trivial task. Here, some comparison results of running of this model on different kind of vector (CRAY C92A, Fujitsu, etc.), parallel computers with distributed memory (IBM SP, CRAY T3E, Beowulf clusters, Macintosh G4 clusters, etc.), parallel computers with shared memory (SGI Origin, SUN, etc.) and parallel computers with two levels of parallelism (IBM SMP, IBM BlueGene/P, clusters of multiprocessor nodes, etc.) will be presented. The main idea in the parallel version of DEM is domain partitioning approach. Discussions according to the effective use of the cache and hierarchical memories of the modern computers as well as the performance, speed-ups and efficiency achieved will be done. The parallel code of DEM, created by using MPI standard library, appears to be highly portable and shows good efficiency and scalability on different kind of vector and parallel computers. Some important applications of the computer model output are presented in short.

A highly efficient 3D level-set grain growth algorithm tailored for ccNUMA architecture

NASA Astrophysics Data System (ADS)

Mießen, C.; Velinov, N.; Gottstein, G.; Barrales-Mora, L. A.

2017-12-01

A highly efficient simulation model for 2D and 3D grain growth was developed based on the level-set method. The model introduces modern computational concepts to achieve excellent performance on parallel computer architectures. Strong scalability was measured on cache-coherent non-uniform memory access (ccNUMA) architectures. To achieve this, the proposed approach considers the application of local level-set functions at the grain level. Ideal and non-ideal grain growth was simulated in 3D with the objective to study the evolution of statistical representative volume elements in polycrystals. In addition, microstructure evolution in an anisotropic magnetic material affected by an external magnetic field was simulated.

Performance evaluation of throughput computing workloads using multi-core processors and graphics processors

NASA Astrophysics Data System (ADS)

Dave, Gaurav P.; Sureshkumar, N.; Blessy Trencia Lincy, S. S.

2017-11-01

Current trend in processor manufacturing focuses on multi-core architectures rather than increasing the clock speed for performance improvement. Graphic processors have become as commodity hardware for providing fast co-processing in computer systems. Developments in IoT, social networking web applications, big data created huge demand for data processing activities and such kind of throughput intensive applications inherently contains data level parallelism which is more suited for SIMD architecture based GPU. This paper reviews the architectural aspects of multi/many core processors and graphics processors. Different case studies are taken to compare performance of throughput computing applications using shared memory programming in OpenMP and CUDA API based programming.

Parallel Finite Element Domain Decomposition for Structural/Acoustic Analysis

NASA Technical Reports Server (NTRS)

Nguyen, Duc T.; Tungkahotara, Siroj; Watson, Willie R.; Rajan, Subramaniam D.

2005-01-01

A domain decomposition (DD) formulation for solving sparse linear systems of equations resulting from finite element analysis is presented. The formulation incorporates mixed direct and iterative equation solving strategics and other novel algorithmic ideas that are optimized to take advantage of sparsity and exploit modern computer architecture, such as memory and parallel computing. The most time consuming part of the formulation is identified and the critical roles of direct sparse and iterative solvers within the framework of the formulation are discussed. Experiments on several computer platforms using several complex test matrices are conducted using software based on the formulation. Small-scale structural examples are used to validate thc steps in the formulation and large-scale (l,000,000+ unknowns) duct acoustic examples are used to evaluate the ORIGIN 2000 processors, and a duster of 6 PCs (running under the Windows environment). Statistics show that the formulation is efficient in both sequential and parallel computing environmental and that the formulation is significantly faster and consumes less memory than that based on one of the best available commercialized parallel sparse solvers.

Parallel rendering

NASA Technical Reports Server (NTRS)

Crockett, Thomas W.

1995-01-01

This article provides a broad introduction to the subject of parallel rendering, encompassing both hardware and software systems. The focus is on the underlying concepts and the issues which arise in the design of parallel rendering algorithms and systems. We examine the different types of parallelism and how they can be applied in rendering applications. Concepts from parallel computing, such as data decomposition, task granularity, scalability, and load balancing, are considered in relation to the rendering problem. We also explore concepts from computer graphics, such as coherence and projection, which have a significant impact on the structure of parallel rendering algorithms. Our survey covers a number of practical considerations as well, including the choice of architectural platform, communication and memory requirements, and the problem of image assembly and display. We illustrate the discussion with numerous examples from the parallel rendering literature, representing most of the principal rendering methods currently used in computer graphics.

Parallel Computation of the Jacobian Matrix for Nonlinear Equation Solvers Using MATLAB

NASA Technical Reports Server (NTRS)

Rose, Geoffrey K.; Nguyen, Duc T.; Newman, Brett A.

2017-01-01

Demonstrating speedup for parallel code on a multicore shared memory PC can be challenging in MATLAB due to underlying parallel operations that are often opaque to the user. This can limit potential for improvement of serial code even for the so-called embarrassingly parallel applications. One such application is the computation of the Jacobian matrix inherent to most nonlinear equation solvers. Computation of this matrix represents the primary bottleneck in nonlinear solver speed such that commercial finite element (FE) and multi-body-dynamic (MBD) codes attempt to minimize computations. A timing study using MATLAB's Parallel Computing Toolbox was performed for numerical computation of the Jacobian. Several approaches for implementing parallel code were investigated while only the single program multiple data (spmd) method using composite objects provided positive results. Parallel code speedup is demonstrated but the goal of linear speedup through the addition of processors was not achieved due to PC architecture.

Final report for the Tera Computer TTI CRADA

DOE Office of Scientific and Technical Information (OSTI.GOV)

Davidson, G.S.; Pavlakos, C.; Silva, C.

1997-01-01

Tera Computer and Sandia National Laboratories have completed a CRADA, which examined the Tera Multi-Threaded Architecture (MTA) for use with large codes of importance to industry and DOE. The MTA is an innovative architecture that uses parallelism to mask latency between memories and processors. The physical implementation is a parallel computer with high cross-section bandwidth and GaAs processors designed by Tera, which support many small computation threads and fast, lightweight context switches between them. When any thread blocks while waiting for memory accesses to complete, another thread immediately begins execution so that high CPU utilization is maintained. The Tera MTAmore » parallel computer has a single, global address space, which is appealing when porting existing applications to a parallel computer. This ease of porting is further enabled by compiler technology that helps break computations into parallel threads. DOE and Sandia National Laboratories were interested in working with Tera to further develop this computing concept. While Tera Computer would continue the hardware development and compiler research, Sandia National Laboratories would work with Tera to ensure that their compilers worked well with important Sandia codes, most particularly CTH, a shock physics code used for weapon safety computations. In addition to that important code, Sandia National Laboratories would complete research on a robotic path planning code, SANDROS, which is important in manufacturing applications, and would evaluate the MTA performance on this code. Finally, Sandia would work directly with Tera to develop 3D visualization codes, which would be appropriate for use with the MTA. Each of these tasks has been completed to the extent possible, given that Tera has just completed the MTA hardware. All of the CRADA work had to be done on simulators.« less

An efficient spectral crystal plasticity solver for GPU architectures

NASA Astrophysics Data System (ADS)

Malahe, Michael

2018-03-01

We present a spectral crystal plasticity (CP) solver for graphics processing unit (GPU) architectures that achieves a tenfold increase in efficiency over prior GPU solvers. The approach makes use of a database containing a spectral decomposition of CP simulations performed using a conventional iterative solver over a parameter space of crystal orientations and applied velocity gradients. The key improvements in efficiency come from reducing global memory transactions, exposing more instruction-level parallelism, reducing integer instructions and performing fast range reductions on trigonometric arguments. The scheme also makes more efficient use of memory than prior work, allowing for larger problems to be solved on a single GPU. We illustrate these improvements with a simulation of 390 million crystal grains on a consumer-grade GPU, which executes at a rate of 2.72 s per strain step.

Modern multicore and manycore architectures: Modelling, optimisation and benchmarking a multiblock CFD code

NASA Astrophysics Data System (ADS)

Hadade, Ioan; di Mare, Luca

2016-08-01

Modern multicore and manycore processors exhibit multiple levels of parallelism through a wide range of architectural features such as SIMD for data parallel execution or threads for core parallelism. The exploitation of multi-level parallelism is therefore crucial for achieving superior performance on current and future processors. This paper presents the performance tuning of a multiblock CFD solver on Intel SandyBridge and Haswell multicore CPUs and the Intel Xeon Phi Knights Corner coprocessor. Code optimisations have been applied on two computational kernels exhibiting different computational patterns: the update of flow variables and the evaluation of the Roe numerical fluxes. We discuss at great length the code transformations required for achieving efficient SIMD computations for both kernels across the selected devices including SIMD shuffles and transpositions for flux stencil computations and global memory transformations. Core parallelism is expressed through threading based on a number of domain decomposition techniques together with optimisations pertaining to alleviating NUMA effects found in multi-socket compute nodes. Results are correlated with the Roofline performance model in order to assert their efficiency for each distinct architecture. We report significant speedups for single thread execution across both kernels: 2-5X on the multicore CPUs and 14-23X on the Xeon Phi coprocessor. Computations at full node and chip concurrency deliver a factor of three speedup on the multicore processors and up to 24X on the Xeon Phi manycore coprocessor.

Parallel Conjugate Gradient: Effects of Ordering Strategies, Programming Paradigms, and Architectural Platforms

NASA Technical Reports Server (NTRS)

Oliker, Leonid; Heber, Gerd; Biswas, Rupak

2000-01-01

The Conjugate Gradient (CG) algorithm is perhaps the best-known iterative technique to solve sparse linear systems that are symmetric and positive definite. A sparse matrix-vector multiply (SPMV) usually accounts for most of the floating-point operations within a CG iteration. In this paper, we investigate the effects of various ordering and partitioning strategies on the performance of parallel CG and SPMV using different programming paradigms and architectures. Results show that for this class of applications, ordering significantly improves overall performance, that cache reuse may be more important than reducing communication, and that it is possible to achieve message passing performance using shared memory constructs through careful data ordering and distribution. However, a multi-threaded implementation of CG on the Tera MTA does not require special ordering or partitioning to obtain high efficiency and scalability.

Scalable Algorithms for Clustering Large Geospatiotemporal Data Sets on Manycore Architectures

NASA Astrophysics Data System (ADS)

Mills, R. T.; Hoffman, F. M.; Kumar, J.; Sreepathi, S.; Sripathi, V.

2016-12-01

The increasing availability of high-resolution geospatiotemporal data sets from sources such as observatory networks, remote sensing platforms, and computational Earth system models has opened new possibilities for knowledge discovery using data sets fused from disparate sources. Traditional algorithms and computing platforms are impractical for the analysis and synthesis of data sets of this size; however, new algorithmic approaches that can effectively utilize the complex memory hierarchies and the extremely high levels of available parallelism in state-of-the-art high-performance computing platforms can enable such analysis. We describe a massively parallel implementation of accelerated k-means clustering and some optimizations to boost computational intensity and utilization of wide SIMD lanes on state-of-the art multi- and manycore processors, including the second-generation Intel Xeon Phi ("Knights Landing") processor based on the Intel Many Integrated Core (MIC) architecture, which includes several new features, including an on-package high-bandwidth memory. We also analyze the code in the context of a few practical applications to the analysis of climatic and remotely-sensed vegetation phenology data sets, and speculate on some of the new applications that such scalable analysis methods may enable.

A Tensor Product Formulation of Strassen's Matrix Multiplication Algorithm with Memory Reduction

DOE PAGES

Kumar, B.; Huang, C. -H.; Sadayappan, P.; ...

1995-01-01

In this article, we present a program generation strategy of Strassen's matrix multiplication algorithm using a programming methodology based on tensor product formulas. In this methodology, block recursive programs such as the fast Fourier Transforms and Strassen's matrix multiplication algorithm are expressed as algebraic formulas involving tensor products and other matrix operations. Such formulas can be systematically translated to high-performance parallel/vector codes for various architectures. In this article, we present a nonrecursive implementation of Strassen's algorithm for shared memory vector processors such as the Cray Y-MP. A previous implementation of Strassen's algorithm synthesized from tensor product formulas required working storagemore » of size O(7 n ) for multiplying 2 n × 2 n matrices. We present a modified formulation in which the working storage requirement is reduced to O(4 n ). The modified formulation exhibits sufficient parallelism for efficient implementation on a shared memory multiprocessor. Performance results on a Cray Y-MP8/64 are presented.« less

A portable approach for PIC on emerging architectures

NASA Astrophysics Data System (ADS)

Decyk, Viktor

2016-03-01

A portable approach for designing Particle-in-Cell (PIC) algorithms on emerging exascale computers, is based on the recognition that 3 distinct programming paradigms are needed. They are: low level vector (SIMD) processing, middle level shared memory parallel programing, and high level distributed memory programming. In addition, there is a memory hierarchy associated with each level. Such algorithms can be initially developed using vectorizing compilers, OpenMP, and MPI. This is the approach recommended by Intel for the Phi processor. These algorithms can then be translated and possibly specialized to other programming models and languages, as needed. For example, the vector processing and shared memory programming might be done with CUDA instead of vectorizing compilers and OpenMP, but generally the algorithm itself is not greatly changed. The UCLA PICKSC web site at http://www.idre.ucla.edu/ contains example open source skeleton codes (mini-apps) illustrating each of these three programming models, individually and in combination. Fortran2003 now supports abstract data types, and design patterns can be used to support a variety of implementations within the same code base. Fortran2003 also supports interoperability with C so that implementations in C languages are also easy to use. Finally, main codes can be translated into dynamic environments such as Python, while still taking advantage of high performing compiled languages. Parallel languages are still evolving with interesting developments in co-Array Fortran, UPC, and OpenACC, among others, and these can also be supported within the same software architecture. Work supported by NSF and DOE Grants.

Design and DSP implementation of star image acquisition and star point fast acquiring and tracking

NASA Astrophysics Data System (ADS)

Zhou, Guohui; Wang, Xiaodong; Hao, Zhihang

2006-02-01

Star sensor is a special high accuracy photoelectric sensor. Attitude acquisition time is an important function index of star sensor. In this paper, the design target is to acquire 10 samples per second dynamic performance. On the basis of analyzing CCD signals timing and star image processing, a new design and a special parallel architecture for improving star image processing are presented in this paper. In the design, the operation moving the data in expanded windows including the star to the on-chip memory of DSP is arranged in the invalid period of CCD frame signal. During the CCD saving the star image to memory, DSP processes the data in the on-chip memory. This parallelism greatly improves the efficiency of processing. The scheme proposed here results in enormous savings of memory normally required. In the scheme, DSP HOLD mode and CPLD technology are used to make a shared memory between CCD and DSP. The efficiency of processing is discussed in numerical tests. Only in 3.5ms is acquired the five lightest stars in the star acquisition stage. In 43us, the data in five expanded windows including stars are moved into the internal memory of DSP, and in 1.6ms, five star coordinates are achieved in the star tracking stage.

Experimental characterization of a binary actuated parallel manipulator

NASA Astrophysics Data System (ADS)

Giuseppe, Carbone

2016-05-01

This paper describes the BAPAMAN (Binary Actuated Parallel MANipulator) series of parallel manipulators that has been conceived at Laboratory of Robotics and Mechatronics (LARM). Basic common characteristics of BAPAMAN series are described. In particular, it is outlined the use of a reduced number of active degrees of freedom, the use of design solutions with flexural joints and Shape Memory Alloy (SMA) actuators for achieving miniaturization, cost reduction and easy operation features. Given the peculiarities of BAPAMAN architecture, specific experimental tests have been proposed and carried out with the aim to validate the proposed design and to evaluate the practical operation performance and the characteristics of a built prototype, in particular, in terms of operation and workspace characteristics.

On some methods for improving time of reachability sets computation for the dynamic system control problem

NASA Astrophysics Data System (ADS)

Zimovets, Artem; Matviychuk, Alexander; Ushakov, Vladimir

2016-12-01

The paper presents two different approaches to reduce the time of computer calculation of reachability sets. First of these two approaches use different data structures for storing the reachability sets in the computer memory for calculation in single-threaded mode. Second approach is based on using parallel algorithms with reference to the data structures from the first approach. Within the framework of this paper parallel algorithm of approximate reachability set calculation on computer with SMP-architecture is proposed. The results of numerical modelling are presented in the form of tables which demonstrate high efficiency of parallel computing technology and also show how computing time depends on the used data structure.

Eigensolver for a Sparse, Large Hermitian Matrix

NASA Technical Reports Server (NTRS)

Tisdale, E. Robert; Oyafuso, Fabiano; Klimeck, Gerhard; Brown, R. Chris

2003-01-01

A parallel-processing computer program finds a few eigenvalues in a sparse Hermitian matrix that contains as many as 100 million diagonal elements. This program finds the eigenvalues faster, using less memory, than do other, comparable eigensolver programs. This program implements a Lanczos algorithm in the American National Standards Institute/ International Organization for Standardization (ANSI/ISO) C computing language, using the Message Passing Interface (MPI) standard to complement an eigensolver in PARPACK. [PARPACK (Parallel Arnoldi Package) is an extension, to parallel-processing computer architectures, of ARPACK (Arnoldi Package), which is a collection of Fortran 77 subroutines that solve large-scale eigenvalue problems.] The eigensolver runs on Beowulf clusters of computers at the Jet Propulsion Laboratory (JPL).

A FAST ITERATIVE METHOD FOR SOLVING THE EIKONAL EQUATION ON TRIANGULATED SURFACES*

PubMed Central

Fu, Zhisong; Jeong, Won-Ki; Pan, Yongsheng; Kirby, Robert M.; Whitaker, Ross T.

2012-01-01

This paper presents an efficient, fine-grained parallel algorithm for solving the Eikonal equation on triangular meshes. The Eikonal equation, and the broader class of Hamilton–Jacobi equations to which it belongs, have a wide range of applications from geometric optics and seismology to biological modeling and analysis of geometry and images. The ability to solve such equations accurately and efficiently provides new capabilities for exploring and visualizing parameter spaces and for solving inverse problems that rely on such equations in the forward model. Efficient solvers on state-of-the-art, parallel architectures require new algorithms that are not, in many cases, optimal, but are better suited to synchronous updates of the solution. In previous work [W. K. Jeong and R. T. Whitaker, SIAM J. Sci. Comput., 30 (2008), pp. 2512–2534], the authors proposed the fast iterative method (FIM) to efficiently solve the Eikonal equation on regular grids. In this paper we extend the fast iterative method to solve Eikonal equations efficiently on triangulated domains on the CPU and on parallel architectures, including graphics processors. We propose a new local update scheme that provides solutions of first-order accuracy for both architectures. We also propose a novel triangle-based update scheme and its corresponding data structure for efficient irregular data mapping to parallel single-instruction multiple-data (SIMD) processors. We provide detailed descriptions of the implementations on a single CPU, a multicore CPU with shared memory, and SIMD architectures with comparative results against state-of-the-art Eikonal solvers. PMID:22641200

Optical interconnection network for parallel access to multi-rank memory in future computing systems.

PubMed

Wang, Kang; Gu, Huaxi; Yang, Yintang; Wang, Kun

2015-08-10

With the number of cores increasing, there is an emerging need for a high-bandwidth low-latency interconnection network, serving core-to-memory communication. In this paper, aiming at the goal of simultaneous access to multi-rank memory, we propose an optical interconnection network for core-to-memory communication. In the proposed network, the wavelength usage is delicately arranged so that cores can communicate with different ranks at the same time and broadcast for flow control can be achieved. A distributed memory controller architecture that works in a pipeline mode is also designed for efficient optical communication and transaction address processes. The scaling method and wavelength assignment for the proposed network are investigated. Compared with traditional electronic bus-based core-to-memory communication, the simulation results based on the PARSEC benchmark show that the bandwidth enhancement and latency reduction are apparent.

High order parallel numerical schemes for solving incompressible flows

NASA Technical Reports Server (NTRS)

Lin, Avi; Milner, Edward J.; Liou, May-Fun; Belch, Richard A.

1992-01-01

The use of parallel computers for numerically solving flow fields has gained much importance in recent years. This paper introduces a new high order numerical scheme for computational fluid dynamics (CFD) specifically designed for parallel computational environments. A distributed MIMD system gives the flexibility of treating different elements of the governing equations with totally different numerical schemes in different regions of the flow field. The parallel decomposition of the governing operator to be solved is the primary parallel split. The primary parallel split was studied using a hypercube like architecture having clusters of shared memory processors at each node. The approach is demonstrated using examples of simple steady state incompressible flows. Future studies should investigate the secondary split because, depending on the numerical scheme that each of the processors applies and the nature of the flow in the specific subdomain, it may be possible for a processor to seek better, or higher order, schemes for its particular subcase.

Parallelization of the FLAPW method

NASA Astrophysics Data System (ADS)

Canning, A.; Mannstadt, W.; Freeman, A. J.

2000-08-01

The FLAPW (full-potential linearized-augmented plane-wave) method is one of the most accurate first-principles methods for determining structural, electronic and magnetic properties of crystals and surfaces. Until the present work, the FLAPW method has been limited to systems of less than about a hundred atoms due to the lack of an efficient parallel implementation to exploit the power and memory of parallel computers. In this work, we present an efficient parallelization of the method by division among the processors of the plane-wave components for each state. The code is also optimized for RISC (reduced instruction set computer) architectures, such as those found on most parallel computers, making full use of BLAS (basic linear algebra subprograms) wherever possible. Scaling results are presented for systems of up to 686 silicon atoms and 343 palladium atoms per unit cell, running on up to 512 processors on a CRAY T3E parallel supercomputer.

Partitioning problems in parallel, pipelined and distributed computing

NASA Technical Reports Server (NTRS)

Bokhari, S.

1985-01-01

The problem of optimally assigning the modules of a parallel program over the processors of a multiple computer system is addressed. A Sum-Bottleneck path algorithm is developed that permits the efficient solution of many variants of this problem under some constraints on the structure of the partitions. In particular, the following problems are solved optimally for a single-host, multiple satellite system: partitioning multiple chain structured parallel programs, multiple arbitrarily structured serial programs and single tree structured parallel programs. In addition, the problems of partitioning chain structured parallel programs across chain connected systems and across shared memory (or shared bus) systems are also solved under certain constraints. All solutions for parallel programs are equally applicable to pipelined programs. These results extend prior research in this area by explicitly taking concurrency into account and permit the efficient utilization of multiple computer architectures for a wide range of problems of practical interest.

The BlueGene/L supercomputer

NASA Astrophysics Data System (ADS)

Bhanota, Gyan; Chen, Dong; Gara, Alan; Vranas, Pavlos

2003-05-01

The architecture of the BlueGene/L massively parallel supercomputer is described. Each computing node consists of a single compute ASIC plus 256 MB of external memory. The compute ASIC integrates two 700 MHz PowerPC 440 integer CPU cores, two 2.8 Gflops floating point units, 4 MB of embedded DRAM as cache, a memory controller for external memory, six 1.4 Gbit/s bi-directional ports for a 3-dimensional torus network connection, three 2.8 Gbit/s bi-directional ports for connecting to a global tree network and a Gigabit Ethernet for I/O. 65,536 of such nodes are connected into a 3-d torus with a geometry of 32×32×64. The total peak performance of the system is 360 Teraflops and the total amount of memory is 16 TeraBytes.

A simple GPU-accelerated two-dimensional MUSCL-Hancock solver for ideal magnetohydrodynamics

NASA Astrophysics Data System (ADS)

Bard, Christopher M.; Dorelli, John C.

2014-02-01

We describe our experience using NVIDIA's CUDA (Compute Unified Device Architecture) C programming environment to implement a two-dimensional second-order MUSCL-Hancock ideal magnetohydrodynamics (MHD) solver on a GTX 480 Graphics Processing Unit (GPU). Taking a simple approach in which the MHD variables are stored exclusively in the global memory of the GTX 480 and accessed in a cache-friendly manner (without further optimizing memory access by, for example, staging data in the GPU's faster shared memory), we achieved a maximum speed-up of ≈126 for a 10242 grid relative to the sequential C code running on a single Intel Nehalem (2.8 GHz) core. This speedup is consistent with simple estimates based on the known floating point performance, memory throughput and parallel processing capacity of the GTX 480.

An Element-Based Concurrent Partitioner for Unstructured Finite Element Meshes

NASA Technical Reports Server (NTRS)

Ding, Hong Q.; Ferraro, Robert D.

1996-01-01

A concurrent partitioner for partitioning unstructured finite element meshes on distributed memory architectures is developed. The partitioner uses an element-based partitioning strategy. Its main advantage over the more conventional node-based partitioning strategy is its modular programming approach to the development of parallel applications. The partitioner first partitions element centroids using a recursive inertial bisection algorithm. Elements and nodes then migrate according to the partitioned centroids, using a data request communication template for unpredictable incoming messages. Our scalable implementation is contrasted to a non-scalable implementation which is a straightforward parallelization of a sequential partitioner.

Exascale Hardware Architectures Working Group

DOE Office of Scientific and Technical Information (OSTI.GOV)

Hemmert, S; Ang, J; Chiang, P

2011-03-15

The ASC Exascale Hardware Architecture working group is challenged to provide input on the following areas impacting the future use and usability of potential exascale computer systems: processor, memory, and interconnect architectures, as well as the power and resilience of these systems. Going forward, there are many challenging issues that will need to be addressed. First, power constraints in processor technologies will lead to steady increases in parallelism within a socket. Additionally, all cores may not be fully independent nor fully general purpose. Second, there is a clear trend toward less balanced machines, in terms of compute capability compared tomore » memory and interconnect performance. In order to mitigate the memory issues, memory technologies will introduce 3D stacking, eventually moving on-socket and likely on-die, providing greatly increased bandwidth but unfortunately also likely providing smaller memory capacity per core. Off-socket memory, possibly in the form of non-volatile memory, will create a complex memory hierarchy. Third, communication energy will dominate the energy required to compute, such that interconnect power and bandwidth will have a significant impact. All of the above changes are driven by the need for greatly increased energy efficiency, as current technology will prove unsuitable for exascale, due to unsustainable power requirements of such a system. These changes will have the most significant impact on programming models and algorithms, but they will be felt across all layers of the machine. There is clear need to engage all ASC working groups in planning for how to deal with technological changes of this magnitude. The primary function of the Hardware Architecture Working Group is to facilitate codesign with hardware vendors to ensure future exascale platforms are capable of efficiently supporting the ASC applications, which in turn need to meet the mission needs of the NNSA Stockpile Stewardship Program. This issue is relatively immediate, as there is only a small window of opportunity to influence hardware design for 2018 machines. Given the short timeline a firm co-design methodology with vendors is of prime importance.« less

Parallelization of a Monte Carlo particle transport simulation code

NASA Astrophysics Data System (ADS)

Hadjidoukas, P.; Bousis, C.; Emfietzoglou, D.

2010-05-01

We have developed a high performance version of the Monte Carlo particle transport simulation code MC4. The original application code, developed in Visual Basic for Applications (VBA) for Microsoft Excel, was first rewritten in the C programming language for improving code portability. Several pseudo-random number generators have been also integrated and studied. The new MC4 version was then parallelized for shared and distributed-memory multiprocessor systems using the Message Passing Interface. Two parallel pseudo-random number generator libraries (SPRNG and DCMT) have been seamlessly integrated. The performance speedup of parallel MC4 has been studied on a variety of parallel computing architectures including an Intel Xeon server with 4 dual-core processors, a Sun cluster consisting of 16 nodes of 2 dual-core AMD Opteron processors and a 200 dual-processor HP cluster. For large problem size, which is limited only by the physical memory of the multiprocessor server, the speedup results are almost linear on all systems. We have validated the parallel implementation against the serial VBA and C implementations using the same random number generator. Our experimental results on the transport and energy loss of electrons in a water medium show that the serial and parallel codes are equivalent in accuracy. The present improvements allow for studying of higher particle energies with the use of more accurate physical models, and improve statistics as more particles tracks can be simulated in low response time.

First experience of vectorizing electromagnetic physics models for detector simulation

NASA Astrophysics Data System (ADS)

Amadio, G.; Apostolakis, J.; Bandieramonte, M.; Bianchini, C.; Bitzes, G.; Brun, R.; Canal, P.; Carminati, F.; de Fine Licht, J.; Duhem, L.; Elvira, D.; Gheata, A.; Jun, S. Y.; Lima, G.; Novak, M.; Presbyterian, M.; Shadura, O.; Seghal, R.; Wenzel, S.

2015-12-01

The recent emergence of hardware architectures characterized by many-core or accelerated processors has opened new opportunities for concurrent programming models taking advantage of both SIMD and SIMT architectures. The GeantV vector prototype for detector simulations has been designed to exploit both the vector capability of mainstream CPUs and multi-threading capabilities of coprocessors including NVidia GPUs and Intel Xeon Phi. The characteristics of these architectures are very different in terms of the vectorization depth, parallelization needed to achieve optimal performance or memory access latency and speed. An additional challenge is to avoid the code duplication often inherent to supporting heterogeneous platforms. In this paper we present the first experience of vectorizing electromagnetic physics models developed for the GeantV project.

An efficient parallel algorithm: Poststack and prestack Kirchhoff 3D depth migration using flexi-depth iterations

NASA Astrophysics Data System (ADS)

Rastogi, Richa; Srivastava, Abhishek; Khonde, Kiran; Sirasala, Kirannmayi M.; Londhe, Ashutosh; Chavhan, Hitesh

2015-07-01

This paper presents an efficient parallel 3D Kirchhoff depth migration algorithm suitable for current class of multicore architecture. The fundamental Kirchhoff depth migration algorithm exhibits inherent parallelism however, when it comes to 3D data migration, as the data size increases the resource requirement of the algorithm also increases. This challenges its practical implementation even on current generation high performance computing systems. Therefore a smart parallelization approach is essential to handle 3D data for migration. The most compute intensive part of Kirchhoff depth migration algorithm is the calculation of traveltime tables due to its resource requirements such as memory/storage and I/O. In the current research work, we target this area and develop a competent parallel algorithm for post and prestack 3D Kirchhoff depth migration, using hybrid MPI+OpenMP programming techniques. We introduce a concept of flexi-depth iterations while depth migrating data in parallel imaging space, using optimized traveltime table computations. This concept provides flexibility to the algorithm by migrating data in a number of depth iterations, which depends upon the available node memory and the size of data to be migrated during runtime. Furthermore, it minimizes the requirements of storage, I/O and inter-node communication, thus making it advantageous over the conventional parallelization approaches. The developed parallel algorithm is demonstrated and analysed on Yuva II, a PARAM series of supercomputers. Optimization, performance and scalability experiment results along with the migration outcome show the effectiveness of the parallel algorithm.

Automatic partitioning of unstructured meshes for the parallel solution of problems in computational mechanics

NASA Technical Reports Server (NTRS)

Farhat, Charbel; Lesoinne, Michel

1993-01-01

Most of the recently proposed computational methods for solving partial differential equations on multiprocessor architectures stem from the 'divide and conquer' paradigm and involve some form of domain decomposition. For those methods which also require grids of points or patches of elements, it is often necessary to explicitly partition the underlying mesh, especially when working with local memory parallel processors. In this paper, a family of cost-effective algorithms for the automatic partitioning of arbitrary two- and three-dimensional finite element and finite difference meshes is presented and discussed in view of a domain decomposed solution procedure and parallel processing. The influence of the algorithmic aspects of a solution method (implicit/explicit computations), and the architectural specifics of a multiprocessor (SIMD/MIMD, startup/transmission time), on the design of a mesh partitioning algorithm are discussed. The impact of the partitioning strategy on load balancing, operation count, operator conditioning, rate of convergence and processor mapping is also addressed. Finally, the proposed mesh decomposition algorithms are demonstrated with realistic examples of finite element, finite volume, and finite difference meshes associated with the parallel solution of solid and fluid mechanics problems on the iPSC/2 and iPSC/860 multiprocessors.

Gregarious Data Re-structuring in a Many Core Architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Shrestha, Sunil; Manzano Franco, Joseph B.; Marquez, Andres

this paper, we have developed a new methodology that takes in consideration the access patterns from a single parallel actor (e.g. a thread), as well as, the access patterns of “grouped” parallel actors that share a resource (e.g. a distributed Level 3 cache). We start with a hierarchical tile code for our target machine and apply a series of transformations at the tile level to improve data residence in a given memory hierarchy level. The contribution of this paper includes (a) collaborative data restructuring for group reuse and (b) low overhead transformation technique to improve access pattern and bring closelymore » connected data elements together. Preliminary results in a many core architecture, Tilera TileGX, shows promising improvements over optimized OpenMP code (up to 31% increase in GFLOPS) and over our own previous work on fine grained runtimes (up to 16%) for selected kernels« less

Performances of multiprocessor multidisk architectures for continuous media storage

NASA Astrophysics Data System (ADS)

Gennart, Benoit A.; Messerli, Vincent; Hersch, Roger D.

1996-03-01

Multimedia interfaces increase the need for large image databases, capable of storing and reading streams of data with strict synchronicity and isochronicity requirements. In order to fulfill these requirements, we consider a parallel image server architecture which relies on arrays of intelligent disk nodes, each disk node being composed of one processor and one or more disks. This contribution analyzes through bottleneck performance evaluation and simulation the behavior of two multi-processor multi-disk architectures: a point-to-point architecture and a shared-bus architecture similar to current multiprocessor workstation architectures. We compare the two architectures on the basis of two multimedia algorithms: the compute-bound frame resizing by resampling and the data-bound disk-to-client stream transfer. The results suggest that the shared bus is a potential bottleneck despite its very high hardware throughput (400Mbytes/s) and that an architecture with addressable local memories located closely to their respective processors could partially remove this bottleneck. The point- to-point architecture is scalable and able to sustain high throughputs for simultaneous compute- bound and data-bound operations.

An FPGA-Based Massively Parallel Neuromorphic Cortex Simulator

PubMed Central

Wang, Runchun M.; Thakur, Chetan S.; van Schaik, André

2018-01-01

This paper presents a massively parallel and scalable neuromorphic cortex simulator designed for simulating large and structurally connected spiking neural networks, such as complex models of various areas of the cortex. The main novelty of this work is the abstraction of a neuromorphic architecture into clusters represented by minicolumns and hypercolumns, analogously to the fundamental structural units observed in neurobiology. Without this approach, simulating large-scale fully connected networks needs prohibitively large memory to store look-up tables for point-to-point connections. Instead, we use a novel architecture, based on the structural connectivity in the neocortex, such that all the required parameters and connections can be stored in on-chip memory. The cortex simulator can be easily reconfigured for simulating different neural networks without any change in hardware structure by programming the memory. A hierarchical communication scheme allows one neuron to have a fan-out of up to 200 k neurons. As a proof-of-concept, an implementation on one Altera Stratix V FPGA was able to simulate 20 million to 2.6 billion leaky-integrate-and-fire (LIF) neurons in real time. We verified the system by emulating a simplified auditory cortex (with 100 million neurons). This cortex simulator achieved a low power dissipation of 1.62 μW per neuron. With the advent of commercially available FPGA boards, our system offers an accessible and scalable tool for the design, real-time simulation, and analysis of large-scale spiking neural networks. PMID:29692702

An FPGA-Based Massively Parallel Neuromorphic Cortex Simulator.

PubMed

Wang, Runchun M; Thakur, Chetan S; van Schaik, André

2018-01-01

This paper presents a massively parallel and scalable neuromorphic cortex simulator designed for simulating large and structurally connected spiking neural networks, such as complex models of various areas of the cortex. The main novelty of this work is the abstraction of a neuromorphic architecture into clusters represented by minicolumns and hypercolumns, analogously to the fundamental structural units observed in neurobiology. Without this approach, simulating large-scale fully connected networks needs prohibitively large memory to store look-up tables for point-to-point connections. Instead, we use a novel architecture, based on the structural connectivity in the neocortex, such that all the required parameters and connections can be stored in on-chip memory. The cortex simulator can be easily reconfigured for simulating different neural networks without any change in hardware structure by programming the memory. A hierarchical communication scheme allows one neuron to have a fan-out of up to 200 k neurons. As a proof-of-concept, an implementation on one Altera Stratix V FPGA was able to simulate 20 million to 2.6 billion leaky-integrate-and-fire (LIF) neurons in real time. We verified the system by emulating a simplified auditory cortex (with 100 million neurons). This cortex simulator achieved a low power dissipation of 1.62 μW per neuron. With the advent of commercially available FPGA boards, our system offers an accessible and scalable tool for the design, real-time simulation, and analysis of large-scale spiking neural networks.

The medial temporal lobe—conduit of parallel connectivity: a model for attention, memory, and perception

PubMed Central

Mozaffari, Brian

2014-01-01

Based on the notion that the brain is equipped with a hierarchical organization, which embodies environmental contingencies across many time scales, this paper suggests that the medial temporal lobe (MTL)—located deep in the hierarchy—serves as a bridge connecting supra- to infra—MTL levels. Bridging the upper and lower regions of the hierarchy provides a parallel architecture that optimizes information flow between upper and lower regions to aid attention, encoding, and processing of quick complex visual phenomenon. Bypassing intermediate hierarchy levels, information conveyed through the MTL “bridge” allows upper levels to make educated predictions about the prevailing context and accordingly select lower representations to increase the efficiency of predictive coding throughout the hierarchy. This selection or activation/deactivation is associated with endogenous attention. In the event that these “bridge” predictions are inaccurate, this architecture enables the rapid encoding of novel contingencies. A review of hierarchical models in relation to memory is provided along with a new theory, Medial-temporal-lobe Conduit for Parallel Connectivity (MCPC). In this scheme, consolidation is considered as a secondary process, occurring after a MTL-bridged connection, which eventually allows upper and lower levels to access each other directly. With repeated reactivations, as contingencies become consolidated, less MTL activity is predicted. Finally, MTL bridging may aid processing transient but structured perceptual events, by allowing communication between upper and lower levels without calling on intermediate levels of representation. PMID:25426036

Stream Processors

NASA Astrophysics Data System (ADS)

Erez, Mattan; Dally, William J.

Stream processors, like other multi core architectures partition their functional units and storage into multiple processing elements. In contrast to typical architectures, which contain symmetric general-purpose cores and a cache hierarchy, stream processors have a significantly leaner design. Stream processors are specifically designed for the stream execution model, in which applications have large amounts of explicit parallel computation, structured and predictable control, and memory accesses that can be performed at a coarse granularity. Applications in the streaming model are expressed in a gather-compute-scatter form, yielding programs with explicit control over transferring data to and from on-chip memory. Relying on these characteristics, which are common to many media processing and scientific computing applications, stream architectures redefine the boundary between software and hardware responsibilities with software bearing much of the complexity required to manage concurrency, locality, and latency tolerance. Thus, stream processors have minimal control consisting of fetching medium- and coarse-grained instructions and executing them directly on the many ALUs. Moreover, the on-chip storage hierarchy of stream processors is under explicit software control, as is all communication, eliminating the need for complex reactive hardware mechanisms.

Coding for parallel execution of hardware-in-the-loop millimeter-wave scene generation models on multicore SIMD processor architectures

NASA Astrophysics Data System (ADS)

Olson, Richard F.

2013-05-01

Rendering of point scatterer based radar scenes for millimeter wave (mmW) seeker tests in real-time hardware-in-the-loop (HWIL) scene generation requires efficient algorithms and vector-friendly computer architectures for complex signal synthesis. New processor technology from Intel implements an extended 256-bit vector SIMD instruction set (AVX, AVX2) in a multi-core CPU design providing peak execution rates of hundreds of GigaFLOPS (GFLOPS) on one chip. Real world mmW scene generation code can approach peak SIMD execution rates only after careful algorithm and source code design. An effective software design will maintain high computing intensity emphasizing register-to-register SIMD arithmetic operations over data movement between CPU caches or off-chip memories. Engineers at the U.S. Army Aviation and Missile Research, Development and Engineering Center (AMRDEC) applied two basic parallel coding methods to assess new 256-bit SIMD multi-core architectures for mmW scene generation in HWIL. These include use of POSIX threads built on vector library functions and more portable, highlevel parallel code based on compiler technology (e.g. OpenMP pragmas and SIMD autovectorization). Since CPU technology is rapidly advancing toward high processor core counts and TeraFLOPS peak SIMD execution rates, it is imperative that coding methods be identified which produce efficient and maintainable parallel code. This paper describes the algorithms used in point scatterer target model rendering, the parallelization of those algorithms, and the execution performance achieved on an AVX multi-core machine using the two basic parallel coding methods. The paper concludes with estimates for scale-up performance on upcoming multi-core technology.

Associative architecture for image processing

NASA Astrophysics Data System (ADS)

Adar, Rutie; Akerib, Avidan

1997-09-01

This article presents a new generation in parallel processing architecture for real-time image processing. The approach is implemented in a real time image processor chip, called the XiumTM-2, based on combining a fully associative array which provides the parallel engine with a serial RISC core on the same die. The architecture is fully programmable and can be programmed to implement a wide range of color image processing, computer vision and media processing functions in real time. The associative part of the chip is based on patented pending methodology of Associative Computing Ltd. (ACL), which condenses 2048 associative processors, each of 128 'intelligent' bits. Each bit can be a processing bit or a memory bit. At only 33 MHz and 0.6 micron manufacturing technology process, the chip has a computational power of 3 billion ALU operations per second and 66 billion string search operations per second. The fully programmable nature of the XiumTM-2 chip enables developers to use ACL tools to write their own proprietary algorithms combined with existing image processing and analysis functions from ACL's extended set of libraries.

An object-oriented approach for parallel self adaptive mesh refinement on block structured grids

NASA Technical Reports Server (NTRS)

Lemke, Max; Witsch, Kristian; Quinlan, Daniel

1993-01-01

Self-adaptive mesh refinement dynamically matches the computational demands of a solver for partial differential equations to the activity in the application's domain. In this paper we present two C++ class libraries, P++ and AMR++, which significantly simplify the development of sophisticated adaptive mesh refinement codes on (massively) parallel distributed memory architectures. The development is based on our previous research in this area. The C++ class libraries provide abstractions to separate the issues of developing parallel adaptive mesh refinement applications into those of parallelism, abstracted by P++, and adaptive mesh refinement, abstracted by AMR++. P++ is a parallel array class library to permit efficient development of architecture independent codes for structured grid applications, and AMR++ provides support for self-adaptive mesh refinement on block-structured grids of rectangular non-overlapping blocks. Using these libraries, the application programmers' work is greatly simplified to primarily specifying the serial single grid application and obtaining the parallel and self-adaptive mesh refinement code with minimal effort. Initial results for simple singular perturbation problems solved by self-adaptive multilevel techniques (FAC, AFAC), being implemented on the basis of prototypes of the P++/AMR++ environment, are presented. Singular perturbation problems frequently arise in large applications, e.g. in the area of computational fluid dynamics. They usually have solutions with layers which require adaptive mesh refinement and fast basic solvers in order to be resolved efficiently.

SCaLeM: A Framework for Characterizing and Analyzing Execution Models

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chavarría-Miranda, Daniel; Manzano Franco, Joseph B.; Krishnamoorthy, Sriram

2014-10-13

As scalable parallel systems evolve towards more complex nodes with many-core architectures and larger trans-petascale & upcoming exascale deployments, there is a need to understand, characterize and quantify the underlying execution models being used on such systems. Execution models are a conceptual layer between applications & algorithms and the underlying parallel hardware and systems software on which those applications run. This paper presents the SCaLeM (Synchronization, Concurrency, Locality, Memory) framework for characterizing and execution models. SCaLeM consists of three basic elements: attributes, compositions and mapping of these compositions to abstract parallel systems. The fundamental Synchronization, Concurrency, Locality and Memory attributesmore » are used to characterize each execution model, while the combinations of those attributes in the form of compositions are used to describe the primitive operations of the execution model. The mapping of the execution model’s primitive operations described by compositions, to an underlying abstract parallel system can be evaluated quantitatively to determine its effectiveness. Finally, SCaLeM also enables the representation and analysis of applications in terms of execution models, for the purpose of evaluating the effectiveness of such mapping.« less

Design and Implementation of a Parallel Multivariate Ensemble Kalman Filter for the Poseidon Ocean General Circulation Model

NASA Technical Reports Server (NTRS)

Keppenne, Christian L.; Rienecker, Michele M.; Koblinsky, Chester (Technical Monitor)

2001-01-01

A multivariate ensemble Kalman filter (MvEnKF) implemented on a massively parallel computer architecture has been implemented for the Poseidon ocean circulation model and tested with a Pacific Basin model configuration. There are about two million prognostic state-vector variables. Parallelism for the data assimilation step is achieved by regionalization of the background-error covariances that are calculated from the phase-space distribution of the ensemble. Each processing element (PE) collects elements of a matrix measurement functional from nearby PEs. To avoid the introduction of spurious long-range covariances associated with finite ensemble sizes, the background-error covariances are given compact support by means of a Hadamard (element by element) product with a three-dimensional canonical correlation function. The methodology and the MvEnKF configuration are discussed. It is shown that the regionalization of the background covariances; has a negligible impact on the quality of the analyses. The parallel algorithm is very efficient for large numbers of observations but does not scale well beyond 100 PEs at the current model resolution. On a platform with distributed memory, memory rather than speed is the limiting factor.

Parallel design of JPEG-LS encoder on graphics processing units

NASA Astrophysics Data System (ADS)

Duan, Hao; Fang, Yong; Huang, Bormin

2012-01-01

With recent technical advances in graphic processing units (GPUs), GPUs have outperformed CPUs in terms of compute capability and memory bandwidth. Many successful GPU applications to high performance computing have been reported. JPEG-LS is an ISO/IEC standard for lossless image compression which utilizes adaptive context modeling and run-length coding to improve compression ratio. However, adaptive context modeling causes data dependency among adjacent pixels and the run-length coding has to be performed in a sequential way. Hence, using JPEG-LS to compress large-volume hyperspectral image data is quite time-consuming. We implement an efficient parallel JPEG-LS encoder for lossless hyperspectral compression on a NVIDIA GPU using the computer unified device architecture (CUDA) programming technology. We use the block parallel strategy, as well as such CUDA techniques as coalesced global memory access, parallel prefix sum, and asynchronous data transfer. We also show the relation between GPU speedup and AVIRIS block size, as well as the relation between compression ratio and AVIRIS block size. When AVIRIS images are divided into blocks, each with 64×64 pixels, we gain the best GPU performance with 26.3x speedup over its original CPU code.

System, methods and apparatus for program optimization for multi-threaded processor architectures

DOEpatents

Bastoul, Cedric; Lethin, Richard A; Leung, Allen K; Meister, Benoit J; Szilagyi, Peter; Vasilache, Nicolas T; Wohlford, David E

2015-01-06

Methods, apparatus and computer software product for source code optimization are provided. In an exemplary embodiment, a first custom computing apparatus is used to optimize the execution of source code on a second computing apparatus. In this embodiment, the first custom computing apparatus contains a memory, a storage medium and at least one processor with at least one multi-stage execution unit. The second computing apparatus contains at least two multi-stage execution units that allow for parallel execution of tasks. The first custom computing apparatus optimizes the code for parallelism, locality of operations and contiguity of memory accesses on the second computing apparatus. This Abstract is provided for the sole purpose of complying with the Abstract requirement rules. This Abstract is submitted with the explicit understanding that it will not be used to interpret or to limit the scope or the meaning of the claims.

Parallelization of MRCI based on hole-particle symmetry.

PubMed

Suo, Bing; Zhai, Gaohong; Wang, Yubin; Wen, Zhenyi; Hu, Xiangqian; Li, Lemin

2005-01-15

The parallel implementation of multireference configuration interaction program based on the hole-particle symmetry is described. The platform to implement the parallelization is an Intel-Architectural cluster consisting of 12 nodes, each of which is equipped with two 2.4-G XEON processors, 3-GB memory, and 36-GB disk, and are connected by a Gigabit Ethernet Switch. The dependence of speedup on molecular symmetries and task granularities is discussed. Test calculations show that the scaling with the number of nodes is about 1.9 (for C1 and Cs), 1.65 (for C2v), and 1.55 (for D2h) when the number of nodes is doubled. The largest calculation performed on this cluster involves 5.6 x 10(8) CSFs.

Multi-mode sensor processing on a dynamically reconfigurable massively parallel processor array

NASA Astrophysics Data System (ADS)

Chen, Paul; Butts, Mike; Budlong, Brad; Wasson, Paul

2008-04-01

This paper introduces a novel computing architecture that can be reconfigured in real time to adapt on demand to multi-mode sensor platforms' dynamic computational and functional requirements. This 1 teraOPS reconfigurable Massively Parallel Processor Array (MPPA) has 336 32-bit processors. The programmable 32-bit communication fabric provides streamlined inter-processor connections with deterministically high performance. Software programmability, scalability, ease of use, and fast reconfiguration time (ranging from microseconds to milliseconds) are the most significant advantages over FPGAs and DSPs. This paper introduces the MPPA architecture, its programming model, and methods of reconfigurability. An MPPA platform for reconfigurable computing is based on a structural object programming model. Objects are software programs running concurrently on hundreds of 32-bit RISC processors and memories. They exchange data and control through a network of self-synchronizing channels. A common application design pattern on this platform, called a work farm, is a parallel set of worker objects, with one input and one output stream. Statically configured work farms with homogeneous and heterogeneous sets of workers have been used in video compression and decompression, network processing, and graphics applications.

A Simple GPU-Accelerated Two-Dimensional MUSCL-Hancock Solver for Ideal Magnetohydrodynamics

NASA Technical Reports Server (NTRS)

Bard, Christopher; Dorelli, John C.

2013-01-01

We describe our experience using NVIDIA's CUDA (Compute Unified Device Architecture) C programming environment to implement a two-dimensional second-order MUSCL-Hancock ideal magnetohydrodynamics (MHD) solver on a GTX 480 Graphics Processing Unit (GPU). Taking a simple approach in which the MHD variables are stored exclusively in the global memory of the GTX 480 and accessed in a cache-friendly manner (without further optimizing memory access by, for example, staging data in the GPU's faster shared memory), we achieved a maximum speed-up of approx. = 126 for a sq 1024 grid relative to the sequential C code running on a single Intel Nehalem (2.8 GHz) core. This speedup is consistent with simple estimates based on the known floating point performance, memory throughput and parallel processing capacity of the GTX 480.

Algorithms for Data Intensive Applications on Intelligent and Smart Memories

DTIC Science & Technology

2003-03-01

editors). Parallel Algorithms and Architectures. North Holland, 1986. [8] P. Diniz . USC ISI, Personal Communication, March, 2001. [9] M. Frigo, C. E ...hierarchy as well as the Translation Lookaside Buer TLB aect the e ectiveness of cache friendly optimizations These penalties vary among...processors and cause large variations in the e ectiveness of cache performance optimizations The area of graph problems is fundamental in a wide variety of

Achieving supercomputer performance for neural net simulation with an array of digital signal processors

DOE Office of Scientific and Technical Information (OSTI.GOV)

Muller, U.A.; Baumle, B.; Kohler, P.

1992-10-01

Music, a DSP-based system with a parallel distributed-memory architecture, provides enormous computing power yet retains the flexibility of a general-purpose computer. Reaching a peak performance of 2.7 Gflops at a significantly lower cost, power consumption, and space requirement than conventional supercomputers, Music is well suited to computationally intensive applications such as neural network simulation. 12 refs., 9 figs., 2 tabs.

Right-Brain/Left-Brain Integrated Associative Processor Employing Convertible Multiple-Instruction-Stream Multiple-Data-Stream Elements

NASA Astrophysics Data System (ADS)

Hayakawa, Hitoshi; Ogawa, Makoto; Shibata, Tadashi

2005-04-01

A very large scale integrated circuit (VLSI) architecture for a multiple-instruction-stream multiple-data-stream (MIMD) associative processor has been proposed. The processor employs an architecture that enables seamless switching from associative operations to arithmetic operations. The MIMD element is convertible to a regular central processing unit (CPU) while maintaining its high performance as an associative processor. Therefore, the MIMD associative processor can perform not only on-chip perception, i.e., searching for the vector most similar to an input vector throughout the on-chip cache memory, but also arithmetic and logic operations similar to those in ordinary CPUs, both simultaneously in parallel processing. Three key technologies have been developed to generate the MIMD element: associative-operation-and-arithmetic-operation switchable calculation units, a versatile register control scheme within the MIMD element for flexible operations, and a short instruction set for minimizing the memory size for program storage. Key circuit blocks were designed and fabricated using 0.18 μm complementary metal-oxide-semiconductor (CMOS) technology. As a result, the full-featured MIMD element is estimated to be 3 mm2, showing the feasibility of an 8-parallel-MIMD-element associative processor in a single chip of 5 mm× 5 mm.

Parallel processing architecture for H.264 deblocking filter on multi-core platforms

NASA Astrophysics Data System (ADS)

Prasad, Durga P.; Sonachalam, Sekar; Kunchamwar, Mangesh K.; Gunupudi, Nageswara Rao

2012-03-01

Massively parallel computing (multi-core) chips offer outstanding new solutions that satisfy the increasing demand for high resolution and high quality video compression technologies such as H.264. Such solutions not only provide exceptional quality but also efficiency, low power, and low latency, previously unattainable in software based designs. While custom hardware and Application Specific Integrated Circuit (ASIC) technologies may achieve lowlatency, low power, and real-time performance in some consumer devices, many applications require a flexible and scalable software-defined solution. The deblocking filter in H.264 encoder/decoder poses difficult implementation challenges because of heavy data dependencies and the conditional nature of the computations. Deblocking filter implementations tend to be fixed and difficult to reconfigure for different needs. The ability to scale up for higher quality requirements such as 10-bit pixel depth or a 4:2:2 chroma format often reduces the throughput of a parallel architecture designed for lower feature set. A scalable architecture for deblocking filtering, created with a massively parallel processor based solution, means that the same encoder or decoder will be deployed in a variety of applications, at different video resolutions, for different power requirements, and at higher bit-depths and better color sub sampling patterns like YUV, 4:2:2, or 4:4:4 formats. Low power, software-defined encoders/decoders may be implemented using a massively parallel processor array, like that found in HyperX technology, with 100 or more cores and distributed memory. The large number of processor elements allows the silicon device to operate more efficiently than conventional DSP or CPU technology. This software programing model for massively parallel processors offers a flexible implementation and a power efficiency close to that of ASIC solutions. This work describes a scalable parallel architecture for an H.264 compliant deblocking filter for multi core platforms such as HyperX technology. Parallel techniques such as parallel processing of independent macroblocks, sub blocks, and pixel row level are examined in this work. The deblocking architecture consists of a basic cell called deblocking filter unit (DFU) and dependent data buffer manager (DFM). The DFU can be used in several instances, catering to different performance needs the DFM serves the data required for the different number of DFUs, and also manages all the neighboring data required for future data processing of DFUs. This approach achieves the scalability, flexibility, and performance excellence required in deblocking filters.

Multi-level Hierarchical Poly Tree computer architectures

NASA Technical Reports Server (NTRS)

Padovan, Joe; Gute, Doug

1990-01-01

Based on the concept of hierarchical substructuring, this paper develops an optimal multi-level Hierarchical Poly Tree (HPT) parallel computer architecture scheme which is applicable to the solution of finite element and difference simulations. Emphasis is given to minimizing computational effort, in-core/out-of-core memory requirements, and the data transfer between processors. In addition, a simplified communications network that reduces the number of I/O channels between processors is presented. HPT configurations that yield optimal superlinearities are also demonstrated. Moreover, to generalize the scope of applicability, special attention is given to developing: (1) multi-level reduction trees which provide an orderly/optimal procedure by which model densification/simplification can be achieved, as well as (2) methodologies enabling processor grading that yields architectures with varying types of multi-level granularity.

FFT Computation with Systolic Arrays, A New Architecture

NASA Technical Reports Server (NTRS)

Boriakoff, Valentin

1994-01-01

The use of the Cooley-Tukey algorithm for computing the l-d FFT lends itself to a particular matrix factorization which suggests direct implementation by linearly-connected systolic arrays. Here we present a new systolic architecture that embodies this algorithm. This implementation requires a smaller number of processors and a smaller number of memory cells than other recent implementations, as well as having all the advantages of systolic arrays. For the implementation of the decimation-in-frequency case, word-serial data input allows continuous real-time operation without the need of a serial-to-parallel conversion device. No control or data stream switching is necessary. Computer simulation of this architecture was done in the context of a 1024 point DFT with a fixed point processor, and CMOS processor implementation has started.

All-memristive neuromorphic computing with level-tuned neurons

NASA Astrophysics Data System (ADS)

Pantazi, Angeliki; Woźniak, Stanisław; Tuma, Tomas; Eleftheriou, Evangelos

2016-09-01

In the new era of cognitive computing, systems will be able to learn and interact with the environment in ways that will drastically enhance the capabilities of current processors, especially in extracting knowledge from vast amount of data obtained from many sources. Brain-inspired neuromorphic computing systems increasingly attract research interest as an alternative to the classical von Neumann processor architecture, mainly because of the coexistence of memory and processing units. In these systems, the basic components are neurons interconnected by synapses. The neurons, based on their nonlinear dynamics, generate spikes that provide the main communication mechanism. The computational tasks are distributed across the neural network, where synapses implement both the memory and the computational units, by means of learning mechanisms such as spike-timing-dependent plasticity. In this work, we present an all-memristive neuromorphic architecture comprising neurons and synapses realized by using the physical properties and state dynamics of phase-change memristors. The architecture employs a novel concept of interconnecting the neurons in the same layer, resulting in level-tuned neuronal characteristics that preferentially process input information. We demonstrate the proposed architecture in the tasks of unsupervised learning and detection of multiple temporal correlations in parallel input streams. The efficiency of the neuromorphic architecture along with the homogenous neuro-synaptic dynamics implemented with nanoscale phase-change memristors represent a significant step towards the development of ultrahigh-density neuromorphic co-processors.

All-memristive neuromorphic computing with level-tuned neurons.

PubMed

Pantazi, Angeliki; Woźniak, Stanisław; Tuma, Tomas; Eleftheriou, Evangelos

2016-09-02

In the new era of cognitive computing, systems will be able to learn and interact with the environment in ways that will drastically enhance the capabilities of current processors, especially in extracting knowledge from vast amount of data obtained from many sources. Brain-inspired neuromorphic computing systems increasingly attract research interest as an alternative to the classical von Neumann processor architecture, mainly because of the coexistence of memory and processing units. In these systems, the basic components are neurons interconnected by synapses. The neurons, based on their nonlinear dynamics, generate spikes that provide the main communication mechanism. The computational tasks are distributed across the neural network, where synapses implement both the memory and the computational units, by means of learning mechanisms such as spike-timing-dependent plasticity. In this work, we present an all-memristive neuromorphic architecture comprising neurons and synapses realized by using the physical properties and state dynamics of phase-change memristors. The architecture employs a novel concept of interconnecting the neurons in the same layer, resulting in level-tuned neuronal characteristics that preferentially process input information. We demonstrate the proposed architecture in the tasks of unsupervised learning and detection of multiple temporal correlations in parallel input streams. The efficiency of the neuromorphic architecture along with the homogenous neuro-synaptic dynamics implemented with nanoscale phase-change memristors represent a significant step towards the development of ultrahigh-density neuromorphic co-processors.

Livermore Big Artificial Neural Network Toolkit

DOE Office of Scientific and Technical Information (OSTI.GOV)

Essen, Brian Van; Jacobs, Sam; Kim, Hyojin

2016-07-01

LBANN is a toolkit that is designed to train artificial neural networks efficiently on high performance computing architectures. It is optimized to take advantages of key High Performance Computing features to accelerate neural network training. Specifically it is optimized for low-latency, high bandwidth interconnects, node-local NVRAM, node-local GPU accelerators, and high bandwidth parallel file systems. It is built on top of the open source Elemental distributed-memory dense and spars-direct linear algebra and optimization library that is released under the BSD license. The algorithms contained within LBANN are drawn from the academic literature and implemented to work within a distributed-memory framework.

Solutions of large-scale electromagnetics problems involving dielectric objects with the parallel multilevel fast multipole algorithm.

PubMed

Ergül, Özgür

2011-11-01

Fast and accurate solutions of large-scale electromagnetics problems involving homogeneous dielectric objects are considered. Problems are formulated with the electric and magnetic current combined-field integral equation and discretized with the Rao-Wilton-Glisson functions. Solutions are performed iteratively by using the multilevel fast multipole algorithm (MLFMA). For the solution of large-scale problems discretized with millions of unknowns, MLFMA is parallelized on distributed-memory architectures using a rigorous technique, namely, the hierarchical partitioning strategy. Efficiency and accuracy of the developed implementation are demonstrated on very large problems involving as many as 100 million unknowns.

A Parallel Cartesian Approach for External Aerodynamics of Vehicles with Complex Geometry

NASA Technical Reports Server (NTRS)

Aftosmis, M. J.; Berger, M. J.; Adomavicius, G.

2001-01-01

This workshop paper presents the current status in the development of a new approach for the solution of the Euler equations on Cartesian meshes with embedded boundaries in three dimensions on distributed and shared memory architectures. The approach uses adaptively refined Cartesian hexahedra to fill the computational domain. Where these cells intersect the geometry, they are cut by the boundary into arbitrarily shaped polyhedra which receive special treatment by the solver. The presentation documents a newly developed multilevel upwind solver based on a flexible domain-decomposition strategy. One novel aspect of the work is its use of space-filling curves (SFC) for memory efficient on-the-fly parallelization, dynamic re-partitioning and automatic coarse mesh generation. Within each subdomain the approach employs a variety reordering techniques so that relevant data are on the same page in memory permitting high-performance on cache-based processors. Details of the on-the-fly SFC based partitioning are presented as are construction rules for the automatic coarse mesh generation. After describing the approach, the paper uses model problems and 3- D configurations to both verify and validate the solver. The model problems demonstrate that second-order accuracy is maintained despite the presence of the irregular cut-cells in the mesh. In addition, it examines both parallel efficiency and convergence behavior. These investigations demonstrate a parallel speed-up in excess of 28 on 32 processors of an SGI Origin 2000 system and confirm that mesh partitioning has no effect on convergence behavior.

Parallelization of the FLAPW method and comparison with the PPW method

NASA Astrophysics Data System (ADS)

Canning, Andrew; Mannstadt, Wolfgang; Freeman, Arthur

2000-03-01

The FLAPW (full-potential linearized-augmented plane-wave) method is one of the most accurate first-principles methods for determining electronic and magnetic properties of crystals and surfaces. In the past the FLAPW method has been limited to systems of about a hundred atoms due to the lack of an efficient parallel implementation to exploit the power and memory of parallel computers. In this work we present an efficient parallelization of the method by division among the processors of the plane-wave components for each state. The code is also optimized for RISC (reduced instruction set computer) architectures, such as those found on most parallel computers, making full use of BLAS (basic linear algebra subprograms) wherever possible. Scaling results are presented for systems of up to 686 silicon atoms and 343 palladium atoms per unit cell running on up to 512 processors on a Cray T3E parallel supercomputer. Some results will also be presented on a comparison of the plane-wave pseudopotential method and the FLAPW method on large systems.

The Metropolis Monte Carlo method with CUDA enabled Graphic Processing Units

DOE Office of Scientific and Technical Information (OSTI.GOV)

Hall, Clifford; School of Physics, Astronomy, and Computational Sciences, George Mason University, 4400 University Dr., Fairfax, VA 22030; Ji, Weixiao

2014-02-01

We present a CPU–GPU system for runtime acceleration of large molecular simulations using GPU computation and memory swaps. The memory architecture of the GPU can be used both as container for simulation data stored on the graphics card and as floating-point code target, providing an effective means for the manipulation of atomistic or molecular data on the GPU. To fully take advantage of this mechanism, efficient GPU realizations of algorithms used to perform atomistic and molecular simulations are essential. Our system implements a versatile molecular engine, including inter-molecule interactions and orientational variables for performing the Metropolis Monte Carlo (MMC) algorithm,more » which is one type of Markov chain Monte Carlo. By combining memory objects with floating-point code fragments we have implemented an MMC parallel engine that entirely avoids the communication time of molecular data at runtime. Our runtime acceleration system is a forerunner of a new class of CPU–GPU algorithms exploiting memory concepts combined with threading for avoiding bus bandwidth and communication. The testbed molecular system used here is a condensed phase system of oligopyrrole chains. A benchmark shows a size scaling speedup of 60 for systems with 210,000 pyrrole monomers. Our implementation can easily be combined with MPI to connect in parallel several CPU–GPU duets. -- Highlights: •We parallelize the Metropolis Monte Carlo (MMC) algorithm on one CPU—GPU duet. •The Adaptive Tempering Monte Carlo employs MMC and profits from this CPU—GPU implementation. •Our benchmark shows a size scaling-up speedup of 62 for systems with 225,000 particles. •The testbed involves a polymeric system of oligopyrroles in the condensed phase. •The CPU—GPU parallelization includes dipole—dipole and Mie—Jones classic potentials.« less

Current implementation and future plans on new code architecture, programming language and user interface

DOE Office of Scientific and Technical Information (OSTI.GOV)

Brun, B.

1997-07-01

Computer technology has improved tremendously during the last years with larger media capacity, memory and more computational power. Visual computing with high-performance graphic interface and desktop computational power have changed the way engineers accomplish everyday tasks, development and safety studies analysis. The emergence of parallel computing will permit simulation over a larger domain. In addition, new development methods, languages and tools have appeared in the last several years.

Fractional Steps methods for transient problems on commodity computer architectures

NASA Astrophysics Data System (ADS)

Krotkiewski, M.; Dabrowski, M.; Podladchikov, Y. Y.

2008-12-01

Fractional Steps methods are suitable for modeling transient processes that are central to many geological applications. Low memory requirements and modest computational complexity facilitates calculations on high-resolution three-dimensional models. An efficient implementation of Alternating Direction Implicit/Locally One-Dimensional schemes for an Opteron-based shared memory system is presented. The memory bandwidth usage, the main bottleneck on modern computer architectures, is specially addressed. High efficiency of above 2 GFlops per CPU is sustained for problems of 1 billion degrees of freedom. The optimized sequential implementation of all 1D sweeps is comparable in execution time to copying the used data in the memory. Scalability of the parallel implementation on up to 8 CPUs is close to perfect. Performing one timestep of the Locally One-Dimensional scheme on a system of 1000 3 unknowns on 8 CPUs takes only 11 s. We validate the LOD scheme using a computational model of an isolated inclusion subject to a constant far field flux. Next, we study numerically the evolution of a diffusion front and the effective thermal conductivity of composites consisting of multiple inclusions and compare the results with predictions based on the differential effective medium approach. Finally, application of the developed parabolic solver is suggested for a real-world problem of fluid transport and reactions inside a reservoir.

Thread-Level Parallelization and Optimization of NWChem for the Intel MIC Architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Shan, Hongzhang; Williams, Samuel; Jong, Wibe de

In the multicore era it was possible to exploit the increase in on-chip parallelism by simply running multiple MPI processes per chip. Unfortunately, manycore processors' greatly increased thread- and data-level parallelism coupled with a reduced memory capacity demand an altogether different approach. In this paper we explore augmenting two NWChem modules, triples correction of the CCSD(T) and Fock matrix construction, with OpenMP in order that they might run efficiently on future manycore architectures. As the next NERSC machine will be a self-hosted Intel MIC (Xeon Phi) based supercomputer, we leverage an existing MIC testbed at NERSC to evaluate our experiments.more » In order to proxy the fact that future MIC machines will not have a host processor, we run all of our experiments in tt native mode. We found that while straightforward application of OpenMP to the deep loop nests associated with the tensor contractions of CCSD(T) was sufficient in attaining high performance, significant effort was required to safely and efficiently thread the TEXAS integral package when constructing the Fock matrix. Ultimately, our new MPI OpenMP hybrid implementations attain up to 65x better performance for the triples part of the CCSD(T) due in large part to the fact that the limited on-card memory limits the existing MPI implementation to a single process per card. Additionally, we obtain up to 1.6x better performance on Fock matrix constructions when compared with the best MPI implementations running multiple processes per card.« less

Thread-level parallelization and optimization of NWChem for the Intel MIC architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Shan, Hongzhang; Williams, Samuel; de Jong, Wibe

In the multicore era it was possible to exploit the increase in on-chip parallelism by simply running multiple MPI processes per chip. Unfortunately, manycore processors' greatly increased thread- and data-level parallelism coupled with a reduced memory capacity demand an altogether different approach. In this paper we explore augmenting two NWChem modules, triples correction of the CCSD(T) and Fock matrix construction, with OpenMP in order that they might run efficiently on future manycore architectures. As the next NERSC machine will be a self-hosted Intel MIC (Xeon Phi) based supercomputer, we leverage an existing MIC testbed at NERSC to evaluate our experiments.more » In order to proxy the fact that future MIC machines will not have a host processor, we run all of our experiments in native mode. We found that while straightforward application of OpenMP to the deep loop nests associated with the tensor contractions of CCSD(T) was sufficient in attaining high performance, significant e ort was required to safely and efeciently thread the TEXAS integral package when constructing the Fock matrix. Ultimately, our new MPI+OpenMP hybrid implementations attain up to 65× better performance for the triples part of the CCSD(T) due in large part to the fact that the limited on-card memory limits the existing MPI implementation to a single process per card. Additionally, we obtain up to 1.6× better performance on Fock matrix constructions when compared with the best MPI implementations running multiple processes per card.« less

Parallel Multivariate Spatio-Temporal Clustering of Large Ecological Datasets on Hybrid Supercomputers

DOE Office of Scientific and Technical Information (OSTI.GOV)

Sreepathi, Sarat; Kumar, Jitendra; Mills, Richard T.

A proliferation of data from vast networks of remote sensing platforms (satellites, unmanned aircraft systems (UAS), airborne etc.), observational facilities (meteorological, eddy covariance etc.), state-of-the-art sensors, and simulation models offer unprecedented opportunities for scientific discovery. Unsupervised classification is a widely applied data mining approach to derive insights from such data. However, classification of very large data sets is a complex computational problem that requires efficient numerical algorithms and implementations on high performance computing (HPC) platforms. Additionally, increasing power, space, cooling and efficiency requirements has led to the deployment of hybrid supercomputing platforms with complex architectures and memory hierarchies like themore » Titan system at Oak Ridge National Laboratory. The advent of such accelerated computing architectures offers new challenges and opportunities for big data analytics in general and specifically, large scale cluster analysis in our case. Although there is an existing body of work on parallel cluster analysis, those approaches do not fully meet the needs imposed by the nature and size of our large data sets. Moreover, they had scaling limitations and were mostly limited to traditional distributed memory computing platforms. We present a parallel Multivariate Spatio-Temporal Clustering (MSTC) technique based on k-means cluster analysis that can target hybrid supercomputers like Titan. We developed a hybrid MPI, CUDA and OpenACC implementation that can utilize both CPU and GPU resources on computational nodes. We describe performance results on Titan that demonstrate the scalability and efficacy of our approach in processing large ecological data sets.« less

Massively parallel quantum computer simulator

NASA Astrophysics Data System (ADS)

De Raedt, K.; Michielsen, K.; De Raedt, H.; Trieu, B.; Arnold, G.; Richter, M.; Lippert, Th.; Watanabe, H.; Ito, N.

2007-01-01

We describe portable software to simulate universal quantum computers on massive parallel computers. We illustrate the use of the simulation software by running various quantum algorithms on different computer architectures, such as a IBM BlueGene/L, a IBM Regatta p690+, a Hitachi SR11000/J1, a Cray X1E, a SGI Altix 3700 and clusters of PCs running Windows XP. We study the performance of the software by simulating quantum computers containing up to 36 qubits, using up to 4096 processors and up to 1 TB of memory. Our results demonstrate that the simulator exhibits nearly ideal scaling as a function of the number of processors and suggest that the simulation software described in this paper may also serve as benchmark for testing high-end parallel computers.

Benchmarking Memory Performance with the Data Cube Operator

NASA Technical Reports Server (NTRS)

Frumkin, Michael A.; Shabanov, Leonid V.

2004-01-01

Data movement across a computer memory hierarchy and across computational grids is known to be a limiting factor for applications processing large data sets. We use the Data Cube Operator on an Arithmetic Data Set, called ADC, to benchmark capabilities of computers and of computational grids to handle large distributed data sets. We present a prototype implementation of a parallel algorithm for computation of the operatol: The algorithm follows a known approach for computing views from the smallest parent. The ADC stresses all levels of grid memory and storage by producing some of 2d views of an Arithmetic Data Set of d-tuples described by a small number of integers. We control data intensity of the ADC by selecting the tuple parameters, the sizes of the views, and the number of realized views. Benchmarking results of memory performance of a number of computer architectures and of a small computational grid are presented.

Parallel volume ray-casting for unstructured-grid data on distributed-memory architectures

NASA Technical Reports Server (NTRS)

Ma, Kwan-Liu

1995-01-01

As computing technology continues to advance, computational modeling of scientific and engineering problems produces data of increasing complexity: large in size and unstructured in shape. Volume visualization of such data is a challenging problem. This paper proposes a distributed parallel solution that makes ray-casting volume rendering of unstructured-grid data practical. Both the data and the rendering process are distributed among processors. At each processor, ray-casting of local data is performed independent of the other processors. The global image composing processes, which require inter-processor communication, are overlapped with the local ray-casting processes to achieve maximum parallel efficiency. This algorithm differs from previous ones in four ways: it is completely distributed, less view-dependent, reasonably scalable, and flexible. Without using dynamic load balancing, test results on the Intel Paragon using from two to 128 processors show, on average, about 60% parallel efficiency.

Discrete sensitivity derivatives of the Navier-Stokes equations with a parallel Krylov solver

NASA Technical Reports Server (NTRS)

Ajmani, Kumud; Taylor, Arthur C., III

1994-01-01

This paper solves an 'incremental' form of the sensitivity equations derived by differentiating the discretized thin-layer Navier Stokes equations with respect to certain design variables of interest. The equations are solved with a parallel, preconditioned Generalized Minimal RESidual (GMRES) solver on a distributed-memory architecture. The 'serial' sensitivity analysis code is parallelized by using the Single Program Multiple Data (SPMD) programming model, domain decomposition techniques, and message-passing tools. Sensitivity derivatives are computed for low and high Reynolds number flows over a NACA 1406 airfoil on a 32-processor Intel Hypercube, and found to be identical to those computed on a single-processor Cray Y-MP. It is estimated that the parallel sensitivity analysis code has to be run on 40-50 processors of the Intel Hypercube in order to match the single-processor processing time of a Cray Y-MP.

Energy-efficient STDP-based learning circuits with memristor synapses

NASA Astrophysics Data System (ADS)

Wu, Xinyu; Saxena, Vishal; Campbell, Kristy A.

2014-05-01

It is now accepted that the traditional von Neumann architecture, with processor and memory separation, is ill suited to process parallel data streams which a mammalian brain can efficiently handle. Moreover, researchers now envision computing architectures which enable cognitive processing of massive amounts of data by identifying spatio-temporal relationships in real-time and solving complex pattern recognition problems. Memristor cross-point arrays, integrated with standard CMOS technology, are expected to result in massively parallel and low-power Neuromorphic computing architectures. Recently, significant progress has been made in spiking neural networks (SNN) which emulate data processing in the cortical brain. These architectures comprise of a dense network of neurons and the synapses formed between the axons and dendrites. Further, unsupervised or supervised competitive learning schemes are being investigated for global training of the network. In contrast to a software implementation, hardware realization of these networks requires massive circuit overhead for addressing and individually updating network weights. Instead, we employ bio-inspired learning rules such as the spike-timing-dependent plasticity (STDP) to efficiently update the network weights locally. To realize SNNs on a chip, we propose to use densely integrating mixed-signal integrate-andfire neurons (IFNs) and cross-point arrays of memristors in back-end-of-the-line (BEOL) of CMOS chips. Novel IFN circuits have been designed to drive memristive synapses in parallel while maintaining overall power efficiency (<1 pJ/spike/synapse), even at spike rate greater than 10 MHz. We present circuit design details and simulation results of the IFN with memristor synapses, its response to incoming spike trains and STDP learning characterization.

Use Computer-Aided Tools to Parallelize Large CFD Applications

NASA Technical Reports Server (NTRS)

Jin, H.; Frumkin, M.; Yan, J.

2000-01-01

Porting applications to high performance parallel computers is always a challenging task. It is time consuming and costly. With rapid progressing in hardware architectures and increasing complexity of real applications in recent years, the problem becomes even more sever. Today, scalability and high performance are mostly involving handwritten parallel programs using message-passing libraries (e.g. MPI). However, this process is very difficult and often error-prone. The recent reemergence of shared memory parallel (SMP) architectures, such as the cache coherent Non-Uniform Memory Access (ccNUMA) architecture used in the SGI Origin 2000, show good prospects for scaling beyond hundreds of processors. Programming on an SMP is simplified by working in a globally accessible address space. The user can supply compiler directives, such as OpenMP, to parallelize the code. As an industry standard for portable implementation of parallel programs for SMPs, OpenMP is a set of compiler directives and callable runtime library routines that extend Fortran, C and C++ to express shared memory parallelism. It promises an incremental path for parallel conversion of existing software, as well as scalability and performance for a complete rewrite or an entirely new development. Perhaps the main disadvantage of programming with directives is that inserted directives may not necessarily enhance performance. In the worst cases, it can create erroneous results. While vendors have provided tools to perform error-checking and profiling, automation in directive insertion is very limited and often failed on large programs, primarily due to the lack of a thorough enough data dependence analysis. To overcome the deficiency, we have developed a toolkit, CAPO, to automatically insert OpenMP directives in Fortran programs and apply certain degrees of optimization. CAPO is aimed at taking advantage of detailed inter-procedural dependence analysis provided by CAPTools, developed by the University of Greenwich, to reduce potential errors made by users. Earlier tests on NAS Benchmarks and ARC3D have demonstrated good success of this tool. In this study, we have applied CAPO to parallelize three large applications in the area of computational fluid dynamics (CFD): OVERFLOW, TLNS3D and INS3D. These codes are widely used for solving Navier-Stokes equations with complicated boundary conditions and turbulence model in multiple zones. Each one comprises of from 50K to 1,00k lines of FORTRAN77. As an example, CAPO took 77 hours to complete the data dependence analysis of OVERFLOW on a workstation (SGI, 175MHz, R10K processor). A fair amount of effort was spent on correcting false dependencies due to lack of necessary knowledge during the analysis. Even so, CAPO provides an easy way for user to interact with the parallelization process. The OpenMP version was generated within a day after the analysis was completed. Due to sequential algorithms involved, code sections in TLNS3D and INS3D need to be restructured by hand to produce more efficient parallel codes. An included figure shows preliminary test results of the generated OVERFLOW with several test cases in single zone. The MPI data points for the small test case were taken from a handcoded MPI version. As we can see, CAPO's version has achieved 18 fold speed up on 32 nodes of the SGI O2K. For the small test case, it outperformed the MPI version. These results are very encouraging, but further work is needed. For example, although CAPO attempts to place directives on the outer- most parallel loops in an interprocedural framework, it does not insert directives based on the best manual strategy. In particular, it lacks the support of parallelization at the multi-zone level. Future work will emphasize on the development of methodology to work in a multi-zone level and with a hybrid approach. Development of tools to perform more complicated code transformation is also needed.

GPU-computing in econophysics and statistical physics

NASA Astrophysics Data System (ADS)

Preis, T.

2011-03-01

A recent trend in computer science and related fields is general purpose computing on graphics processing units (GPUs), which can yield impressive performance. With multiple cores connected by high memory bandwidth, today's GPUs offer resources for non-graphics parallel processing. This article provides a brief introduction into the field of GPU computing and includes examples. In particular computationally expensive analyses employed in financial market context are coded on a graphics card architecture which leads to a significant reduction of computing time. In order to demonstrate the wide range of possible applications, a standard model in statistical physics - the Ising model - is ported to a graphics card architecture as well, resulting in large speedup values.

Mapping implicit spectral methods to distributed memory architectures

NASA Technical Reports Server (NTRS)

Overman, Andrea L.; Vanrosendale, John

1991-01-01

Spectral methods were proven invaluable in numerical simulation of PDEs (Partial Differential Equations), but the frequent global communication required raises a fundamental barrier to their use on highly parallel architectures. To explore this issue, a 3-D implicit spectral method was implemented on an Intel hypercube. Utilization of about 50 percent was achieved on a 32 node iPSC/860 hypercube, for a 64 x 64 x 64 Fourier-spectral grid; finer grids yield higher utilizations. Chebyshev-spectral grids are more problematic, since plane-relaxation based multigrid is required. However, by using a semicoarsening multigrid algorithm, and by relaxing all multigrid levels concurrently, relatively high utilizations were also achieved in this harder case.

Implementing An Image Understanding System Architecture Using Pipe

NASA Astrophysics Data System (ADS)

Luck, Randall L.

1988-03-01

This paper will describe PIPE and how it can be used to implement an image understanding system. Image understanding is the process of developing a description of an image in order to make decisions about its contents. The tasks of image understanding are generally split into low level vision and high level vision. Low level vision is performed by PIPE -a high performance parallel processor with an architecture specifically designed for processing video images at up to 60 fields per second. High level vision is performed by one of several types of serial or parallel computers - depending on the application. An additional processor called ISMAP performs the conversion from iconic image space to symbolic feature space. ISMAP plugs into one of PIPE's slots and is memory mapped into the high level processor. Thus it forms the high speed link between the low and high level vision processors. The mechanisms for bottom-up, data driven processing and top-down, model driven processing are discussed.

Optimization of sparse matrix-vector multiplication on emerging multicore platforms

DOE Office of Scientific and Technical Information (OSTI.GOV)

Williams, Samuel; Oliker, Leonid; Vuduc, Richard

2007-01-01

We are witnessing a dramatic change in computer architecture due to the multicore paradigm shift, as every electronic device from cell phones to supercomputers confronts parallelism of unprecedented scale. To fully unleash the potential of these systems, the HPC community must develop multicore specific optimization methodologies for important scientific computations. In this work, we examine sparse matrix-vector multiply (SpMV) - one of the most heavily used kernels in scientific computing - across a broad spectrum of multicore designs. Our experimental platform includes the homogeneous AMD dual-core and Intel quad-core designs, the heterogeneous STI Cell, as well as the first scientificmore » study of the highly multithreaded Sun Niagara2. We present several optimization strategies especially effective for the multicore environment, and demonstrate significant performance improvements compared to existing state-of-the-art serial and parallel SpMV implementations. Additionally, we present key insights into the architectural tradeoffs of leading multicore design strategies, in the context of demanding memory-bound numerical algorithms.« less

Memory for time distinguishes between perception and action.

PubMed

Bueti, Domenica; Walsh, Vincent

2010-01-01

Our experience of time is unlike that of other features of the sensory world such as colour, movement, touch, or sound because there is no unique receptor system through which it is received. However, since time can be perceived, remembered, estimated, and compared in a way analogous to other sensory experiences, it should perhaps be subject to some of the same architectures or principles that have advanced understanding in these other domains. By adapting a task designed to test visual memory within a perception/action framework we investigated whether memory for time is affected by the use to which temporal information is put. When remembering a visual or auditory duration for subsequent motor production, storage is biased by a delay of up to 8 s. When the same duration is remembered for subsequent perception, however, there is no such effect of delay on memory. The results suggest a distinction in temporal memory that parallels the perception/action dichotomy in vision.

Hypercube matrix computation task

NASA Technical Reports Server (NTRS)

Calalo, R.; Imbriale, W.; Liewer, P.; Lyons, J.; Manshadi, F.; Patterson, J.

1987-01-01

The Hypercube Matrix Computation (Year 1986-1987) task investigated the applicability of a parallel computing architecture to the solution of large scale electromagnetic scattering problems. Two existing electromagnetic scattering codes were selected for conversion to the Mark III Hypercube concurrent computing environment. They were selected so that the underlying numerical algorithms utilized would be different thereby providing a more thorough evaluation of the appropriateness of the parallel environment for these types of problems. The first code was a frequency domain method of moments solution, NEC-2, developed at Lawrence Livermore National Laboratory. The second code was a time domain finite difference solution of Maxwell's equations to solve for the scattered fields. Once the codes were implemented on the hypercube and verified to obtain correct solutions by comparing the results with those from sequential runs, several measures were used to evaluate the performance of the two codes. First, a comparison was provided of the problem size possible on the hypercube with 128 megabytes of memory for a 32-node configuration with that available in a typical sequential user environment of 4 to 8 megabytes. Then, the performance of the codes was anlyzed for the computational speedup attained by the parallel architecture.

Parallel hyperspectral compressive sensing method on GPU

NASA Astrophysics Data System (ADS)

Bernabé, Sergio; Martín, Gabriel; Nascimento, José M. P.

2015-10-01

Remote hyperspectral sensors collect large amounts of data per flight usually with low spatial resolution. It is known that the bandwidth connection between the satellite/airborne platform and the ground station is reduced, thus a compression onboard method is desirable to reduce the amount of data to be transmitted. This paper presents a parallel implementation of an compressive sensing method, called parallel hyperspectral coded aperture (P-HYCA), for graphics processing units (GPU) using the compute unified device architecture (CUDA). This method takes into account two main properties of hyperspectral dataset, namely the high correlation existing among the spectral bands and the generally low number of endmembers needed to explain the data, which largely reduces the number of measurements necessary to correctly reconstruct the original data. Experimental results conducted using synthetic and real hyperspectral datasets on two different GPU architectures by NVIDIA: GeForce GTX 590 and GeForce GTX TITAN, reveal that the use of GPUs can provide real-time compressive sensing performance. The achieved speedup is up to 20 times when compared with the processing time of HYCA running on one core of the Intel i7-2600 CPU (3.4GHz), with 16 Gbyte memory.

Massively parallel processor computer

NASA Technical Reports Server (NTRS)

Fung, L. W. (Inventor)

1983-01-01

An apparatus for processing multidimensional data with strong spatial characteristics, such as raw image data, characterized by a large number of parallel data streams in an ordered array is described. It comprises a large number (e.g., 16,384 in a 128 x 128 array) of parallel processing elements operating simultaneously and independently on single bit slices of a corresponding array of incoming data streams under control of a single set of instructions. Each of the processing elements comprises a bidirectional data bus in communication with a register for storing single bit slices together with a random access memory unit and associated circuitry, including a binary counter/shift register device, for performing logical and arithmetical computations on the bit slices, and an I/O unit for interfacing the bidirectional data bus with the data stream source. The massively parallel processor architecture enables very high speed processing of large amounts of ordered parallel data, including spatial translation by shifting or sliding of bits vertically or horizontally to neighboring processing elements.

Parallel solution of high-order numerical schemes for solving incompressible flows

NASA Technical Reports Server (NTRS)

Milner, Edward J.; Lin, Avi; Liou, May-Fun; Blech, Richard A.

1993-01-01

A new parallel numerical scheme for solving incompressible steady-state flows is presented. The algorithm uses a finite-difference approach to solving the Navier-Stokes equations. The algorithms are scalable and expandable. They may be used with only two processors or with as many processors as are available. The code is general and expandable. Any size grid may be used. Four processors of the NASA LeRC Hypercluster were used to solve for steady-state flow in a driven square cavity. The Hypercluster was configured in a distributed-memory, hypercube-like architecture. By using a 50-by-50 finite-difference solution grid, an efficiency of 74 percent (a speedup of 2.96) was obtained.

Real-time autocorrelator for fluorescence correlation spectroscopy based on graphical-processor-unit architecture: method, implementation, and comparative studies

NASA Astrophysics Data System (ADS)

Laracuente, Nicholas; Grossman, Carl

2013-03-01

We developed an algorithm and software to calculate autocorrelation functions from real-time photon-counting data using the fast, parallel capabilities of graphical processor units (GPUs). Recent developments in hardware and software have allowed for general purpose computing with inexpensive GPU hardware. These devices are more suited for emulating hardware autocorrelators than traditional CPU-based software applications by emphasizing parallel throughput over sequential speed. Incoming data are binned in a standard multi-tau scheme with configurable points-per-bin size and are mapped into a GPU memory pattern to reduce time-expensive memory access. Applications include dynamic light scattering (DLS) and fluorescence correlation spectroscopy (FCS) experiments. We ran the software on a 64-core graphics pci card in a 3.2 GHz Intel i5 CPU based computer running Linux. FCS measurements were made on Alexa-546 and Texas Red dyes in a standard buffer (PBS). Software correlations were compared to hardware correlator measurements on the same signals. Supported by HHMI and Swarthmore College

Testing New Programming Paradigms with NAS Parallel Benchmarks

NASA Technical Reports Server (NTRS)

Jin, H.; Frumkin, M.; Schultz, M.; Yan, J.

2000-01-01

Over the past decade, high performance computing has evolved rapidly, not only in hardware architectures but also with increasing complexity of real applications. Technologies have been developing to aim at scaling up to thousands of processors on both distributed and shared memory systems. Development of parallel programs on these computers is always a challenging task. Today, writing parallel programs with message passing (e.g. MPI) is the most popular way of achieving scalability and high performance. However, writing message passing programs is difficult and error prone. Recent years new effort has been made in defining new parallel programming paradigms. The best examples are: HPF (based on data parallelism) and OpenMP (based on shared memory parallelism). Both provide simple and clear extensions to sequential programs, thus greatly simplify the tedious tasks encountered in writing message passing programs. HPF is independent of memory hierarchy, however, due to the immaturity of compiler technology its performance is still questionable. Although use of parallel compiler directives is not new, OpenMP offers a portable solution in the shared-memory domain. Another important development involves the tremendous progress in the internet and its associated technology. Although still in its infancy, Java promisses portability in a heterogeneous environment and offers possibility to "compile once and run anywhere." In light of testing these new technologies, we implemented new parallel versions of the NAS Parallel Benchmarks (NPBs) with HPF and OpenMP directives, and extended the work with Java and Java-threads. The purpose of this study is to examine the effectiveness of alternative programming paradigms. NPBs consist of five kernels and three simulated applications that mimic the computation and data movement of large scale computational fluid dynamics (CFD) applications. We started with the serial version included in NPB2.3. Optimization of memory and cache usage was applied to several benchmarks, noticeably BT and SP, resulting in better sequential performance. In order to overcome the lack of an HPF performance model and guide the development of the HPF codes, we employed an empirical performance model for several primitives found in the benchmarks. We encountered a few limitations of HPF, such as lack of supporting the "REDISTRIBUTION" directive and no easy way to handle irregular computation. The parallelization with OpenMP directives was done at the outer-most loop level to achieve the largest granularity. The performance of six HPF and OpenMP benchmarks is compared with their MPI counterparts for the Class-A problem size in the figure in next page. These results were obtained on an SGI Origin2000 (195MHz) with MIPSpro-f77 compiler 7.2.1 for OpenMP and MPI codes and PGI pghpf-2.4.3 compiler with MPI interface for HPF programs.

A Programming Model Performance Study Using the NAS Parallel Benchmarks

DOE PAGES

Shan, Hongzhang; Blagojević, Filip; Min, Seung-Jai; ...

2010-01-01

Harnessing the power of multicore platforms is challenging due to the additional levels of parallelism present. In this paper we use the NAS Parallel Benchmarks to study three programming models, MPI, OpenMP and PGAS to understand their performance and memory usage characteristics on current multicore architectures. To understand these characteristics we use the Integrated Performance Monitoring tool and other ways to measure communication versus computation time, as well as the fraction of the run time spent in OpenMP. The benchmarks are run on two different Cray XT5 systems and an Infiniband cluster. Our results show that in general the threemore » programming models exhibit very similar performance characteristics. In a few cases, OpenMP is significantly faster because it explicitly avoids communication. For these particular cases, we were able to re-write the UPC versions and achieve equal performance to OpenMP. Using OpenMP was also the most advantageous in terms of memory usage. Also we compare performance differences between the two Cray systems, which have quad-core and hex-core processors. We show that at scale the performance is almost always slower on the hex-core system because of increased contention for network resources.« less

Mapping of H.264 decoding on a multiprocessor architecture

NASA Astrophysics Data System (ADS)

van der Tol, Erik B.; Jaspers, Egbert G.; Gelderblom, Rob H.

2003-05-01

Due to the increasing significance of development costs in the competitive domain of high-volume consumer electronics, generic solutions are required to enable reuse of the design effort and to increase the potential market volume. As a result from this, Systems-on-Chip (SoCs) contain a growing amount of fully programmable media processing devices as opposed to application-specific systems, which offered the most attractive solutions due to a high performance density. The following motivates this trend. First, SoCs are increasingly dominated by their communication infrastructure and embedded memory, thereby making the cost of the functional units less significant. Moreover, the continuously growing design costs require generic solutions that can be applied over a broad product range. Hence, powerful programmable SoCs are becoming increasingly attractive. However, to enable power-efficient designs, that are also scalable over the advancing VLSI technology, parallelism should be fully exploited. Both task-level and instruction-level parallelism can be provided by means of e.g. a VLIW multiprocessor architecture. To provide the above-mentioned scalability, we propose to partition the data over the processors, instead of traditional functional partitioning. An advantage of this approach is the inherent locality of data, which is extremely important for communication-efficient software implementations. Consequently, a software implementation is discussed, enabling e.g. SD resolution H.264 decoding with a two-processor architecture, whereas High-Definition (HD) decoding can be achieved with an eight-processor system, executing the same software. Experimental results show that the data communication considerably reduces up to 65% directly improving the overall performance. Apart from considerable improvement in memory bandwidth, this novel concept of partitioning offers a natural approach for optimally balancing the load of all processors, thereby further improving the overall speedup.

Evaluating architecture impact on system energy efficiency

PubMed Central

Yu, Shijie; Wang, Rui; Luan, Zhongzhi; Qian, Depei

2017-01-01

As the energy consumption has been surging in an unsustainable way, it is important to understand the impact of existing architecture designs from energy efficiency perspective, which is especially valuable for High Performance Computing (HPC) and datacenter environment hosting tens of thousands of servers. One obstacle hindering the advance of comprehensive evaluation on energy efficiency is the deficient power measuring approach. Most of the energy study relies on either external power meters or power models, both of these two methods contain intrinsic drawbacks in their practical adoption and measuring accuracy. Fortunately, the advent of Intel Running Average Power Limit (RAPL) interfaces has promoted the power measurement ability into next level, with higher accuracy and finer time resolution. Therefore, we argue it is the exact time to conduct an in-depth evaluation of the existing architecture designs to understand their impact on system energy efficiency. In this paper, we leverage representative benchmark suites including serial and parallel workloads from diverse domains to evaluate the architecture features such as Non Uniform Memory Access (NUMA), Simultaneous Multithreading (SMT) and Turbo Boost. The energy is tracked at subcomponent level such as Central Processing Unit (CPU) cores, uncore components and Dynamic Random-Access Memory (DRAM) through exploiting the power measurement ability exposed by RAPL. The experiments reveal non-intuitive results: 1) the mismatch between local compute and remote memory node caused by NUMA effect not only generates dramatic power and energy surge but also deteriorates the energy efficiency significantly; 2) for multithreaded application such as the Princeton Application Repository for Shared-Memory Computers (PARSEC), most of the workloads benefit a notable increase of energy efficiency using SMT, with more than 40% decline in average power consumption; 3) Turbo Boost is effective to accelerate the workload execution and further preserve the energy, however it may not be applicable on system with tight power budget. PMID:29161317

Evaluating architecture impact on system energy efficiency.

PubMed

Yu, Shijie; Yang, Hailong; Wang, Rui; Luan, Zhongzhi; Qian, Depei

2017-01-01

As the energy consumption has been surging in an unsustainable way, it is important to understand the impact of existing architecture designs from energy efficiency perspective, which is especially valuable for High Performance Computing (HPC) and datacenter environment hosting tens of thousands of servers. One obstacle hindering the advance of comprehensive evaluation on energy efficiency is the deficient power measuring approach. Most of the energy study relies on either external power meters or power models, both of these two methods contain intrinsic drawbacks in their practical adoption and measuring accuracy. Fortunately, the advent of Intel Running Average Power Limit (RAPL) interfaces has promoted the power measurement ability into next level, with higher accuracy and finer time resolution. Therefore, we argue it is the exact time to conduct an in-depth evaluation of the existing architecture designs to understand their impact on system energy efficiency. In this paper, we leverage representative benchmark suites including serial and parallel workloads from diverse domains to evaluate the architecture features such as Non Uniform Memory Access (NUMA), Simultaneous Multithreading (SMT) and Turbo Boost. The energy is tracked at subcomponent level such as Central Processing Unit (CPU) cores, uncore components and Dynamic Random-Access Memory (DRAM) through exploiting the power measurement ability exposed by RAPL. The experiments reveal non-intuitive results: 1) the mismatch between local compute and remote memory node caused by NUMA effect not only generates dramatic power and energy surge but also deteriorates the energy efficiency significantly; 2) for multithreaded application such as the Princeton Application Repository for Shared-Memory Computers (PARSEC), most of the workloads benefit a notable increase of energy efficiency using SMT, with more than 40% decline in average power consumption; 3) Turbo Boost is effective to accelerate the workload execution and further preserve the energy, however it may not be applicable on system with tight power budget.

Contention Modeling for Multithreaded Distributed Shared Memory Machines: The Cray XMT

DOE Office of Scientific and Technical Information (OSTI.GOV)

Secchi, Simone; Tumeo, Antonino; Villa, Oreste

Distributed Shared Memory (DSM) machines are a wide class of multi-processor computing systems where a large virtually-shared address space is mapped on a network of physically distributed memories. High memory latency and network contention are two of the main factors that limit performance scaling of such architectures. Modern high-performance computing DSM systems have evolved toward exploitation of massive hardware multi-threading and fine-grained memory hashing to tolerate irregular latencies, avoid network hot-spots and enable high scaling. In order to model the performance of such large-scale machines, parallel simulation has been proved to be a promising approach to achieve good accuracy inmore » reasonable times. One of the most critical factors in solving the simulation speed-accuracy trade-off is network modeling. The Cray XMT is a massively multi-threaded supercomputing architecture that belongs to the DSM class, since it implements a globally-shared address space abstraction on top of a physically distributed memory substrate. In this paper, we discuss the development of a contention-aware network model intended to be integrated in a full-system XMT simulator. We start by measuring the effects of network contention in a 128-processor XMT machine and then investigate the trade-off that exists between simulation accuracy and speed, by comparing three network models which operate at different levels of accuracy. The comparison and model validation is performed by executing a string-matching algorithm on the full-system simulator and on the XMT, using three datasets that generate noticeably different contention patterns.« less

Geocomputation over Hybrid Computer Architecture and Systems: Prior Works and On-going Initiatives at UARK

NASA Astrophysics Data System (ADS)

Shi, X.

2015-12-01

As NSF indicated - "Theory and experimentation have for centuries been regarded as two fundamental pillars of science. It is now widely recognized that computational and data-enabled science forms a critical third pillar." Geocomputation is the third pillar of GIScience and geosciences. With the exponential growth of geodata, the challenge of scalable and high performance computing for big data analytics become urgent because many research activities are constrained by the inability of software or tool that even could not complete the computation process. Heterogeneous geodata integration and analytics obviously magnify the complexity and operational time frame. Many large-scale geospatial problems may be not processable at all if the computer system does not have sufficient memory or computational power. Emerging computer architectures, such as Intel's Many Integrated Core (MIC) Architecture and Graphics Processing Unit (GPU), and advanced computing technologies provide promising solutions to employ massive parallelism and hardware resources to achieve scalability and high performance for data intensive computing over large spatiotemporal and social media data. Exploring novel algorithms and deploying the solutions in massively parallel computing environment to achieve the capability for scalable data processing and analytics over large-scale, complex, and heterogeneous geodata with consistent quality and high-performance has been the central theme of our research team in the Department of Geosciences at the University of Arkansas (UARK). New multi-core architectures combined with application accelerators hold the promise to achieve scalability and high performance by exploiting task and data levels of parallelism that are not supported by the conventional computing systems. Such a parallel or distributed computing environment is particularly suitable for large-scale geocomputation over big data as proved by our prior works, while the potential of such advanced infrastructure remains unexplored in this domain. Within this presentation, our prior and on-going initiatives will be summarized to exemplify how we exploit multicore CPUs, GPUs, and MICs, and clusters of CPUs, GPUs and MICs, to accelerate geocomputation in different applications.

PIC codes for plasma accelerators on emerging computer architectures (GPUS, Multicore/Manycore CPUS)

NASA Astrophysics Data System (ADS)

Vincenti, Henri

2016-03-01

The advent of exascale computers will enable 3D simulations of a new laser-plasma interaction regimes that were previously out of reach of current Petasale computers. However, the paradigm used to write current PIC codes will have to change in order to fully exploit the potentialities of these new computing architectures. Indeed, achieving Exascale computing facilities in the next decade will be a great challenge in terms of energy consumption and will imply hardware developments directly impacting our way of implementing PIC codes. As data movement (from die to network) is by far the most energy consuming part of an algorithm future computers will tend to increase memory locality at the hardware level and reduce energy consumption related to data movement by using more and more cores on each compute nodes (''fat nodes'') that will have a reduced clock speed to allow for efficient cooling. To compensate for frequency decrease, CPU machine vendors are making use of long SIMD instruction registers that are able to process multiple data with one arithmetic operator in one clock cycle. SIMD register length is expected to double every four years. GPU's also have a reduced clock speed per core and can process Multiple Instructions on Multiple Datas (MIMD). At the software level Particle-In-Cell (PIC) codes will thus have to achieve both good memory locality and vectorization (for Multicore/Manycore CPU) to fully take advantage of these upcoming architectures. In this talk, we present the portable solutions we implemented in our high performance skeleton PIC code PICSAR to both achieve good memory locality and cache reuse as well as good vectorization on SIMD architectures. We also present the portable solutions used to parallelize the Pseudo-sepctral quasi-cylindrical code FBPIC on GPUs using the Numba python compiler.

Parallelization of elliptic solver for solving 1D Boussinesq model

NASA Astrophysics Data System (ADS)

Tarwidi, D.; Adytia, D.

2018-03-01

In this paper, a parallel implementation of an elliptic solver in solving 1D Boussinesq model is presented. Numerical solution of Boussinesq model is obtained by implementing a staggered grid scheme to continuity, momentum, and elliptic equation of Boussinesq model. Tridiagonal system emerging from numerical scheme of elliptic equation is solved by cyclic reduction algorithm. The parallel implementation of cyclic reduction is executed on multicore processors with shared memory architectures using OpenMP. To measure the performance of parallel program, large number of grids is varied from 28 to 214. Two test cases of numerical experiment, i.e. propagation of solitary and standing wave, are proposed to evaluate the parallel program. The numerical results are verified with analytical solution of solitary and standing wave. The best speedup of solitary and standing wave test cases is about 2.07 with 214 of grids and 1.86 with 213 of grids, respectively, which are executed by using 8 threads. Moreover, the best efficiency of parallel program is 76.2% and 73.5% for solitary and standing wave test cases, respectively.

Efficient Numeric and Geometric Computations using Heterogeneous Shared Memory Architectures

DTIC Science & Technology

2017-10-04

Report: Efficient Numeric and Geometric Computations using Heterogeneous Shared Memory Architectures The views, opinions and/or findings contained in this...Chapel Hill Title: Efficient Numeric and Geometric Computations using Heterogeneous Shared Memory Architectures Report Term: 0-Other Email: dm...algorithms for scientific and geometric computing by exploiting the power and performance efficiency of heterogeneous shared memory architectures . These

A Neural Network Architecture For Rapid Model Indexing In Computer Vision Systems

NASA Astrophysics Data System (ADS)

Pawlicki, Ted

1988-03-01

Models of objects stored in memory have been shown to be useful for guiding the processing of computer vision systems. A major consideration in such systems, however, is how stored models are initially accessed and indexed by the system. As the number of stored models increases, the time required to search memory for the correct model becomes high. Parallel distributed, connectionist, neural networks' have been shown to have appealing content addressable memory properties. This paper discusses an architecture for efficient storage and reference of model memories stored as stable patterns of activity in a parallel, distributed, connectionist, neural network. The emergent properties of content addressability and resistance to noise are exploited to perform indexing of the appropriate object centered model from image centered primitives. The system consists of three network modules each of which represent information relative to a different frame of reference. The model memory network is a large state space vector where fields in the vector correspond to ordered component objects and relative, object based spatial relationships between the component objects. The component assertion network represents evidence about the existence of object primitives in the input image. It establishes local frames of reference for object primitives relative to the image based frame of reference. The spatial relationship constraint network is an intermediate representation which enables the association between the object based and the image based frames of reference. This intermediate level represents information about possible object orderings and establishes relative spatial relationships from the image based information in the component assertion network below. It is also constrained by the lawful object orderings in the model memory network above. The system design is consistent with current psychological theories of recognition by component. It also seems to support Marr's notions of hierarchical indexing. (i.e. the specificity, adjunct, and parent indices) It supports the notion that multiple canonical views of an object may have to be stored in memory to enable its efficient identification. The use of variable fields in the state space vectors appears to keep the number of required nodes in the network down to a tractable number while imposing a semantic value on different areas of the state space. This semantic imposition supports an interface between the analogical aspects of neural networks and the propositional paradigms of symbolic processing.

Ultrafast Target Recognition via Super-Parallel Holograph Based Correlator, RAM and Associative Memory

DTIC Science & Technology

2008-03-11

JTC) 2𔃾 based on a dynamic material answers the challenge of fast correlation with large databases. Images retrieved from the SPHRAM and used as the...transform (JTC) and matched spatial filter or VanderLugt ( VLC ) correlators, either of which can be implemented in real-time by degenerate four wave-mixing in...proposed system, consisting of the SPHROM coupled with a shift-invariant real-time VLC . The correlation is performed in the VLC architecture to

A model for the distributed storage and processing of large arrays

NASA Technical Reports Server (NTRS)

Mehrota, P.; Pratt, T. W.

1983-01-01

A conceptual model for parallel computations on large arrays is developed. The model provides a set of language concepts appropriate for processing arrays which are generally too large to fit in the primary memories of a multiprocessor system. The semantic model is used to represent arrays on a concurrent architecture in such a way that the performance realities inherent in the distributed storage and processing can be adequately represented. An implementation of the large array concept as an Ada package is also described.

SCELib2: the new revision of SCELib, the parallel computational library of molecular properties in the single center approach

NASA Astrophysics Data System (ADS)

Sanna, N.; Morelli, G.

2004-09-01

In this paper we present the new version of the SCELib program (CPC Catalogue identifier ADMG) a full numerical implementation of the Single Center Expansion (SCE) method. The physics involved is that of producing the SCE description of molecular electronic densities, of molecular electrostatic potentials and of molecular perturbed potentials due to a point negative or positive charge. This new revision of the program has been optimized to run in serial as well as in parallel execution mode, to support a larger set of molecular symmetries and to permit the restart of long-lasting calculations. To measure the performance of this new release, a comparative study has been carried out on the most powerful computing architectures in serial and parallel runs. The results of the calculations reported in this paper refer to real cases medium to large molecular systems and they are reported in full details to benchmark at best the parallel architectures the new SCELib code will run on. Program summaryTitle of program: SCELib2 Catalogue identifier: ADGU Program summary URL:http://cpc.cs.qub.ac.uk/summaries/ADGU Program obtainable from: CPC Program Library, Queen's University of Belfast, N. Ireland Reference to previous versions: Comput. Phys. Commun. 128 (2) (2000) 139 (CPC catalogue identifier: ADMG) Does the new version supersede the original program?: Yes Computer for which the program is designed and others on which it has been tested: HP ES45 and rx2600, SUN ES4500, IBM SP and any single CPU workstation based on Alpha, SPARC, POWER, Itanium2 and X86 processors Installations: CASPUR, local Operating systems under which the program has been tested: HP Tru64 V5.X, SUNOS V5.8, IBM AIX V5.X, Linux RedHat V8.0 Programming language used: C Memory required to execute with typical data: 10 Mwords. Up to 2000 Mwords depending on the molecular system and runtime parameters No. of bits in a word: 64 No. of processors used: 1 to 32 Has the code been vectorized or parallelized?: Yes No. of bytes in distributed program, including test data, etc.: 3 798 507 No. of lines in distributed program, including test data, etc.: 187 226 Distribution format: tar.gz Nature of physical problem: In this set of codes an efficient procedure is implemented to describe the wavefunction and related molecular properties of a polyatomic molecular system within the Single Center of Expansion (SCE) approximation. The resulting SCE wavefunction, electron density, electrostatic and exchange/correlation potentials can then be used via a proper Application Programming Interface (API) to describe the target molecular system which can be employed in electron-molecule scattering calculations. The molecular properties expanded over a single center turn out to also be of more general application and some possible uses in quantum chemistry, biomodelling and drug design are also outlined. Method of solution: The polycentre Hartee-Fock solution for a molecule of arbitrary geometry, based on linear combination of Gaussian-Type Orbital (GTO), is expanded over a single center, typically the Center Of Mass (C.O.M.), by means of a Gauss-Legendre/Chebyschev quadrature over the θ, φ angular coordinates. The resulting SCE numerical wavefunction is then used to calculate the one-particle electron density, the electrostatic potential and two different models for the correlation/polarization potentials induced by the impinging electron, which have the correct asymptotic behaviour for the leading dipole molecular polarizabilities. Restrictions on the complexity of the problem: Depending on the molecular system under study and on the operating conditions the program may or may not fit into available RAM memory. In this case a feature of the program is to memory map a disk file in order to efficiently access the memory data through a disk device. Typical running time: The execution time strongly depends on the molecular target description and on the hardware/OS chosen, it is directly proportional to the ( r, θ, φ) grid size and to the number of angular basis functions used. Thus, from the program printout of the main arrays memory occupancy, the user can approximately derive the expected computer time needed for a given calculation executed in serial mode. For parallel executions the overall efficiency must be further taken into account, and this depends on the no. of processors used as well as on the parallel architecture chosen, so a simple general law is at present not determinable. Unusual features of the program: The code has been engineered to use dynamical, runtime determined, global parameters with the aim to have all the data fitted in the RAM memory. Some unusual circumstances, e.g., when using large values of those parameters, may cause the program to run with unexpected performance reductions due to runtime bottlenecks like those caused by memory swap operations which strongly depend on the hardware used. In such cases, a parallel execution of the code is generally sufficient to fix the problem since the data size is partitioned over the available processors. When a suitable parallel system is not available for execution, a mechanism of memory mapped file can be used; with this option on, all the available memory will be used as a buffer for a disk file which contains the whole data set, thus having a better throughput with respect to the traditional swapping/paging of the Unix OS.

The prefrontal landscape: implications of functional architecture for understanding human mentation and the central executive.

PubMed

Goldman-Rakic, P S

1996-10-29

The functional architecture of prefrontal cortex is central to our understanding of human mentation and cognitive prowess. This region of the brain is often treated as an undifferentiated structure, on the one hand, or as a mosaic of psychological faculties, on the other. This paper focuses on the working memory processor as a specialization of prefrontal cortex and argues that the different areas within prefrontal cortex represent iterations of this function for different information domains, including spatial cognition, object cognition and additionally, in humans, semantic processing. According to this parallel processing architecture, the 'central executive' could be considered an emergent property of multiple domain-specific processors operating interactively. These processors are specializations of different prefrontal cortical areas, each interconnected both with the domain-relevant long-term storage sites in posterior regions of the cortex and with appropriate output pathways.

Scalable parallel communications

NASA Technical Reports Server (NTRS)

Maly, K.; Khanna, S.; Overstreet, C. M.; Mukkamala, R.; Zubair, M.; Sekhar, Y. S.; Foudriat, E. C.

1992-01-01

Coarse-grain parallelism in networking (that is, the use of multiple protocol processors running replicated software sending over several physical channels) can be used to provide gigabit communications for a single application. Since parallel network performance is highly dependent on real issues such as hardware properties (e.g., memory speeds and cache hit rates), operating system overhead (e.g., interrupt handling), and protocol performance (e.g., effect of timeouts), we have performed detailed simulations studies of both a bus-based multiprocessor workstation node (based on the Sun Galaxy MP multiprocessor) and a distributed-memory parallel computer node (based on the Touchstone DELTA) to evaluate the behavior of coarse-grain parallelism. Our results indicate: (1) coarse-grain parallelism can deliver multiple 100 Mbps with currently available hardware platforms and existing networking protocols (such as Transmission Control Protocol/Internet Protocol (TCP/IP) and parallel Fiber Distributed Data Interface (FDDI) rings); (2) scale-up is near linear in n, the number of protocol processors, and channels (for small n and up to a few hundred Mbps); and (3) since these results are based on existing hardware without specialized devices (except perhaps for some simple modifications of the FDDI boards), this is a low cost solution to providing multiple 100 Mbps on current machines. In addition, from both the performance analysis and the properties of these architectures, we conclude: (1) multiple processors providing identical services and the use of space division multiplexing for the physical channels can provide better reliability than monolithic approaches (it also provides graceful degradation and low-cost load balancing); (2) coarse-grain parallelism supports running several transport protocols in parallel to provide different types of service (for example, one TCP handles small messages for many users, other TCP's running in parallel provide high bandwidth service to a single application); and (3) coarse grain parallelism will be able to incorporate many future improvements from related work (e.g., reduced data movement, fast TCP, fine-grain parallelism) also with near linear speed-ups.

Code Parallelization with CAPO: A User Manual

NASA Technical Reports Server (NTRS)

Jin, Hao-Qiang; Frumkin, Michael; Yan, Jerry; Biegel, Bryan (Technical Monitor)

2001-01-01

A software tool has been developed to assist the parallelization of scientific codes. This tool, CAPO, extends an existing parallelization toolkit, CAPTools developed at the University of Greenwich, to generate OpenMP parallel codes for shared memory architectures. This is an interactive toolkit to transform a serial Fortran application code to an equivalent parallel version of the software - in a small fraction of the time normally required for a manual parallelization. We first discuss the way in which loop types are categorized and how efficient OpenMP directives can be defined and inserted into the existing code using the in-depth interprocedural analysis. The use of the toolkit on a number of application codes ranging from benchmark to real-world application codes is presented. This will demonstrate the great potential of using the toolkit to quickly parallelize serial programs as well as the good performance achievable on a large number of toolkit to quickly parallelize serial programs as well as the good performance achievable on a large number of processors. The second part of the document gives references to the parameters and the graphic user interface implemented in the toolkit. Finally a set of tutorials is included for hands-on experiences with this toolkit.

The Visualization Toolkit (VTK): Rewriting the rendering code for modern graphics cards

NASA Astrophysics Data System (ADS)

Hanwell, Marcus D.; Martin, Kenneth M.; Chaudhary, Aashish; Avila, Lisa S.

2015-09-01

The Visualization Toolkit (VTK) is an open source, permissively licensed, cross-platform toolkit for scientific data processing, visualization, and data analysis. It is over two decades old, originally developed for a very different graphics card architecture. Modern graphics cards feature fully programmable, highly parallelized architectures with large core counts. VTK's rendering code was rewritten to take advantage of modern graphics cards, maintaining most of the toolkit's programming interfaces. This offers the opportunity to compare the performance of old and new rendering code on the same systems/cards. Significant improvements in rendering speeds and memory footprints mean that scientific data can be visualized in greater detail than ever before. The widespread use of VTK means that these improvements will reap significant benefits.

Octree-based, GPU implementation of a continuous cellular automaton for the simulation of complex, evolving surfaces

NASA Astrophysics Data System (ADS)

Ferrando, N.; Gosálvez, M. A.; Cerdá, J.; Gadea, R.; Sato, K.

2011-03-01

Presently, dynamic surface-based models are required to contain increasingly larger numbers of points and to propagate them over longer time periods. For large numbers of surface points, the octree data structure can be used as a balance between low memory occupation and relatively rapid access to the stored data. For evolution rules that depend on neighborhood states, extended simulation periods can be obtained by using simplified atomistic propagation models, such as the Cellular Automata (CA). This method, however, has an intrinsic parallel updating nature and the corresponding simulations are highly inefficient when performed on classical Central Processing Units (CPUs), which are designed for the sequential execution of tasks. In this paper, a series of guidelines is presented for the efficient adaptation of octree-based, CA simulations of complex, evolving surfaces into massively parallel computing hardware. A Graphics Processing Unit (GPU) is used as a cost-efficient example of the parallel architectures. For the actual simulations, we consider the surface propagation during anisotropic wet chemical etching of silicon as a computationally challenging process with a wide-spread use in microengineering applications. A continuous CA model that is intrinsically parallel in nature is used for the time evolution. Our study strongly indicates that parallel computations of dynamically evolving surfaces simulated using CA methods are significantly benefited by the incorporation of octrees as support data structures, substantially decreasing the overall computational time and memory usage.

Implementation of a fully-balanced periodic tridiagonal solver on a parallel distributed memory architecture

NASA Technical Reports Server (NTRS)

Eidson, T. M.; Erlebacher, G.

1994-01-01

While parallel computers offer significant computational performance, it is generally necessary to evaluate several programming strategies. Two programming strategies for a fairly common problem - a periodic tridiagonal solver - are developed and evaluated. Simple model calculations as well as timing results are presented to evaluate the various strategies. The particular tridiagonal solver evaluated is used in many computational fluid dynamic simulation codes. The feature that makes this algorithm unique is that these simulation codes usually require simultaneous solutions for multiple right-hand-sides (RHS) of the system of equations. Each RHS solutions is independent and thus can be computed in parallel. Thus a Gaussian elimination type algorithm can be used in a parallel computation and the more complicated approaches such as cyclic reduction are not required. The two strategies are a transpose strategy and a distributed solver strategy. For the transpose strategy, the data is moved so that a subset of all the RHS problems is solved on each of the several processors. This usually requires significant data movement between processor memories across a network. The second strategy attempts to have the algorithm allow the data across processor boundaries in a chained manner. This usually requires significantly less data movement. An approach to accomplish this second strategy in a near-perfect load-balanced manner is developed. In addition, an algorithm will be shown to directly transform a sequential Gaussian elimination type algorithm into the parallel chained, load-balanced algorithm.

Parallel architectures for iterative methods on adaptive, block structured grids

NASA Technical Reports Server (NTRS)

Gannon, D.; Vanrosendale, J.

1983-01-01

A parallel computer architecture well suited to the solution of partial differential equations in complicated geometries is proposed. Algorithms for partial differential equations contain a great deal of parallelism. But this parallelism can be difficult to exploit, particularly on complex problems. One approach to extraction of this parallelism is the use of special purpose architectures tuned to a given problem class. The architecture proposed here is tuned to boundary value problems on complex domains. An adaptive elliptic algorithm which maps effectively onto the proposed architecture is considered in detail. Two levels of parallelism are exploited by the proposed architecture. First, by making use of the freedom one has in grid generation, one can construct grids which are locally regular, permitting a one to one mapping of grids to systolic style processor arrays, at least over small regions. All local parallelism can be extracted by this approach. Second, though there may be a regular global structure to the grids constructed, there will be parallelism at this level. One approach to finding and exploiting this parallelism is to use an architecture having a number of processor clusters connected by a switching network. The use of such a network creates a highly flexible architecture which automatically configures to the problem being solved.

A site oriented supercomputer for theoretical physics: The Fermilab Advanced Computer Program Multi Array Processor System (ACMAPS)

DOE Office of Scientific and Technical Information (OSTI.GOV)

Nash, T.; Atac, R.; Cook, A.

1989-03-06

The ACPMAPS multipocessor is a highly cost effective, local memory parallel computer with a hypercube or compound hypercube architecture. Communication requires the attention of only the two communicating nodes. The design is aimed at floating point intensive, grid like problems, particularly those with extreme computing requirements. The processing nodes of the system are single board array processors, each with a peak power of 20 Mflops, supported by 8 Mbytes of data and 2 Mbytes of instruction memory. The system currently being assembled has a peak power of 5 Gflops. The nodes are based on the Weitek XL Chip set. Themore » system delivers performance at approximately $300/Mflop. 8 refs., 4 figs.« less

Programmable synaptic devices for electronic neural nets

NASA Technical Reports Server (NTRS)

Moopenn, A.; Thakoor, A. P.

1990-01-01

The architecture, design, and operational characteristics of custom VLSI and thin film synaptic devices are described. The devices include CMOS-based synaptic chips containing 1024 reprogrammable synapses with a 6-bit dynamic range, and nonvolatile, write-once, binary synaptic arrays based on memory switching in hydrogenated amorphous silicon films. Their suitability for embodiment of fully parallel and analog neural hardware is discussed. Specifically, a neural network solution to an assignment problem of combinatorial global optimization, implemented in fully parallel hardware using the synaptic chips, is described. The network's ability to provide optimal and near optimal solutions over a time scale of few neuron time constants has been demonstrated and suggests a speedup improvement of several orders of magnitude over conventional search methods.

Simulating spin models on GPU

NASA Astrophysics Data System (ADS)

Weigel, Martin

2011-09-01

Over the last couple of years it has been realized that the vast computational power of graphics processing units (GPUs) could be harvested for purposes other than the video game industry. This power, which at least nominally exceeds that of current CPUs by large factors, results from the relative simplicity of the GPU architectures as compared to CPUs, combined with a large number of parallel processing units on a single chip. To benefit from this setup for general computing purposes, the problems at hand need to be prepared in a way to profit from the inherent parallelism and hierarchical structure of memory accesses. In this contribution I discuss the performance potential for simulating spin models, such as the Ising model, on GPU as compared to conventional simulations on CPU.

SAR processing on the MPP

NASA Technical Reports Server (NTRS)

Batcher, K. E.; Eddey, E. E.; Faiss, R. O.; Gilmore, P. A.

1981-01-01

The processing of synthetic aperture radar (SAR) signals using the massively parallel processor (MPP) is discussed. The fast Fourier transform convolution procedures employed in the algorithms are described. The MPP architecture comprises an array unit (ARU) which processes arrays of data; an array control unit which controls the operation of the ARU and performs scalar arithmetic; a program and data management unit which controls the flow of data; and a unique staging memory (SM) which buffers and permutes data. The ARU contains a 128 by 128 array of bit-serial processing elements (PE). Two-by-four surarrays of PE's are packaged in a custom VLSI HCMOS chip. The staging memory is a large multidimensional-access memory which buffers and permutes data flowing with the system. Efficient SAR processing is achieved via ARU communication paths and SM data manipulation. Real time processing capability can be realized via a multiple ARU, multiple SM configuration.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Zhao, Luning; Neuscamman, Eric

We present a modification to variational Monte Carlo’s linear method optimization scheme that addresses a critical memory bottleneck while maintaining compatibility with both the traditional ground state variational principle and our recently-introduced variational principle for excited states. For wave function ansatzes with tens of thousands of variables, our modification reduces the required memory per parallel process from tens of gigabytes to hundreds of megabytes, making the methodology a much better fit for modern supercomputer architectures in which data communication and per-process memory consumption are primary concerns. We verify the efficacy of the new optimization scheme in small molecule tests involvingmore » both the Hilbert space Jastrow antisymmetric geminal power ansatz and real space multi-Slater Jastrow expansions. Satisfied with its performance, we have added the optimizer to the QMCPACK software package, with which we demonstrate on a hydrogen ring a prototype approach for making systematically convergent, non-perturbative predictions of Mott-insulators’ optical band gaps.« less

Parallel Fortran-MPI software for numerical inversion of the Laplace transform and its application to oscillatory water levels in groundwater environments

USGS Publications Warehouse

Zhan, X.

2005-01-01

A parallel Fortran-MPI (Message Passing Interface) software for numerical inversion of the Laplace transform based on a Fourier series method is developed to meet the need of solving intensive computational problems involving oscillatory water level's response to hydraulic tests in a groundwater environment. The software is a parallel version of ACM (The Association for Computing Machinery) Transactions on Mathematical Software (TOMS) Algorithm 796. Running 38 test examples indicated that implementation of MPI techniques with distributed memory architecture speedups the processing and improves the efficiency. Applications to oscillatory water levels in a well during aquifer tests are presented to illustrate how this package can be applied to solve complicated environmental problems involved in differential and integral equations. The package is free and is easy to use for people with little or no previous experience in using MPI but who wish to get off to a quick start in parallel computing. ?? 2004 Elsevier Ltd. All rights reserved.

The design and implementation of a parallel unstructured Euler solver using software primitives

NASA Technical Reports Server (NTRS)

Das, R.; Mavriplis, D. J.; Saltz, J.; Gupta, S.; Ponnusamy, R.

1992-01-01

This paper is concerned with the implementation of a three-dimensional unstructured grid Euler-solver on massively parallel distributed-memory computer architectures. The goal is to minimize solution time by achieving high computational rates with a numerically efficient algorithm. An unstructured multigrid algorithm with an edge-based data structure has been adopted, and a number of optimizations have been devised and implemented in order to accelerate the parallel communication rates. The implementation is carried out by creating a set of software tools, which provide an interface between the parallelization issues and the sequential code, while providing a basis for future automatic run-time compilation support. Large practical unstructured grid problems are solved on the Intel iPSC/860 hypercube and Intel Touchstone Delta machine. The quantitative effect of the various optimizations are demonstrated, and we show that the combined effect of these optimizations leads to roughly a factor of three performance improvement. The overall solution efficiency is compared with that obtained on the CRAY-YMP vector supercomputer.

Parallel Discrete Molecular Dynamics Simulation With Speculation and In-Order Commitment*†

PubMed Central

Khan, Md. Ashfaquzzaman; Herbordt, Martin C.

2011-01-01

Discrete molecular dynamics simulation (DMD) uses simplified and discretized models enabling simulations to advance by event rather than by timestep. DMD is an instance of discrete event simulation and so is difficult to scale: even in this multi-core era, all reported DMD codes are serial. In this paper we discuss the inherent difficulties of scaling DMD and present our method of parallelizing DMD through event-based decomposition. Our method is microarchitecture inspired: speculative processing of events exposes parallelism, while in-order commitment ensures correctness. We analyze the potential of this parallelization method for shared-memory multiprocessors. Achieving scalability required extensive experimentation with scheduling and synchronization methods to mitigate serialization. The speed-up achieved for a variety of system sizes and complexities is nearly 6× on an 8-core and over 9× on a 12-core processor. We present and verify analytical models that account for the achieved performance as a function of available concurrency and architectural limitations. PMID:21822327

Parallel Discrete Molecular Dynamics Simulation With Speculation and In-Order Commitment.

PubMed

Khan, Md Ashfaquzzaman; Herbordt, Martin C

2011-07-20

Discrete molecular dynamics simulation (DMD) uses simplified and discretized models enabling simulations to advance by event rather than by timestep. DMD is an instance of discrete event simulation and so is difficult to scale: even in this multi-core era, all reported DMD codes are serial. In this paper we discuss the inherent difficulties of scaling DMD and present our method of parallelizing DMD through event-based decomposition. Our method is microarchitecture inspired: speculative processing of events exposes parallelism, while in-order commitment ensures correctness. We analyze the potential of this parallelization method for shared-memory multiprocessors. Achieving scalability required extensive experimentation with scheduling and synchronization methods to mitigate serialization. The speed-up achieved for a variety of system sizes and complexities is nearly 6× on an 8-core and over 9× on a 12-core processor. We present and verify analytical models that account for the achieved performance as a function of available concurrency and architectural limitations.

The cost of conservative synchronization in parallel discrete event simulations

NASA Technical Reports Server (NTRS)

Nicol, David M.

1990-01-01

The performance of a synchronous conservative parallel discrete-event simulation protocol is analyzed. The class of simulation models considered is oriented around a physical domain and possesses a limited ability to predict future behavior. A stochastic model is used to show that as the volume of simulation activity in the model increases relative to a fixed architecture, the complexity of the average per-event overhead due to synchronization, event list manipulation, lookahead calculations, and processor idle time approach the complexity of the average per-event overhead of a serial simulation. The method is therefore within a constant factor of optimal. The analysis demonstrates that on large problems--those for which parallel processing is ideally suited--there is often enough parallel workload so that processors are not usually idle. The viability of the method is also demonstrated empirically, showing how good performance is achieved on large problems using a thirty-two node Intel iPSC/2 distributed memory multiprocessor.

Paradigms and strategies for scientific computing on distributed memory concurrent computers

DOE Office of Scientific and Technical Information (OSTI.GOV)

Foster, I.T.; Walker, D.W.

1994-06-01

In this work we examine recent advances in parallel languages and abstractions that have the potential for improving the programmability and maintainability of large-scale, parallel, scientific applications running on high performance architectures and networks. This paper focuses on Fortran M, a set of extensions to Fortran 77 that supports the modular design of message-passing programs. We describe the Fortran M implementation of a particle-in-cell (PIC) plasma simulation application, and discuss issues in the optimization of the code. The use of two other methodologies for parallelizing the PIC application are considered. The first is based on the shared object abstraction asmore » embodied in the Orca language. The second approach is the Split-C language. In Fortran M, Orca, and Split-C the ability of the programmer to control the granularity of communication is important is designing an efficient implementation.« less

Multicore Challenges and Benefits for High Performance Scientific Computing

DOE PAGES

Nielsen, Ida M. B.; Janssen, Curtis L.

2008-01-01

Until recently, performance gains in processors were achieved largely by improvements in clock speeds and instruction level parallelism. Thus, applications could obtain performance increases with relatively minor changes by upgrading to the latest generation of computing hardware. Currently, however, processor performance improvements are realized by using multicore technology and hardware support for multiple threads within each core, and taking full advantage of this technology to improve the performance of applications requires exposure of extreme levels of software parallelism. We will here discuss the architecture of parallel computers constructed from many multicore chips as well as techniques for managing the complexitymore » of programming such computers, including the hybrid message-passing/multi-threading programming model. We will illustrate these ideas with a hybrid distributed memory matrix multiply and a quantum chemistry algorithm for energy computation using Møller–Plesset perturbation theory.« less

Tolerant (parallel) Programming

NASA Technical Reports Server (NTRS)

DiNucci, David C.; Bailey, David H. (Technical Monitor)

1997-01-01

In order to be truly portable, a program must be tolerant of a wide range of development and execution environments, and a parallel program is just one which must be tolerant of a very wide range. This paper first defines the term "tolerant programming", then describes many layers of tools to accomplish it. The primary focus is on F-Nets, a formal model for expressing computation as a folded partial-ordering of operations, thereby providing an architecture-independent expression of tolerant parallel algorithms. For implementing F-Nets, Cooperative Data Sharing (CDS) is a subroutine package for implementing communication efficiently in a large number of environments (e.g. shared memory and message passing). Software Cabling (SC), a very-high-level graphical programming language for building large F-Nets, possesses many of the features normally expected from today's computer languages (e.g. data abstraction, array operations). Finally, L2(sup 3) is a CASE tool which facilitates the construction, compilation, execution, and debugging of SC programs.

User-Defined Data Distributions in High-Level Programming Languages

NASA Technical Reports Server (NTRS)

Diaconescu, Roxana E.; Zima, Hans P.

2006-01-01

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.

Contemporary Spaces of Memory - Towards Transdisciplinarity in Architecture

NASA Astrophysics Data System (ADS)

Kabrońska, Joanna

2017-10-01

The paper explores new phenomena in the contemporary practice of commemoration implemented through architecture. Architectural objects related to memory can be a place where new trends and phenomena appear earlier than in other architectural objects. The text is an attempt to prove that these new spaces of memory are a kind of laboratory where new ideas taking place in architecture and related disciplines are being tested. Research focuses on the bond between the complex and difficult problem of memory and the issue of transdisciplinarity in architecture. Over the last few decades architecture has been - in comparison to other areas - a relatively closed domain of knowledge. Contemporary places of memory - different from the traditional - may be the evidence of changes. On the basis of theoretical approaches, interdisciplinary surveys, in-field analyses and case studies the paper give insight into the relationships between architecture and other areas, emerging in the recently created spaces of memory of different types. The text indicates that today both the study and the design of such places is difficult without going beyond the field of architecture. There is a need for further extensive research, but the paper confirms the potential of this research direction. Spaces of memory offer the opportunity to capture the transformation of the discipline at the moment when the process begins.

A Case Study on Neural Inspired Dynamic Memory Management Strategies for High Performance Computing.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Vineyard, Craig Michael; Verzi, Stephen Joseph

As high performance computing architectures pursue more computational power there is a need for increased memory capacity and bandwidth as well. A multi-level memory (MLM) architecture addresses this need by combining multiple memory types with different characteristics as varying levels of the same architecture. How to efficiently utilize this memory infrastructure is an unknown challenge, and in this research we sought to investigate whether neural inspired approaches can meaningfully help with memory management. In particular we explored neurogenesis inspired re- source allocation, and were able to show a neural inspired mixed controller policy can beneficially impact how MLM architectures utilizemore » memory.« less

A GPU-Based Architecture for Real-Time Data Assessment at Synchrotron Experiments

NASA Astrophysics Data System (ADS)

Chilingaryan, Suren; Mirone, Alessandro; Hammersley, Andrew; Ferrero, Claudio; Helfen, Lukas; Kopmann, Andreas; Rolo, Tomy dos Santos; Vagovic, Patrik

2011-08-01

Advances in digital detector technology leads presently to rapidly increasing data rates in imaging experiments. Using fast two-dimensional detectors in computed tomography, the data acquisition can be much faster than the reconstruction if no adequate measures are taken, especially when a high photon flux at synchrotron sources is used. We have optimized the reconstruction software employed at the micro-tomography beamlines of our synchrotron facilities to use the computational power of modern graphic cards. The main paradigm of our approach is the full utilization of all system resources. We use a pipelined architecture, where the GPUs are used as compute coprocessors to reconstruct slices, while the CPUs are preparing the next ones. Special attention is devoted to minimize data transfers between the host and GPU memory and to execute memory transfers in parallel with the computations. We were able to reduce the reconstruction time by a factor 30 and process a typical data set of 20 GB in 40 seconds. The time needed for the first evaluation of the reconstructed sample is reduced significantly and quasi real-time visualization is now possible.

Two-dimensional shape recognition using sparse distributed memory

NASA Technical Reports Server (NTRS)

Kanerva, Pentti; Olshausen, Bruno

1990-01-01

Researchers propose a method for recognizing two-dimensional shapes (hand-drawn characters, for example) with an associative memory. The method consists of two stages: first, the image is preprocessed to extract tangents to the contour of the shape; second, the set of tangents is converted to a long bit string for recognition with sparse distributed memory (SDM). SDM provides a simple, massively parallel architecture for an associative memory. Long bit vectors (256 to 1000 bits, for example) serve as both data and addresses to the memory, and patterns are grouped or classified according to similarity in Hamming distance. At the moment, tangents are extracted in a simple manner by progressively blurring the image and then using a Canny-type edge detector (Canny, 1986) to find edges at each stage of blurring. This results in a grid of tangents. While the technique used for obtaining the tangents is at present rather ad hoc, researchers plan to adopt an existing framework for extracting edge orientation information over a variety of resolutions, such as suggested by Watson (1987, 1983), Marr and Hildreth (1980), or Canny (1986).

Optimization of Sparse Matrix-Vector Multiplication on Emerging Multicore Platforms

DOE Office of Scientific and Technical Information (OSTI.GOV)

Williams, Samuel; Oliker, Leonid; Vuduc, Richard

2008-10-16

We are witnessing a dramatic change in computer architecture due to the multicore paradigm shift, as every electronic device from cell phones to supercomputers confronts parallelism of unprecedented scale. To fully unleash the potential of these systems, the HPC community must develop multicore specific-optimization methodologies for important scientific computations. In this work, we examine sparse matrix-vector multiply (SpMV) - one of the most heavily used kernels in scientific computing - across a broad spectrum of multicore designs. Our experimental platform includes the homogeneous AMD quad-core, AMD dual-core, and Intel quad-core designs, the heterogeneous STI Cell, as well as one ofmore » the first scientific studies of the highly multithreaded Sun Victoria Falls (a Niagara2 SMP). We present several optimization strategies especially effective for the multicore environment, and demonstrate significant performance improvements compared to existing state-of-the-art serial and parallel SpMV implementations. Additionally, we present key insights into the architectural trade-offs of leading multicore design strategies, in the context of demanding memory-bound numerical algorithms.« less

LDRD project final report : hybrid AI/cognitive tactical behavior framework for LVC.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Djordjevich, Donna D.; Xavier, Patrick Gordon; Brannon, Nathan Gregory

This Lab-Directed Research and Development (LDRD) sought to develop technology that enhances scenario construction speed, entity behavior robustness, and scalability in Live-Virtual-Constructive (LVC) simulation. We investigated issues in both simulation architecture and behavior modeling. We developed path-planning technology that improves the ability to express intent in the planning task while still permitting an efficient search algorithm. An LVC simulation demonstrated how this enables 'one-click' layout of squad tactical paths, as well as dynamic re-planning for simulated squads and for real and simulated mobile robots. We identified human response latencies that can be exploited in parallel/distributed architectures. We did an experimentalmore » study to determine where parallelization would be productive in Umbra-based force-on-force (FOF) simulations. We developed and implemented a data-driven simulation composition approach that solves entity class hierarchy issues and supports assurance of simulation fairness. Finally, we proposed a flexible framework to enable integration of multiple behavior modeling components that model working memory phenomena with different degrees of sophistication.« less

A portable MPI-based parallel vector template library

NASA Technical Reports Server (NTRS)

Sheffler, Thomas J.

1995-01-01

This paper discusses the design and implementation of a polymorphic collection library for distributed address-space parallel computers. The library provides a data-parallel programming model for C++ by providing three main components: a single generic collection class, generic algorithms over collections, and generic algebraic combining functions. Collection elements are the fourth component of a program written using the library and may be either of the built-in types of C or of user-defined types. Many ideas are borrowed from the Standard Template Library (STL) of C++, although a restricted programming model is proposed because of the distributed address-space memory model assumed. Whereas the STL provides standard collections and implementations of algorithms for uniprocessors, this paper advocates standardizing interfaces that may be customized for different parallel computers. Just as the STL attempts to increase programmer productivity through code reuse, a similar standard for parallel computers could provide programmers with a standard set of algorithms portable across many different architectures. The efficacy of this approach is verified by examining performance data collected from an initial implementation of the library running on an IBM SP-2 and an Intel Paragon.

A Portable MPI-Based Parallel Vector Template Library

NASA Technical Reports Server (NTRS)

Sheffler, Thomas J.

1995-01-01

This paper discusses the design and implementation of a polymorphic collection library for distributed address-space parallel computers. The library provides a data-parallel programming model for C + + by providing three main components: a single generic collection class, generic algorithms over collections, and generic algebraic combining functions. Collection elements are the fourth component of a program written using the library and may be either of the built-in types of c or of user-defined types. Many ideas are borrowed from the Standard Template Library (STL) of C++, although a restricted programming model is proposed because of the distributed address-space memory model assumed. Whereas the STL provides standard collections and implementations of algorithms for uniprocessors, this paper advocates standardizing interfaces that may be customized for different parallel computers. Just as the STL attempts to increase programmer productivity through code reuse, a similar standard for parallel computers could provide programmers with a standard set of algorithms portable across many different architectures. The efficacy of this approach is verified by examining performance data collected from an initial implementation of the library running on an IBM SP-2 and an Intel Paragon.

Dust Dynamics in Protoplanetary Disks: Parallel Computing with PVM

NASA Astrophysics Data System (ADS)

de La Fuente Marcos, Carlos; Barge, Pierre; de La Fuente Marcos, Raúl

2002-03-01

We describe a parallel version of our high-order-accuracy particle-mesh code for the simulation of collisionless protoplanetary disks. We use this code to carry out a massively parallel, two-dimensional, time-dependent, numerical simulation, which includes dust particles, to study the potential role of large-scale, gaseous vortices in protoplanetary disks. This noncollisional problem is easy to parallelize on message-passing multicomputer architectures. We performed the simulations on a cache-coherent nonuniform memory access Origin 2000 machine, using both the parallel virtual machine (PVM) and message-passing interface (MPI) message-passing libraries. Our performance analysis suggests that, for our problem, PVM is about 25% faster than MPI. Using PVM and MPI made it possible to reduce CPU time and increase code performance. This allows for simulations with a large number of particles (N ~ 105-106) in reasonable CPU times. The performances of our implementation of the pa! rallel code on an Origin 2000 supercomputer are presented and discussed. They exhibit very good speedup behavior and low load unbalancing. Our results confirm that giant gaseous vortices can play a dominant role in giant planet formation.

SciSpark: Highly Interactive and Scalable Model Evaluation and Climate Metrics

NASA Astrophysics Data System (ADS)

Wilson, B. D.; Palamuttam, R. S.; Mogrovejo, R. M.; Whitehall, K. D.; Mattmann, C. A.; Verma, R.; Waliser, D. E.; Lee, H.

2015-12-01

Remote sensing data and climate model output are multi-dimensional arrays of massive sizes locked away in heterogeneous file formats (HDF5/4, NetCDF 3/4) and metadata models (HDF-EOS, CF) making it difficult to perform multi-stage, iterative science processing since each stage requires writing and reading data to and from disk. We are developing a lightning fast Big Data technology called SciSpark based on ApacheTM Spark under a NASA AIST grant (PI Mattmann). Spark implements the map-reduce paradigm for parallel computing on a cluster, but emphasizes in-memory computation, "spilling" to disk only as needed, and so outperforms the disk-based ApacheTM Hadoop by 100x in memory and by 10x on disk. SciSpark will enable scalable model evaluation by executing large-scale comparisons of A-Train satellite observations to model grids on a cluster of 10 to 1000 compute nodes. This 2nd generation capability for NASA's Regional Climate Model Evaluation System (RCMES) will compute simple climate metrics at interactive speeds, and extend to quite sophisticated iterative algorithms such as machine-learning based clustering of temperature PDFs, and even graph-based algorithms for searching for Mesocale Convective Complexes. We have implemented a parallel data ingest capability in which the user specifies desired variables (arrays) as several time-sorted lists of URL's (i.e. using OPeNDAP model.nc?varname, or local files). The specified variables are partitioned by time/space and then each Spark node pulls its bundle of arrays into memory to begin a computation pipeline. We also investigated the performance of several N-dim. array libraries (scala breeze, java jblas & netlib-java, and ND4J). We are currently developing science codes using ND4J and studying memory behavior on the JVM. On the pyspark side, many of our science codes already use the numpy and SciPy ecosystems. The talk will cover: the architecture of SciSpark, the design of the scientific RDD (sRDD) data structure, our efforts to integrate climate science algorithms in Python and Scala, parallel ingest and partitioning of A-Train satellite observations from HDF files and model grids from netCDF files, first parallel runs to compute comparison statistics and PDF's, and first metrics quantifying parallel speedups and memory & disk usage.

VASP-4096: a very high performance programmable device for digital media processing applications

NASA Astrophysics Data System (ADS)

Krikelis, Argy

2001-03-01

Over the past few years, technology drivers for microprocessors have changed significantly. Media data delivery and processing--such as telecommunications, networking, video processing, speech recognition and 3D graphics--is increasing in importance and will soon dominate the processing cycles consumed in computer-based systems. This paper presents the architecture of the VASP-4096 processor. VASP-4096 provides high media performance with low energy consumption by integrating associative SIMD parallel processing with embedded microprocessor technology. The major innovations in the VASP-4096 is the integration of thousands of processing units in a single chip that are capable of support software programmable high-performance mathematical functions as well as abstract data processing. In addition to 4096 processing units, VASP-4096 integrates on a single chip a RISC controller that is an implementation of the SPARC architecture, 128 Kbytes of Data Memory, and I/O interfaces. The SIMD processing in VASP-4096 implements the ASProCore architecture, which is a proprietary implementation of SIMD processing, operates at 266 MHz with program instructions issued by the RISC controller. The device also integrates a 64-bit synchronous main memory interface operating at 133 MHz (double-data rate), and a 64- bit 66 MHz PCI interface. VASP-4096, compared with other processors architectures that support media processing, offers true performance scalability, support for deterministic and non-deterministic data processing on a single device, and software programmability that can be re- used in future chip generations.

Concurrent Collections (CnC): A new approach to parallel programming

DOE Office of Scientific and Technical Information (OSTI.GOV)

Knobe, Kathleen

2010-05-07

A common approach in designing parallel languages is to provide some high level handles to manipulate the use of the parallel platform. This exposes some aspects of the target platform, for example, shared vs. distributed memory. It may expose some but not all types of parallelism, for example, data parallelism but not task parallelism. This approach must find a balance between the desire to provide a simple view for the domain expert and provide sufficient power for tuning. This is hard for any given architecture and harder if the language is to apply to a range of architectures. Either simplicitymore » or power is lost. Instead of viewing the language design problem as one of providing the programmer with high level handles, we view the problem as one of designing an interface. On one side of this interface is the programmer (domain expert) who knows the application but needs no knowledge of any aspects of the platform. On the other side of the interface is the performance expert (programmer or program) who demands maximal flexibility for optimizing the mapping to a wide range of target platforms (parallel / serial, shared / distributed, homogeneous / heterogeneous, etc.) but needs no knowledge of the domain. Concurrent Collections (CnC) is based on this separation of concerns. The talk will present CnC and its benefits. About the speaker. Kathleen Knobe has focused throughout her career on parallelism especially compiler technology, runtime system design and language design. She worked at Compass (aka Massachusetts Computer Associates) from 1980 to 1991 designing compilers for a wide range of parallel platforms for Thinking Machines, MasPar, Alliant, Numerix, and several government projects. In 1991 she decided to finish her education. After graduating from MIT in 1997, she joined Digital Equipment’s Cambridge Research Lab (CRL). She stayed through the DEC/Compaq/HP mergers and when CRL was acquired and absorbed by Intel. She currently works in the Software and Services Group / Technology Pathfinding and Innovation.« less

Concurrent Collections (CnC): A new approach to parallel programming

ScienceCinema

Knobe, Kathleen

2018-04-16

A common approach in designing parallel languages is to provide some high level handles to manipulate the use of the parallel platform. This exposes some aspects of the target platform, for example, shared vs. distributed memory. It may expose some but not all types of parallelism, for example, data parallelism but not task parallelism. This approach must find a balance between the desire to provide a simple view for the domain expert and provide sufficient power for tuning. This is hard for any given architecture and harder if the language is to apply to a range of architectures. Either simplicity or power is lost. Instead of viewing the language design problem as one of providing the programmer with high level handles, we view the problem as one of designing an interface. On one side of this interface is the programmer (domain expert) who knows the application but needs no knowledge of any aspects of the platform. On the other side of the interface is the performance expert (programmer or program) who demands maximal flexibility for optimizing the mapping to a wide range of target platforms (parallel / serial, shared / distributed, homogeneous / heterogeneous, etc.) but needs no knowledge of the domain. Concurrent Collections (CnC) is based on this separation of concerns. The talk will present CnC and its benefits. About the speaker. Kathleen Knobe has focused throughout her career on parallelism especially compiler technology, runtime system design and language design. She worked at Compass (aka Massachusetts Computer Associates) from 1980 to 1991 designing compilers for a wide range of parallel platforms for Thinking Machines, MasPar, Alliant, Numerix, and several government projects. In 1991 she decided to finish her education. After graduating from MIT in 1997, she joined Digital Equipment’s Cambridge Research Lab (CRL). She stayed through the DEC/Compaq/HP mergers and when CRL was acquired and absorbed by Intel. She currently works in the Software and Services Group / Technology Pathfinding and Innovation.

Parallelization and checkpointing of GPU applications through program transformation

DOE Office of Scientific and Technical Information (OSTI.GOV)

Solano-Quinde, Lizandro Damian

2012-01-01

GPUs have emerged as a powerful tool for accelerating general-purpose applications. The availability of programming languages that makes writing general-purpose applications for running on GPUs tractable have consolidated GPUs as an alternative for accelerating general purpose applications. Among the areas that have benefited from GPU acceleration are: signal and image processing, computational fluid dynamics, quantum chemistry, and, in general, the High Performance Computing (HPC) Industry. In order to continue to exploit higher levels of parallelism with GPUs, multi-GPU systems are gaining popularity. In this context, single-GPU applications are parallelized for running in multi-GPU systems. Furthermore, multi-GPU systems help to solvemore » the GPU memory limitation for applications with large application memory footprint. Parallelizing single-GPU applications has been approached by libraries that distribute the workload at runtime, however, they impose execution overhead and are not portable. On the other hand, on traditional CPU systems, parallelization has been approached through application transformation at pre-compile time, which enhances the application to distribute the workload at application level and does not have the issues of library-based approaches. Hence, a parallelization scheme for GPU systems based on application transformation is needed. Like any computing engine of today, reliability is also a concern in GPUs. GPUs are vulnerable to transient and permanent failures. Current checkpoint/restart techniques are not suitable for systems with GPUs. Checkpointing for GPU systems present new and interesting challenges, primarily due to the natural differences imposed by the hardware design, the memory subsystem architecture, the massive number of threads, and the limited amount of synchronization among threads. Therefore, a checkpoint/restart technique suitable for GPU systems is needed. The goal of this work is to exploit higher levels of parallelism and to develop support for application-level fault tolerance in applications using multiple GPUs. Our techniques reduce the burden of enhancing single-GPU applications to support these features. To achieve our goal, this work designs and implements a framework for enhancing a single-GPU OpenCL application through application transformation.« less

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ibrahim, Khaled Z.; Epifanovsky, Evgeny; Williams, Samuel W.

Coupled-cluster methods provide highly accurate models of molecular structure by explicit numerical calculation of tensors representing the correlation between electrons. These calculations are dominated by a sequence of tensor contractions, motivating the development of numerical libraries for such operations. While based on matrix-matrix multiplication, these libraries are specialized to exploit symmetries in the molecular structure and in electronic interactions, and thus reduce the size of the tensor representation and the complexity of contractions. The resulting algorithms are irregular and their parallelization has been previously achieved via the use of dynamic scheduling or specialized data decompositions. We introduce our efforts tomore » extend the Libtensor framework to work in the distributed memory environment in a scalable and energy efficient manner. We achieve up to 240 speedup compared with the best optimized shared memory implementation. We attain scalability to hundreds of thousands of compute cores on three distributed-memory architectures, (Cray XC30&XC40, BlueGene/Q), and on a heterogeneous GPU-CPU system (Cray XK7). As the bottlenecks shift from being compute-bound DGEMM's to communication-bound collectives as the size of the molecular system scales, we adopt two radically different parallelization approaches for handling load-imbalance. Nevertheless, we preserve a uni ed interface to both programming models to maintain the productivity of computational quantum chemists.« less

Energy scaling advantages of resistive memory crossbar based computation and its application to sparse coding

DOE Office of Scientific and Technical Information (OSTI.GOV)

Agarwal, Sapan; Quach, Tu -Thach; Parekh, Ojas

In this study, the exponential increase in data over the last decade presents a significant challenge to analytics efforts that seek to process and interpret such data for various applications. Neural-inspired computing approaches are being developed in order to leverage the computational properties of the analog, low-power data processing observed in biological systems. Analog resistive memory crossbars can perform a parallel read or a vector-matrix multiplication as well as a parallel write or a rank-1 update with high computational efficiency. For an N × N crossbar, these two kernels can be O(N) more energy efficient than a conventional digital memory-basedmore » architecture. If the read operation is noise limited, the energy to read a column can be independent of the crossbar size (O(1)). These two kernels form the basis of many neuromorphic algorithms such as image, text, and speech recognition. For instance, these kernels can be applied to a neural sparse coding algorithm to give an O(N) reduction in energy for the entire algorithm when run with finite precision. Sparse coding is a rich problem with a host of applications including computer vision, object tracking, and more generally unsupervised learning.« less

Energy scaling advantages of resistive memory crossbar based computation and its application to sparse coding

DOE PAGES

Agarwal, Sapan; Quach, Tu -Thach; Parekh, Ojas; ...

2016-01-06

In this study, the exponential increase in data over the last decade presents a significant challenge to analytics efforts that seek to process and interpret such data for various applications. Neural-inspired computing approaches are being developed in order to leverage the computational properties of the analog, low-power data processing observed in biological systems. Analog resistive memory crossbars can perform a parallel read or a vector-matrix multiplication as well as a parallel write or a rank-1 update with high computational efficiency. For an N × N crossbar, these two kernels can be O(N) more energy efficient than a conventional digital memory-basedmore » architecture. If the read operation is noise limited, the energy to read a column can be independent of the crossbar size (O(1)). These two kernels form the basis of many neuromorphic algorithms such as image, text, and speech recognition. For instance, these kernels can be applied to a neural sparse coding algorithm to give an O(N) reduction in energy for the entire algorithm when run with finite precision. Sparse coding is a rich problem with a host of applications including computer vision, object tracking, and more generally unsupervised learning.« less

Multi-variants synthesis of Petri nets for FPGA devices

NASA Astrophysics Data System (ADS)

Bukowiec, Arkadiusz; Doligalski, Michał

2015-09-01

There is presented new method of synthesis of application specific logic controllers for FPGA devices. The specification of control algorithm is made with use of control interpreted Petri net (PT type). It allows specifying parallel processes in easy way. The Petri net is decomposed into state-machine type subnets. In this case, each subnet represents one parallel process. For this purpose there are applied algorithms of coloring of Petri nets. There are presented two approaches of such decomposition: with doublers of macroplaces or with one global wait place. Next, subnets are implemented into two-level logic circuit of the controller. The levels of logic circuit are obtained as a result of its architectural decomposition. The first level combinational circuit is responsible for generation of next places and second level decoder is responsible for generation output symbols. There are worked out two variants of such circuits: with one shared operational memory or with many flexible distributed memories as a decoder. Variants of Petri net decomposition and structures of logic circuits can be combined together without any restrictions. It leads to existence of four variants of multi-variants synthesis.

Effects of Ordering Strategies and Programming Paradigms on Sparse Matrix Computations

NASA Technical Reports Server (NTRS)

Oliker, Leonid; Li, Xiaoye; Husbands, Parry; Biswas, Rupak; Biegel, Bryan (Technical Monitor)

2002-01-01

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.

GPU and APU computations of Finite Time Lyapunov Exponent fields

NASA Astrophysics Data System (ADS)

Conti, Christian; Rossinelli, Diego; Koumoutsakos, Petros

2012-03-01

We present GPU and APU accelerated computations of Finite-Time Lyapunov Exponent (FTLE) fields. The calculation of FTLEs is a computationally intensive process, as in order to obtain the sharp ridges associated with the Lagrangian Coherent Structures an extensive resampling of the flow field is required. The computational performance of this resampling is limited by the memory bandwidth of the underlying computer architecture. The present technique harnesses data-parallel execution of many-core architectures and relies on fast and accurate evaluations of moment conserving functions for the mesh to particle interpolations. We demonstrate how the computation of FTLEs can be efficiently performed on a GPU and on an APU through OpenCL and we report over one order of magnitude improvements over multi-threaded executions in FTLE computations of bluff body flows.

Memory, reasoning, and categorization: parallels and common mechanisms

PubMed Central

Hayes, Brett K.; Heit, Evan; Rotello, Caren M.

2014-01-01

Traditionally, memory, reasoning, and categorization have been treated as separate components of human cognition. We challenge this distinction, arguing that there is broad scope for crossover between the methods and theories developed for each task. The links between memory and reasoning are illustrated in a review of two lines of research. The first takes theoretical ideas (two-process accounts) and methodological tools (signal detection analysis, receiver operating characteristic curves) from memory research and applies them to important issues in reasoning research: relations between induction and deduction, and the belief bias effect. The second line of research introduces a task in which subjects make either memory or reasoning judgments for the same set of stimuli. Other than broader generalization for reasoning than memory, the results were similar for the two tasks, across a variety of experimental stimuli and manipulations. It was possible to simultaneously explain performance on both tasks within a single cognitive architecture, based on exemplar-based comparisons of similarity. The final sections explore evidence for empirical and processing links between inductive reasoning and categorization and between categorization and recognition. An important implication is that progress in all three of these fields will be expedited by further investigation of the many commonalities between these tasks. PMID:24987380

Memory, reasoning, and categorization: parallels and common mechanisms.

PubMed

Hayes, Brett K; Heit, Evan; Rotello, Caren M

2014-01-01

Traditionally, memory, reasoning, and categorization have been treated as separate components of human cognition. We challenge this distinction, arguing that there is broad scope for crossover between the methods and theories developed for each task. The links between memory and reasoning are illustrated in a review of two lines of research. The first takes theoretical ideas (two-process accounts) and methodological tools (signal detection analysis, receiver operating characteristic curves) from memory research and applies them to important issues in reasoning research: relations between induction and deduction, and the belief bias effect. The second line of research introduces a task in which subjects make either memory or reasoning judgments for the same set of stimuli. Other than broader generalization for reasoning than memory, the results were similar for the two tasks, across a variety of experimental stimuli and manipulations. It was possible to simultaneously explain performance on both tasks within a single cognitive architecture, based on exemplar-based comparisons of similarity. The final sections explore evidence for empirical and processing links between inductive reasoning and categorization and between categorization and recognition. An important implication is that progress in all three of these fields will be expedited by further investigation of the many commonalities between these tasks.

STEMsalabim: A high-performance computing cluster friendly code for scanning transmission electron microscopy image simulations of thin specimens.

PubMed

Oelerich, Jan Oliver; Duschek, Lennart; Belz, Jürgen; Beyer, Andreas; Baranovskii, Sergei D; Volz, Kerstin

2017-06-01

We present a new multislice code for the computer simulation of scanning transmission electron microscope (STEM) images based on the frozen lattice approximation. Unlike existing software packages, the code is optimized to perform well on highly parallelized computing clusters, combining distributed and shared memory architectures. This enables efficient calculation of large lateral scanning areas of the specimen within the frozen lattice approximation and fine-grained sweeps of parameter space. Copyright © 2017 Elsevier B.V. All rights reserved.

Efficient Synthesis of Graph Methods: a Dynamically Scheduled Architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Minutoli, Marco; Castellana, Vito G.; Tumeo, Antonino

RDF databases naturally map to a graph representation and employ languages, such as SPARQL, that implements queries as graph pattern matching routines. Graph methods exhibit an irregular behavior: they present unpredictable, fine-grained data accesses, and are synchronization inten- sive. Graph data structures expose large amounts of dy- namic parallelism, but are difficult to partition without gen- erating load unbalance. In this paper, we present a novel ar- chitecture to improve the synthesis of graph methods. Our design addresses the issues of these algorithms with two com- ponents: a Dynamic Task Scheduler (DTS), which reduces load unbalance and maximize resource utilization,more » and a Hi- erarchical Memory Interface controller (HMI), which pro- vides support for concurrent memory operations on multi- ported/multi-banked shared memories. We evaluate our ap- proach by generating the accelerators for a set of SPARQL queries from the Lehigh University Benchmark (LUBM). We first analyze the load unbalance of these queries, showing that execution time among tasks can differ even of order of magnitudes. We then synthesize the queries and com- pare the performance of the resulting accelerators against the current state of the art. Experimental results show that our solution provides a speedup over the serial implementa- tion close to the theoretical maximum and a speedup up to 3.45 over a baseline parallel implementation. We conclude our study by exploring the design space to achieve maximum memory channels utilization. The best design used at least three of the four memory channels for more than 90% of the execution time.« less

Parallelization of Lower-Upper Symmetric Gauss-Seidel Method for Chemically Reacting Flow

NASA Technical Reports Server (NTRS)

Yoon, Seokkwan; Jost, Gabriele; Chang, Sherry

2005-01-01

Development of technologies for exploration of the solar system has revived an interest in computational simulation of chemically reacting flows since planetary probe vehicles exhibit non-equilibrium phenomena during the atmospheric entry of a planet or a moon as well as the reentry to the Earth. Stability in combustion is essential for new propulsion systems. Numerical solution of real-gas flows often increases computational work by an order-of-magnitude compared to perfect gas flow partly because of the increased complexity of equations to solve. Recently, as part of Project Columbia, NASA has integrated a cluster of interconnected SGI Altix systems to provide a ten-fold increase in current supercomputing capacity that includes an SGI Origin system. Both the new and existing machines are based on cache coherent non-uniform memory access architecture. Lower-Upper Symmetric Gauss-Seidel (LU-SGS) relaxation method has been implemented into both perfect and real gas flow codes including Real-Gas Aerodynamic Simulator (RGAS). However, the vectorized RGAS code runs inefficiently on cache-based shared-memory machines such as SGI system. Parallelization of a Gauss-Seidel method is nontrivial due to its sequential nature. The LU-SGS method has been vectorized on an oblique plane in INS3D-LU code that has been one of the base codes for NAS Parallel benchmarks. The oblique plane has been called a hyperplane by computer scientists. It is straightforward to parallelize a Gauss-Seidel method by partitioning the hyperplanes once they are formed. Another way of parallelization is to schedule processors like a pipeline using software. Both hyperplane and pipeline methods have been implemented using openMP directives. The present paper reports the performance of the parallelized RGAS code on SGI Origin and Altix systems.

Intelligent holographic databases

NASA Astrophysics Data System (ADS)

Barbastathis, George

Memory is a key component of intelligence. In the human brain, physical structure and functionality jointly provide diverse memory modalities at multiple time scales. How could we engineer artificial memories with similar faculties? In this thesis, we attack both hardware and algorithmic aspects of this problem. A good part is devoted to holographic memory architectures, because they meet high capacity and parallelism requirements. We develop and fully characterize shift multiplexing, a novel storage method that simplifies disk head design for holographic disks. We develop and optimize the design of compact refreshable holographic random access memories, showing several ways that 1 Tbit can be stored holographically in volume less than 1 m3, with surface density more than 20 times higher than conventional silicon DRAM integrated circuits. To address the issue of photorefractive volatility, we further develop the two-lambda (dual wavelength) method for shift multiplexing, and combine electrical fixing with angle multiplexing to demonstrate 1,000 multiplexed fixed holograms. Finally, we propose a noise model and an information theoretic metric to optimize the imaging system of a holographic memory, in terms of storage density and error rate. Motivated by the problem of interfacing sensors and memories to a complex system with limited computational resources, we construct a computer game of Desert Survival, built as a high-dimensional non-stationary virtual environment in a competitive setting. The efficacy of episodic learning, implemented as a reinforced Nearest Neighbor scheme, and the probability of winning against a control opponent improve significantly by concentrating the algorithmic effort to the virtual desert neighborhood that emerges as most significant at any time. The generalized computational model combines the autonomous neural network and von Neumann paradigms through a compact, dynamic central representation, which contains the most salient features of the sensory inputs, fused with relevant recollections, reminiscent of the hypothesized cognitive function of awareness. The Declarative Memory is searched both by content and address, suggesting a holographic implementation. The proposed computer architecture may lead to a novel paradigm that solves 'hard' cognitive problems at low cost.

HTM Spatial Pooler With Memristor Crossbar Circuits for Sparse Biometric Recognition.

PubMed

James, Alex Pappachen; Fedorova, Irina; Ibrayev, Timur; Kudithipudi, Dhireesha

2017-06-01

Hierarchical Temporal Memory (HTM) is an online machine learning algorithm that emulates the neo-cortex. The development of a scalable on-chip HTM architecture is an open research area. The two core substructures of HTM are spatial pooler and temporal memory. In this work, we propose a new Spatial Pooler circuit design with parallel memristive crossbar arrays for the 2D columns. The proposed design was validated on two different benchmark datasets, face recognition, and speech recognition. The circuits are simulated and analyzed using a practical memristor device model and 0.18 μm IBM CMOS technology model. The databases AR, YALE, ORL, and UFI, are used to test the performance of the design in face recognition. TIMIT dataset is used for the speech recognition.

High-Throughput, Adaptive FFT Architecture for FPGA-Based Spaceborne Data Processors

NASA Technical Reports Server (NTRS)

NguyenKobayashi, Kayla; Zheng, Jason X.; He, Yutao; Shah, Biren N.

2011-01-01

Exponential growth in microelectronics technology such as field-programmable gate arrays (FPGAs) has enabled high-performance spaceborne instruments with increasing onboard data processing capabilities. As a commonly used digital signal processing (DSP) building block, fast Fourier transform (FFT) has been of great interest in onboard data processing applications, which needs to strike a reasonable balance between high-performance (throughput, block size, etc.) and low resource usage (power, silicon footprint, etc.). It is also desirable to be designed so that a single design can be reused and adapted into instruments with different requirements. The Multi-Pass Wide Kernel FFT (MPWK-FFT) architecture was developed, in which the high-throughput benefits of the parallel FFT structure and the low resource usage of Singleton s single butterfly method is exploited. The result is a wide-kernel, multipass, adaptive FFT architecture. The 32K-point MPWK-FFT architecture includes 32 radix-2 butterflies, 64 FIFOs to store the real inputs, 64 FIFOs to store the imaginary inputs, complex twiddle factor storage, and FIFO logic to route the outputs to the correct FIFO. The inputs are stored in sequential fashion into the FIFOs, and the outputs of each butterfly are sequentially written first into the even FIFO, then the odd FIFO. Because of the order of the outputs written into the FIFOs, the depth of the even FIFOs, which are 768 each, are 1.5 times larger than the odd FIFOs, which are 512 each. The total memory needed for data storage, assuming that each sample is 36 bits, is 2.95 Mbits. The twiddle factors are stored in internal ROM inside the FPGA for fast access time. The total memory size to store the twiddle factors is 589.9Kbits. This FFT structure combines the benefits of high throughput from the parallel FFT kernels and low resource usage from the multi-pass FFT kernels with desired adaptability. Space instrument missions that need onboard FFT capabilities such as the proposed DESDynl, SWOT (Surface Water Ocean Topography), and Europa sounding radar missions would greatly benefit from this technology with significant reductions in non-recurring cost and risk.

Architecture-Adaptive Computing Environment: A Tool for Teaching Parallel Programming

NASA Technical Reports Server (NTRS)

Dorband, John E.; Aburdene, Maurice F.

2002-01-01

Recently, networked and cluster computation have become very popular. This paper is an introduction to a new C based parallel language for architecture-adaptive programming, aCe C. The primary purpose of aCe (Architecture-adaptive Computing Environment) is to encourage programmers to implement applications on parallel architectures by providing them the assurance that future architectures will be able to run their applications with a minimum of modification. A secondary purpose is to encourage computer architects to develop new types of architectures by providing an easily implemented software development environment and a library of test applications. This new language should be an ideal tool to teach parallel programming. In this paper, we will focus on some fundamental features of aCe C.

An 81.6 μW FastICA processor for epileptic seizure detection.

PubMed

Yang, Chia-Hsiang; Shih, Yi-Hsin; Chiueh, Herming

2015-02-01

To improve the performance of epileptic seizure detection, independent component analysis (ICA) is applied to multi-channel signals to separate artifacts and signals of interest. FastICA is an efficient algorithm to compute ICA. To reduce the energy dissipation, eigenvalue decomposition (EVD) is utilized in the preprocessing stage to reduce the convergence time of iterative calculation of ICA components. EVD is computed efficiently through an array structure of processing elements running in parallel. Area-efficient EVD architecture is realized by leveraging the approximate Jacobi algorithm, leading to a 77.2% area reduction. By choosing proper memory element and reduced wordlength, the power and area of storage memory are reduced by 95.6% and 51.7%, respectively. The chip area is minimized through fixed-point implementation and architectural transformations. Given a latency constraint of 0.1 s, an 86.5% area reduction is achieved compared to the direct-mapped architecture. Fabricated in 90 nm CMOS, the core area of the chip is 0.40 mm(2). The FastICA processor, part of an integrated epileptic control SoC, dissipates 81.6 μW at 0.32 V. The computation delay of a frame of 256 samples for 8 channels is 84.2 ms. Compared to prior work, 0.5% power dissipation, 26.7% silicon area, and 3.4 × computation speedup are achieved. The performance of the chip was verified by human dataset.

Embedded ensemble propagation for improving performance, portability, and scalability of uncertainty quantification on emerging computational architectures

DOE PAGES

Phipps, Eric T.; D'Elia, Marta; Edwards, Harold C.; ...

2017-04-18

In this study, quantifying simulation uncertainties is a critical component of rigorous predictive simulation. A key component of this is forward propagation of uncertainties in simulation input data to output quantities of interest. Typical approaches involve repeated sampling of the simulation over the uncertain input data, and can require numerous samples when accurately propagating uncertainties from large numbers of sources. Often simulation processes from sample to sample are similar and much of the data generated from each sample evaluation could be reused. We explore a new method for implementing sampling methods that simultaneously propagates groups of samples together in anmore » embedded fashion, which we call embedded ensemble propagation. We show how this approach takes advantage of properties of modern computer architectures to improve performance by enabling reuse between samples, reducing memory bandwidth requirements, improving memory access patterns, improving opportunities for fine-grained parallelization, and reducing communication costs. We describe a software technique for implementing embedded ensemble propagation based on the use of C++ templates and describe its integration with various scientific computing libraries within Trilinos. We demonstrate improved performance, portability and scalability for the approach applied to the simulation of partial differential equations on a variety of CPU, GPU, and accelerator architectures, including up to 131,072 cores on a Cray XK7 (Titan).« less

Software Design for Real-Time Systems on Parallel Computers: Formal Specifications.

DTIC Science & Technology

1996-04-01

This research investigated the important issues related to the analysis and design of real - time systems targeted to parallel architectures. In...particular, the software specification models for real - time systems on parallel architectures were evaluated. A survey of current formal methods for...uniprocessor real - time systems specifications was conducted to determine their extensibility in specifying real - time systems on parallel architectures. In

Parallel Simulation of Unsteady Turbulent Flames

NASA Technical Reports Server (NTRS)

Menon, Suresh

1996-01-01

Time-accurate simulation of turbulent flames in high Reynolds number flows is a challenging task since both fluid dynamics and combustion must be modeled accurately. To numerically simulate this phenomenon, very large computer resources (both time and memory) are required. Although current vector supercomputers are capable of providing adequate resources for simulations of this nature, the high cost and their limited availability, makes practical use of such machines less than satisfactory. At the same time, the explicit time integration algorithms used in unsteady flow simulations often possess a very high degree of parallelism, making them very amenable to efficient implementation on large-scale parallel computers. Under these circumstances, distributed memory parallel computers offer an excellent near-term solution for greatly increased computational speed and memory, at a cost that may render the unsteady simulations of the type discussed above more feasible and affordable.This paper discusses the study of unsteady turbulent flames using a simulation algorithm that is capable of retaining high parallel efficiency on distributed memory parallel architectures. Numerical studies are carried out using large-eddy simulation (LES). In LES, the scales larger than the grid are computed using a time- and space-accurate scheme, while the unresolved small scales are modeled using eddy viscosity based subgrid models. This is acceptable for the moment/energy closure since the small scales primarily provide a dissipative mechanism for the energy transferred from the large scales. However, for combustion to occur, the species must first undergo mixing at the small scales and then come into molecular contact. Therefore, global models cannot be used. Recently, a new model for turbulent combustion was developed, in which the combustion is modeled, within the subgrid (small-scales) using a methodology that simulates the mixing and the molecular transport and the chemical kinetics within each LES grid cell. Finite-rate kinetics can be included without any closure and this approach actually provides a means to predict the turbulent rates and the turbulent flame speed. The subgrid combustion model requires resolution of the local time scales associated with small-scale mixing, molecular diffusion and chemical kinetics and, therefore, within each grid cell, a significant amount of computations must be carried out before the large-scale (LES resolved) effects are incorporated. Therefore, this approach is uniquely suited for parallel processing and has been implemented on various systems such as: Intel Paragon, IBM SP-2, Cray T3D and SGI Power Challenge (PC) using the system independent Message Passing Interface (MPI) compiler. In this paper, timing data on these machines is reported along with some characteristic results.

Parallelized multi–graphics processing unit framework for high-speed Gabor-domain optical coherence microscopy

PubMed Central

Tankam, Patrice; Santhanam, Anand P.; Lee, Kye-Sung; Won, Jungeun; Canavesi, Cristina; Rolland, Jannick P.

2014-01-01

Abstract. Gabor-domain optical coherence microscopy (GD-OCM) is a volumetric high-resolution technique capable of acquiring three-dimensional (3-D) skin images with histological resolution. Real-time image processing is needed to enable GD-OCM imaging in a clinical setting. We present a parallelized and scalable multi-graphics processing unit (GPU) computing framework for real-time GD-OCM image processing. A parallelized control mechanism was developed to individually assign computation tasks to each of the GPUs. For each GPU, the optimal number of amplitude-scans (A-scans) to be processed in parallel was selected to maximize GPU memory usage and core throughput. We investigated five computing architectures for computational speed-up in processing 1000×1000 A-scans. The proposed parallelized multi-GPU computing framework enables processing at a computational speed faster than the GD-OCM image acquisition, thereby facilitating high-speed GD-OCM imaging in a clinical setting. Using two parallelized GPUs, the image processing of a 1×1×0.6  mm3 skin sample was performed in about 13 s, and the performance was benchmarked at 6.5 s with four GPUs. This work thus demonstrates that 3-D GD-OCM data may be displayed in real-time to the examiner using parallelized GPU processing. PMID:24695868

Parallelized multi-graphics processing unit framework for high-speed Gabor-domain optical coherence microscopy.

PubMed

Tankam, Patrice; Santhanam, Anand P; Lee, Kye-Sung; Won, Jungeun; Canavesi, Cristina; Rolland, Jannick P

2014-07-01

Gabor-domain optical coherence microscopy (GD-OCM) is a volumetric high-resolution technique capable of acquiring three-dimensional (3-D) skin images with histological resolution. Real-time image processing is needed to enable GD-OCM imaging in a clinical setting. We present a parallelized and scalable multi-graphics processing unit (GPU) computing framework for real-time GD-OCM image processing. A parallelized control mechanism was developed to individually assign computation tasks to each of the GPUs. For each GPU, the optimal number of amplitude-scans (A-scans) to be processed in parallel was selected to maximize GPU memory usage and core throughput. We investigated five computing architectures for computational speed-up in processing 1000×1000 A-scans. The proposed parallelized multi-GPU computing framework enables processing at a computational speed faster than the GD-OCM image acquisition, thereby facilitating high-speed GD-OCM imaging in a clinical setting. Using two parallelized GPUs, the image processing of a 1×1×0.6  mm3 skin sample was performed in about 13 s, and the performance was benchmarked at 6.5 s with four GPUs. This work thus demonstrates that 3-D GD-OCM data may be displayed in real-time to the examiner using parallelized GPU processing.

Considerations on the Use of Custom Accelerators for Big Data Analytics

DOE Office of Scientific and Technical Information (OSTI.GOV)

Castellana, Vito G.; Tumeo, Antonino; Minutoli, Marco

Accelerators, including Graphic Processing Units (GPUs) for gen- eral purpose computation, many-core designs with wide vector units (e.g., Intel Phi), have become a common component of many high performance clusters. The appearance of more stable and reliable tools tools that can automatically convert code written in high-level specifications with annotations (such as C or C++) to hardware de- scription languages (High-Level Synthesis - HLS), is also setting the stage for a broader use of reconfigurable devices (e.g., Field Pro- grammable Gate Arrays - FPGAs) in high performance system for the implementation of custom accelerators, helped by the fact that newmore » processors include advanced cache-coherent interconnects for these components. In this chapter, we briefly survey the status of the use of accelerators in high performance systems targeted at big data analytics applications. We argue that, although the progress in the use of accelerators for this class of applications has been sig- nificant, differently from scientific simulations there still are gaps to close. This is particularly true for the ”irregular” behaviors exhibited by no-SQL, graph databases. We focus our attention on the limits of HLS tools for data analytics and graph methods, and discuss a new architectural template that better fits the requirement of this class of applications. We validate the new architectural templates by mod- ifying the Graph Engine for Multithreaded System (GEMS) frame- work to support accelerators generated with such a methodology, and testing with queries coming from the Lehigh University Benchmark (LUBM). The architectural template enables better supporting the task and memory level parallelism present in graph methods by sup- porting a new control model and a enhanced memory interface. We show that out solution allows generating parallel accelerators, pro- viding speed ups with respect to conventional HLS flows. We finally draw conclusions and present a perspective on the use of reconfig- urable devices and Design Automation tools for data analytics.« less

Parallel Architectures and Parallel Algorithms for Integrated Vision Systems. Ph.D. Thesis

NASA Technical Reports Server (NTRS)

Choudhary, Alok Nidhi

1989-01-01

Computer vision is regarded as one of the most complex and computationally intensive problems. An integrated vision system (IVS) is a system that uses vision algorithms from all levels of processing to perform for a high level application (e.g., object recognition). An IVS normally involves algorithms from low level, intermediate level, and high level vision. Designing parallel architectures for vision systems is of tremendous interest to researchers. Several issues are addressed in parallel architectures and parallel algorithms for integrated vision systems.

Parallel machine architecture and compiler design facilities

NASA Technical Reports Server (NTRS)

Kuck, David J.; Yew, Pen-Chung; Padua, David; Sameh, Ahmed; Veidenbaum, Alex

1990-01-01

The objective is to provide an integrated simulation environment for studying and evaluating various issues in designing parallel systems, including machine architectures, parallelizing compiler techniques, and parallel algorithms. The status of Delta project (which objective is to provide a facility to allow rapid prototyping of parallelized compilers that can target toward different machine architectures) is summarized. Included are the surveys of the program manipulation tools developed, the environmental software supporting Delta, and the compiler research projects in which Delta has played a role.

Metal oxide resistive random access memory based synaptic devices for brain-inspired computing

NASA Astrophysics Data System (ADS)

Gao, Bin; Kang, Jinfeng; Zhou, Zheng; Chen, Zhe; Huang, Peng; Liu, Lifeng; Liu, Xiaoyan

2016-04-01

The traditional Boolean computing paradigm based on the von Neumann architecture is facing great challenges for future information technology applications such as big data, the Internet of Things (IoT), and wearable devices, due to the limited processing capability issues such as binary data storage and computing, non-parallel data processing, and the buses requirement between memory units and logic units. The brain-inspired neuromorphic computing paradigm is believed to be one of the promising solutions for realizing more complex functions with a lower cost. To perform such brain-inspired computing with a low cost and low power consumption, novel devices for use as electronic synapses are needed. Metal oxide resistive random access memory (ReRAM) devices have emerged as the leading candidate for electronic synapses. This paper comprehensively addresses the recent work on the design and optimization of metal oxide ReRAM-based synaptic devices. A performance enhancement methodology and optimized operation scheme to achieve analog resistive switching and low-energy training behavior are provided. A three-dimensional vertical synapse network architecture is proposed for high-density integration and low-cost fabrication. The impacts of the ReRAM synaptic device features on the performances of neuromorphic systems are also discussed on the basis of a constructed neuromorphic visual system with a pattern recognition function. Possible solutions to achieve the high recognition accuracy and efficiency of neuromorphic systems are presented.

Aho-Corasick String Matching on Shared and Distributed Memory Parallel Architectures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Tumeo, Antonino; Villa, Oreste; Chavarría-Miranda, Daniel

String matching is at the core of many critical applications, including network intrusion detection systems, search engines, virus scanners, spam filters, DNA and protein sequencing, and data mining. For all of these applications string matching requires a combination of (sometimes all) the following characteristics: high and/or predictable performance, support for large data sets and flexibility of integration and customization. Many software based implementations targeting conventional cache-based microprocessors fail to achieve high and predictable performance requirements, while Field-Programmable Gate Array (FPGA) implementations and dedicated hardware solutions fail to support large data sets (dictionary sizes) and are difficult to integrate and customize.more » The advent of multicore, multithreaded, and GPU-based systems is opening the possibility for software based solutions to reach very high performance at a sustained rate. This paper compares several software-based implementations of the Aho-Corasick string searching algorithm for high performance systems. We discuss the implementation of the algorithm on several types of shared-memory high-performance architectures (Niagara 2, large x86 SMPs and Cray XMT), distributed memory with homogeneous processing elements (InfiniBand cluster of x86 multicores) and heterogeneous processing elements (InfiniBand cluster of x86 multicores with NVIDIA Tesla C10 GPUs). We describe in detail how each solution achieves the objectives of supporting large dictionaries, sustaining high performance, and enabling customization and flexibility using various data sets.« less

Geospace simulations on the Cell BE processor

NASA Astrophysics Data System (ADS)

Germaschewski, K.; Raeder, J.; Larson, D.

2008-12-01

OpenGGCM (Open Geospace General circulation Model) is an established numerical code that simulates the Earth's space environment. The most computing intensive part is the MHD (magnetohydrodynamics) solver that models the plasma surrounding Earth and its interaction with Earth's magnetic field and the solar wind flowing in from the sun. Like other global magnetosphere codes, OpenGGCM's realism is limited by computational constraints on grid resolution. We investigate porting of the MHD solver to the Cell BE architecture, a novel inhomogeneous multicore architecture capable of up to 230 GFlops per processor. Realizing this high performance on the Cell processor is a programming challenge, though. We implemented the MHD solver using a multi-level parallel approach: On the coarsest level, the problem is distributed to processors based upon the usual domain decomposition approach. Then, on each processor, the problem is divided into 3D columns, each of which is handled by the memory limited SPEs (synergistic processing elements) slice by slice. Finally, SIMD instructions are used to fully exploit the vector/SIMD FPUs in each SPE. Memory management needs to be handled explicitly by the code, using DMA to move data from main memory to the per-SPE local store and vice versa. We obtained excellent performance numbers, a speed-up of a factor of 25 compared to just using the main processor, while still keeping the numerical implementation details of the code maintainable.

PCM-Based Durable Write Cache for Fast Disk I/O

DOE Office of Scientific and Technical Information (OSTI.GOV)

Liu, Zhuo; Wang, Bin; Carpenter, Patrick

2012-01-01

Flash based solid-state devices (FSSDs) have been adopted within the memory hierarchy to improve the performance of hard disk drive (HDD) based storage system. However, with the fast development of storage-class memories, new storage technologies with better performance and higher write endurance than FSSDs are emerging, e.g., phase-change memory (PCM). Understanding how to leverage these state-of-the-art storage technologies for modern computing systems is important to solve challenging data intensive computing problems. In this paper, we propose to leverage PCM for a hybrid PCM-HDD storage architecture. We identify the limitations of traditional LRU caching algorithms for PCM-based caches, and develop amore » novel hash-based write caching scheme called HALO to improve random write performance of hard disks. To address the limited durability of PCM devices and solve the degraded spatial locality in traditional wear-leveling techniques, we further propose novel PCM management algorithms that provide effective wear-leveling while maximizing access parallelism. We have evaluated this PCM-based hybrid storage architecture using applications with a diverse set of I/O access patterns. Our experimental results demonstrate that the HALO caching scheme leads to an average reduction of 36.8% in execution time compared to the LRU caching scheme, and that the SFC wear leveling extends the lifetime of PCM by a factor of 21.6.« less

Parallel multireference configuration interaction calculations on mini-β-carotenes and β-carotene

NASA Astrophysics Data System (ADS)

Kleinschmidt, Martin; Marian, Christel M.; Waletzke, Mirko; Grimme, Stefan

2009-01-01

We present a parallelized version of a direct selecting multireference configuration interaction (MRCI) code [S. Grimme and M. Waletzke, J. Chem. Phys. 111, 5645 (1999)]. The program can be run either in ab initio mode or as semiempirical procedure combined with density functional theory (DFT/MRCI). We have investigated the efficiency of the parallelization in case studies on carotenoids and porphyrins. The performance is found to depend heavily on the cluster architecture. While the speed-up on the older Intel Netburst technology is close to linear for up to 12-16 processes, our results indicate that it is not favorable to use all cores of modern Intel Dual Core or Quad Core processors simultaneously for memory intensive tasks. Due to saturation of the memory bandwidth, we recommend to run less demanding tasks on the latter architectures in parallel to two (Dual Core) or four (Quad Core) MRCI processes per node. The DFT/MRCI branch has been employed to study the low-lying singlet and triplet states of mini-n-β-carotenes (n =3, 5, 7, 9) and β-carotene (n =11) at the geometries of the ground state, the first excited triplet state, and the optically bright singlet state. The order of states depends heavily on the conjugation length and the nuclear geometry. The B1u+ state constitutes the S1 state in the vertical absorption spectrum of mini-3-β-carotene but switches order with the 2 A1g- state upon excited state relaxation. In the longer carotenes, near degeneracy or even root flipping between the B1u+ and B1u- states is observed whereas the 3 A1g- state is found to remain energetically above the optically bright B1u+ state at all nuclear geometries investigated here. The DFT/MRCI method is seen to underestimate the absolute excitation energies of the longer mini-β-carotenes but the energy gaps between the excited states are reproduced well. In addition to singlet data, triplet-triplet absorption energies are presented. For β-carotene, where these transition energies are known from experiment, excellent agreement with our calculations is observed.

Parallel multireference configuration interaction calculations on mini-beta-carotenes and beta-carotene.

PubMed

Kleinschmidt, Martin; Marian, Christel M; Waletzke, Mirko; Grimme, Stefan

2009-01-28

We present a parallelized version of a direct selecting multireference configuration interaction (MRCI) code [S. Grimme and M. Waletzke, J. Chem. Phys. 111, 5645 (1999)]. The program can be run either in ab initio mode or as semiempirical procedure combined with density functional theory (DFT/MRCI). We have investigated the efficiency of the parallelization in case studies on carotenoids and porphyrins. The performance is found to depend heavily on the cluster architecture. While the speed-up on the older Intel Netburst technology is close to linear for up to 12-16 processes, our results indicate that it is not favorable to use all cores of modern Intel Dual Core or Quad Core processors simultaneously for memory intensive tasks. Due to saturation of the memory bandwidth, we recommend to run less demanding tasks on the latter architectures in parallel to two (Dual Core) or four (Quad Core) MRCI processes per node. The DFT/MRCI branch has been employed to study the low-lying singlet and triplet states of mini-n-beta-carotenes (n=3, 5, 7, 9) and beta-carotene (n=11) at the geometries of the ground state, the first excited triplet state, and the optically bright singlet state. The order of states depends heavily on the conjugation length and the nuclear geometry. The (1)B(u) (+) state constitutes the S(1) state in the vertical absorption spectrum of mini-3-beta-carotene but switches order with the 2 (1)A(g) (-) state upon excited state relaxation. In the longer carotenes, near degeneracy or even root flipping between the (1)B(u) (+) and (1)B(u) (-) states is observed whereas the 3 (1)A(g) (-) state is found to remain energetically above the optically bright (1)B(u) (+) state at all nuclear geometries investigated here. The DFT/MRCI method is seen to underestimate the absolute excitation energies of the longer mini-beta-carotenes but the energy gaps between the excited states are reproduced well. In addition to singlet data, triplet-triplet absorption energies are presented. For beta-carotene, where these transition energies are known from experiment, excellent agreement with our calculations is observed.

Cognitive Architectures for Multimedia Learning

ERIC Educational Resources Information Center

Reed, Stephen K.

2006-01-01

This article provides a tutorial overview of cognitive architectures that can form a theoretical foundation for designing multimedia instruction. Cognitive architectures include a description of memory stores, memory codes, and cognitive operations. Architectures that are relevant to multimedia learning include Paivio's dual coding theory,…

The path toward HEP High Performance Computing

NASA Astrophysics Data System (ADS)

Apostolakis, John; Brun, René; Carminati, Federico; Gheata, Andrei; Wenzel, Sandro

2014-06-01

High Energy Physics code has been known for making poor use of high performance computing architectures. Efforts in optimising HEP code on vector and RISC architectures have yield limited results and recent studies have shown that, on modern architectures, it achieves a performance between 10% and 50% of the peak one. Although several successful attempts have been made to port selected codes on GPUs, no major HEP code suite has a "High Performance" implementation. With LHC undergoing a major upgrade and a number of challenging experiments on the drawing board, HEP cannot any longer neglect the less-than-optimal performance of its code and it has to try making the best usage of the hardware. This activity is one of the foci of the SFT group at CERN, which hosts, among others, the Root and Geant4 project. The activity of the experiments is shared and coordinated via a Concurrency Forum, where the experience in optimising HEP code is presented and discussed. Another activity is the Geant-V project, centred on the development of a highperformance prototype for particle transport. Achieving a good concurrency level on the emerging parallel architectures without a complete redesign of the framework can only be done by parallelizing at event level, or with a much larger effort at track level. Apart the shareable data structures, this typically implies a multiplication factor in terms of memory consumption compared to the single threaded version, together with sub-optimal handling of event processing tails. Besides this, the low level instruction pipelining of modern processors cannot be used efficiently to speedup the program. We have implemented a framework that allows scheduling vectors of particles to an arbitrary number of computing resources in a fine grain parallel approach. The talk will review the current optimisation activities within the SFT group with a particular emphasis on the development perspectives towards a simulation framework able to profit best from the recent technology evolution in computing.

Data Movement Dominates: Advanced Memory Technology to Address the Real Exascale Power Problem

DOE Office of Scientific and Technical Information (OSTI.GOV)

Bergman, Keren

Energy is the fundamental barrier to Exascale supercomputing and is dominated by the cost of moving data from one point to another, not computation. Similarly, performance is dominated by data movement, not computation. The solution to this problem requires three critical technologies: 3D integration, optical chip-to-chip communication, and a new communication model. The central goal of the Sandia led "Data Movement Dominates" project aimed to develop memory systems and new architectures based on these technologies that have the potential to lower the cost of local memory accesses by orders of magnitude and provide substantially more bandwidth. Only through these transformationalmore » advances can future systems reach the goals of Exascale computing with a manageable power budgets. The Sandia led team included co-PIs from Columbia University, Lawrence Berkeley Lab, and the University of Maryland. The Columbia effort of Data Movement Dominates focused on developing a physically accurate simulation environment and experimental verification for optically-connected memory (OCM) systems that can enable continued performance scaling through high-bandwidth capacity, energy-efficient bit-rate transparency, and time-of-flight latency. With OCM, memory device parallelism and total capacity can scale to match future high-performance computing requirements without sacrificing data-movement efficiency. When we consider systems with integrated photonics, links to memory can be seamlessly integrated with the interconnection network-in a sense, memory becomes a primary aspect of the interconnection network. At the core of the Columbia effort, toward expanding our understanding of OCM enabled computing we have created an integrated modeling and simulation environment that uniquely integrates the physical behavior of the optical layer. The PhoenxSim suite of design and software tools developed under this effort has enabled the co-design of and performance evaluation photonics-enabled OCM architectures on Exascale computing systems.« less

Development of an Implicit, Charge and Energy Conserving 2D Electromagnetic PIC Code on Advanced Architectures

NASA Astrophysics Data System (ADS)

Payne, Joshua; Taitano, William; Knoll, Dana; Liebs, Chris; Murthy, Karthik; Feltman, Nicolas; Wang, Yijie; McCarthy, Colleen; Cieren, Emanuel

2012-10-01

In order to solve problems such as the ion coalescence and slow MHD shocks fully kinetically we developed a fully implicit 2D energy and charge conserving electromagnetic PIC code, PlasmaApp2D. PlasmaApp2D differs from previous implicit PIC implementations in that it will utilize advanced architectures such as GPUs and shared memory CPU systems, with problems too large to fit into cache. PlasmaApp2D will be a hybrid CPU-GPU code developed primarily to run on the DARWIN cluster at LANL utilizing four 12-core AMD Opteron CPUs and two NVIDIA Tesla GPUs per node. MPI will be used for cross-node communication, OpenMP will be used for on-node parallelism, and CUDA will be used for the GPUs. Development progress and initial results will be presented.

Exploring performance and energy tradeoffs for irregular applications: A case study on the Tilera many-core architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Panyala, Ajay; Chavarría-Miranda, Daniel; Manzano, Joseph B.

High performance, parallel applications with irregular data accesses are becoming a critical workload class for modern systems. In particular, the execution of such workloads on emerging many-core systems is expected to be a significant component of applications in data mining, machine learning, scientific computing and graph analytics. However, power and energy constraints limit the capabilities of individual cores, memory hierarchy and on-chip interconnect of such systems, thus leading to architectural and software trade-os that must be understood in the context of the intended application’s behavior. Irregular applications are notoriously hard to optimize given their data-dependent access patterns, lack of structuredmore » locality and complex data structures and code patterns. We have ported two irregular applications, graph community detection using the Louvain method (Grappolo) and high-performance conjugate gradient (HPCCG), to the Tilera many-core system and have conducted a detailed study of platform-independent and platform-specific optimizations that improve their performance as well as reduce their overall energy consumption. To conduct this study, we employ an auto-tuning based approach that explores the optimization design space along three dimensions - memory layout schemes, GCC compiler flag choices and OpenMP loop scheduling options. We leverage MIT’s OpenTuner auto-tuning framework to explore and recommend energy optimal choices for different combinations of parameters. We then conduct an in-depth architectural characterization to understand the memory behavior of the selected workloads. Finally, we perform a correlation study to demonstrate the interplay between the hardware behavior and application characteristics. Using auto-tuning, we demonstrate whole-node energy savings and performance improvements of up to 49:6% and 60% relative to a baseline instantiation, and up to 31% and 45:4% relative to manually optimized variants.« less

Geospace simulations using modern accelerator processor technology

NASA Astrophysics Data System (ADS)

Germaschewski, K.; Raeder, J.; Larson, D. J.

2009-12-01

OpenGGCM (Open Geospace General Circulation Model) is a well-established numerical code simulating the Earth's space environment. The most computing intensive part is the MHD (magnetohydrodynamics) solver that models the plasma surrounding Earth and its interaction with Earth's magnetic field and the solar wind flowing in from the sun. Like other global magnetosphere codes, OpenGGCM's realism is currently limited by computational constraints on grid resolution. OpenGGCM has been ported to make use of the added computational powerof modern accelerator based processor architectures, in particular the Cell processor. The Cell architecture is a novel inhomogeneous multicore architecture capable of achieving up to 230 GFLops on a single chip. The University of New Hampshire recently acquired a PowerXCell 8i based computing cluster, and here we will report initial performance results of OpenGGCM. Realizing the high theoretical performance of the Cell processor is a programming challenge, though. We implemented the MHD solver using a multi-level parallelization approach: On the coarsest level, the problem is distributed to processors based upon the usual domain decomposition approach. Then, on each processor, the problem is divided into 3D columns, each of which is handled by the memory limited SPEs (synergistic processing elements) slice by slice. Finally, SIMD instructions are used to fully exploit the SIMD FPUs in each SPE. Memory management needs to be handled explicitly by the code, using DMA to move data from main memory to the per-SPE local store and vice versa. We use a modern technique, automatic code generation, which shields the application programmer from having to deal with all of the implementation details just described, keeping the code much more easily maintainable. Our preliminary results indicate excellent performance, a speed-up of a factor of 30 compared to the unoptimized version.

Some Problems and Solutions in Transferring Ecosystem Simulation Codes to Supercomputers

NASA Technical Reports Server (NTRS)

Skiles, J. W.; Schulbach, C. H.

1994-01-01

Many computer codes for the simulation of ecological systems 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. Recent recognition of ecosystem science as a High Performance Computing and Communications Program Grand Challenge area emphasizes supercomputers (both parallel and distributed systems) as the next set of tools for ecological simulation. Transferring ecosystem simulation codes to such systems is not a matter of simply compiling and executing existing code on the supercomputer since there are significant differences in the system architectures of sequential, scalar computers and parallel and/or vector supercomputers. To more appropriately match the application to the architecture (necessary to achieve reasonable performance), the parallelism (if it exists) of the original application must be exploited. We discuss our work in transferring a general grassland simulation model (developed on a VAX in the FORTRAN computer programming language) to a Cray Y-MP. We show the Cray shared-memory vector-architecture, and discuss our rationale for selecting the Cray. We describe porting the model to the Cray and executing and verifying a baseline version, and we discuss the changes we made to exploit the parallelism in the application and to improve code execution. As a result, the Cray executed the model 30 times faster than the VAX 11/785 and 10 times faster than a Sun 4 workstation. We achieved an additional speed-up of approximately 30 percent over the original Cray run by using the compiler's vectorizing capabilities and the machine's ability to put subroutines and functions "in-line" in the code. With the modifications, the code still runs at only about 5% of the Cray's peak speed because it makes ineffective use of the vector processing capabilities of the Cray. We conclude with a discussion and future plans.

A Blocked Linear Method for Optimizing Large Parameter Sets in Variational Monte Carlo

DOE PAGES

Zhao, Luning; Neuscamman, Eric

2017-05-17

We present a modification to variational Monte Carlo’s linear method optimization scheme that addresses a critical memory bottleneck while maintaining compatibility with both the traditional ground state variational principle and our recently-introduced variational principle for excited states. For wave function ansatzes with tens of thousands of variables, our modification reduces the required memory per parallel process from tens of gigabytes to hundreds of megabytes, making the methodology a much better fit for modern supercomputer architectures in which data communication and per-process memory consumption are primary concerns. We verify the efficacy of the new optimization scheme in small molecule tests involvingmore » both the Hilbert space Jastrow antisymmetric geminal power ansatz and real space multi-Slater Jastrow expansions. Satisfied with its performance, we have added the optimizer to the QMCPACK software package, with which we demonstrate on a hydrogen ring a prototype approach for making systematically convergent, non-perturbative predictions of Mott-insulators’ optical band gaps.« less

Logic gates realized by nonvolatile GeTe/Sb2Te3 super lattice phase-change memory with a magnetic field input

NASA Astrophysics Data System (ADS)

Lu, Bin; Cheng, Xiaomin; Feng, Jinlong; Guan, Xiawei; Miao, Xiangshui

2016-07-01

Nonvolatile memory devices or circuits that can implement both storage and calculation are a crucial requirement for the efficiency improvement of modern computer. In this work, we realize logic functions by using [GeTe/Sb2Te3]n super lattice phase change memory (PCM) cell in which higher threshold voltage is needed for phase change with a magnetic field applied. First, the [GeTe/Sb2Te3]n super lattice cells were fabricated and the R-V curve was measured. Then we designed the logic circuits with the super lattice PCM cell verified by HSPICE simulation and experiments. Seven basic logic functions are first demonstrated in this letter; then several multi-input logic gates are presented. The proposed logic devices offer the advantages of simple structures and low power consumption, indicating that the super lattice PCM has the potential in the future nonvolatile central processing unit design, facilitating the development of massive parallel computing architecture.

Proceedings of the second SISAL users` conference

DOE Office of Scientific and Technical Information (OSTI.GOV)

Feo, J T; Frerking, C; Miller, P J

1992-12-01

This report contains papers on the following topics: A sisal code for computing the fourier transform on S{sub N}; five ways to fill your knapsack; simulating material dislocation motion in sisal; candis as an interface for sisal; parallelisation and performance of the burg algorithm on a shared-memory multiprocessor; use of genetic algorithm in sisal to solve the file design problem; implementing FFT`s in sisal; programming and evaluating the performance of signal processing applications in the sisal programming environment; sisal and Von Neumann-based languages: translation and intercommunication; an IF2 code generator for ADAM architecture; program partitioning for NUMA multiprocessor computer systems;more » mapping functional parallelism on distributed memory machines; implicit array copying: prevention is better than cure ; mathematical syntax for sisal; an approach for optimizing recursive functions; implementing arrays in sisal 2.0; Fol: an object oriented extension to the sisal language; twine: a portable, extensible sisal execution kernel; and investigating the memory performance of the optimizing sisal compiler.« less

Numerical study of the vortex tube reconnection using vortex particle method on many graphics cards

NASA Astrophysics Data System (ADS)

Kudela, Henryk; Kosior, Andrzej

2014-08-01

Vortex Particle Methods are one of the most convenient ways of tracking the vorticity evolution. In the article we presented numerical recreation of the real life experiment concerning head-on collision of two vortex rings. In the experiment the evolution and reconnection of the vortex structures is tracked with passive markers (paint particles) which in viscous fluid does not follow the evolution of vorticity field. In numerical computations we showed the difference between vorticity evolution and movement of passive markers. The agreement with the experiment was very good. Due to problems with very long time of computations on a single processor the Vortex-in-Cell method was implemented on the multicore architecture of the graphics cards (GPUs). Vortex Particle Methods are very well suited for parallel computations. As there are myriads of particles in the flow and for each of them the same equations of motion have to be solved the SIMD architecture used in GPUs seems to be perfect. The main disadvantage in this case is the small amount of the RAM memory. To overcome this problem we created a multiGPU implementation of the VIC method. Some remarks on parallel computing are given in the article.

On some Aitken-like acceleration of the Schwarz method

NASA Astrophysics Data System (ADS)

Garbey, M.; Tromeur-Dervout, D.

2002-12-01

In this paper we present a family of domain decomposition based on Aitken-like acceleration of the Schwarz method seen as an iterative procedure with a linear rate of convergence. We first present the so-called Aitken-Schwarz procedure for linear differential operators. The solver can be a direct solver when applied to the Helmholtz problem with five-point finite difference scheme on regular grids. We then introduce the Steffensen-Schwarz variant which is an iterative domain decomposition solver that can be applied to linear and nonlinear problems. We show that these solvers have reasonable numerical efficiency compared to classical fast solvers for the Poisson problem or multigrids for more general linear and nonlinear elliptic problems. However, the salient feature of our method is that our algorithm has high tolerance to slow network in the context of distributed parallel computing and is attractive, generally speaking, to use with computer architecture for which performance is limited by the memory bandwidth rather than the flop performance of the CPU. This is nowadays the case for most parallel. computer using the RISC processor architecture. We will illustrate this highly desirable property of our algorithm with large-scale computing experiments.

Medical image denoising via optimal implementation of non-local means on hybrid parallel architecture.

PubMed

Nguyen, Tuan-Anh; Nakib, Amir; Nguyen, Huy-Nam

2016-06-01

The Non-local means denoising filter has been established as gold standard for image denoising problem in general and particularly in medical imaging due to its efficiency. However, its computation time limited its applications in real world application, especially in medical imaging. In this paper, a distributed version on parallel hybrid architecture is proposed to solve the computation time problem and a new method to compute the filters' coefficients is also proposed, where we focused on the implementation and the enhancement of filters' parameters via taking the neighborhood of the current voxel more accurately into account. In terms of implementation, our key contribution consists in reducing the number of shared memory accesses. The different tests of the proposed method were performed on the brain-web database for different levels of noise. Performances and the sensitivity were quantified in terms of speedup, peak signal to noise ratio, execution time, the number of floating point operations. The obtained results demonstrate the efficiency of the proposed method. Moreover, the implementation is compared to that of other techniques, recently published in the literature. Copyright © 2016 Elsevier Ireland Ltd. All rights reserved.

Feasibility of using the Massively Parallel Processor for large eddy simulations and other Computational Fluid Dynamics applications

NASA Technical Reports Server (NTRS)

Bruno, John

1984-01-01

The results of an investigation into the feasibility of using the MPP for direct and large eddy simulations of the Navier-Stokes equations is presented. A major part of this study was devoted to the implementation of two of the standard numerical algorithms for CFD. These implementations were not run on the Massively Parallel Processor (MPP) since the machine delivered to NASA Goddard does not have sufficient capacity. Instead, a detailed implementation plan was designed and from these were derived estimates of the time and space requirements of the algorithms on a suitably configured MPP. In addition, other issues related to the practical implementation of these algorithms on an MPP-like architecture were considered; namely, adaptive grid generation, zonal boundary conditions, the table lookup problem, and the software interface. Performance estimates show that the architectural components of the MPP, the Staging Memory and the Array Unit, appear to be well suited to the numerical algorithms of CFD. This combined with the prospect of building a faster and larger MMP-like machine holds the promise of achieving sustained gigaflop rates that are required for the numerical simulations in CFD.

Cross-scale efficient tensor contractions for coupled cluster computations through multiple programming model backends

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ibrahim, Khaled Z.; Epifanovsky, Evgeny; Williams, Samuel

Coupled-cluster methods provide highly accurate models of molecular structure through explicit numerical calculation of tensors representing the correlation between electrons. These calculations are dominated by a sequence of tensor contractions, motivating the development of numerical libraries for such operations. While based on matrix–matrix multiplication, these libraries are specialized to exploit symmetries in the molecular structure and in electronic interactions, and thus reduce the size of the tensor representation and the complexity of contractions. The resulting algorithms are irregular and their parallelization has been previously achieved via the use of dynamic scheduling or specialized data decompositions. We introduce our efforts tomore » extend the Libtensor framework to work in the distributed memory environment in a scalable and energy-efficient manner. We achieve up to 240× speedup compared with the optimized shared memory implementation of Libtensor. We attain scalability to hundreds of thousands of compute cores on three distributed-memory architectures (Cray XC30 and XC40, and IBM Blue Gene/Q), and on a heterogeneous GPU-CPU system (Cray XK7). As the bottlenecks shift from being compute-bound DGEMM's to communication-bound collectives as the size of the molecular system scales, we adopt two radically different parallelization approaches for handling load-imbalance, tasking and bulk synchronous models. Nevertheless, we preserve a unified interface to both programming models to maintain the productivity of computational quantum chemists.« less

Cross-scale efficient tensor contractions for coupled cluster computations through multiple programming model backends

DOE PAGES

Ibrahim, Khaled Z.; Epifanovsky, Evgeny; Williams, Samuel; ...

2017-03-08

Coupled-cluster methods provide highly accurate models of molecular structure through explicit numerical calculation of tensors representing the correlation between electrons. These calculations are dominated by a sequence of tensor contractions, motivating the development of numerical libraries for such operations. While based on matrix–matrix multiplication, these libraries are specialized to exploit symmetries in the molecular structure and in electronic interactions, and thus reduce the size of the tensor representation and the complexity of contractions. The resulting algorithms are irregular and their parallelization has been previously achieved via the use of dynamic scheduling or specialized data decompositions. We introduce our efforts tomore » extend the Libtensor framework to work in the distributed memory environment in a scalable and energy-efficient manner. We achieve up to 240× speedup compared with the optimized shared memory implementation of Libtensor. We attain scalability to hundreds of thousands of compute cores on three distributed-memory architectures (Cray XC30 and XC40, and IBM Blue Gene/Q), and on a heterogeneous GPU-CPU system (Cray XK7). As the bottlenecks shift from being compute-bound DGEMM's to communication-bound collectives as the size of the molecular system scales, we adopt two radically different parallelization approaches for handling load-imbalance, tasking and bulk synchronous models. Nevertheless, we preserve a unified interface to both programming models to maintain the productivity of computational quantum chemists.« less

Kinetic Inductance Memory Cell and Architecture for Superconducting Computers

NASA Astrophysics Data System (ADS)

Chen, George J.

Josephson memory devices typically use a superconducting loop containing one or more Josephson junctions to store information. The magnetic inductance of the loop in conjunction with the Josephson junctions provides multiple states to store data. This thesis shows that replacing the magnetic inductor in a memory cell with a kinetic inductor can lead to a smaller cell size. However, magnetic control of the cells is lost. Thus, a current-injection based architecture for a memory array has been designed to work around this problem. The isolation between memory cells that magnetic control provides is provided through resistors in this new architecture. However, these resistors allow leakage current to flow which ultimately limits the size of the array due to power considerations. A kinetic inductance memory array will be limited to 4K bits with a read access time of 320 ps for a 1 um linewidth technology. If a power decoder could be developed, the memory architecture could serve as the blueprint for a fast (<1 ns), large scale (>1 Mbit) superconducting memory array.

Bit-parallel arithmetic in a massively-parallel associative processor

NASA Technical Reports Server (NTRS)

Scherson, Isaac D.; Kramer, David A.; Alleyne, Brian D.

1992-01-01

A simple but powerful new architecture based on a classical associative processor model is presented. Algorithms for performing the four basic arithmetic operations both for integer and floating point operands are described. For m-bit operands, the proposed architecture makes it possible to execute complex operations in O(m) cycles as opposed to O(m exp 2) for bit-serial machines. A word-parallel, bit-parallel, massively-parallel computing system can be constructed using this architecture with VLSI technology. The operation of this system is demonstrated for the fast Fourier transform and matrix multiplication.

Unobtrusive Software and System Health Management with R2U2 on a Parallel MIMD Coprocessor

NASA Technical Reports Server (NTRS)

Schumann, Johann; Moosbrugger, Patrick

2017-01-01

Dynamic monitoring of software and system health of a complex cyber-physical system requires observers that continuously monitor variables of the embedded software in order to detect anomalies and reason about root causes. There exists a variety of techniques for code instrumentation, but instrumentation might change runtime behavior and could require costly software re-certification. In this paper, we present R2U2E, a novel realization of our real-time, Realizable, Responsive, and Unobtrusive Unit (R2U2). The R2U2E observers are executed in parallel on a dedicated 16-core EPIPHANY co-processor, thereby avoiding additional computational overhead to the system under observation. A DMA-based shared memory access architecture allows R2U2E to operate without any code instrumentation or program interference.

Efficient partitioning and assignment on programs for multiprocessor execution

NASA Technical Reports Server (NTRS)

Standley, Hilda M.

1993-01-01

The general problem studied is that of segmenting or partitioning programs for distribution across a multiprocessor system. Efficient partitioning and the assignment of program elements are of great importance since the time consumed in this overhead activity may easily dominate the computation, effectively eliminating any gains made by the use of the parallelism. In this study, the partitioning of sequentially structured programs (written in FORTRAN) is evaluated. Heuristics, developed for similar applications are examined. Finally, a model for queueing networks with finite queues is developed which may be used to analyze multiprocessor system architectures with a shared memory approach to the problem of partitioning. The properties of sequentially written programs form obstacles to large scale (at the procedure or subroutine level) parallelization. Data dependencies of even the minutest nature, reflecting the sequential development of the program, severely limit parallelism. The design of heuristic algorithms is tied to the experience gained in the parallel splitting. Parallelism obtained through the physical separation of data has seen some success, especially at the data element level. Data parallelism on a grander scale requires models that accurately reflect the effects of blocking caused by finite queues. A model for the approximation of the performance of finite queueing networks is developed. This model makes use of the decomposition approach combined with the efficiency of product form solutions.

Optimizing Irregular Applications for Energy and Performance on the Tilera Many-core Architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chavarría-Miranda, Daniel; Panyala, Ajay R.; Halappanavar, Mahantesh

Optimizing applications simultaneously for energy and performance is a complex problem. High performance, parallel, irregular applications are notoriously hard to optimize due to their data-dependent memory accesses, lack of structured locality and complex data structures and code patterns. Irregular kernels are growing in importance in applications such as machine learning, graph analytics and combinatorial scientific computing. Performance- and energy-efficient implementation of these kernels on modern, energy efficient, multicore and many-core platforms is therefore an important and challenging problem. We present results from optimizing two irregular applications { the Louvain method for community detection (Grappolo), and high-performance conjugate gradient (HPCCG) {more » on the Tilera many-core system. We have significantly extended MIT's OpenTuner auto-tuning framework to conduct a detailed study of platform-independent and platform-specific optimizations to improve performance as well as reduce total energy consumption. We explore the optimization design space along three dimensions: memory layout schemes, compiler-based code transformations, and optimization of parallel loop schedules. Using auto-tuning, we demonstrate whole node energy savings of up to 41% relative to a baseline instantiation, and up to 31% relative to manually optimized variants.« less

Efficient parallel linear scaling construction of the density matrix for Born-Oppenheimer molecular dynamics.

PubMed

Mniszewski, S M; Cawkwell, M J; Wall, M E; Mohd-Yusof, J; Bock, N; Germann, T C; Niklasson, A M N

2015-10-13

We present an algorithm for the calculation of the density matrix that for insulators scales linearly with system size and parallelizes efficiently on multicore, shared memory platforms with small and controllable numerical errors. The algorithm is based on an implementation of the second-order spectral projection (SP2) algorithm [ Niklasson, A. M. N. Phys. Rev. B 2002 , 66 , 155115 ] in sparse matrix algebra with the ELLPACK-R data format. We illustrate the performance of the algorithm within self-consistent tight binding theory by total energy calculations of gas phase poly(ethylene) molecules and periodic liquid water systems containing up to 15,000 atoms on up to 16 CPU cores. We consider algorithm-specific performance aspects, such as local vs nonlocal memory access and the degree of matrix sparsity. Comparisons to sparse matrix algebra implementations using off-the-shelf libraries on multicore CPUs, graphics processing units (GPUs), and the Intel many integrated core (MIC) architecture are also presented. The accuracy and stability of the algorithm are illustrated with long duration Born-Oppenheimer molecular dynamics simulations of 1000 water molecules and a 303 atom Trp cage protein solvated by 2682 water molecules.

Optical interconnection networks for high-performance computing systems

NASA Astrophysics Data System (ADS)

Biberman, Aleksandr; Bergman, Keren

2012-04-01

Enabled by silicon photonic technology, optical interconnection networks have the potential to be a key disruptive technology in computing and communication industries. The enduring pursuit of performance gains in computing, combined with stringent power constraints, has fostered the ever-growing computational parallelism associated with chip multiprocessors, memory systems, high-performance computing systems and data centers. Sustaining these parallelism growths introduces unique challenges for on- and off-chip communications, shifting the focus toward novel and fundamentally different communication approaches. Chip-scale photonic interconnection networks, enabled by high-performance silicon photonic devices, offer unprecedented bandwidth scalability with reduced power consumption. We demonstrate that the silicon photonic platforms have already produced all the high-performance photonic devices required to realize these types of networks. Through extensive empirical characterization in much of our work, we demonstrate such feasibility of waveguides, modulators, switches and photodetectors. We also demonstrate systems that simultaneously combine many functionalities to achieve more complex building blocks. We propose novel silicon photonic devices, subsystems, network topologies and architectures to enable unprecedented performance of these photonic interconnection networks. Furthermore, the advantages of photonic interconnection networks extend far beyond the chip, offering advanced communication environments for memory systems, high-performance computing systems, and data centers.

Iterative load-balancing method with multigrid level relaxation for particle simulation with short-range interactions

NASA Astrophysics Data System (ADS)

Furuichi, Mikito; Nishiura, Daisuke

2017-10-01

We developed dynamic load-balancing algorithms for Particle Simulation Methods (PSM) involving short-range interactions, such as Smoothed Particle Hydrodynamics (SPH), Moving Particle Semi-implicit method (MPS), and Discrete Element method (DEM). These are needed to handle billions of particles modeled in large distributed-memory computer systems. Our method utilizes flexible orthogonal domain decomposition, allowing the sub-domain boundaries in the column to be different for each row. The imbalances in the execution time between parallel logical processes are treated as a nonlinear residual. Load-balancing is achieved by minimizing the residual within the framework of an iterative nonlinear solver, combined with a multigrid technique in the local smoother. Our iterative method is suitable for adjusting the sub-domain frequently by monitoring the performance of each computational process because it is computationally cheaper in terms of communication and memory costs than non-iterative methods. Numerical tests demonstrated the ability of our approach to handle workload imbalances arising from a non-uniform particle distribution, differences in particle types, or heterogeneous computer architecture which was difficult with previously proposed methods. We analyzed the parallel efficiency and scalability of our method using Earth simulator and K-computer supercomputer systems.

Fluid/Structure Interaction Studies of Aircraft Using High Fidelity Equations on Parallel Computers

NASA Technical Reports Server (NTRS)

Guruswamy, Guru; VanDalsem, William (Technical Monitor)

1994-01-01

Abstract Aeroelasticity which involves strong coupling of fluids, structures and controls is an important element in designing an aircraft. Computational aeroelasticity using low fidelity methods such as the linear aerodynamic flow equations coupled with the modal structural equations are well advanced. Though these low fidelity approaches are computationally less intensive, they are not adequate for the analysis of modern aircraft such as High Speed Civil Transport (HSCT) and Advanced Subsonic Transport (AST) which can experience complex flow/structure interactions. HSCT can experience vortex induced aeroelastic oscillations whereas AST can experience transonic buffet associated structural oscillations. Both aircraft may experience a dip in the flutter speed at the transonic regime. For accurate aeroelastic computations at these complex fluid/structure interaction situations, high fidelity equations such as the Navier-Stokes for fluids and the finite-elements for structures are needed. Computations using these high fidelity equations require large computational resources both in memory and speed. Current conventional super computers have reached their limitations both in memory and speed. As a result, parallel computers have evolved to overcome the limitations of conventional computers. This paper will address the transition that is taking place in computational aeroelasticity from conventional computers to parallel computers. The paper will address special techniques needed to take advantage of the architecture of new parallel computers. Results will be illustrated from computations made on iPSC/860 and IBM SP2 computer by using ENSAERO code that directly couples the Euler/Navier-Stokes flow equations with high resolution finite-element structural equations.

Experimental investigation of active rib stitch knitted architecture for flow control applications

NASA Astrophysics Data System (ADS)

Abel, Julianna M.; Mane, Poorna; Pascoe, Benjamin; Luntz, Jonathan; Brei, Diann

2010-04-01

Actively manipulating flow characteristics around the wing can enhance the high-lift capability and reduce drag; thereby, increasing fuel economy, improving maneuverability and operation over diverse flight conditions which enables longer, more varied missions. Active knits, a novel class of cellular structural smart material actuator architectures created by continuous, interlocked loops of stranded active material, produce distributed actuation that can actively manipulate the local surface of the aircraft wing to improve flow characteristics. Rib stitch active knits actuate normal to the surface, producing span-wise discrete periodic arrays that can withstand aerodynamic forces while supplying the necessary displacement for flow control. This paper presents a preliminary experimental investigation of the pressuredisplacement actuation performance capabilities of a rib stitch active knit based upon shape memory alloy (SMA) wire. SMA rib stitch prototypes in both individual form and in stacked and nestled architectures were experimentally tested for their quasi-static load-displacement characteristics, verifying the parallel and series relationships of the architectural configurations. The various configurations tested demonstrated the potential of active knits to generate the required level of distributed surface displacements while under aerodynamic level loads for various forms of flow control.

Parallel processing architecture for computing inverse differential kinematic equations of the PUMA arm

NASA Technical Reports Server (NTRS)

Hsia, T. C.; Lu, G. Z.; Han, W. H.

1987-01-01

In advanced robot control problems, on-line computation of inverse Jacobian solution is frequently required. Parallel processing architecture is an effective way to reduce computation time. A parallel processing architecture is developed for the inverse Jacobian (inverse differential kinematic equation) of the PUMA arm. The proposed pipeline/parallel algorithm can be inplemented on an IC chip using systolic linear arrays. This implementation requires 27 processing cells and 25 time units. Computation time is thus significantly reduced.

Automatic Generation of Directive-Based Parallel Programs for Shared Memory Parallel Systems

NASA Technical Reports Server (NTRS)

Jin, Hao-Qiang; Yan, Jerry; Frumkin, Michael

2000-01-01

The shared-memory programming model is a very effective way to achieve parallelism on shared memory parallel computers. As great progress was made in hardware and software technologies, performance of parallel programs with compiler directives has demonstrated large improvement. The introduction of OpenMP directives, the industrial standard for shared-memory programming, has minimized the issue of portability. Due to its ease of programming and its good performance, the technique has become very popular. In this study, we have extended CAPTools, a computer-aided parallelization toolkit, to automatically generate directive-based, OpenMP, parallel programs. We outline techniques used in the implementation of the tool and present test results on the NAS parallel benchmarks and ARC3D, a CFD application. This work demonstrates the great potential of using computer-aided tools to quickly port parallel programs and also achieve good performance.

A general purpose subroutine for fast fourier transform on a distributed memory parallel machine

NASA Technical Reports Server (NTRS)

Dubey, A.; Zubair, M.; Grosch, C. E.

1992-01-01

One issue which is central in developing a general purpose Fast Fourier Transform (FFT) subroutine on a distributed memory parallel machine is the data distribution. It is possible that different users would like to use the FFT routine with different data distributions. Thus, there is a need to design FFT schemes on distributed memory parallel machines which can support a variety of data distributions. An FFT implementation on a distributed memory parallel machine which works for a number of data distributions commonly encountered in scientific applications is presented. The problem of rearranging the data after computing the FFT is also addressed. The performance of the implementation on a distributed memory parallel machine Intel iPSC/860 is evaluated.

Framework for analysis of guaranteed QOS systems

NASA Astrophysics Data System (ADS)

Chaudhry, Shailender; Choudhary, Alok

1997-01-01

Multimedia data is isochronous in nature and entails managing and delivering high volumes of data. Multiprocessors with their large processing power, vast memory, and fast interconnects, are an ideal candidate for the implementation of multimedia applications. Initially, multiprocessors were designed to execute scientific programs and thus their architecture was optimized to provide low message latency and efficiently support regular communication patterns. Hence, they have a regular network topology and most use wormhole routing. The design offers the benefits of a simple router, small buffer size, and network latency that is almost independent of path length. Among the various multimedia applications, video on demand (VOD) server is well-suited for implementation using parallel multiprocessors. Logical models for VOD servers are presently mapped onto multiprocessors. Our paper provides a framework for calculating bounds on utilization of system resources with which QoS parameters for each isochronous stream can be guaranteed. Effects of the architecture of multiprocessors, and efficiency of various local models and mapping on particular architectures can be investigated within our framework. Our framework is based on rigorous proofs and provides tight bounds. The results obtained may be used as the basis for admission control tests. To illustrate the versatility of our framework, we provide bounds on utilization for various logical models applied to mesh connected architectures for a video on demand server. Our results show that worm hole routing can lead to packets waiting for transmission of other packets that apparently share no common resources. This situation is analogous to head-of-the-line blocking. We find that the provision of multiple VCs per link and multiple flit buffers improves utilization (even under guaranteed QoS parameters). This analogous to parallel iterative matching.

Fast multipurpose Monte Carlo simulation for proton therapy using multi- and many-core CPU architectures.

PubMed

Souris, Kevin; Lee, John Aldo; Sterpin, Edmond

2016-04-01

Accuracy in proton therapy treatment planning can be improved using Monte Carlo (MC) simulations. However the long computation time of such methods hinders their use in clinical routine. This work aims to develop a fast multipurpose Monte Carlo simulation tool for proton therapy using massively parallel central processing unit (CPU) architectures. A new Monte Carlo, called MCsquare (many-core Monte Carlo), has been designed and optimized for the last generation of Intel Xeon processors and Intel Xeon Phi coprocessors. These massively parallel architectures offer the flexibility and the computational power suitable to MC methods. The class-II condensed history algorithm of MCsquare provides a fast and yet accurate method of simulating heavy charged particles such as protons, deuterons, and alphas inside voxelized geometries. Hard ionizations, with energy losses above a user-specified threshold, are simulated individually while soft events are regrouped in a multiple scattering theory. Elastic and inelastic nuclear interactions are sampled from ICRU 63 differential cross sections, thereby allowing for the computation of prompt gamma emission profiles. MCsquare has been benchmarked with the gate/geant4 Monte Carlo application for homogeneous and heterogeneous geometries. Comparisons with gate/geant4 for various geometries show deviations within 2%-1 mm. In spite of the limited memory bandwidth of the coprocessor simulation time is below 25 s for 10(7) primary 200 MeV protons in average soft tissues using all Xeon Phi and CPU resources embedded in a single desktop unit. MCsquare exploits the flexibility of CPU architectures to provide a multipurpose MC simulation tool. Optimized code enables the use of accurate MC calculation within a reasonable computation time, adequate for clinical practice. MCsquare also simulates prompt gamma emission and can thus be used also for in vivo range verification.

Parallel implementation of the particle simulation method with dynamic load balancing: Toward realistic geodynamical simulation

NASA Astrophysics Data System (ADS)

Furuichi, M.; Nishiura, D.

2015-12-01

Fully Lagrangian methods such as Smoothed Particle Hydrodynamics (SPH) and Discrete Element Method (DEM) have been widely used to solve the continuum and particles motions in the computational geodynamics field. These mesh-free methods are suitable for the problems with the complex geometry and boundary. In addition, their Lagrangian nature allows non-diffusive advection useful for tracking history dependent properties (e.g. rheology) of the material. These potential advantages over the mesh-based methods offer effective numerical applications to the geophysical flow and tectonic processes, which are for example, tsunami with free surface and floating body, magma intrusion with fracture of rock, and shear zone pattern generation of granular deformation. In order to investigate such geodynamical problems with the particle based methods, over millions to billion particles are required for the realistic simulation. Parallel computing is therefore important for handling such huge computational cost. An efficient parallel implementation of SPH and DEM methods is however known to be difficult especially for the distributed-memory architecture. Lagrangian methods inherently show workload imbalance problem for parallelization with the fixed domain in space, because particles move around and workloads change during the simulation. Therefore dynamic load balance is key technique to perform the large scale SPH and DEM simulation. In this work, we present the parallel implementation technique of SPH and DEM method utilizing dynamic load balancing algorithms toward the high resolution simulation over large domain using the massively parallel super computer system. Our method utilizes the imbalances of the executed time of each MPI process as the nonlinear term of parallel domain decomposition and minimizes them with the Newton like iteration method. In order to perform flexible domain decomposition in space, the slice-grid algorithm is used. Numerical tests show that our approach is suitable for solving the particles with different calculation costs (e.g. boundary particles) as well as the heterogeneous computer architecture. We analyze the parallel efficiency and scalability on the super computer systems (K-computer, Earth simulator 3, etc.).

Evaluation of fault-tolerant parallel-processor architectures over long space missions

NASA Technical Reports Server (NTRS)

Johnson, Sally C.

1989-01-01

The impact of a five year space mission environment on fault-tolerant parallel processor architectures is examined. The target application is a Strategic Defense Initiative (SDI) satellite requiring 256 parallel processors to provide the computation throughput. The reliability requirements are that the system still be operational after five years with .99 probability and that the probability of system failure during one-half hour of full operation be less than 10(-7). The fault tolerance features an architecture must possess to meet these reliability requirements are presented, many potential architectures are briefly evaluated, and one candidate architecture, the Charles Stark Draper Laboratory's Fault-Tolerant Parallel Processor (FTPP) is evaluated in detail. A methodology for designing a preliminary system configuration to meet the reliability and performance requirements of the mission is then presented and demonstrated by designing an FTPP configuration.

CMOL: A New Concept for Nanoelectronics

NASA Astrophysics Data System (ADS)

Likharev, Konstantin

2005-03-01

I will review the recent work on devices and architectures for future hybrid semiconductor/molecular integrated circuits, in particular those of ``CMOL'' variety [1]. Such circuits would combine an advanced CMOS subsystem fabricated by the usual lithographic patterning, two layers of parallel metallic nanowires formed, e.g., by nanoimprint, and two-terminal molecular devices self-assembled on the nanowire crosspoints. Estimates show that this powerful combination may allow CMOL circuits to reach an unparalleled density (up to 10^12 functions per cm^2) and ultrahigh rate of information processing (up to 10^20 operations per second on a single chip), at acceptable power dissipation. The main challenges on the way toward practical CMOL technology are: (i) reliable chemically-directed self-assembly of mid-size organic molecules, and (ii) the development of efficient defect-tolerant architectures for CMOL circuits. Our recent work has shown that such architectures may be developed not only for terabit-scale memories and naturally defect-tolerant mixed-signal neuromorphic networks, but (rather unexpectedly) also for FPGA-style digital Boolean circuits. [1] For details, see http://rsfq1.physics.sunysb.edu/˜likharev/nano/Springer04.pdf

Two-dimensional nonsteady viscous flow simulation on the Navier-Stokes computer miniNode

NASA Technical Reports Server (NTRS)

Nosenchuck, Daniel M.; Littman, Michael G.; Flannery, William

1986-01-01

The needs of large-scale scientific computation are outpacing the growth in performance of mainframe supercomputers. In particular, problems in fluid mechanics involving complex flow simulations require far more speed and capacity than that provided by current and proposed Class VI supercomputers. To address this concern, the Navier-Stokes Computer (NSC) was developed. The NSC is a parallel-processing machine, comprised of individual Nodes, each comparable in performance to current supercomputers. The global architecture is that of a hypercube, and a 128-Node NSC has been designed. New architectural features, such as a reconfigurable many-function ALU pipeline and a multifunction memory-ALU switch, have provided the capability to efficiently implement a wide range of algorithms. Efficient algorithms typically involve numerically intensive tasks, which often include conditional operations. These operations may be efficiently implemented on the NSC without, in general, sacrificing vector-processing speed. To illustrate the architecture, programming, and several of the capabilities of the NSC, the simulation of two-dimensional, nonsteady viscous flows on a prototype Node, called the miniNode, is presented.

A pervasive parallel framework for visualization: final report for FWP 10-014707

DOE Office of Scientific and Technical Information (OSTI.GOV)

Moreland, Kenneth D.

2014-01-01

We are on the threshold of a transformative change in the basic architecture of highperformance computing. The use of accelerator processors, characterized by large core counts, shared but asymmetrical memory, and heavy thread loading, is quickly becoming the norm in high performance computing. These accelerators represent significant challenges in updating our existing base of software. An intrinsic problem with this transition is a fundamental programming shift from message passing processes to much more fine thread scheduling with memory sharing. Another problem is the lack of stability in accelerator implementation; processor and compiler technology is currently changing rapidly. This report documentsmore » the results of our three-year ASCR project to address these challenges. Our project includes the development of the Dax toolkit, which contains the beginnings of new algorithms for a new generation of computers and the underlying infrastructure to rapidly prototype and build further algorithms as necessary.« less

Acoustic simulation in architecture with parallel algorithm

NASA Astrophysics Data System (ADS)

Li, Xiaohong; Zhang, Xinrong; Li, Dan

2004-03-01

In allusion to complexity of architecture environment and Real-time simulation of architecture acoustics, a parallel radiosity algorithm was developed. The distribution of sound energy in scene is solved with this method. And then the impulse response between sources and receivers at frequency segment, which are calculated with multi-process, are combined into whole frequency response. The numerical experiment shows that parallel arithmetic can improve the acoustic simulating efficiency of complex scene.

Support for non-locking parallel reception of packets belonging to a single memory reception FIFO

DOEpatents

Chen, Dong [Yorktown Heights, NY; Heidelberger, Philip [Yorktown Heights, NY; Salapura, Valentina [Yorktown Heights, NY; Senger, Robert M [Yorktown Heights, NY; Steinmacher-Burow, Burkhard [Boeblingen, DE; Sugawara, Yutaka [Yorktown Heights, NY

2011-01-27

A method and apparatus for distributed parallel messaging in a parallel computing system. A plurality of DMA engine units are configured in a multiprocessor system to operate in parallel, one DMA engine unit for transferring a current packet received at a network reception queue to a memory location in a memory FIFO (rmFIFO) region of a memory. A control unit implements logic to determine whether any prior received packet destined for that rmFIFO is still in a process of being stored in the associated memory by another DMA engine unit of the plurality, and prevent the one DMA engine unit from indicating completion of storing the current received packet in the reception memory FIFO (rmFIFO) until all prior received packets destined for that rmFIFO are completely stored by the other DMA engine units. Thus, there is provided non-locking support so that multiple packets destined for a single rmFIFO are transferred and stored in parallel to predetermined locations in a memory.

Power-Aware Compiler Controllable Chip Multiprocessor

NASA Astrophysics Data System (ADS)

Shikano, Hiroaki; Shirako, Jun; Wada, Yasutaka; Kimura, Keiji; Kasahara, Hironori

A power-aware compiler controllable chip multiprocessor (CMP) is presented and its performance and power consumption are evaluated with the optimally scheduled advanced multiprocessor (OSCAR) parallelizing compiler. The CMP is equipped with power control registers that change clock frequency and power supply voltage to functional units including processor cores, memories, and an interconnection network. The OSCAR compiler carries out coarse-grain task parallelization of programs and reduces power consumption using architectural power control support and the compiler's power saving scheme. The performance evaluation shows that MPEG-2 encoding on the proposed CMP with four CPUs results in 82.6% power reduction in real-time execution mode with a deadline constraint on its sequential execution time. Furthermore, MP3 encoding on a heterogeneous CMP with four CPUs and four accelerators results in 53.9% power reduction at 21.1-fold speed-up in performance against its sequential execution in the fastest execution mode.

Genealogical correspondence of a forebrain centre implies an executive brain in the protostome–deuterostome bilaterian ancestor

PubMed Central

2016-01-01

Orthologous genes involved in the formation of proteins associated with memory acquisition are similarly expressed in forebrain centres that exhibit similar cognitive properties. These proteins include cAMP-dependent protein kinase A catalytic subunit (PKA-Cα) and phosphorylated Ca2+/calmodulin-dependent protein kinase II (pCaMKII), both required for long-term memory formation which is enriched in rodent hippocampus and insect mushroom bodies, both implicated in allocentric memory and both possessing corresponding neuronal architectures. Antibodies against these proteins resolve forebrain centres, or their equivalents, having the same ground pattern of neuronal organization in species across five phyla. The ground pattern is defined by olfactory or chemosensory afferents supplying systems of parallel fibres of intrinsic neurons intersected by orthogonal domains of afferent and efferent arborizations with local interneurons providing feedback loops. The totality of shared characters implies a deep origin in the protostome–deuterostome bilaterian ancestor of elements of a learning and memory circuit. Proxies for such an ancestral taxon are simple extant bilaterians, particularly acoels that express PKA-Cα and pCaMKII in discrete anterior domains that can be properly referred to as brains. PMID:26598732

Design and testing of the first 2D Prototype Vertically Integrated Pattern Recognition Associative Memory

NASA Astrophysics Data System (ADS)

Liu, T.; Deptuch, G.; Hoff, J.; Jindariani, S.; Joshi, S.; Olsen, J.; Tran, N.; Trimpl, M.

2015-02-01

An associative memory-based track finding approach has been proposed for a Level 1 tracking trigger to cope with increasing luminosities at the LHC. The associative memory uses a massively parallel architecture to tackle the intrinsically complex combinatorics of track finding algorithms, thus avoiding the typical power law dependence of execution time on occupancy and solving the pattern recognition in times roughly proportional to the number of hits. This is of crucial importance given the large occupancies typical of hadronic collisions. The design of an associative memory system capable of dealing with the complexity of HL-LHC collisions and with the short latency required by Level 1 triggering poses significant, as yet unsolved, technical challenges. For this reason, an aggressive R&D program has been launched at Fermilab to advance state of-the-art associative memory technology, the so called VIPRAM (Vertically Integrated Pattern Recognition Associative Memory) project. The VIPRAM leverages emerging 3D vertical integration technology to build faster and denser Associative Memory devices. The first step is to implement in conventional VLSI the associative memory building blocks that can be used in 3D stacking; in other words, the building blocks are laid out as if it is a 3D design. In this paper, we report on the first successful implementation of a 2D VIPRAM demonstrator chip (protoVIPRAM00). The results show that these building blocks are ready for 3D stacking.

Design and testing of the first 2D Prototype Vertically Integrated Pattern Recognition Associative Memory

DOE Office of Scientific and Technical Information (OSTI.GOV)

Liu, T.; Deptuch, G.; Hoff, J.

An associative memory-based track finding approach has been proposed for a Level 1 tracking trigger to cope with increasing luminosities at the LHC. The associative memory uses a massively parallel architecture to tackle the intrinsically complex combinatorics of track finding algorithms, thus avoiding the typical power law dependence of execution time on occupancy and solving the pattern recognition in times roughly proportional to the number of hits. This is of crucial importance given the large occupancies typical of hadronic collisions. The design of an associative memory system capable of dealing with the complexity of HL-LHC collisions and with the shortmore » latency required by Level 1 triggering poses significant, as yet unsolved, technical challenges. For this reason, an aggressive R&D program has been launched at Fermilab to advance state of-the-art associative memory technology, the so called VIPRAM (Vertically Integrated Pattern Recognition Associative Memory) project. The VIPRAM leverages emerging 3D vertical integration technology to build faster and denser Associative Memory devices. The first step is to implement in conventional VLSI the associative memory building blocks that can be used in 3D stacking, in other words, the building blocks are laid out as if it is a 3D design. In this paper, we report on the first successful implementation of a 2D VIPRAM demonstrator chip (protoVIPRAM00). The results show that these building blocks are ready for 3D stacking.« less

Recent Progress on the Parallel Implementation of Moving-Body Overset Grid Schemes

NASA Technical Reports Server (NTRS)

Wissink, Andrew; Allen, Edwin (Technical Monitor)

1998-01-01

Viscous calculations about geometrically complex bodies in which there is relative motion between component parts is one of the most computationally demanding problems facing CFD researchers today. This presentation documents results from the first two years of a CHSSI-funded effort within the U.S. Army AFDD to develop scalable dynamic overset grid methods for unsteady viscous calculations with moving-body problems. The first pan of the presentation will focus on results from OVERFLOW-D1, a parallelized moving-body overset grid scheme that employs traditional Chimera methodology. The two processes that dominate the cost of such problems are the flow solution on each component and the intergrid connectivity solution. Parallel implementations of the OVERFLOW flow solver and DCF3D connectivity software are coupled with a proposed two-part static-dynamic load balancing scheme and tested on the IBM SP and Cray T3E multi-processors. The second part of the presentation will cover some recent results from OVERFLOW-D2, a new flow solver that employs Cartesian grids with various levels of refinement, facilitating solution adaption. A study of the parallel performance of the scheme on large distributed- memory multiprocessor computer architectures will be reported.

Dynamic Load-Balancing for Distributed Heterogeneous Computing of Parallel CFD Problems

NASA Technical Reports Server (NTRS)

Ecer, A.; Chien, Y. P.; Boenisch, T.; Akay, H. U.

2000-01-01

The developed methodology is aimed at improving the efficiency of executing block-structured algorithms on parallel, distributed, heterogeneous computers. The basic approach of these algorithms is to divide the flow domain into many sub- domains called blocks, and solve the governing equations over these blocks. Dynamic load balancing problem is defined as the efficient distribution of the blocks among the available processors over a period of several hours of computations. In environments with computers of different architecture, operating systems, CPU speed, memory size, load, and network speed, balancing the loads and managing the communication between processors becomes crucial. Load balancing software tools for mutually dependent parallel processes have been created to efficiently utilize an advanced computation environment and algorithms. These tools are dynamic in nature because of the chances in the computer environment during execution time. More recently, these tools were extended to a second operating system: NT. In this paper, the problems associated with this application will be discussed. Also, the developed algorithms were combined with the load sharing capability of LSF to efficiently utilize workstation clusters for parallel computing. Finally, results will be presented on running a NASA based code ADPAC to demonstrate the developed tools for dynamic load balancing.

Auditory short-term memory in the primate auditory cortex

PubMed Central

Scott, Brian H.; Mishkin, Mortimer

2015-01-01

Sounds are fleeting, and assembling the sequence of inputs at the ear into a coherent percept requires auditory memory across various time scales. Auditory short-term memory comprises at least two components: an active ‘working memory’ bolstered by rehearsal, and a sensory trace that may be passively retained. Working memory relies on representations recalled from long-term memory, and their rehearsal may require phonological mechanisms unique to humans. The sensory component, passive short-term memory (pSTM), is tractable to study in nonhuman primates, whose brain architecture and behavioral repertoire are comparable to our own. This review discusses recent advances in the behavioral and neurophysiological study of auditory memory with a focus on single-unit recordings from macaque monkeys performing delayed-match-to-sample (DMS) tasks. Monkeys appear to employ pSTM to solve these tasks, as evidenced by the impact of interfering stimuli on memory performance. In several regards, pSTM in monkeys resembles pitch memory in humans, and may engage similar neural mechanisms. Neural correlates of DMS performance have been observed throughout the auditory and prefrontal cortex, defining a network of areas supporting auditory STM with parallels to that supporting visual STM. These correlates include persistent neural firing, or a suppression of firing, during the delay period of the memory task, as well as suppression or (less commonly) enhancement of sensory responses when a sound is repeated as a ‘match’ stimulus. Auditory STM is supported by a distributed temporo-frontal network in which sensitivity to stimulus history is an intrinsic feature of auditory processing. PMID:26541581

Efficiently modeling neural networks on massively parallel computers

NASA Technical Reports Server (NTRS)

Farber, Robert M.

1993-01-01

Neural networks are a very useful tool for analyzing and modeling complex real world systems. Applying neural network simulations to real world problems generally involves large amounts of data and massive amounts of computation. To efficiently handle the computational requirements of large problems, we have implemented at Los Alamos a highly efficient neural network compiler for serial computers, vector computers, vector parallel computers, and fine grain SIMD computers such as the CM-2 connection machine. This paper describes the mapping used by the compiler to implement feed-forward backpropagation neural networks for a SIMD (Single Instruction Multiple Data) architecture parallel computer. Thinking Machines Corporation has benchmarked our code at 1.3 billion interconnects per second (approximately 3 gigaflops) on a 64,000 processor CM-2 connection machine (Singer 1990). This mapping is applicable to other SIMD computers and can be implemented on MIMD computers such as the CM-5 connection machine. Our mapping has virtually no communications overhead with the exception of the communications required for a global summation across the processors (which has a sub-linear runtime growth on the order of O(log(number of processors)). We can efficiently model very large neural networks which have many neurons and interconnects and our mapping can extend to arbitrarily large networks (within memory limitations) by merging the memory space of separate processors with fast adjacent processor interprocessor communications. This paper will consider the simulation of only feed forward neural network although this method is extendable to recurrent networks.

Supporting shared data structures on distributed memory architectures

NASA Technical Reports Server (NTRS)

Koelbel, Charles; Mehrotra, Piyush; Vanrosendale, John

1990-01-01

Programming nonshared memory systems is more difficult than programming shared memory systems, since there is no support for shared data structures. Current programming languages for distributed memory architectures force the user to decompose all data structures into separate pieces, with each piece owned by one of the processors in the machine, and with all communication explicitly specified by low-level message-passing primitives. A new programming environment is presented for distributed memory architectures, providing a global name space and allowing direct access to remote parts of data values. The analysis and program transformations required to implement this environment are described, and the efficiency of the resulting code on the NCUBE/7 and IPSC/2 hypercubes are described.

Linguistic representations and memory architectures: The devil is in the details.

PubMed

Chacón, Dustin Alfonso; Momma, Shota; Phillips, Colin

2016-01-01

Attempts to explain linguistic phenomena as consequences of memory constraints require detailed specification of linguistic representations and memory architectures alike. We discuss examples of supposed locality biases in language comprehension and production, and their link to memory constraints. Findings do not generally favor Christiansen & Chater's (C&C's) approach. We discuss connections to debates that stretch back to the nineteenth century.

CLOCS (Computer with Low Context-Switching Time) Architecture Reference Documents

DTIC Science & Technology

1988-05-06

Peculiarities The only state inside the central processing unit(CPU) is a program status word. All data operations are memory to memory. One result of this... to the challenge "if I whore to design RISC, this is how I would do it." The architecture was designed by Mark Davis and Bill Gallmeister. 1.2...are memory to memory. Any special devices added should be memory mapped. The program counter is even memory mapped. 1.3.1 Working storage There is no

Electromagnetic Physics Models for Parallel Computing Architectures

NASA Astrophysics Data System (ADS)

Amadio, G.; Ananya, A.; Apostolakis, J.; Aurora, A.; Bandieramonte, M.; Bhattacharyya, A.; Bianchini, C.; Brun, R.; Canal, P.; Carminati, F.; Duhem, L.; Elvira, D.; Gheata, A.; Gheata, M.; Goulas, I.; Iope, R.; Jun, S. Y.; Lima, G.; Mohanty, A.; Nikitina, T.; Novak, M.; Pokorski, W.; Ribon, A.; Seghal, R.; Shadura, O.; Vallecorsa, S.; Wenzel, S.; Zhang, Y.

2016-10-01

The recent emergence of hardware architectures characterized by many-core or accelerated processors has opened new opportunities for concurrent programming models taking advantage of both SIMD and SIMT architectures. GeantV, a next generation detector simulation, has been designed to exploit both the vector capability of mainstream CPUs and multi-threading capabilities of coprocessors including NVidia GPUs and Intel Xeon Phi. The characteristics of these architectures are very different in terms of the vectorization depth and type of parallelization needed to achieve optimal performance. In this paper we describe implementation of electromagnetic physics models developed for parallel computing architectures as a part of the GeantV project. Results of preliminary performance evaluation and physics validation are presented as well.

Modelling parallel programs and multiprocessor architectures with AXE

NASA Technical Reports Server (NTRS)

Yan, Jerry C.; Fineman, Charles E.

1991-01-01

AXE, An Experimental Environment for Parallel Systems, was designed to model and simulate for parallel systems at the process level. It provides an integrated environment for specifying computation models, multiprocessor architectures, data collection, and performance visualization. AXE is being used at NASA-Ames for developing resource management strategies, parallel problem formulation, multiprocessor architectures, and operating system issues related to the High Performance Computing and Communications Program. AXE's simple, structured user-interface enables the user to model parallel programs and machines precisely and efficiently. Its quick turn-around time keeps the user interested and productive. AXE models multicomputers. The user may easily modify various architectural parameters including the number of sites, connection topologies, and overhead for operating system activities. Parallel computations in AXE are represented as collections of autonomous computing objects known as players. Their use and behavior is described. Performance data of the multiprocessor model can be observed on a color screen. These include CPU and message routing bottlenecks, and the dynamic status of the software.

Performance evaluation of canny edge detection on a tiled multicore architecture

NASA Astrophysics Data System (ADS)

Brethorst, Andrew Z.; Desai, Nehal; Enright, Douglas P.; Scrofano, Ronald

2011-01-01

In the last few years, a variety of multicore architectures have been used to parallelize image processing applications. In this paper, we focus on assessing the parallel speed-ups of different Canny edge detection parallelization strategies on the Tile64, a tiled multicore architecture developed by the Tilera Corporation. Included in these strategies are different ways Canny edge detection can be parallelized, as well as differences in data management. The two parallelization strategies examined were loop-level parallelism and domain decomposition. Loop-level parallelism is achieved through the use of OpenMP,1 and it is capable of parallelization across the range of values over which a loop iterates. Domain decomposition is the process of breaking down an image into subimages, where each subimage is processed independently, in parallel. The results of the two strategies show that for the same number of threads, programmer implemented, domain decomposition exhibits higher speed-ups than the compiler managed, loop-level parallelism implemented with OpenMP.

Fast quantum Monte Carlo on a GPU

NASA Astrophysics Data System (ADS)

Lutsyshyn, Y.

2015-02-01

We present a scheme for the parallelization of quantum Monte Carlo method on graphical processing units, focusing on variational Monte Carlo simulation of bosonic systems. We use asynchronous execution schemes with shared memory persistence, and obtain an excellent utilization of the accelerator. The CUDA code is provided along with a package that simulates liquid helium-4. The program was benchmarked on several models of Nvidia GPU, including Fermi GTX560 and M2090, and the Kepler architecture K20 GPU. Special optimization was developed for the Kepler cards, including placement of data structures in the register space of the Kepler GPUs. Kepler-specific optimization is discussed.

Digital and optical shape representation and pattern recognition; Proceedings of the Meeting, Orlando, FL, Apr. 4-6, 1988

NASA Technical Reports Server (NTRS)

Juday, Richard D. (Editor)

1988-01-01

The present conference discusses topics in pattern-recognition correlator architectures, digital stereo systems, geometric image transformations and their applications, topics in pattern recognition, filter algorithms, object detection and classification, shape representation techniques, and model-based object recognition methods. Attention is given to edge-enhancement preprocessing using liquid crystal TVs, massively-parallel optical data base management, three-dimensional sensing with polar exponential sensor arrays, the optical processing of imaging spectrometer data, hybrid associative memories and metric data models, the representation of shape primitives in neural networks, and the Monte Carlo estimation of moment invariants for pattern recognition.

Computation of Engine Noise Propagation and Scattering Off an Aircraft

NASA Technical Reports Server (NTRS)

Xu, J.; Stanescu, D.; Hussaini, M. Y.; Farassat, F.

2003-01-01

The paper presents a comparison of experimental noise data measured in flight on a two-engine business jet aircraft with Kulite microphones placed on the suction surface of the wing with computational results. Both a time-domain discontinuous Galerkin spectral method and a frequency-domain spectral element method are used to simulate the radiation of the dominant spinning mode from the engine and its reflection and scattering by the fuselage and the wing. Both methods are implemented in computer codes that use the distributed memory model to make use of large parallel architectures. The results show that trends of the noise field are well predicted by both methods.

Study of the mapping of Navier-Stokes algorithms onto multiple-instruction/multiple-data-stream computers

NASA Technical Reports Server (NTRS)

Eberhardt, D. S.; Baganoff, D.; Stevens, K.

1984-01-01

Implicit approximate-factored algorithms have certain properties that are suitable for parallel processing. A particular computational fluid dynamics (CFD) code, using this algorithm, is mapped onto a multiple-instruction/multiple-data-stream (MIMD) computer architecture. An explanation of this mapping procedure is presented, as well as some of the difficulties encountered when trying to run the code concurrently. Timing results are given for runs on the Ames Research Center's MIMD test facility which consists of two VAX 11/780's with a common MA780 multi-ported memory. Speedups exceeding 1.9 for characteristic CFD runs were indicated by the timing results.

Shared direct memory access on the Explorer 2-LX

NASA Technical Reports Server (NTRS)

Musgrave, Jeffrey L.

1990-01-01

Advances in Expert System technology and Artificial Intelligence have provided a framework for applying automated Intelligence to the solution of problems which were generally perceived as intractable using more classical approaches. As a result, hybrid architectures and parallel processing capability have become more common in computing environments. The Texas Instruments Explorer II-LX is an example of a machine which combines a symbolic processing environment, and a computationally oriented environment in a single chassis for integrated problem solutions. This user's manual is an attempt to make these capabilities more accessible to a wider range of engineers and programmers with problems well suited to solution in such an environment.

Using architecture information and real-time resource state to reduce power consumption and communication costs in parallel applications.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Brandt, James M.; Devine, Karen Dragon; Gentile, Ann C.

2014-09-01

As computer systems grow in both size and complexity, the need for applications and run-time systems to adjust to their dynamic environment also grows. The goal of the RAAMP LDRD was to combine static architecture information and real-time system state with algorithms to conserve power, reduce communication costs, and avoid network contention. We devel- oped new data collection and aggregation tools to extract static hardware information (e.g., node/core hierarchy, network routing) as well as real-time performance data (e.g., CPU uti- lization, power consumption, memory bandwidth saturation, percentage of used bandwidth, number of network stalls). We created application interfaces that allowedmore » this data to be used easily by algorithms. Finally, we demonstrated the benefit of integrating system and application information for two use cases. The first used real-time power consumption and memory bandwidth saturation data to throttle concurrency to save power without increasing application execution time. The second used static or real-time network traffic information to reduce or avoid network congestion by remapping MPI tasks to allocated processors. Results from our work are summarized in this report; more details are available in our publications [2, 6, 14, 16, 22, 29, 38, 44, 51, 54].« less

Algorithms and Libraries

NASA Technical Reports Server (NTRS)

Dongarra, Jack

1998-01-01

This exploratory study initiated our inquiry into algorithms and applications that would benefit by latency tolerant approach to algorithm building, including the construction of new algorithms where appropriate. In a multithreaded execution, when a processor reaches a point where remote memory access is necessary, the request is sent out on the network and a context--switch occurs to a new thread of computation. This effectively masks a long and unpredictable latency due to remote loads, thereby providing tolerance to remote access latency. We began to develop standards to profile various algorithm and application parameters, such as the degree of parallelism, granularity, precision, instruction set mix, interprocessor communication, latency etc. These tools will continue to develop and evolve as the Information Power Grid environment matures. To provide a richer context for this research, the project also focused on issues of fault-tolerance and computation migration of numerical algorithms and software. During the initial phase we tried to increase our understanding of the bottlenecks in single processor performance. Our work began by developing an approach for the automatic generation and optimization of numerical software for processors with deep memory hierarchies and pipelined functional units. Based on the results we achieved in this study we are planning to study other architectures of interest, including development of cost models, and developing code generators appropriate to these architectures.

The BLAZE language: A parallel language for scientific programming

NASA Technical Reports Server (NTRS)

Mehrotra, P.; Vanrosendale, J.

1985-01-01

A Pascal-like scientific programming language, Blaze, is described. Blaze contains array arithmetic, forall loops, and APL-style accumulation operators, which allow natural expression of fine grained parallelism. It also employs an applicative or functional procedure invocation mechanism, which makes it easy for compilers to extract coarse grained parallelism using machine specific program restructuring. Thus Blaze should allow one to achieve highly parallel execution on multiprocessor architectures, while still providing the user with onceptually sequential control flow. A central goal in the design of Blaze is portability across a broad range of parallel architectures. The multiple levels of parallelism present in Blaze code, in principle, allow a compiler to extract the types of parallelism appropriate for the given architecture while neglecting the remainder. The features of Blaze are described and shows how this language would be used in typical scientific programming.

Parallel Signal Processing and System Simulation using aCe

NASA Technical Reports Server (NTRS)

Dorband, John E.; Aburdene, Maurice F.

2003-01-01

Recently, networked and cluster computation have become very popular for both signal processing and system simulation. A new language is ideally suited for parallel signal processing applications and system simulation since it allows the programmer to explicitly express the computations that can be performed concurrently. In addition, the new C based parallel language (ace C) for architecture-adaptive programming allows programmers to implement algorithms and system simulation applications on parallel architectures by providing them with the assurance that future parallel architectures will be able to run their applications with a minimum of modification. In this paper, we will focus on some fundamental features of ace C and present a signal processing application (FFT).

GPU-accelerated adjoint algorithmic differentiation

NASA Astrophysics Data System (ADS)

Gremse, Felix; Höfter, Andreas; Razik, Lukas; Kiessling, Fabian; Naumann, Uwe

2016-03-01

Many scientific problems such as classifier training or medical image reconstruction can be expressed as minimization of differentiable real-valued cost functions and solved with iterative gradient-based methods. Adjoint algorithmic differentiation (AAD) enables automated computation of gradients of such cost functions implemented as computer programs. To backpropagate adjoint derivatives, excessive memory is potentially required to store the intermediate partial derivatives on a dedicated data structure, referred to as the ;tape;. Parallelization is difficult because threads need to synchronize their accesses during taping and backpropagation. This situation is aggravated for many-core architectures, such as Graphics Processing Units (GPUs), because of the large number of light-weight threads and the limited memory size in general as well as per thread. We show how these limitations can be mediated if the cost function is expressed using GPU-accelerated vector and matrix operations which are recognized as intrinsic functions by our AAD software. We compare this approach with naive and vectorized implementations for CPUs. We use four increasingly complex cost functions to evaluate the performance with respect to memory consumption and gradient computation times. Using vectorization, CPU and GPU memory consumption could be substantially reduced compared to the naive reference implementation, in some cases even by an order of complexity. The vectorization allowed usage of optimized parallel libraries during forward and reverse passes which resulted in high speedups for the vectorized CPU version compared to the naive reference implementation. The GPU version achieved an additional speedup of 7.5 ± 4.4, showing that the processing power of GPUs can be utilized for AAD using this concept. Furthermore, we show how this software can be systematically extended for more complex problems such as nonlinear absorption reconstruction for fluorescence-mediated tomography.

GPU-Accelerated Adjoint Algorithmic Differentiation.

PubMed

Gremse, Felix; Höfter, Andreas; Razik, Lukas; Kiessling, Fabian; Naumann, Uwe

2016-03-01

Many scientific problems such as classifier training or medical image reconstruction can be expressed as minimization of differentiable real-valued cost functions and solved with iterative gradient-based methods. Adjoint algorithmic differentiation (AAD) enables automated computation of gradients of such cost functions implemented as computer programs. To backpropagate adjoint derivatives, excessive memory is potentially required to store the intermediate partial derivatives on a dedicated data structure, referred to as the "tape". Parallelization is difficult because threads need to synchronize their accesses during taping and backpropagation. This situation is aggravated for many-core architectures, such as Graphics Processing Units (GPUs), because of the large number of light-weight threads and the limited memory size in general as well as per thread. We show how these limitations can be mediated if the cost function is expressed using GPU-accelerated vector and matrix operations which are recognized as intrinsic functions by our AAD software. We compare this approach with naive and vectorized implementations for CPUs. We use four increasingly complex cost functions to evaluate the performance with respect to memory consumption and gradient computation times. Using vectorization, CPU and GPU memory consumption could be substantially reduced compared to the naive reference implementation, in some cases even by an order of complexity. The vectorization allowed usage of optimized parallel libraries during forward and reverse passes which resulted in high speedups for the vectorized CPU version compared to the naive reference implementation. The GPU version achieved an additional speedup of 7.5 ± 4.4, showing that the processing power of GPUs can be utilized for AAD using this concept. Furthermore, we show how this software can be systematically extended for more complex problems such as nonlinear absorption reconstruction for fluorescence-mediated tomography.

GPU-Accelerated Adjoint Algorithmic Differentiation

PubMed Central

Gremse, Felix; Höfter, Andreas; Razik, Lukas; Kiessling, Fabian; Naumann, Uwe

2015-01-01

Many scientific problems such as classifier training or medical image reconstruction can be expressed as minimization of differentiable real-valued cost functions and solved with iterative gradient-based methods. Adjoint algorithmic differentiation (AAD) enables automated computation of gradients of such cost functions implemented as computer programs. To backpropagate adjoint derivatives, excessive memory is potentially required to store the intermediate partial derivatives on a dedicated data structure, referred to as the “tape”. Parallelization is difficult because threads need to synchronize their accesses during taping and backpropagation. This situation is aggravated for many-core architectures, such as Graphics Processing Units (GPUs), because of the large number of light-weight threads and the limited memory size in general as well as per thread. We show how these limitations can be mediated if the cost function is expressed using GPU-accelerated vector and matrix operations which are recognized as intrinsic functions by our AAD software. We compare this approach with naive and vectorized implementations for CPUs. We use four increasingly complex cost functions to evaluate the performance with respect to memory consumption and gradient computation times. Using vectorization, CPU and GPU memory consumption could be substantially reduced compared to the naive reference implementation, in some cases even by an order of complexity. The vectorization allowed usage of optimized parallel libraries during forward and reverse passes which resulted in high speedups for the vectorized CPU version compared to the naive reference implementation. The GPU version achieved an additional speedup of 7.5 ± 4.4, showing that the processing power of GPUs can be utilized for AAD using this concept. Furthermore, we show how this software can be systematically extended for more complex problems such as nonlinear absorption reconstruction for fluorescence-mediated tomography. PMID:26941443

Adaptation of a Multi-Block Structured Solver for Effective Use in a Hybrid CPU/GPU Massively Parallel Environment

NASA Astrophysics Data System (ADS)

Gutzwiller, David; Gontier, Mathieu; Demeulenaere, Alain

2014-11-01

Multi-Block structured solvers hold many advantages over their unstructured counterparts, such as a smaller memory footprint and efficient serial performance. Historically, multi-block structured solvers have not been easily adapted for use in a High Performance Computing (HPC) environment, and the recent trend towards hybrid GPU/CPU architectures has further complicated the situation. This paper will elaborate on developments and innovations applied to the NUMECA FINE/Turbo solver that have allowed near-linear scalability with real-world problems on over 250 hybrid GPU/GPU cluster nodes. Discussion will focus on the implementation of virtual partitioning and load balancing algorithms using a novel meta-block concept. This implementation is transparent to the user, allowing all pre- and post-processing steps to be performed using a simple, unpartitioned grid topology. Additional discussion will elaborate on developments that have improved parallel performance, including fully parallel I/O with the ADIOS API and the GPU porting of the computationally heavy CPUBooster convergence acceleration module. Head of HPC and Release Management, Numeca International.

On the Development of an Efficient Parallel Hybrid Solver with Application to Acoustically Treated Aero-Engine Nacelles

NASA Technical Reports Server (NTRS)

Watson, Willie R.; Nark, Douglas M.; Nguyen, Duc T.; Tungkahotara, Siroj

2006-01-01

A finite element solution to the convected Helmholtz equation in a nonuniform flow is used to model the noise field within 3-D acoustically treated aero-engine nacelles. Options to select linear or cubic Hermite polynomial basis functions and isoparametric elements are included. However, the key feature of the method is a domain decomposition procedure that is based upon the inter-mixing of an iterative and a direct solve strategy for solving the discrete finite element equations. This procedure is optimized to take full advantage of sparsity and exploit the increased memory and parallel processing capability of modern computer architectures. Example computations are presented for the Langley Flow Impedance Test facility and a rectangular mapping of a full scale, generic aero-engine nacelle. The accuracy and parallel performance of this new solver are tested on both model problems using a supercomputer that contains hundreds of central processing units. Results show that the method gives extremely accurate attenuation predictions, achieves super-linear speedup over hundreds of CPUs, and solves upward of 25 million complex equations in a quarter of an hour.

Auditory short-term memory in the primate auditory cortex.

PubMed

Scott, Brian H; Mishkin, Mortimer

2016-06-01

Sounds are fleeting, and assembling the sequence of inputs at the ear into a coherent percept requires auditory memory across various time scales. Auditory short-term memory comprises at least two components: an active ׳working memory' bolstered by rehearsal, and a sensory trace that may be passively retained. Working memory relies on representations recalled from long-term memory, and their rehearsal may require phonological mechanisms unique to humans. The sensory component, passive short-term memory (pSTM), is tractable to study in nonhuman primates, whose brain architecture and behavioral repertoire are comparable to our own. This review discusses recent advances in the behavioral and neurophysiological study of auditory memory with a focus on single-unit recordings from macaque monkeys performing delayed-match-to-sample (DMS) tasks. Monkeys appear to employ pSTM to solve these tasks, as evidenced by the impact of interfering stimuli on memory performance. In several regards, pSTM in monkeys resembles pitch memory in humans, and may engage similar neural mechanisms. Neural correlates of DMS performance have been observed throughout the auditory and prefrontal cortex, defining a network of areas supporting auditory STM with parallels to that supporting visual STM. These correlates include persistent neural firing, or a suppression of firing, during the delay period of the memory task, as well as suppression or (less commonly) enhancement of sensory responses when a sound is repeated as a ׳match' stimulus. Auditory STM is supported by a distributed temporo-frontal network in which sensitivity to stimulus history is an intrinsic feature of auditory processing. This article is part of a Special Issue entitled SI: Auditory working memory. Published by Elsevier B.V.

A distributed parallel storage architecture and its potential application within EOSDIS

NASA Technical Reports Server (NTRS)

Johnston, William E.; Tierney, Brian; Feuquay, Jay; Butzer, Tony

1994-01-01

We describe the architecture, implementation, use of a scalable, high performance, distributed-parallel data storage system developed in the ARPA funded MAGIC gigabit testbed. A collection of wide area distributed disk servers operate in parallel to provide logical block level access to large data sets. Operated primarily as a network-based cache, the architecture supports cooperation among independently owned resources to provide fast, large-scale, on-demand storage to support data handling, simulation, and computation.

NETRA: A parallel architecture for integrated vision systems. 1: Architecture and organization

NASA Technical Reports Server (NTRS)

Choudhary, Alok N.; Patel, Janak H.; Ahuja, Narendra

1989-01-01

Computer vision is regarded as one of the most complex and computationally intensive problems. An integrated vision system (IVS) is considered to be a system that uses vision algorithms from all levels of processing for a high level application (such as object recognition). A model of computation is presented for parallel processing for an IVS. Using the model, desired features and capabilities of a parallel architecture suitable for IVSs are derived. Then a multiprocessor architecture (called NETRA) is presented. This architecture is highly flexible without the use of complex interconnection schemes. The topology of NETRA is recursively defined and hence is easily scalable from small to large systems. Homogeneity of NETRA permits fault tolerance and graceful degradation under faults. It is a recursively defined tree-type hierarchical architecture where each of the leaf nodes consists of a cluster of processors connected with a programmable crossbar with selective broadcast capability to provide for desired flexibility. A qualitative evaluation of NETRA is presented. Then general schemes are described to map parallel algorithms onto NETRA. Algorithms are classified according to their communication requirements for parallel processing. An extensive analysis of inter-cluster communication strategies in NETRA is presented, and parameters affecting performance of parallel algorithms when mapped on NETRA are discussed. Finally, a methodology to evaluate performance of algorithms on NETRA is described.

Electromagnetic physics models for parallel computing architectures

DOE PAGES

Amadio, G.; Ananya, A.; Apostolakis, J.; ...

2016-11-21

The recent emergence of hardware architectures characterized by many-core or accelerated processors has opened new opportunities for concurrent programming models taking advantage of both SIMD and SIMT architectures. GeantV, a next generation detector simulation, has been designed to exploit both the vector capability of mainstream CPUs and multi-threading capabilities of coprocessors including NVidia GPUs and Intel Xeon Phi. The characteristics of these architectures are very different in terms of the vectorization depth and type of parallelization needed to achieve optimal performance. In this paper we describe implementation of electromagnetic physics models developed for parallel computing architectures as a part ofmore » the GeantV project. Finally, the results of preliminary performance evaluation and physics validation are presented as well.« less

Kanerva's sparse distributed memory: An associative memory algorithm well-suited to the Connection Machine

NASA Technical Reports Server (NTRS)

Rogers, David

1988-01-01

The advent of the Connection Machine profoundly changes the world of supercomputers. The highly nontraditional architecture makes possible the exploration of algorithms that were impractical for standard Von Neumann architectures. Sparse distributed memory (SDM) is an example of such an algorithm. Sparse distributed memory is a particularly simple and elegant formulation for an associative memory. The foundations for sparse distributed memory are described, and some simple examples of using the memory are presented. The relationship of sparse distributed memory to three important computational systems is shown: random-access memory, neural networks, and the cerebellum of the brain. Finally, the implementation of the algorithm for sparse distributed memory on the Connection Machine is discussed.

Parallel architecture for rapid image generation and analysis

DOE Office of Scientific and Technical Information (OSTI.GOV)

Nerheim, R.J.

1987-01-01

A multiprocessor architecture inspired by the Disney multiplane camera is proposed. For many applications, this approach produces a natural mapping of processors to objects in a scene. Such a mapping promotes parallelism and reduces the hidden-surface work with minimal interprocessor communication and low-overhead cost. Existing graphics architectures store the final picture as a monolithic entity. The architecture here stores each object's image separately. It assembles the final composite picture from component images only when the video display needs to be refreshed. This organization simplifies the work required to animate moving objects that occlude other objects. In addition, the architecture hasmore » multiple processors that generate the component images in parallel. This further shortens the time needed to create a composite picture. In addition to generating images for animation, the architecture has the ability to decompose images.« less

Parallel Logic Programming and Parallel Systems Software and Hardware

DTIC Science & Technology

1989-07-29

Conference, Dallas TX. January 1985. (55) [Rous75] Roussel, P., "PROLOG: Manuel de Reference et d’Uilisation", Group d’ Intelligence Artificielle , Universite d...completed. Tools were provided for software development using artificial intelligence techniques. Al software for massively parallel architectures was...using artificial intelligence tech- niques. Al software for massively parallel architectures was started. 1. Introduction We describe research conducted

Standardization and validation of a parallel form of the verbal and non-verbal recognition memory test in an Italian population sample.

PubMed

Smirni, Daniela; Smirni, Pietro; Di Martino, Giovanni; Cipolotti, Lisa; Oliveri, Massimiliano; Turriziani, Patrizia

2018-05-04

In the neuropsychological assessment of several neurological conditions, recognition memory evaluation is requested. Recognition seems to be more appropriate than recall to study verbal and non-verbal memory, because interferences of psychological and emotional disorders are less relevant in the recognition than they are in recall memory paradigms. In many neurological disorders, longitudinal repeated assessments are needed to monitor the effectiveness of rehabilitation programs or pharmacological treatments on the recovery of memory. In order to contain the practice effect in repeated neuropsychological evaluations, it is necessary the use of parallel forms of the tests. Having two parallel forms of the same test, that kept administration procedures and scoring constant, is a great advantage in both clinical practice, for the monitoring of memory disorder, and in experimental practice, to allow the repeated evaluation of memory on healthy and neurological subjects. First aim of the present study was to provide normative values in an Italian sample (n = 160) for a parallel form of a verbal and non-verbal recognition memory battery. Multiple regression analysis revealed significant effects of age and education on recognition memory performance, whereas sex did not reach a significant probability level. Inferential cutoffs have been determined and equivalent scores computed. Secondly, the study aimed to validate the equivalence of the two parallel forms of the Recognition Memory Test. The correlations analyses between the total scores of the two versions of the test and correlation between the three subtasks revealed that the two forms are parallel and the subtasks are equivalent for difficulty.

Energy reduction through voltage scaling and lightweight checking

NASA Astrophysics Data System (ADS)

Kadric, Edin

As the semiconductor roadmap reaches smaller feature sizes and the end of Dennard Scaling, design goals change, and managing the power envelope often dominates delay minimization. Voltage scaling remains a powerful tool to reduce energy. We find that it results in about 60% geomean energy reduction on top of other common low-energy optimizations with 22nm CMOS technology. However, when voltage is reduced, it becomes easier for noise and particle strikes to upset a node, potentially causing Silent Data Corruption (SDC). The 60% energy reduction, therefore, comes with a significant drop in reliability. Duplication with checking and triple-modular redundancy are traditional approaches used to combat transient errors, but spending 2--3x the energy for redundant computation can diminish or reverse the benefits of voltage scaling. As an alternative, we explore the opportunity to use checking operations that are cheaper than the base computation they are guarding. We devise a classification system for applications and their lightweight checking characteristics. In particular, we identify and evaluate the effectiveness of lightweight checks in a broad set of common tasks in scientific computing and signal processing. We find that the lightweight checks cost only a fraction of the base computation (0-25%) and allow us to recover the reliability losses from voltage scaling. Overall, we show about 50% net energy reduction without compromising reliability compared to operation at the nominal voltage. We use FPGAs (Field-Programmable Gate Arrays) in our work, although the same ideas can be applied to different systems. On top of voltage scaling, we explore other common low-energy techniques for FPGAs: transmission gates, gate boosting, power gating, low-leakage (high-Vth) processes, and dual-V dd architectures. We do not scale voltage for memories, so lower voltages help us reduce logic and interconnect energy, but not memory energy. At lower voltages, memories become dominant, and we get diminishing returns from continuing to scale voltage. To ensure that memories do not become a bottleneck, we also design an energy-robust FPGA memory architecture, which attempts to minimize communication energy due to mismatches between application and architecture. We do this alongside application parallelism tuning. We show our techniques on a wide range of applications, including a large real-time system used for Wide-Area Motion Imaging (WAMI).

Anatomically constrained neural network models for the categorization of facial expression

NASA Astrophysics Data System (ADS)

McMenamin, Brenton W.; Assadi, Amir H.

2004-12-01

The ability to recognize facial expression in humans is performed with the amygdala which uses parallel processing streams to identify the expressions quickly and accurately. Additionally, it is possible that a feedback mechanism may play a role in this process as well. Implementing a model with similar parallel structure and feedback mechanisms could be used to improve current facial recognition algorithms for which varied expressions are a source for error. An anatomically constrained artificial neural-network model was created that uses this parallel processing architecture and feedback to categorize facial expressions. The presence of a feedback mechanism was not found to significantly improve performance for models with parallel architecture. However the use of parallel processing streams significantly improved accuracy over a similar network that did not have parallel architecture. Further investigation is necessary to determine the benefits of using parallel streams and feedback mechanisms in more advanced object recognition tasks.

Anatomically constrained neural network models for the categorization of facial expression

NASA Astrophysics Data System (ADS)

McMenamin, Brenton W.; Assadi, Amir H.

2005-01-01

The ability to recognize facial expression in humans is performed with the amygdala which uses parallel processing streams to identify the expressions quickly and accurately. Additionally, it is possible that a feedback mechanism may play a role in this process as well. Implementing a model with similar parallel structure and feedback mechanisms could be used to improve current facial recognition algorithms for which varied expressions are a source for error. An anatomically constrained artificial neural-network model was created that uses this parallel processing architecture and feedback to categorize facial expressions. The presence of a feedback mechanism was not found to significantly improve performance for models with parallel architecture. However the use of parallel processing streams significantly improved accuracy over a similar network that did not have parallel architecture. Further investigation is necessary to determine the benefits of using parallel streams and feedback mechanisms in more advanced object recognition tasks.

Parallel digital forensics infrastructure.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Liebrock, Lorie M.; Duggan, David Patrick

2009-10-01

This report documents the architecture and implementation of a Parallel Digital Forensics infrastructure. This infrastructure is necessary for supporting the design, implementation, and testing of new classes of parallel digital forensics tools. Digital Forensics has become extremely difficult with data sets of one terabyte and larger. The only way to overcome the processing time of these large sets is to identify and develop new parallel algorithms for performing the analysis. To support algorithm research, a flexible base infrastructure is required. A candidate architecture for this base infrastructure was designed, instantiated, and tested by this project, in collaboration with New Mexicomore » Tech. Previous infrastructures were not designed and built specifically for the development and testing of parallel algorithms. With the size of forensics data sets only expected to increase significantly, this type of infrastructure support is necessary for continued research in parallel digital forensics. This report documents the implementation of the parallel digital forensics (PDF) infrastructure architecture and implementation.« less

The BLAZE language - A parallel language for scientific programming

NASA Technical Reports Server (NTRS)

Mehrotra, Piyush; Van Rosendale, John

1987-01-01

A Pascal-like scientific programming language, BLAZE, is described. BLAZE contains array arithmetic, forall loops, and APL-style accumulation operators, which allow natural expression of fine grained parallelism. It also employs an applicative or functional procedure invocation mechanism, which makes it easy for compilers to extract coarse grained parallelism using machine specific program restructuring. Thus BLAZE should allow one to achieve highly parallel execution on multiprocessor architectures, while still providing the user with conceptually sequential control flow. A central goal in the design of BLAZE is portability across a broad range of parallel architectures. The multiple levels of parallelism present in BLAZE code, in principle, allow a compiler to extract the types of parallelism appropriate for the given architecture while neglecting the remainder. The features of BLAZE are described and it is shown how this language would be used in typical scientific programming.

Optical RAM-enabled cache memory and optical routing for chip multiprocessors: technologies and architectures

NASA Astrophysics Data System (ADS)

Pleros, Nikos; Maniotis, Pavlos; Alexoudi, Theonitsa; Fitsios, Dimitris; Vagionas, Christos; Papaioannou, Sotiris; Vyrsokinos, K.; Kanellos, George T.

2014-03-01

The processor-memory performance gap, commonly referred to as "Memory Wall" problem, owes to the speed mismatch between processor and electronic RAM clock frequencies, forcing current Chip Multiprocessor (CMP) configurations to consume more than 50% of the chip real-estate for caching purposes. In this article, we present our recent work spanning from Si-based integrated optical RAM cell architectures up to complete optical cache memory architectures for Chip Multiprocessor configurations. Moreover, we discuss on e/o router subsystems with up to Tb/s routing capacity for cache interconnection purposes within CMP configurations, currently pursued within the FP7 PhoxTrot project.

Scaling deep learning workloads: NVIDIA DGX-1/Pascal and Intel Knights Landing

DOE Office of Scientific and Technical Information (OSTI.GOV)

Gawande, Nitin A.; Landwehr, Joshua B.; Daily, Jeffrey A.

Deep Learning (DL) algorithms have become ubiquitous in data analytics. As a result, major computing vendors --- including NVIDIA, Intel, AMD, and IBM --- have architectural road-maps influenced by DL workloads. Furthermore, several vendors have recently advertised new computing products as accelerating large DL workloads. Unfortunately, it is difficult for data scientists to quantify the potential of these different products. This paper provides a performance and power analysis of important DL workloads on two major parallel architectures: NVIDIA DGX-1 (eight Pascal P100 GPUs interconnected with NVLink) and Intel Knights Landing (KNL) CPUs interconnected with Intel Omni-Path or Cray Aries. Ourmore » evaluation consists of a cross section of convolutional neural net workloads: CifarNet, AlexNet, GoogLeNet, and ResNet50 topologies using the Cifar10 and ImageNet datasets. The workloads are vendor-optimized for each architecture. Our analysis indicates that although GPUs provide the highest overall performance, the gap can close for some convolutional networks; and the KNL can be competitive in performance/watt. We find that NVLink facilitates scaling efficiency on GPUs. However, its importance is heavily dependent on neural network architecture. Furthermore, for weak-scaling --- sometimes encouraged by restricted GPU memory --- NVLink is less important.« less

A parallel solver for huge dense linear systems

NASA Astrophysics Data System (ADS)

Badia, J. M.; Movilla, J. L.; Climente, J. I.; Castillo, M.; Marqués, M.; Mayo, R.; Quintana-Ortí, E. S.; Planelles, J.

2011-11-01

HDSS (Huge Dense Linear System Solver) is a Fortran Application Programming Interface (API) to facilitate the parallel solution of very large dense systems to scientists and engineers. The API makes use of parallelism to yield an efficient solution of the systems on a wide range of parallel platforms, from clusters of processors to massively parallel multiprocessors. It exploits out-of-core strategies to leverage the secondary memory in order to solve huge linear systems O(100.000). The API is based on the parallel linear algebra library PLAPACK, and on its Out-Of-Core (OOC) extension POOCLAPACK. Both PLAPACK and POOCLAPACK use the Message Passing Interface (MPI) as the communication layer and BLAS to perform the local matrix operations. The API provides a friendly interface to the users, hiding almost all the technical aspects related to the parallel execution of the code and the use of the secondary memory to solve the systems. In particular, the API can automatically select the best way to store and solve the systems, depending of the dimension of the system, the number of processes and the main memory of the platform. Experimental results on several parallel platforms report high performance, reaching more than 1 TFLOP with 64 cores to solve a system with more than 200 000 equations and more than 10 000 right-hand side vectors. New version program summaryProgram title: Huge Dense System Solver (HDSS) Catalogue identifier: AEHU_v1_1 Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEHU_v1_1.html Program obtainable from: CPC Program Library, Queen's University, Belfast, N. Ireland Licensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.html No. of lines in distributed program, including test data, etc.: 87 062 No. of bytes in distributed program, including test data, etc.: 1 069 110 Distribution format: tar.gz Programming language: Fortran90, C Computer: Parallel architectures: multiprocessors, computer clusters Operating system: Linux/Unix Has the code been vectorized or parallelized?: Yes, includes MPI primitives. RAM: Tested for up to 190 GB Classification: 6.5 External routines: MPI ( http://www.mpi-forum.org/), BLAS ( http://www.netlib.org/blas/), PLAPACK ( http://www.cs.utexas.edu/~plapack/), POOCLAPACK ( ftp://ftp.cs.utexas.edu/pub/rvdg/PLAPACK/pooclapack.ps) (code for PLAPACK and POOCLAPACK is included in the distribution). Catalogue identifier of previous version: AEHU_v1_0 Journal reference of previous version: Comput. Phys. Comm. 182 (2011) 533 Does the new version supersede the previous version?: Yes Nature of problem: Huge scale dense systems of linear equations, Ax=B, beyond standard LAPACK capabilities. Solution method: The linear systems are solved by means of parallelized routines based on the LU factorization, using efficient secondary storage algorithms when the available main memory is insufficient. Reasons for new version: In many applications we need to guarantee a high accuracy in the solution of very large linear systems and we can do it by using double-precision arithmetic. Summary of revisions: Version 1.1 Can be used to solve linear systems using double-precision arithmetic. New version of the initialization routine. The user can choose the kind of arithmetic and the values of several parameters of the environment. Running time: About 5 hours to solve a system with more than 200 000 equations and more than 10 000 right-hand side vectors using double-precision arithmetic on an eight-node commodity cluster with a total of 64 Intel cores.

A numerical differentiation library exploiting parallel architectures

NASA Astrophysics Data System (ADS)

Voglis, C.; Hadjidoukas, P. E.; Lagaris, I. E.; Papageorgiou, D. G.

2009-08-01

We present a software library for numerically estimating first and second order partial derivatives of a function by finite differencing. Various truncation schemes are offered resulting in corresponding formulas that are accurate to order O(h), O(h), and O(h), h being the differencing step. The derivatives are calculated via forward, backward and central differences. Care has been taken that only feasible points are used in the case where bound constraints are imposed on the variables. The Hessian may be approximated either from function or from gradient values. There are three versions of the software: a sequential version, an OpenMP version for shared memory architectures and an MPI version for distributed systems (clusters). The parallel versions exploit the multiprocessing capability offered by computer clusters, as well as modern multi-core systems and due to the independent character of the derivative computation, the speedup scales almost linearly with the number of available processors/cores. Program summaryProgram title: NDL (Numerical Differentiation Library) Catalogue identifier: AEDG_v1_0 Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEDG_v1_0.html Program obtainable from: CPC Program Library, Queen's University, Belfast, N. Ireland Licensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.html No. of lines in distributed program, including test data, etc.: 73 030 No. of bytes in distributed program, including test data, etc.: 630 876 Distribution format: tar.gz Programming language: ANSI FORTRAN-77, ANSI C, MPI, OPENMP Computer: Distributed systems (clusters), shared memory systems Operating system: Linux, Solaris Has the code been vectorised or parallelized?: Yes RAM: The library uses O(N) internal storage, N being the dimension of the problem Classification: 4.9, 4.14, 6.5 Nature of problem: The numerical estimation of derivatives at several accuracy levels is a common requirement in many computational tasks, such as optimization, solution of nonlinear systems, etc. The parallel implementation that exploits systems with multiple CPUs is very important for large scale and computationally expensive problems. Solution method: Finite differencing is used with carefully chosen step that minimizes the sum of the truncation and round-off errors. The parallel versions employ both OpenMP and MPI libraries. Restrictions: The library uses only double precision arithmetic. Unusual features: The software takes into account bound constraints, in the sense that only feasible points are used to evaluate the derivatives, and given the level of the desired accuracy, the proper formula is automatically employed. Running time: Running time depends on the function's complexity. The test run took 15 ms for the serial distribution, 0.6 s for the OpenMP and 4.2 s for the MPI parallel distribution on 2 processors.

Impact of Cognitive Architectures on Human-Computer Interaction

DTIC Science & Technology

2014-09-01

activation, reinforced learning, emotion, semantic memory , episodic memory , and visual imagery.12 In 2010 Rosenbloom created a variant of the Soar...being added to almost every new version. In 2004 Nuxoll and Laird added episodic memory to the Soar architecture.11 In 2008 Laird presented...York (NY): Psychology Press; 2014; p. 1–50. 11. Nuxoll A, Laird JE. A cognitive model of episodic memory integrated with a general cognitive

Memory Reconsolidation and Computational Learning

DTIC Science & Technology

2010-03-01

Cooper and H.T. Siegelmann, "Memory Reconsolidation for Natural Language Processing," Cognitive Neurodynamics , 3, 2009: 365-372. M.M. Olsen, N...computerized memories and other state of the art cognitive architectures, our memory system has the ability to process on-line and in real-time as...on both continuous and binary inputs, unlike state of the art methods in case based reasoning and in cognitive architectures, which are bound to

WARP3D-Release 10.8: Dynamic Nonlinear Analysis of Solids using a Preconditioned Conjugate Gradient Software Architecture

NASA Technical Reports Server (NTRS)

Koppenhoefer, Kyle C.; Gullerud, Arne S.; Ruggieri, Claudio; Dodds, Robert H., Jr.; Healy, Brian E.

1998-01-01

This report describes theoretical background material and commands necessary to use the WARP3D finite element code. WARP3D is under continuing development as a research code for the solution of very large-scale, 3-D solid models subjected to static and dynamic loads. Specific features in the code oriented toward the investigation of ductile fracture in metals include a robust finite strain formulation, a general J-integral computation facility (with inertia, face loading), an element extinction facility to model crack growth, nonlinear material models including viscoplastic effects, and the Gurson-Tver-gaard dilatant plasticity model for void growth. The nonlinear, dynamic equilibrium equations are solved using an incremental-iterative, implicit formulation with full Newton iterations to eliminate residual nodal forces. The history integration of the nonlinear equations of motion is accomplished with Newmarks Beta method. A central feature of WARP3D involves the use of a linear-preconditioned conjugate gradient (LPCG) solver implemented in an element-by-element format to replace a conventional direct linear equation solver. This software architecture dramatically reduces both the memory requirements and CPU time for very large, nonlinear solid models since formation of the assembled (dynamic) stiffness matrix is avoided. Analyses thus exhibit the numerical stability for large time (load) steps provided by the implicit formulation coupled with the low memory requirements characteristic of an explicit code. In addition to the much lower memory requirements of the LPCG solver, the CPU time required for solution of the linear equations during each Newton iteration is generally one-half or less of the CPU time required for a traditional direct solver. All other computational aspects of the code (element stiffnesses, element strains, stress updating, element internal forces) are implemented in the element-by- element, blocked architecture. This greatly improves vectorization of the code on uni-processor hardware and enables straightforward parallel-vector processing of element blocks on multi-processor hardware.

Neural networks and applications tutorial

NASA Astrophysics Data System (ADS)

Guyon, I.

1991-09-01

The importance of neural networks has grown dramatically during this decade. While only a few years ago they were primarily of academic interest, now dozens of companies and many universities are investigating the potential use of these systems and products are beginning to appear. The idea of building a machine whose architecture is inspired by that of the brain has roots which go far back in history. Nowadays, technological advances of computers and the availability of custom integrated circuits, permit simulations of hundreds or even thousands of neurons. In conjunction, the growing interest in learning machines, non-linear dynamics and parallel computation spurred renewed attention in artificial neural networks. Many tentative applications have been proposed, including decision systems (associative memories, classifiers, data compressors and optimizers), or parametric models for signal processing purposes (system identification, automatic control, noise canceling, etc.). While they do not always outperform standard methods, neural network approaches are already used in some real world applications for pattern recognition and signal processing tasks. The tutorial is divided into six lectures, that where presented at the Third Graduate Summer Course on Computational Physics (September 3-7, 1990) on Parallel Architectures and Applications, organized by the European Physical Society: (1) Introduction: machine learning and biological computation. (2) Adaptive artificial neurons (perceptron, ADALINE, sigmoid units, etc.): learning rules and implementations. (3) Neural network systems: architectures, learning algorithms. (4) Applications: pattern recognition, signal processing, etc. (5) Elements of learning theory: how to build networks which generalize. (6) A case study: a neural network for on-line recognition of handwritten alphanumeric characters.

Super and parallel computers and their impact on civil engineering

DOE Office of Scientific and Technical Information (OSTI.GOV)

Kamat, M.P.

1986-01-01

This book presents the papers given at a conference on the use of supercomputers in civil engineering. Topics considered at the conference included solving nonlinear equations on a hypercube, a custom architectured parallel processing system, distributed data processing, algorithms, computer architecture, parallel processing, vector processing, computerized simulation, and cost benefit analysis.

Design of a massively parallel computer using bit serial processing elements

NASA Technical Reports Server (NTRS)

Aburdene, Maurice F.; Khouri, Kamal S.; Piatt, Jason E.; Zheng, Jianqing

1995-01-01

A 1-bit serial processor designed for a parallel computer architecture is described. This processor is used to develop a massively parallel computational engine, with a single instruction-multiple data (SIMD) architecture. The computer is simulated and tested to verify its operation and to measure its performance for further development.

[CMACPAR an modified parallel neuro-controller for control processes].

PubMed

Ramos, E; Surós, R

1999-01-01

CMACPAR is a Parallel Neurocontroller oriented to real time systems as for example Control Processes. Its characteristics are mainly a fast learning algorithm, a reduced number of calculations, great generalization capacity, local learning and intrinsic parallelism. This type of neurocontroller is used in real time applications required by refineries, hydroelectric centers, factories, etc. In this work we present the analysis and the parallel implementation of a modified scheme of the Cerebellar Model CMAC for the n-dimensional space projection using a mean granularity parallel neurocontroller. The proposed memory management allows for a significant memory reduction in training time and required memory size.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Janjusic, Tommy; Kartsaklis, Christos

Memory scalability is an enduring problem and bottleneck that plagues many parallel codes. Parallel codes designed for High Performance Systems are typically designed over the span of several, and in some instances 10+, years. As a result, optimization practices which were appropriate for earlier systems may no longer be valid and thus require careful optimization consideration. Specifically, parallel codes whose memory footprint is a function of their scalability must be carefully considered for future exa-scale systems. In this paper we present a methodology and tool to study the memory scalability of parallel codes. Using our methodology we evaluate an applicationmore » s memory footprint as a function of scalability, which we coined memory efficiency, and describe our results. In particular, using our in-house tools we can pinpoint the specific application components which contribute to the application s overall memory foot-print (application data- structures, libraries, etc.).« less

ATDM LANL FleCSI: Topology and Execution Framework

DOE Office of Scientific and Technical Information (OSTI.GOV)

Bergen, Benjamin Karl

FleCSI is a compile-time configurable C++ framework designed to support multi-physics application development. As such, FleCSI attempts to provide a very general set of infrastructure design patterns that can be specialized and extended to suit the needs of a broad variety of solver and data requirements. This means that FleCSI is potentially useful to many different ECP projects. Current support includes multidimensional mesh topology, mesh geometry, and mesh adjacency information, n-dimensional hashed-tree data structures, graph partitioning interfaces, and dependency closures (to identify data dependencies between distributed-memory address spaces). FleCSI introduces a functional programming model with control, execution, and data abstractionsmore » that are consistent with state-of-the-art task-based runtimes such as Legion and Charm++. The model also provides support for fine-grained, data-parallel execution with backend support for runtimes such as OpenMP and C++17. The FleCSI abstraction layer provides the developer with insulation from the underlying runtimes, while allowing support for multiple runtime systems, including conventional models like asynchronous MPI. The intent is to give developers a concrete set of user-friendly programming tools that can be used now, while allowing flexibility in choosing runtime implementations and optimizations that can be applied to architectures and runtimes that arise in the future. This project is essential to the ECP Ristra Next-Generation Code project, part of ASC ATDM, because it provides a hierarchically parallel programming model that is consistent with the design of modern system architectures, but which allows for the straightforward expression of algorithmic parallelism in a portably performant manner.« less

Application of CHAD hydrodynamics to shock-wave problems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Trease, H.E.; O`Rourke, P.J.; Sahota, M.S.

1997-12-31

CHAD is the latest in a sequence of continually evolving computer codes written to effectively utilize massively parallel computer architectures and the latest grid generators for unstructured meshes. Its applications range from automotive design issues such as in-cylinder and manifold flows of internal combustion engines, vehicle aerodynamics, underhood cooling and passenger compartment heating, ventilation, and air conditioning to shock hydrodynamics and materials modeling. CHAD solves the full unsteady Navier-Stoke equations with the k-epsilon turbulence model in three space dimensions. The code has four major features that distinguish it from the earlier KIVA code, also developed at Los Alamos. First, itmore » is based on a node-centered, finite-volume method in which, like finite element methods, all fluid variables are located at computational nodes. The computational mesh efficiently and accurately handles all element shapes ranging from tetrahedra to hexahedra. Second, it is written in standard Fortran 90 and relies on automatic domain decomposition and a universal communication library written in standard C and MPI for unstructured grids to effectively exploit distributed-memory parallel architectures. Thus the code is fully portable to a variety of computing platforms such as uniprocessor workstations, symmetric multiprocessors, clusters of workstations, and massively parallel platforms. Third, CHAD utilizes a variable explicit/implicit upwind method for convection that improves computational efficiency in flows that have large velocity Courant number variations due to velocity of mesh size variations. Fourth, CHAD is designed to also simulate shock hydrodynamics involving multimaterial anisotropic behavior under high shear. The authors will discuss CHAD capabilities and show several sample calculations showing the strengths and weaknesses of CHAD.« less

PIMS: Memristor-Based Processing-in-Memory-and-Storage.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Cook, Jeanine

Continued progress in computing has augmented the quest for higher performance with a new quest for higher energy efficiency. This has led to the re-emergence of Processing-In-Memory (PIM) ar- chitectures that offer higher density and performance with some boost in energy efficiency. Past PIM work either integrated a standard CPU with a conventional DRAM to improve the CPU- memory link, or used a bit-level processor with Single Instruction Multiple Data (SIMD) control, but neither matched the energy consumption of the memory to the computation. We originally proposed to develop a new architecture derived from PIM that more effectively addressed energymore » efficiency for high performance scientific, data analytics, and neuromorphic applications. We also originally planned to implement a von Neumann architecture with arithmetic/logic units (ALUs) that matched the power consumption of an advanced storage array to maximize energy efficiency. Implementing this architecture in storage was our original idea, since by augmenting storage (in- stead of memory), the system could address both in-memory computation and applications that accessed larger data sets directly from storage, hence Processing-in-Memory-and-Storage (PIMS). However, as our research matured, we discovered several things that changed our original direc- tion, the most important being that a PIM that implements a standard von Neumann-type archi- tecture results in significant energy efficiency improvement, but only about a O(10) performance improvement. In addition to this, the emergence of new memory technologies moved us to propos- ing a non-von Neumann architecture, called Superstrider, implemented not in storage, but in a new DRAM technology called High Bandwidth Memory (HBM). HBM is a stacked DRAM tech- nology that includes a logic layer where an architecture such as Superstrider could potentially be implemented.« less

Concurrent extensions to the FORTRAN language for parallel programming of computational fluid dynamics algorithms

NASA Technical Reports Server (NTRS)

Weeks, Cindy Lou

1986-01-01

Experiments were conducted at NASA Ames Research Center to define multi-tasking software requirements for multiple-instruction, multiple-data stream (MIMD) computer architectures. The focus was on specifying solutions for algorithms in the field of computational fluid dynamics (CFD). The program objectives were to allow researchers to produce usable parallel application software as soon as possible after acquiring MIMD computer equipment, to provide researchers with an easy-to-learn and easy-to-use parallel software language which could be implemented on several different MIMD machines, and to enable researchers to list preferred design specifications for future MIMD computer architectures. Analysis of CFD algorithms indicated that extensions of an existing programming language, adaptable to new computer architectures, provided the best solution to meeting program objectives. The CoFORTRAN Language was written in response to these objectives and to provide researchers a means to experiment with parallel software solutions to CFD algorithms on machines with parallel architectures.

3-DIMENSIONAL Optoelectronic

NASA Astrophysics Data System (ADS)

Krishnamoorthy, Ashok Venketaraman

This thesis covers the design, analysis, optimization, and implementation of optoelectronic (N,M,F) networks. (N,M,F) networks are generic space-division networks that are well suited to implementation using optoelectronic integrated circuits and free-space optical interconnects. An (N,M,F) networks consists of N input channels each having a fanout F_{rm o}, M output channels each having a fanin F_{rm i}, and Log_{rm K}(N/F) stages of K x K switches. The functionality of the fanout, switching, and fanin stages depends on the specific application. Three applications of optoelectronic (N,M,F) networks are considered. The first is an optoelectronic (N,1,1) content -addressable memory system that achieves associative recall on two-dimensional images retrieved from a parallel-access optical memory. The design and simulation of the associative memory are discussed, and an experimental emulation of a prototype system using images from a parallel-readout optical disk is presented. The system design provides superior performance to existing electronic content-addressable memory chips in terms of capacity and search rate, and uses readily available optical disk and VLSI technologies. Next, a scalable optoelectronic (N,M,F) neural network that uses free-space holographic optical interconnects is presented. The neural architecture minimizes the number of optical transmitters needed, and provides accurate electronic fanin with low signal skew, and dendritic-type fan-in processing capability in a compact layout. Optimal data-encoding methods and circuit techniques are discussed. The implementation of an prototype optoelectronic neural system, and its application to a simple recognition task is demonstrated. Finally, the design, analysis, and optimization of a (N,N,F) self-routing, packet-switched multistage interconnection network is described. The network is suitable for parallel computing and broadband switching applications. The tradeoff between optical and electronic interconnects is examined quantitatively by varying the electronic switch size K. The performance of the (N,N,F) network versus the fanning parameter F, is also analyzed. It is shown that the optoelectronic (N,N,F) networks provide a range of performance-cost alternatives, and offer superior performance-per-cost to fully electronic switching networks and to previous networks designs.

SciSpark: Highly Interactive and Scalable Model Evaluation and Climate Metrics

NASA Astrophysics Data System (ADS)

Wilson, B. D.; Mattmann, C. A.; Waliser, D. E.; Kim, J.; Loikith, P.; Lee, H.; McGibbney, L. J.; Whitehall, K. D.

2014-12-01

Remote sensing data and climate model output are multi-dimensional arrays of massive sizes locked away in heterogeneous file formats (HDF5/4, NetCDF 3/4) and metadata models (HDF-EOS, CF) making it difficult to perform multi-stage, iterative science processing since each stage requires writing and reading data to and from disk. We are developing a lightning fast Big Data technology called SciSpark based on ApacheTM Spark. Spark implements the map-reduce paradigm for parallel computing on a cluster, but emphasizes in-memory computation, "spilling" to disk only as needed, and so outperforms the disk-based ApacheTM Hadoop by 100x in memory and by 10x on disk, and makes iterative algorithms feasible. SciSpark will enable scalable model evaluation by executing large-scale comparisons of A-Train satellite observations to model grids on a cluster of 100 to 1000 compute nodes. This 2nd generation capability for NASA's Regional Climate Model Evaluation System (RCMES) will compute simple climate metrics at interactive speeds, and extend to quite sophisticated iterative algorithms such as machine-learning (ML) based clustering of temperature PDFs, and even graph-based algorithms for searching for Mesocale Convective Complexes. The goals of SciSpark are to: (1) Decrease the time to compute comparison statistics and plots from minutes to seconds; (2) Allow for interactive exploration of time-series properties over seasons and years; (3) Decrease the time for satellite data ingestion into RCMES to hours; (4) Allow for Level-2 comparisons with higher-order statistics or PDF's in minutes to hours; and (5) Move RCMES into a near real time decision-making platform. We will report on: the architecture and design of SciSpark, our efforts to integrate climate science algorithms in Python and Scala, parallel ingest and partitioning (sharding) of A-Train satellite observations from HDF files and model grids from netCDF files, first parallel runs to compute comparison statistics and PDF's, and first metrics quantifying parallel speedups and memory & disk usage.

Syntactic Change in the Parallel Architecture: The Case of Parasitic Gaps

ERIC Educational Resources Information Center

Culicover, Peter W.

2017-01-01

In Jackendoff's Parallel Architecture, the well-formed expressions of a language are licensed by correspondences between phonology, syntax, and conceptual structure. I show how this architecture can be used to make sense of the existence of parasitic gap constructions. A parasitic gap is one that is rendered acceptable because of the presence of…

Address tracing for parallel machines

NASA Technical Reports Server (NTRS)

Stunkel, Craig B.; Janssens, Bob; Fuchs, W. Kent

1991-01-01

Recently implemented parallel system address-tracing methods based on several metrics are surveyed. The issues specific to collection of traces for both shared and distributed memory parallel computers are highlighted. Five general categories of address-trace collection methods are examined: hardware-captured, interrupt-based, simulation-based, altered microcode-based, and instrumented program-based traces. The problems unique to shared memory and distributed memory multiprocessors are examined separately.

OpenCL: A Parallel Programming Standard for Heterogeneous Computing Systems.

PubMed

Stone, John E; Gohara, David; Shi, Guochun

2010-05-01

We provide an overview of the key architectural features of recent microprocessor designs and describe the programming model and abstractions provided by OpenCL, a new parallel programming standard targeting these architectures.

SciSpark: In-Memory Map-Reduce for Earth Science Algorithms

NASA Astrophysics Data System (ADS)

Ramirez, P.; Wilson, B. D.; Whitehall, K. D.; Palamuttam, R. S.; Mattmann, C. A.; Shah, S.; Goodman, A.; Burke, W.

2016-12-01

We are developing a lightning fast Big Data technology called SciSpark based on ApacheTM Spark under a NASA AIST grant (PI Mattmann). Spark implements the map-reduce paradigm for parallel computing on a cluster, but emphasizes in-memory computation, "spilling" to disk only as needed, and so outperforms the disk-based Apache Hadoop by 100x in memory and by 10x on disk. SciSpark extends Spark to support Earth Science use in three ways: Efficient ingest of N-dimensional geo-located arrays (physical variables) from netCDF3/4, HDF4/5, and/or OPeNDAP URLS; Array operations for dense arrays in scala and Java using the ND4S/ND4J or Breeze libraries; Operations to "split" datasets across a Spark cluster by time or space or both. For example, a decade-long time-series of geo-variables can be split across time to enable parallel "speedups" of analysis by day, month, or season. Similarly, very high-resolution climate grids can be partitioned into spatial tiles for parallel operations across rows, columns, or blocks. In addition, using Spark's gateway into python, PySpark, one can utilize the entire ecosystem of numpy, scipy, etc. Finally, SciSpark Notebooks provide a modern eNotebook technology in which scala, python, or spark-sql codes are entered into cells in the Notebook and executed on the cluster, with results, plots, or graph visualizations displayed in "live widgets". We have exercised SciSpark by implementing three complex Use Cases: discovery and evolution of Mesoscale Convective Complexes (MCCs) in storms, yielding a graph of connected components; PDF Clustering of atmospheric state using parallel K-Means; and statistical "rollups" of geo-variables or model-to-obs. differences (i.e. mean, stddev, skewness, & kurtosis) by day, month, season, year, and multi-year. Geo-variables are ingested and split across the cluster using methods on the sciSparkContext object including netCDFVariables() for spatial decomposition and wholeNetCDFVariables() for time-series. The presentation will cover the architecture of SciSpark, the design of the scientific RDD (sRDD) data structures for N-dim. arrays, results from the three science Use Cases, example Notebooks, lessons learned from the algorithm implementations, and parallel performance metrics.

PEM-PCA: a parallel expectation-maximization PCA face recognition architecture.

PubMed

Rujirakul, Kanokmon; So-In, Chakchai; Arnonkijpanich, Banchar

2014-01-01

Principal component analysis or PCA has been traditionally used as one of the feature extraction techniques in face recognition systems yielding high accuracy when requiring a small number of features. However, the covariance matrix and eigenvalue decomposition stages cause high computational complexity, especially for a large database. Thus, this research presents an alternative approach utilizing an Expectation-Maximization algorithm to reduce the determinant matrix manipulation resulting in the reduction of the stages' complexity. To improve the computational time, a novel parallel architecture was employed to utilize the benefits of parallelization of matrix computation during feature extraction and classification stages including parallel preprocessing, and their combinations, so-called a Parallel Expectation-Maximization PCA architecture. Comparing to a traditional PCA and its derivatives, the results indicate lower complexity with an insignificant difference in recognition precision leading to high speed face recognition systems, that is, the speed-up over nine and three times over PCA and Parallel PCA.

Deployment of the OSIRIS EM-PIC code on the Intel Knights Landing architecture

NASA Astrophysics Data System (ADS)

Fonseca, Ricardo

2017-10-01

Electromagnetic particle-in-cell (EM-PIC) codes such as OSIRIS have found widespread use in modelling the highly nonlinear and kinetic processes that occur in several relevant plasma physics scenarios, ranging from astrophysical settings to high-intensity laser plasma interaction. Being computationally intensive, these codes require large scale HPC systems, and a continuous effort in adapting the algorithm to new hardware and computing paradigms. In this work, we report on our efforts on deploying the OSIRIS code on the new Intel Knights Landing (KNL) architecture. Unlike the previous generation (Knights Corner), these boards are standalone systems, and introduce several new features, include the new AVX-512 instructions and on-package MCDRAM. We will focus on the parallelization and vectorization strategies followed, as well as memory management, and present a detailed performance evaluation of code performance in comparison with the CPU code. This work was partially supported by Fundaçã para a Ciência e Tecnologia (FCT), Portugal, through Grant No. PTDC/FIS-PLA/2940/2014.

Activity-dependent synaptic plasticity of a chalcogenide electronic synapse for neuromorphic systems.

PubMed

Li, Yi; Zhong, Yingpeng; Zhang, Jinjian; Xu, Lei; Wang, Qing; Sun, Huajun; Tong, Hao; Cheng, Xiaoming; Miao, Xiangshui

2014-05-09

Nanoscale inorganic electronic synapses or synaptic devices, which are capable of emulating the functions of biological synapses of brain neuronal systems, are regarded as the basic building blocks for beyond-Von Neumann computing architecture, combining information storage and processing. Here, we demonstrate a Ag/AgInSbTe/Ag structure for chalcogenide memristor-based electronic synapses. The memristive characteristics with reproducible gradual resistance tuning are utilised to mimic the activity-dependent synaptic plasticity that serves as the basis of memory and learning. Bidirectional long-term Hebbian plasticity modulation is implemented by the coactivity of pre- and postsynaptic spikes, and the sign and degree are affected by assorted factors including the temporal difference, spike rate and voltage. Moreover, synaptic saturation is observed to be an adjustment of Hebbian rules to stabilise the growth of synaptic weights. Our results may contribute to the development of highly functional plastic electronic synapses and the further construction of next-generation parallel neuromorphic computing architecture.

New On-board Microprocessors

NASA Astrophysics Data System (ADS)

Weigand, R.

Two new processor devices have been developed for the use on board of spacecrafts. An 8-bit 8032-microcontroller targets typical controlling applications in instruments and sub-systems, or could be used as a main processor on small satellites, whereas the LEON 32-bit SPARC processor can be used for high performance controlling and data processing tasks. The ADV80S32 is fully compliant to the Intel 80x1 architecture and instruction set, extended by additional peripherals, 512 bytes on-chip RAM and a bootstrap PROM, which allows downloading the application software using the CCSDS PacketWire pro- tocol. The memory controller provides a de-multiplexed address/data bus, and allows to access up to 16 MB data and 8 MB program RAM. The peripherals have been de- signed for the specific needs of a spacecraft, such as serial interfaces compatible to RS232, PacketWire and TTC-B-01, counters/timers for extended duration and a CRC calculation unit accelerating the CCSDS TM/TC protocol. The 0.5 um Atmel manu- facturing technology (MG2RT) provides latch-up and total dose immunity; SEU fault immunity is implemented by using SEU hardened Flip-Flops and EDAC protection of internal and external memories. The maximum clock frequency of 20 MHz allows a processing power of 3 MIPS. Engineering samples are available. For SW develop- ment, various SW packages for the 8051 architecture are on the market. The LEON processor implements a 32-bit SPARC V8 architecture, including all the multiply and divide instructions, complemented by a floating-point unit (FPU). It includes several standard peripherals, such as timers/watchdog, interrupt controller, UARTs, parallel I/Os and a memory controller, allowing to use 8, 16 and 32 bit PROM, SRAM or memory mapped I/O. With on-chip separate instruction and data caches, almost one instruction per clock cycle can be reached in some applications. A 33-MHz 32-bit PCI master/target interface and a PCI arbiter allow operating the device in a plug-in card (for SW development on PC etc.), or to consider using it as a PCI master controller in an on-board system. Advanced SEU fault tolerance is in- troduced by design, using triple modular redundancy (TMR) flip-flops for all registers and EDAC protection for all memories. The device will be manufactured in a radia- tion hard Atmel 0.25 um technology, targeting 100 MHz processor clock frequency. The non fault-tolerant LEON processor VHDL model is available as free source code, and the SPARC architecture is a well-known industry standard. Therefore, know-how, software tools and operating systems are widely available.

Optical information-processing systems and architectures II; Proceedings of the Meeting, San Diego, CA, July 9-13, 1990

NASA Astrophysics Data System (ADS)

Javidi, Bahram

The present conference discusses topics in the fields of neural networks, acoustooptic signal processing, pattern recognition, phase-only processing, nonlinear signal processing, image processing, optical computing, and optical information processing. Attention is given to the optical implementation of an inner-product neural associative memory, optoelectronic associative recall via motionless-head/parallel-readout optical disk, a compact real-time acoustooptic image correlator, a multidimensional synthetic estimation filter, and a light-efficient joint transform optical correlator. Also discussed are a high-resolution spatial light modulator, compact real-time interferometric Fourier-transform processors, a fast decorrelation algorithm for permutation arrays, the optical interconnection of optical modules, and carry-free optical binary adders.

HEP - A semaphore-synchronized multiprocessor with central control. [Heterogeneous Element Processor

NASA Technical Reports Server (NTRS)

Gilliland, M. C.; Smith, B. J.; Calvert, W.

1976-01-01

The paper describes the design concept of the Heterogeneous Element Processor (HEP), a system tailored to the special needs of scientific simulation. In order to achieve high-speed computation required by simulation, HEP features a hierarchy of processes executing in parallel on a number of processors, with synchronization being largely accomplished by hardware. A full-empty-reserve scheme of synchronization is realized by zero-one-valued hardware semaphores. A typical system has, besides the control computer and the scheduler, an algebraic module, a memory module, a first-in first-out (FIFO) module, an integrator module, and an I/O module. The architecture of the scheduler and the algebraic module is examined in detail.

Arranging computer architectures to create higher-performance controllers

NASA Technical Reports Server (NTRS)

Jacklin, Stephen A.

1988-01-01

Techniques for integrating microprocessors, array processors, and other intelligent devices in control systems are reviewed, with an emphasis on the (re)arrangement of components to form distributed or parallel processing systems. Consideration is given to the selection of the host microprocessor, increasing the power and/or memory capacity of the host, multitasking software for the host, array processors to reduce computation time, the allocation of real-time and non-real-time events to different computer subsystems, intelligent devices to share the computational burden for real-time events, and intelligent interfaces to increase communication speeds. The case of a helicopter vibration-suppression and stabilization controller is analyzed as an example, and significant improvements in computation and throughput rates are demonstrated.

Automatic Data Traffic Control on DSM Architecture

NASA Technical Reports Server (NTRS)

Frumkin, Michael; Jin, Hao-Qiang; Yan, Jerry; Kwak, Dochan (Technical Monitor)

2000-01-01

We study data traffic on distributed shared memory machines and conclude that data placement and grouping improve performance of scientific codes. We present several methods which user can employ to improve data traffic in his code. We report on implementation of a tool which detects the code fragments causing data congestions and advises user on improvements of data routing in these fragments. The capabilities of the tool include deduction of data alignment and affinity from the source code; detection of the code constructs having abnormally high cache or TLB misses; generation of data placement constructs. We demonstrate the capabilities of the tool on experiments with NAS parallel benchmarks and with a simple computational fluid dynamics application ARC3D.

Extreme-scale Algorithms and Solver Resilience

DOE Office of Scientific and Technical Information (OSTI.GOV)

Dongarra, Jack

A widening gap exists between the peak performance of high-performance computers and the performance achieved by complex applications running on these platforms. Over the next decade, extreme-scale systems will present major new challenges to algorithm development that could amplify this mismatch in such a way that it prevents the productive use of future DOE Leadership computers due to the following; Extreme levels of parallelism due to multicore processors; An increase in system fault rates requiring algorithms to be resilient beyond just checkpoint/restart; Complex memory hierarchies and costly data movement in both energy and performance; Heterogeneous system architectures (mixing CPUs, GPUs,more » etc.); and Conflicting goals of performance, resilience, and power requirements.« less

Reactor Dosimetry Applications Using RAPTOR-M3G:. a New Parallel 3-D Radiation Transport Code

NASA Astrophysics Data System (ADS)

Longoni, Gianluca; Anderson, Stanwood L.

2009-08-01

The numerical solution of the Linearized Boltzmann Equation (LBE) via the Discrete Ordinates method (SN) requires extensive computational resources for large 3-D neutron and gamma transport applications due to the concurrent discretization of the angular, spatial, and energy domains. This paper will discuss the development RAPTOR-M3G (RApid Parallel Transport Of Radiation - Multiple 3D Geometries), a new 3-D parallel radiation transport code, and its application to the calculation of ex-vessel neutron dosimetry responses in the cavity of a commercial 2-loop Pressurized Water Reactor (PWR). RAPTOR-M3G is based domain decomposition algorithms, where the spatial and angular domains are allocated and processed on multi-processor computer architectures. As compared to traditional single-processor applications, this approach reduces the computational load as well as the memory requirement per processor, yielding an efficient solution methodology for large 3-D problems. Measured neutron dosimetry responses in the reactor cavity air gap will be compared to the RAPTOR-M3G predictions. This paper is organized as follows: Section 1 discusses the RAPTOR-M3G methodology; Section 2 describes the 2-loop PWR model and the numerical results obtained. Section 3 addresses the parallel performance of the code, and Section 4 concludes this paper with final remarks and future work.

A Family of ACO Routing Protocols for Mobile Ad Hoc Networks.

PubMed

Rupérez Cañas, Delfín; Sandoval Orozco, Ana Lucila; García Villalba, Luis Javier; Kim, Tai-Hoon

2017-05-22

In this work, an ACO routing protocol for mobile ad hoc networks based on AntHocNet is specified. As its predecessor, this new protocol, called AntOR, is hybrid in the sense that it contains elements from both reactive and proactive routing. Specifically, it combines a reactive route setup process with a proactive route maintenance and improvement process. Key aspects of the AntOR protocol are the disjoint-link and disjoint-node routes, separation between the regular pheromone and the virtual pheromone in the diffusion process and the exploration of routes, taking into consideration the number of hops in the best routes. In this work, a family of ACO routing protocols based on AntOR is also specified. These protocols are based on protocol successive refinements. In this work, we also present a parallelized version of AntOR that we call PAntOR. Using programming multiprocessor architectures based on the shared memory protocol, PAntOR allows running tasks in parallel using threads. This parallelization is applicable in the route setup phase, route local repair process and link failure notification. In addition, a variant of PAntOR that consists of having more than one interface, which we call PAntOR-MI (PAntOR-Multiple Interface), is specified. This approach parallelizes the sending of broadcast messages by interface through threads.

Chrestenson transform FPGA embedded factorizations.

PubMed

Corinthios, Michael J

2016-01-01

Chrestenson generalized Walsh transform factorizations for parallel processing imbedded implementations on field programmable gate arrays are presented. This general base transform, sometimes referred to as the Discrete Chrestenson transform, has received special attention in recent years. In fact, the Discrete Fourier transform and Walsh-Hadamard transform are but special cases of the Chrestenson generalized Walsh transform. Rotations of a base-p hypercube, where p is an arbitrary integer, are shown to produce dynamic contention-free memory allocation, in processor architecture. The approach is illustrated by factorizations involving the processing of matrices of the transform which are function of four variables. Parallel operations are implemented matrix multiplications. Each matrix, of dimension N × N, where N = p (n) , n integer, has a structure that depends on a variable parameter k that denotes the iteration number in the factorization process. The level of parallelism, in the form of M = p (m) processors can be chosen arbitrarily by varying m between zero to its maximum value of n - 1. The result is an equation describing the generalised parallelism factorization as a function of the four variables n, p, k and m. Applications of the approach are shown in relation to configuring field programmable gate arrays for digital signal processing applications.

OpenCL: A Parallel Programming Standard for Heterogeneous Computing Systems

PubMed Central

Stone, John E.; Gohara, David; Shi, Guochun

2010-01-01

We provide an overview of the key architectural features of recent microprocessor designs and describe the programming model and abstractions provided by OpenCL, a new parallel programming standard targeting these architectures. PMID:21037981

Developing Information Power Grid Based Algorithms and Software

NASA Technical Reports Server (NTRS)

Dongarra, Jack

1998-01-01

This exploratory study initiated our effort to understand performance modeling on parallel systems. The basic goal of performance modeling is to understand and predict the performance of a computer program or set of programs on a computer system. Performance modeling has numerous applications, including evaluation of algorithms, optimization of code implementations, parallel library development, comparison of system architectures, parallel system design, and procurement of new systems. Our work lays the basis for the construction of parallel libraries that allow for the reconstruction of application codes on several distinct architectures so as to assure performance portability. Following our strategy, once the requirements of applications are well understood, one can then construct a library in a layered fashion. The top level of this library will consist of architecture-independent geometric, numerical, and symbolic algorithms that are needed by the sample of applications. These routines should be written in a language that is portable across the targeted architectures.

Runtime support for parallelizing data mining algorithms

NASA Astrophysics Data System (ADS)

Jin, Ruoming; Agrawal, Gagan

2002-03-01

With recent technological advances, shared memory parallel machines have become more scalable, and offer large main memories and high bus bandwidths. They are emerging as good platforms for data warehousing and data mining. In this paper, we focus on shared memory parallelization of data mining algorithms. We have developed a series of techniques for parallelization of data mining algorithms, including full replication, full locking, fixed locking, optimized full locking, and cache-sensitive locking. Unlike previous work on shared memory parallelization of specific data mining algorithms, all of our techniques apply to a large number of common data mining algorithms. In addition, we propose a reduction-object based interface for specifying a data mining algorithm. We show how our runtime system can apply any of the technique we have developed starting from a common specification of the algorithm.

GPU-based parallel algorithm for blind image restoration using midfrequency-based methods

NASA Astrophysics Data System (ADS)

Xie, Lang; Luo, Yi-han; Bao, Qi-liang

2013-08-01

GPU-based general-purpose computing is a new branch of modern parallel computing, so the study of parallel algorithms specially designed for GPU hardware architecture is of great significance. In order to solve the problem of high computational complexity and poor real-time performance in blind image restoration, the midfrequency-based algorithm for blind image restoration was analyzed and improved in this paper. Furthermore, a midfrequency-based filtering method is also used to restore the image hardly with any recursion or iteration. Combining the algorithm with data intensiveness, data parallel computing and GPU execution model of single instruction and multiple threads, a new parallel midfrequency-based algorithm for blind image restoration is proposed in this paper, which is suitable for stream computing of GPU. In this algorithm, the GPU is utilized to accelerate the estimation of class-G point spread functions and midfrequency-based filtering. Aiming at better management of the GPU threads, the threads in a grid are scheduled according to the decomposition of the filtering data in frequency domain after the optimization of data access and the communication between the host and the device. The kernel parallelism structure is determined by the decomposition of the filtering data to ensure the transmission rate to get around the memory bandwidth limitation. The results show that, with the new algorithm, the operational speed is significantly increased and the real-time performance of image restoration is effectively improved, especially for high-resolution images.

PFLOTRAN User Manual: A Massively Parallel Reactive Flow and Transport Model for Describing Surface and Subsurface Processes

DOE Office of Scientific and Technical Information (OSTI.GOV)

Lichtner, Peter C.; Hammond, Glenn E.; Lu, Chuan

PFLOTRAN solves a system of generally nonlinear partial differential equations describing multi-phase, multicomponent and multiscale reactive flow and transport in porous materials. The code is designed to run on massively parallel computing architectures as well as workstations and laptops (e.g. Hammond et al., 2011). Parallelization is achieved through domain decomposition using the PETSc (Portable Extensible Toolkit for Scientific Computation) libraries for the parallelization framework (Balay et al., 1997). PFLOTRAN has been developed from the ground up for parallel scalability and has been run on up to 218 processor cores with problem sizes up to 2 billion degrees of freedom. Writtenmore » in object oriented Fortran 90, the code requires the latest compilers compatible with Fortran 2003. At the time of this writing this requires gcc 4.7.x, Intel 12.1.x and PGC compilers. As a requirement of running problems with a large number of degrees of freedom, PFLOTRAN allows reading input data that is too large to fit into memory allotted to a single processor core. The current limitation to the problem size PFLOTRAN can handle is the limitation of the HDF5 file format used for parallel IO to 32 bit integers. Noting that 2 32 = 4; 294; 967; 296, this gives an estimate of the maximum problem size that can be currently run with PFLOTRAN. Hopefully this limitation will be remedied in the near future.« less

CMOS VLSI Active-Pixel Sensor for Tracking

NASA Technical Reports Server (NTRS)

Pain, Bedabrata; Sun, Chao; Yang, Guang; Heynssens, Julie

2004-01-01

An architecture for a proposed active-pixel sensor (APS) and a design to implement the architecture in a complementary metal oxide semiconductor (CMOS) very-large-scale integrated (VLSI) circuit provide for some advanced features that are expected to be especially desirable for tracking pointlike features of stars. The architecture would also make this APS suitable for robotic- vision and general pointing and tracking applications. CMOS imagers in general are well suited for pointing and tracking because they can be configured for random access to selected pixels and to provide readout from windows of interest within their fields of view. However, until now, the architectures of CMOS imagers have not supported multiwindow operation or low-noise data collection. Moreover, smearing and motion artifacts in collected images have made prior CMOS imagers unsuitable for tracking applications. The proposed CMOS imager (see figure) would include an array of 1,024 by 1,024 pixels containing high-performance photodiode-based APS circuitry. The pixel pitch would be 9 m. The operations of the pixel circuits would be sequenced and otherwise controlled by an on-chip timing and control block, which would enable the collection of image data, during a single frame period, from either the full frame (that is, all 1,024 1,024 pixels) or from within as many as 8 different arbitrarily placed windows as large as 8 by 8 pixels each. A typical prior CMOS APS operates in a row-at-a-time ( grolling-shutter h) readout mode, which gives rise to exposure skew. In contrast, the proposed APS would operate in a sample-first/readlater mode, suppressing rolling-shutter effects. In this mode, the analog readout signals from the pixels corresponding to the windows of the interest (which windows, in the star-tracking application, would presumably contain guide stars) would be sampled rapidly by routing them through a programmable diagonal switch array to an on-chip parallel analog memory array. The diagonal-switch and memory addresses would be generated by the on-chip controller. The memory array would be large enough to hold differential signals acquired from all 8 windows during a frame period. Following the rapid sampling from all the windows, the contents of the memory array would be read out sequentially by use of a capacitive transimpedance amplifier (CTIA) at a maximum data rate of 10 MHz. This data rate is compatible with an update rate of almost 10 Hz, even in full-frame operation

Parallel hyperbolic PDE simulation on clusters: Cell versus GPU

NASA Astrophysics Data System (ADS)

Rostrup, Scott; De Sterck, Hans

2010-12-01

Increasingly, high-performance computing is looking towards data-parallel computational devices to enhance computational performance. Two technologies that have received significant attention are IBM's Cell Processor and NVIDIA's CUDA programming model for graphics processing unit (GPU) computing. In this paper we investigate the acceleration of parallel hyperbolic partial differential equation simulation on structured grids with explicit time integration on clusters with Cell and GPU backends. The message passing interface (MPI) is used for communication between nodes at the coarsest level of parallelism. Optimizations of the simulation code at the several finer levels of parallelism that the data-parallel devices provide are described in terms of data layout, data flow and data-parallel instructions. Optimized Cell and GPU performance are compared with reference code performance on a single x86 central processing unit (CPU) core in single and double precision. We further compare the CPU, Cell and GPU platforms on a chip-to-chip basis, and compare performance on single cluster nodes with two CPUs, two Cell processors or two GPUs in a shared memory configuration (without MPI). We finally compare performance on clusters with 32 CPUs, 32 Cell processors, and 32 GPUs using MPI. Our GPU cluster results use NVIDIA Tesla GPUs with GT200 architecture, but some preliminary results on recently introduced NVIDIA GPUs with the next-generation Fermi architecture are also included. This paper provides computational scientists and engineers who are considering porting their codes to accelerator environments with insight into how structured grid based explicit algorithms can be optimized for clusters with Cell and GPU accelerators. It also provides insight into the speed-up that may be gained on current and future accelerator architectures for this class of applications. Program summaryProgram title: SWsolver Catalogue identifier: AEGY_v1_0 Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEGY_v1_0.html Program obtainable from: CPC Program Library, Queen's University, Belfast, N. Ireland Licensing provisions: GPL v3 No. of lines in distributed program, including test data, etc.: 59 168 No. of bytes in distributed program, including test data, etc.: 453 409 Distribution format: tar.gz Programming language: C, CUDA Computer: Parallel Computing Clusters. Individual compute nodes may consist of x86 CPU, Cell processor, or x86 CPU with attached NVIDIA GPU accelerator. Operating system: Linux Has the code been vectorised or parallelized?: Yes. Tested on 1-128 x86 CPU cores, 1-32 Cell Processors, and 1-32 NVIDIA GPUs. RAM: Tested on Problems requiring up to 4 GB per compute node. Classification: 12 External routines: MPI, CUDA, IBM Cell SDK Nature of problem: MPI-parallel simulation of Shallow Water equations using high-resolution 2D hyperbolic equation solver on regular Cartesian grids for x86 CPU, Cell Processor, and NVIDIA GPU using CUDA. Solution method: SWsolver provides 3 implementations of a high-resolution 2D Shallow Water equation solver on regular Cartesian grids, for CPU, Cell Processor, and NVIDIA GPU. Each implementation uses MPI to divide work across a parallel computing cluster. Additional comments: Sub-program numdiff is used for the test run.

Fault tolerant onboard packet switch architecture for communication satellites: Shared memory per beam approach

NASA Technical Reports Server (NTRS)

Shalkhauser, Mary JO; Quintana, Jorge A.; Soni, Nitin J.

1994-01-01

The NASA Lewis Research Center is developing a multichannel communication signal processing satellite (MCSPS) system which will provide low data rate, direct to user, commercial communications services. The focus of current space segment developments is a flexible, high-throughput, fault tolerant onboard information switching processor. This information switching processor (ISP) is a destination-directed packet switch which performs both space and time switching to route user information among numerous user ground terminals. Through both industry study contracts and in-house investigations, several packet switching architectures were examined. A contention-free approach, the shared memory per beam architecture, was selected for implementation. The shared memory per beam architecture, fault tolerance insertion, implementation, and demonstration plans are described.

Efficient parallel implementation of active appearance model fitting algorithm on GPU.

PubMed

Wang, Jinwei; Ma, Xirong; Zhu, Yuanping; Sun, Jizhou

2014-01-01

The active appearance model (AAM) is one of the most powerful model-based object detecting and tracking methods which has been widely used in various situations. However, the high-dimensional texture representation causes very time-consuming computations, which makes the AAM difficult to apply to real-time systems. The emergence of modern graphics processing units (GPUs) that feature a many-core, fine-grained parallel architecture provides new and promising solutions to overcome the computational challenge. In this paper, we propose an efficient parallel implementation of the AAM fitting algorithm on GPUs. Our design idea is fine grain parallelism in which we distribute the texture data of the AAM, in pixels, to thousands of parallel GPU threads for processing, which makes the algorithm fit better into the GPU architecture. We implement our algorithm using the compute unified device architecture (CUDA) on the Nvidia's GTX 650 GPU, which has the latest Kepler architecture. To compare the performance of our algorithm with different data sizes, we built sixteen face AAM models of different dimensional textures. The experiment results show that our parallel AAM fitting algorithm can achieve real-time performance for videos even on very high-dimensional textures.

Efficient Parallel Implementation of Active Appearance Model Fitting Algorithm on GPU

PubMed Central

Wang, Jinwei; Ma, Xirong; Zhu, Yuanping; Sun, Jizhou

2014-01-01

The active appearance model (AAM) is one of the most powerful model-based object detecting and tracking methods which has been widely used in various situations. However, the high-dimensional texture representation causes very time-consuming computations, which makes the AAM difficult to apply to real-time systems. The emergence of modern graphics processing units (GPUs) that feature a many-core, fine-grained parallel architecture provides new and promising solutions to overcome the computational challenge. In this paper, we propose an efficient parallel implementation of the AAM fitting algorithm on GPUs. Our design idea is fine grain parallelism in which we distribute the texture data of the AAM, in pixels, to thousands of parallel GPU threads for processing, which makes the algorithm fit better into the GPU architecture. We implement our algorithm using the compute unified device architecture (CUDA) on the Nvidia's GTX 650 GPU, which has the latest Kepler architecture. To compare the performance of our algorithm with different data sizes, we built sixteen face AAM models of different dimensional textures. The experiment results show that our parallel AAM fitting algorithm can achieve real-time performance for videos even on very high-dimensional textures. PMID:24723812

Affinity-aware checkpoint restart

DOE PAGES

Saini, Ajay; Rezaei, Arash; Mueller, Frank; ...

2014-12-08

Current checkpointing techniques employed to overcome faults for HPC applications result in inferior application performance after restart from a checkpoint for a number of applications. This is due to a lack of page and core affinity awareness of the checkpoint/restart (C/R) mechanism, i.e., application tasks originally pinned to cores may be restarted on different cores, and in case of non-uniform memory architectures (NUMA), quite common today, memory pages associated with tasks on a NUMA node may be associated with a different NUMA node after restart. Here, this work contributes a novel design technique for C/R mechanisms to preserve task-to-core mapsmore » and NUMA node specific page affinities across restarts. Experimental results with BLCR, a C/R mechanism, enhanced with affinity awareness demonstrate significant performance benefits of 37%-73% for the NAS Parallel Benchmark codes and 6-12% for NAMD with negligible overheads instead of up to nearly four times longer an execution times without affinity-aware restarts on 16 cores.« less

Solution of large nonlinear quasistatic structural mechanics problems on distributed-memory multiprocessor computers

DOE Office of Scientific and Technical Information (OSTI.GOV)

Blanford, M.

1997-12-31

Most commercially-available quasistatic finite element programs assemble element stiffnesses into a global stiffness matrix, then use a direct linear equation solver to obtain nodal displacements. However, for large problems (greater than a few hundred thousand degrees of freedom), the memory size and computation time required for this approach becomes prohibitive. Moreover, direct solution does not lend itself to the parallel processing needed for today`s multiprocessor systems. This talk gives an overview of the iterative solution strategy of JAS3D, the nonlinear large-deformation quasistatic finite element program. Because its architecture is derived from an explicit transient-dynamics code, it does not ever assemblemore » a global stiffness matrix. The author describes the approach he used to implement the solver on multiprocessor computers, and shows examples of problems run on hundreds of processors and more than a million degrees of freedom. Finally, he describes some of the work he is presently doing to address the challenges of iterative convergence for ill-conditioned problems.« less

Affinity-aware checkpoint restart

DOE Office of Scientific and Technical Information (OSTI.GOV)

Saini, Ajay; Rezaei, Arash; Mueller, Frank

Current checkpointing techniques employed to overcome faults for HPC applications result in inferior application performance after restart from a checkpoint for a number of applications. This is due to a lack of page and core affinity awareness of the checkpoint/restart (C/R) mechanism, i.e., application tasks originally pinned to cores may be restarted on different cores, and in case of non-uniform memory architectures (NUMA), quite common today, memory pages associated with tasks on a NUMA node may be associated with a different NUMA node after restart. Here, this work contributes a novel design technique for C/R mechanisms to preserve task-to-core mapsmore » and NUMA node specific page affinities across restarts. Experimental results with BLCR, a C/R mechanism, enhanced with affinity awareness demonstrate significant performance benefits of 37%-73% for the NAS Parallel Benchmark codes and 6-12% for NAMD with negligible overheads instead of up to nearly four times longer an execution times without affinity-aware restarts on 16 cores.« less

The Fermilab lattice supercomputer project

DOE Office of Scientific and Technical Information (OSTI.GOV)

Fischler, M.; Atac, R.; Cook, A.

1989-02-01

The ACPMAPS system is a highly cost effective, local memory MIMD computer targeted at algorithm development and production running for gauge theory on the lattice. The machine consists of a compound hypercube of crates, each of which is a full crossbar switch containing several processors. The processing nodes are single board array processors based on the Weitek XL chip set, each with a peak power of 20 MFLOPS and supported by 8 MBytes of data memory. The system currently being assembled has a peak power of 5 GFLOPS, delivering performance at approximately $250/MFLOP. The system is programmable in C andmore » Fortran. An underpinning of software routines (CANOPY) provides an easy and natural way of coding lattice problems, such that the details of parallelism, and communication and system architecture are transparent to the user. CANOPY can easily be ported to any single CPU or MIMD system which supports C, and allows the coding of typical applications with very little effort. 3 refs., 1 fig.« less

Development of the Large-Scale Statistical Analysis System of Satellites Observations Data with Grid Datafarm Architecture

NASA Astrophysics Data System (ADS)

Yamamoto, K.; Murata, K.; Kimura, E.; Honda, R.

2006-12-01

In the Solar-Terrestrial Physics (STP) field, the amount of satellite observation data has been increasing every year. It is necessary to solve the following three problems to achieve large-scale statistical analyses of plenty of data. (i) More CPU power and larger memory and disk size are required. However, total powers of personal computers are not enough to analyze such amount of data. Super-computers provide a high performance CPU and rich memory area, but they are usually separated from the Internet or connected only for the purpose of programming or data file transfer. (ii) Most of the observation data files are managed at distributed data sites over the Internet. Users have to know where the data files are located. (iii) Since no common data format in the STP field is available now, users have to prepare reading program for each data by themselves. To overcome the problems (i) and (ii), we constructed a parallel and distributed data analysis environment based on the Gfarm reference implementation of the Grid Datafarm architecture. The Gfarm shares both computational resources and perform parallel distributed processings. In addition, the Gfarm provides the Gfarm filesystem which can be as virtual directory tree among nodes. The Gfarm environment is composed of three parts; a metadata server to manage distributed files information, filesystem nodes to provide computational resources and a client to throw a job into metadata server and manages data processing schedulings. In the present study, both data files and data processes are parallelized on the Gfarm with 6 file system nodes: CPU clock frequency of each node is Pentium V 1GHz, 256MB memory and40GB disk. To evaluate performances of the present Gfarm system, we scanned plenty of data files, the size of which is about 300MB for each, in three processing methods: sequential processing in one node, sequential processing by each node and parallel processing by each node. As a result, in comparison between the number of files and the elapsed time, parallel and distributed processing shorten the elapsed time to 1/5 than sequential processing. On the other hand, sequential processing times were shortened in another experiment, whose file size is smaller than 100KB. In this case, the elapsed time to scan one file is within one second. It implies that disk swap took place in case of parallel processing by each node. We note that the operation became unstable when the number of the files exceeded 1000. To overcome the problem (iii), we developed an original data class. This class supports our reading of data files with various data formats since it converts them into an original data format since it defines schemata for every type of data and encapsulates the structure of data files. In addition, since this class provides a function of time re-sampling, users can easily convert multiple data (array) with different time resolution into the same time resolution array. Finally, using the Gfarm, we achieved a high performance environment for large-scale statistical data analyses. It should be noted that the present method is effective only when one data file size is large enough. At present, we are restructuring the new Gfarm environment with 8 nodes: CPU is Athlon 64 x2 Dual Core 2GHz, 2GB memory and 1.2TB disk (using RAID0) for each node. Our original class is to be implemented on the new Gfarm environment. In the present talk, we show the latest results with applying the present system for data analyses with huge number of satellite observation data files.

Scalable Motion Estimation Processor Core for Multimedia System-on-Chip Applications

NASA Astrophysics Data System (ADS)

Lai, Yeong-Kang; Hsieh, Tian-En; Chen, Lien-Fei

2007-04-01

In this paper, we describe a high-throughput and scalable motion estimation processor architecture for multimedia system-on-chip applications. The number of processing elements (PEs) is scalable according to the variable algorithm parameters and the performance required for different applications. Using the PE rings efficiently and an intelligent memory-interleaving organization, the efficiency of the architecture can be increased. Moreover, using efficient on-chip memories and a data management technique can effectively decrease the power consumption and memory bandwidth. Techniques for reducing the number of interconnections and external memory accesses are also presented. Our results demonstrate that the proposed scalable PE-ringed architecture is a flexible and high-performance processor core in multimedia system-on-chip applications.

Automatic Generation of OpenMP Directives and Its Application to Computational Fluid Dynamics Codes

NASA Technical Reports Server (NTRS)

Yan, Jerry; Jin, Haoqiang; Frumkin, Michael; Yan, Jerry (Technical Monitor)

2000-01-01

The shared-memory programming model is a very effective way to achieve parallelism on shared memory parallel computers. As great progress was made in hardware and software technologies, performance of parallel programs with compiler directives has demonstrated large improvement. The introduction of OpenMP directives, the industrial standard for shared-memory programming, has minimized the issue of portability. In this study, we have extended CAPTools, a computer-aided parallelization toolkit, to automatically generate OpenMP-based parallel programs with nominal user assistance. We outline techniques used in the implementation of the tool and discuss the application of this tool on the NAS Parallel Benchmarks and several computational fluid dynamics codes. This work demonstrates the great potential of using the tool to quickly port parallel programs and also achieve good performance that exceeds some of the commercial tools.

A 0.13-µm implementation of 5 Gb/s and 3-mW folded parallel architecture for AES algorithm

NASA Astrophysics Data System (ADS)

Rahimunnisa, K.; Karthigaikumar, P.; Kirubavathy, J.; Jayakumar, J.; Kumar, S. Suresh

2014-02-01

A new architecture for encrypting and decrypting the confidential data using Advanced Encryption Standard algorithm is presented in this article. This structure combines the folded structure with parallel architecture to increase the throughput. The whole architecture achieved high throughput with less power. The proposed architecture is implemented in 0.13-µm Complementary metal-oxide-semiconductor (CMOS) technology. The proposed structure is compared with different existing structures, and from the result it is proved that the proposed structure gives higher throughput and less power compared to existing works.

Performance Evaluation of Remote Memory Access (RMA) Programming on Shared Memory Parallel Computers

NASA Technical Reports Server (NTRS)

Jin, Hao-Qiang; Jost, Gabriele; Biegel, Bryan A. (Technical Monitor)

2002-01-01

The purpose of this study is to evaluate the feasibility of remote memory access (RMA) programming on shared memory parallel computers. We discuss different RMA based implementations of selected CFD application benchmark kernels and compare them to corresponding message passing based codes. For the message-passing implementation we use MPI point-to-point and global communication routines. For the RMA based approach we consider two different libraries supporting this programming model. One is a shared memory parallelization library (SMPlib) developed at NASA Ames, the other is the MPI-2 extensions to the MPI Standard. We give timing comparisons for the different implementation strategies and discuss the performance.

Data communications in a parallel active messaging interface of a parallel computer

DOEpatents

Blocksome, Michael A.; Ratterman, Joseph D.; Smith, Brian E.

2014-09-02

Eager send data communications in a parallel active messaging interface (`PAMI`) of a parallel computer, the PAMI composed of data communications endpoints that specify a client, a context, and a task, including receiving an eager send data communications instruction with transfer data disposed in a send buffer characterized by a read/write send buffer memory address in a read/write virtual address space of the origin endpoint; determining for the send buffer a read-only send buffer memory address in a read-only virtual address space, the read-only virtual address space shared by both the origin endpoint and the target endpoint, with all frames of physical memory mapped to pages of virtual memory in the read-only virtual address space; and communicating by the origin endpoint to the target endpoint an eager send message header that includes the read-only send buffer memory address.

Data communications in a parallel active messaging interface of a parallel computer

DOEpatents

Blocksome, Michael A.; Ratterman, Joseph D.; Smith, Brian E.

2014-09-16

Eager send data communications in a parallel active messaging interface (`PAMI`) of a parallel computer, the PAMI composed of data communications endpoints that specify a client, a context, and a task, including receiving an eager send data communications instruction with transfer data disposed in a send buffer characterized by a read/write send buffer memory address in a read/write virtual address space of the origin endpoint; determining for the send buffer a read-only send buffer memory address in a read-only virtual address space, the read-only virtual address space shared by both the origin endpoint and the target endpoint, with all frames of physical memory mapped to pages of virtual memory in the read-only virtual address space; and communicating by the origin endpoint to the target endpoint an eager send message header that includes the read-only send buffer memory address.

Dynamically programmable cache

NASA Astrophysics Data System (ADS)

Nakkar, Mouna; Harding, John A.; Schwartz, David A.; Franzon, Paul D.; Conte, Thomas

1998-10-01

Reconfigurable machines have recently been used as co- processors to accelerate the execution of certain algorithms or program subroutines. The problems with the above approach include high reconfiguration time and limited partial reconfiguration. By far the most critical problems are: (1) the small on-chip memory which results in slower execution time, and (2) small FPGA areas that cannot implement large subroutines. Dynamically Programmable Cache (DPC) is a novel architecture for embedded processors which offers solutions to the above problems. To solve memory access problems, DPC processors merge reconfigurable arrays with the data cache at various cache levels to create a multi-level reconfigurable machines. As a result DPC machines have both higher data accessibility and FPGA memory bandwidth. To solve the limited FPGA resource problem, DPC processors implemented multi-context switching (Virtualization) concept. Virtualization allows implementation of large subroutines with fewer FPGA cells. Additionally, DPC processors can parallelize the execution of several operations resulting in faster execution time. In this paper, the speedup improvement for DPC machines are shown to be 5X faster than an Altera FLEX10K FPGA chip and 2X faster than a Sun Ultral SPARC station for two different algorithms (convolution and motion estimation).

A parallel architecture of interpolated timing recovery for high- speed data transfer rate and wide capture-range

NASA Astrophysics Data System (ADS)

Higashino, Satoru; Kobayashi, Shoei; Yamagami, Tamotsu

2007-06-01

High data transfer rate has been demanded for data storage devices along increasing the storage capacity. In order to increase the transfer rate, high-speed data processing techniques in read-channel devices are required. Generally, parallel architecture is utilized for the high-speed digital processing. We have developed a new architecture of Interpolated Timing Recovery (ITR) to achieve high-speed data transfer rate and wide capture-range in read-channel devices for the information storage channels. It facilitates the parallel implementation on large-scale-integration (LSI) devices.

Support for Debugging Automatically Parallelized Programs

NASA Technical Reports Server (NTRS)

Hood, Robert; Jost, Gabriele; Biegel, Bryan (Technical Monitor)

2001-01-01

This viewgraph presentation provides information on the technical aspects of debugging computer code that has been automatically converted for use in a parallel computing system. Shared memory parallelization and distributed memory parallelization entail separate and distinct challenges for a debugging program. A prototype system has been developed which integrates various tools for the debugging of automatically parallelized programs including the CAPTools Database which provides variable definition information across subroutines as well as array distribution information.

Fast multipurpose Monte Carlo simulation for proton therapy using multi- and many-core CPU architectures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Souris, Kevin, E-mail: kevin.souris@uclouvain.be; Lee, John Aldo; Sterpin, Edmond

2016-04-15

Purpose: Accuracy in proton therapy treatment planning can be improved using Monte Carlo (MC) simulations. However the long computation time of such methods hinders their use in clinical routine. This work aims to develop a fast multipurpose Monte Carlo simulation tool for proton therapy using massively parallel central processing unit (CPU) architectures. Methods: A new Monte Carlo, called MCsquare (many-core Monte Carlo), has been designed and optimized for the last generation of Intel Xeon processors and Intel Xeon Phi coprocessors. These massively parallel architectures offer the flexibility and the computational power suitable to MC methods. The class-II condensed history algorithmmore » of MCsquare provides a fast and yet accurate method of simulating heavy charged particles such as protons, deuterons, and alphas inside voxelized geometries. Hard ionizations, with energy losses above a user-specified threshold, are simulated individually while soft events are regrouped in a multiple scattering theory. Elastic and inelastic nuclear interactions are sampled from ICRU 63 differential cross sections, thereby allowing for the computation of prompt gamma emission profiles. MCsquare has been benchmarked with the GATE/GEANT4 Monte Carlo application for homogeneous and heterogeneous geometries. Results: Comparisons with GATE/GEANT4 for various geometries show deviations within 2%–1 mm. In spite of the limited memory bandwidth of the coprocessor simulation time is below 25 s for 10{sup 7} primary 200 MeV protons in average soft tissues using all Xeon Phi and CPU resources embedded in a single desktop unit. Conclusions: MCsquare exploits the flexibility of CPU architectures to provide a multipurpose MC simulation tool. Optimized code enables the use of accurate MC calculation within a reasonable computation time, adequate for clinical practice. MCsquare also simulates prompt gamma emission and can thus be used also for in vivo range verification.« less

Runtime and Architecture Support for Efficient Data Exchange in Multi-Accelerator Applications.

PubMed

Cabezas, Javier; Gelado, Isaac; Stone, John E; Navarro, Nacho; Kirk, David B; Hwu, Wen-Mei

2015-05-01

Heterogeneous parallel computing applications often process large data sets that require multiple GPUs to jointly meet their needs for physical memory capacity and compute throughput. However, the lack of high-level abstractions in previous heterogeneous parallel programming models force programmers to resort to multiple code versions, complex data copy steps and synchronization schemes when exchanging data between multiple GPU devices, which results in high software development cost, poor maintainability, and even poor performance. This paper describes the HPE runtime system, and the associated architecture support, which enables a simple, efficient programming interface for exchanging data between multiple GPUs through either interconnects or cross-node network interfaces. The runtime and architecture support presented in this paper can also be used to support other types of accelerators. We show that the simplified programming interface reduces programming complexity. The research presented in this paper started in 2009. It has been implemented and tested extensively in several generations of HPE runtime systems as well as adopted into the NVIDIA GPU hardware and drivers for CUDA 4.0 and beyond since 2011. The availability of real hardware that support key HPE features gives rise to a rare opportunity for studying the effectiveness of the hardware support by running important benchmarks on real runtime and hardware. Experimental results show that in a exemplar heterogeneous system, peer DMA and double-buffering, pinned buffers, and software techniques can improve the inter-accelerator data communication bandwidth by 2×. They can also improve the execution speed by 1.6× for a 3D finite difference, 2.5× for 1D FFT, and 1.6× for merge sort, all measured on real hardware. The proposed architecture support enables the HPE runtime to transparently deploy these optimizations under simple portable user code, allowing system designers to freely employ devices of different capabilities. We further argue that simple interfaces such as HPE are needed for most applications to benefit from advanced hardware features in practice.

Runtime and Architecture Support for Efficient Data Exchange in Multi-Accelerator Applications

PubMed Central

Cabezas, Javier; Gelado, Isaac; Stone, John E.; Navarro, Nacho; Kirk, David B.; Hwu, Wen-mei

2014-01-01

Heterogeneous parallel computing applications often process large data sets that require multiple GPUs to jointly meet their needs for physical memory capacity and compute throughput. However, the lack of high-level abstractions in previous heterogeneous parallel programming models force programmers to resort to multiple code versions, complex data copy steps and synchronization schemes when exchanging data between multiple GPU devices, which results in high software development cost, poor maintainability, and even poor performance. This paper describes the HPE runtime system, and the associated architecture support, which enables a simple, efficient programming interface for exchanging data between multiple GPUs through either interconnects or cross-node network interfaces. The runtime and architecture support presented in this paper can also be used to support other types of accelerators. We show that the simplified programming interface reduces programming complexity. The research presented in this paper started in 2009. It has been implemented and tested extensively in several generations of HPE runtime systems as well as adopted into the NVIDIA GPU hardware and drivers for CUDA 4.0 and beyond since 2011. The availability of real hardware that support key HPE features gives rise to a rare opportunity for studying the effectiveness of the hardware support by running important benchmarks on real runtime and hardware. Experimental results show that in a exemplar heterogeneous system, peer DMA and double-buffering, pinned buffers, and software techniques can improve the inter-accelerator data communication bandwidth by 2×. They can also improve the execution speed by 1.6× for a 3D finite difference, 2.5× for 1D FFT, and 1.6× for merge sort, all measured on real hardware. The proposed architecture support enables the HPE runtime to transparently deploy these optimizations under simple portable user code, allowing system designers to freely employ devices of different capabilities. We further argue that simple interfaces such as HPE are needed for most applications to benefit from advanced hardware features in practice. PMID:26180487

Flexible and twistable non-volatile memory cell array with all-organic one diode-one resistor architecture.

PubMed

Ji, Yongsung; Zeigler, David F; Lee, Dong Su; Choi, Hyejung; Jen, Alex K-Y; Ko, Heung Cho; Kim, Tae-Wook

2013-01-01

Flexible organic memory devices are one of the integral components for future flexible organic electronics. However, high-density all-organic memory cell arrays on malleable substrates without cross-talk have not been demonstrated because of difficulties in their fabrication and relatively poor performances to date. Here we demonstrate the first flexible all-organic 64-bit memory cell array possessing one diode-one resistor architectures. Our all-organic one diode-one resistor cell exhibits excellent rewritable switching characteristics, even during and after harsh physical stresses. The write-read-erase-read output sequence of the cells perfectly correspond to the external pulse signal regardless of substrate deformation. The one diode-one resistor cell array is clearly addressed at the specified cells and encoded letters based on the standard ASCII character code. Our study on integrated organic memory cell arrays suggests that the all-organic one diode-one resistor cell architecture is suitable for high-density flexible organic memory applications in the future.

Data preprocessing for determining outer/inner parallelization in the nested loop problem using OpenMP

NASA Astrophysics Data System (ADS)

Handhika, T.; Bustamam, A.; Ernastuti, Kerami, D.

2017-07-01

Multi-thread programming using OpenMP on the shared-memory architecture with hyperthreading technology allows the resource to be accessed by multiple processors simultaneously. Each processor can execute more than one thread for a certain period of time. However, its speedup depends on the ability of the processor to execute threads in limited quantities, especially the sequential algorithm which contains a nested loop. The number of the outer loop iterations is greater than the maximum number of threads that can be executed by a processor. The thread distribution technique that had been found previously only be applied by the high-level programmer. This paper generates a parallelization procedure for low-level programmer in dealing with 2-level nested loop problems with the maximum number of threads that can be executed by a processor is smaller than the number of the outer loop iterations. Data preprocessing which is related to the number of the outer loop and the inner loop iterations, the computational time required to execute each iteration and the maximum number of threads that can be executed by a processor are used as a strategy to determine which parallel region that will produce optimal speedup.

Massively Multithreaded Maxflow for Image Segmentation on the Cray XMT-2

PubMed Central

Bokhari, Shahid H.; Çatalyürek, Ümit V.; Gurcan, Metin N.

2014-01-01

SUMMARY Image segmentation is a very important step in the computerized analysis of digital images. The maxflow mincut approach has been successfully used to obtain minimum energy segmentations of images in many fields. Classical algorithms for maxflow in networks do not directly lend themselves to efficient parallel implementations on contemporary parallel processors. We present the results of an implementation of Goldberg-Tarjan preflow-push algorithm on the Cray XMT-2 massively multithreaded supercomputer. This machine has hardware support for 128 threads in each physical processor, a uniformly accessible shared memory of up to 4 TB and hardware synchronization for each 64 bit word. It is thus well-suited to the parallelization of graph theoretic algorithms, such as preflow-push. We describe the implementation of the preflow-push code on the XMT-2 and present the results of timing experiments on a series of synthetically generated as well as real images. Our results indicate very good performance on large images and pave the way for practical applications of this machine architecture for image analysis in a production setting. The largest images we have run are 320002 pixels in size, which are well beyond the largest previously reported in the literature. PMID:25598745

Parallelization of NAS Benchmarks for Shared Memory Multiprocessors

NASA Technical Reports Server (NTRS)

Waheed, Abdul; Yan, Jerry C.; Saini, Subhash (Technical Monitor)

1998-01-01

This paper presents our experiences of parallelizing the sequential implementation of NAS benchmarks using compiler directives on SGI Origin2000 distributed shared memory (DSM) system. Porting existing applications to new high performance parallel and distributed computing platforms is a challenging task. Ideally, a user develops a sequential version of the application, leaving the task of porting to new generations of high performance computing systems to parallelization tools and compilers. Due to the simplicity of programming shared-memory multiprocessors, compiler developers have provided various facilities to allow the users to exploit parallelism. Native compilers on SGI Origin2000 support multiprocessing directives to allow users to exploit loop-level parallelism in their programs. Additionally, supporting tools can accomplish this process automatically and present the results of parallelization to the users. We experimented with these compiler directives and supporting tools by parallelizing sequential implementation of NAS benchmarks. Results reported in this paper indicate that with minimal effort, the performance gain is comparable with the hand-parallelized, carefully optimized, message-passing implementations of the same benchmarks.

Constrained Multipoint Aerodynamic Shape Optimization Using an Adjoint Formulation and Parallel Computers

NASA Technical Reports Server (NTRS)

Reuther, James; Jameson, Antony; Alonso, Juan Jose; Rimlinger, Mark J.; Saunders, David

1997-01-01

An aerodynamic shape optimization method that treats the design of complex aircraft configurations subject to high fidelity computational fluid dynamics (CFD), geometric constraints and multiple design points is described. The design process will be greatly accelerated through the use of both control theory and distributed memory computer architectures. Control theory is employed to derive the adjoint differential equations whose solution allows for the evaluation of design gradient information at a fraction of the computational cost required by previous design methods. The resulting problem is implemented on parallel distributed memory architectures using a domain decomposition approach, an optimized communication schedule, and the MPI (Message Passing Interface) standard for portability and efficiency. The final result achieves very rapid aerodynamic design based on a higher order CFD method. In order to facilitate the integration of these high fidelity CFD approaches into future multi-disciplinary optimization (NW) applications, new methods must be developed which are capable of simultaneously addressing complex geometries, multiple objective functions, and geometric design constraints. In our earlier studies, we coupled the adjoint based design formulations with unconstrained optimization algorithms and showed that the approach was effective for the aerodynamic design of airfoils, wings, wing-bodies, and complex aircraft configurations. In many of the results presented in these earlier works, geometric constraints were satisfied either by a projection into feasible space or by posing the design space parameterization such that it automatically satisfied constraints. Furthermore, with the exception of reference 9 where the second author initially explored the use of multipoint design in conjunction with adjoint formulations, our earlier works have focused on single point design efforts. Here we demonstrate that the same methodology may be extended to treat complete configuration designs subject to multiple design points and geometric constraints. Examples are presented for both transonic and supersonic configurations ranging from wing alone designs to complex configuration designs involving wing, fuselage, nacelles and pylons.

The effect of earthquake on architecture geometry with non-parallel system irregularity configuration

NASA Astrophysics Data System (ADS)

Teddy, Livian; Hardiman, Gagoek; Nuroji; Tudjono, Sri

2017-12-01

Indonesia is an area prone to earthquake that may cause casualties and damage to buildings. The fatalities or the injured are not largely caused by the earthquake, but by building collapse. The collapse of the building is resulted from the building behaviour against the earthquake, and it depends on many factors, such as architectural design, geometry configuration of structural elements in horizontal and vertical plans, earthquake zone, geographical location (distance to earthquake center), soil type, material quality, and construction quality. One of the geometry configurations that may lead to the collapse of the building is irregular configuration of non-parallel system. In accordance with FEMA-451B, irregular configuration in non-parallel system is defined to have existed if the vertical lateral force-retaining elements are neither parallel nor symmetric with main orthogonal axes of the earthquake-retaining axis system. Such configuration may lead to torque, diagonal translation and local damage to buildings. It does not mean that non-parallel irregular configuration should not be formed on architectural design; however the designer must know the consequence of earthquake behaviour against buildings with irregular configuration of non-parallel system. The present research has the objective to identify earthquake behaviour in architectural geometry with irregular configuration of non-parallel system. The present research was quantitative with simulation experimental method. It consisted of 5 models, where architectural data and model structure data were inputted and analyzed using the software SAP2000 in order to find out its performance, and ETAB2015 to determine the eccentricity occurred. The output of the software analysis was tabulated, graphed, compared and analyzed with relevant theories. For areas of strong earthquake zones, avoid designing buildings which wholly form irregular configuration of non-parallel system. If it is inevitable to design a building with building parts containing irregular configuration of non-parallel system, make it more rigid by forming a triangle module, and use the formula.A good collaboration is needed between architects and structural experts in creating earthquake architecture.

A processing architecture for associative short-term memory in electronic noses

NASA Astrophysics Data System (ADS)

Pioggia, G.; Ferro, M.; Di Francesco, F.; DeRossi, D.

2006-11-01

Electronic nose (e-nose) architectures usually consist of several modules that process various tasks such as control, data acquisition, data filtering, feature selection and pattern analysis. Heterogeneous techniques derived from chemometrics, neural networks, and fuzzy rules used to implement such tasks may lead to issues concerning module interconnection and cooperation. Moreover, a new learning phase is mandatory once new measurements have been added to the dataset, thus causing changes in the previously derived model. Consequently, if a loss in the previous learning occurs (catastrophic interference), real-time applications of e-noses are limited. To overcome these problems this paper presents an architecture for dynamic and efficient management of multi-transducer data processing techniques and for saving an associative short-term memory of the previously learned model. The architecture implements an artificial model of a hippocampus-based working memory, enabling the system to be ready for real-time applications. Starting from the base models available in the architecture core, dedicated models for neurons, maps and connections were tailored to an artificial olfactory system devoted to analysing olive oil. In order to verify the ability of the processing architecture in associative and short-term memory, a paired-associate learning test was applied. The avoidance of catastrophic interference was observed.

Geospatial Applications on Different Parallel and Distributed Systems in enviroGRIDS Project

NASA Astrophysics Data System (ADS)

Rodila, D.; Bacu, V.; Gorgan, D.

2012-04-01

The execution of Earth Science applications and services on parallel and distributed systems has become a necessity especially due to the large amounts of Geospatial data these applications require and the large geographical areas they cover. The parallelization of these applications comes to solve important performance issues and can spread from task parallelism to data parallelism as well. Parallel and distributed architectures such as Grid, Cloud, Multicore, etc. seem to offer the necessary functionalities to solve important problems in the Earth Science domain: storing, distribution, management, processing and security of Geospatial data, execution of complex processing through task and data parallelism, etc. A main goal of the FP7-funded project enviroGRIDS (Black Sea Catchment Observation and Assessment System supporting Sustainable Development) [1] is the development of a Spatial Data Infrastructure targeting this catchment region but also the development of standardized and specialized tools for storing, analyzing, processing and visualizing the Geospatial data concerning this area. For achieving these objectives, the enviroGRIDS deals with the execution of different Earth Science applications, such as hydrological models, Geospatial Web services standardized by the Open Geospatial Consortium (OGC) and others, on parallel and distributed architecture to maximize the obtained performance. This presentation analysis the integration and execution of Geospatial applications on different parallel and distributed architectures and the possibility of choosing among these architectures based on application characteristics and user requirements through a specialized component. Versions of the proposed platform have been used in enviroGRIDS project on different use cases such as: the execution of Geospatial Web services both on Web and Grid infrastructures [2] and the execution of SWAT hydrological models both on Grid and Multicore architectures [3]. The current focus is to integrate in the proposed platform the Cloud infrastructure, which is still a paradigm with critical problems to be solved despite the great efforts and investments. Cloud computing comes as a new way of delivering resources while using a large set of old as well as new technologies and tools for providing the necessary functionalities. The main challenges in the Cloud computing, most of them identified also in the Open Cloud Manifesto 2009, address resource management and monitoring, data and application interoperability and portability, security, scalability, software licensing, etc. We propose a platform able to execute different Geospatial applications on different parallel and distributed architectures such as Grid, Cloud, Multicore, etc. with the possibility of choosing among these architectures based on application characteristics and complexity, user requirements, necessary performances, cost support, etc. The execution redirection on a selected architecture is realized through a specialized component and has the purpose of offering a flexible way in achieving the best performances considering the existing restrictions.

Integral Images: Efficient Algorithms for Their Computation and Storage in Resource-Constrained Embedded Vision Systems

PubMed Central

Ehsan, Shoaib; Clark, Adrian F.; ur Rehman, Naveed; McDonald-Maier, Klaus D.

2015-01-01

The integral image, an intermediate image representation, has found extensive use in multi-scale local feature detection algorithms, such as Speeded-Up Robust Features (SURF), allowing fast computation of rectangular features at constant speed, independent of filter size. For resource-constrained real-time embedded vision systems, computation and storage of integral image presents several design challenges due to strict timing and hardware limitations. Although calculation of the integral image only consists of simple addition operations, the total number of operations is large owing to the generally large size of image data. Recursive equations allow substantial decrease in the number of operations but require calculation in a serial fashion. This paper presents two new hardware algorithms that are based on the decomposition of these recursive equations, allowing calculation of up to four integral image values in a row-parallel way without significantly increasing the number of operations. An efficient design strategy is also proposed for a parallel integral image computation unit to reduce the size of the required internal memory (nearly 35% for common HD video). Addressing the storage problem of integral image in embedded vision systems, the paper presents two algorithms which allow substantial decrease (at least 44.44%) in the memory requirements. Finally, the paper provides a case study that highlights the utility of the proposed architectures in embedded vision systems. PMID:26184211

Accelerating solutions of one-dimensional unsteady PDEs with GPU-based swept time-space decomposition

NASA Astrophysics Data System (ADS)

Magee, Daniel J.; Niemeyer, Kyle E.

2018-03-01

The expedient design of precision components in aerospace and other high-tech industries requires simulations of physical phenomena often described by partial differential equations (PDEs) without exact solutions. Modern design problems require simulations with a level of resolution difficult to achieve in reasonable amounts of time-even in effectively parallelized solvers. Though the scale of the problem relative to available computing power is the greatest impediment to accelerating these applications, significant performance gains can be achieved through careful attention to the details of memory communication and access. The swept time-space decomposition rule reduces communication between sub-domains by exhausting the domain of influence before communicating boundary values. Here we present a GPU implementation of the swept rule, which modifies the algorithm for improved performance on this processing architecture by prioritizing use of private (shared) memory, avoiding interblock communication, and overwriting unnecessary values. It shows significant improvement in the execution time of finite-difference solvers for one-dimensional unsteady PDEs, producing speedups of 2 - 9 × for a range of problem sizes, respectively, compared with simple GPU versions and 7 - 300 × compared with parallel CPU versions. However, for a more sophisticated one-dimensional system of equations discretized with a second-order finite-volume scheme, the swept rule performs 1.2 - 1.9 × worse than a standard implementation for all problem sizes.

Integral Images: Efficient Algorithms for Their Computation and Storage in Resource-Constrained Embedded Vision Systems.

PubMed

Ehsan, Shoaib; Clark, Adrian F; Naveed ur Rehman; McDonald-Maier, Klaus D

2015-07-10

The integral image, an intermediate image representation, has found extensive use in multi-scale local feature detection algorithms, such as Speeded-Up Robust Features (SURF), allowing fast computation of rectangular features at constant speed, independent of filter size. For resource-constrained real-time embedded vision systems, computation and storage of integral image presents several design challenges due to strict timing and hardware limitations. Although calculation of the integral image only consists of simple addition operations, the total number of operations is large owing to the generally large size of image data. Recursive equations allow substantial decrease in the number of operations but require calculation in a serial fashion. This paper presents two new hardware algorithms that are based on the decomposition of these recursive equations, allowing calculation of up to four integral image values in a row-parallel way without significantly increasing the number of operations. An efficient design strategy is also proposed for a parallel integral image computation unit to reduce the size of the required internal memory (nearly 35% for common HD video). Addressing the storage problem of integral image in embedded vision systems, the paper presents two algorithms which allow substantial decrease (at least 44.44%) in the memory requirements. Finally, the paper provides a case study that highlights the utility of the proposed architectures in embedded vision systems.

MUTILS - a set of efficient modeling tools for multi-core CPUs implemented in MEX

NASA Astrophysics Data System (ADS)

Krotkiewski, Marcin; Dabrowski, Marcin

2013-04-01

The need for computational performance is common in scientific applications, and in particular in numerical simulations, where high resolution models require efficient processing of large amounts of data. Especially in the context of geological problems the need to increase the model resolution to resolve physical and geometrical complexities seems to have no limits. Alas, the performance of new generations of CPUs does not improve any longer by simply increasing clock speeds. Current industrial trends are to increase the number of computational cores. As a result, parallel implementations are required in order to fully utilize the potential of new processors, and to study more complex models. We target simulations on small to medium scale shared memory computers: laptops and desktop PCs with ~8 CPU cores and up to tens of GB of memory to high-end servers with ~50 CPU cores and hundereds of GB of memory. In this setting MATLAB is often the environment of choice for scientists that want to implement their own models with little effort. It is a useful general purpose mathematical software package, but due to its versatility some of its functionality is not as efficient as it could be. In particular, the challanges of modern multi-core architectures are not fully addressed. We have developed MILAMIN 2 - an efficient FEM modeling environment written in native MATLAB. Amongst others, MILAMIN provides functions to define model geometry, generate and convert structured and unstructured meshes (also through interfaces to external mesh generators), compute element and system matrices, apply boundary conditions, solve the system of linear equations, address non-linear and transient problems, and perform post-processing. MILAMIN strives to combine the ease of code development and the computational efficiency. Where possible, the code is optimized and/or parallelized within the MATLAB framework. Native MATLAB is augmented with the MUTILS library - a set of MEX functions that implement the computationally intensive, performance critical parts of the code, which we have identified to be bottlenecks. Here, we discuss the functionality and performance of the MUTILS library. Currently, it includes: 1. time and memory efficient assembly of sparse matrices for FEM simulations 2. parallel sparse matrix - vector product with optimizations speficic to symmetric matrices and multiple degrees of freedom per node 3. parallel point in triangle location and point in tetrahedron location for unstructured, adaptive 2D and 3D meshes (useful for 'marker in cell' type of methods) 4. parallel FEM interpolation for 2D and 3D meshes of elements of different types and orders, and for different number of degrees of freedom per node 5. a stand-alone, MEX implementation of the Conjugate Gradients iterative solver 6. interface to METIS graph partitioning and a fast implementation of RCM reordering

A Core Knowledge Architecture of Visual Working Memory

ERIC Educational Resources Information Center

Wood, Justin N.

2011-01-01

Visual working memory (VWM) is widely thought to contain specialized buffers for retaining spatial and object information: a "spatial-object architecture." However, studies of adults, infants, and nonhuman animals show that visual cognition builds on core knowledge systems that retain more specialized representations: (1) spatiotemporal…

Xyce Parallel Electronic Simulator : users' guide, version 2.0.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Hoekstra, Robert John; Waters, Lon J.; Rankin, Eric Lamont

2004-06-01

This manual describes the use of the Xyce Parallel Electronic Simulator. Xyce has been designed as a SPICE-compatible, high-performance analog circuit simulator capable of simulating electrical circuits at a variety of abstraction levels. Primarily, Xyce has been written to support the simulation needs of the Sandia National Laboratories electrical designers. This development has focused on improving capability the current state-of-the-art in the following areas: {sm_bullet} Capability to solve extremely large circuit problems by supporting large-scale parallel computing platforms (up to thousands of processors). Note that this includes support for most popular parallel and serial computers. {sm_bullet} Improved performance for allmore » numerical kernels (e.g., time integrator, nonlinear and linear solvers) through state-of-the-art algorithms and novel techniques. {sm_bullet} Device models which are specifically tailored to meet Sandia's needs, including many radiation-aware devices. {sm_bullet} A client-server or multi-tiered operating model wherein the numerical kernel can operate independently of the graphical user interface (GUI). {sm_bullet} Object-oriented code design and implementation using modern coding practices that ensure that the Xyce Parallel Electronic Simulator will be maintainable and extensible far into the future. Xyce is a parallel code in the most general sense of the phrase - a message passing of computing platforms. These include serial, shared-memory and distributed-memory parallel implementation - which allows it to run efficiently on the widest possible number parallel as well as heterogeneous platforms. Careful attention has been paid to the specific nature of circuit-simulation problems to ensure that optimal parallel efficiency is achieved as the number of processors grows. One feature required by designers is the ability to add device models, many specific to the needs of Sandia, to the code. To this end, the device package in the Xyce These input formats include standard analytical models, behavioral models look-up Parallel Electronic Simulator is designed to support a variety of device model inputs. tables, and mesh-level PDE device models. Combined with this flexible interface is an architectural design that greatly simplifies the addition of circuit models. One of the most important feature of Xyce is in providing a platform for computational research and development aimed specifically at the needs of the Laboratory. With Xyce, Sandia now has an 'in-house' capability with which both new electrical (e.g., device model development) and algorithmic (e.g., faster time-integration methods) research and development can be performed. Ultimately, these capabilities are migrated to end users.« less

Automatic disease diagnosis using optimised weightless neural networks for low-power wearable devices

PubMed Central

Edla, Damodar Reddy; Kuppili, Venkatanareshbabu; Dharavath, Ramesh; Beechu, Nareshkumar Reddy

2017-01-01

Low-power wearable devices for disease diagnosis are used at anytime and anywhere. These are non-invasive and pain-free for the better quality of life. However, these devices are resource constrained in terms of memory and processing capability. Memory constraint allows these devices to store a limited number of patterns and processing constraint provides delayed response. It is a challenging task to design a robust classification system under above constraints with high accuracy. In this Letter, to resolve this problem, a novel architecture for weightless neural networks (WNNs) has been proposed. It uses variable sized random access memories to optimise the memory usage and a modified binary TRIE data structure for reducing the test time. In addition, a bio-inspired-based genetic algorithm has been employed to improve the accuracy. The proposed architecture is experimented on various disease datasets using its software and hardware realisations. The experimental results prove that the proposed architecture achieves better performance in terms of accuracy, memory saving and test time as compared to standard WNNs. It also outperforms in terms of accuracy as compared to conventional neural network-based classifiers. The proposed architecture is a powerful part of most of the low-power wearable devices for the solution of memory, accuracy and time issues. PMID:28868148