Search NASA⌕ Search

SEARCH · Search NASA

Results for “MULTIPROCESSORS”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 55 records · Page 3

On-package multiprocessor ground-referenced single-ended interconnect

A system of interconnected chips comprising a multi-chip module (MCM) includes a first processor chip, a second processor chip, and an MCM package configured to include the first processor chip, the second processor chip, and an interconnect circuit. The first processor chip is configured to include a first ground-referenced single-ended signaling (GRS) interface circuit. A first set of electrical traces fabricated within the MCM package and configured to couple the first GRS interface circuit to the interconnect circuit. The second processor chip is configured to include a second GRS interface circuit. A second set of electrical traces fabricated within the MCM package and configured to coupled the second GRS interface circuit to the interconnect circuit.

Dally, William J.↗

Efficient Approximation Algorithms for Weighted $b$-Matching

We describe a half-approximation algorithm, b-Suitor, for computing a b-Matching of maximum weight in a graph with weights on the edges. b-Matching is a generalization of the well-known Matching problem in graphs, where the objective is to choose a subset of M edges in the graph such that at most a speci ed number b(v) of edges in M are incident on each vertex v. Subject to this restriction we maximize the sum of the weights of the edges in M. We prove that the b-Suitor algorithm computes the same b-Matching as the one obtained by the greedy algorithm for the problem. We implement the algorithm on serial and shared-memory parallel processors, and compare its performance against a collection of approximation algorithms that have been proposed earlier. Our results show that the b-Suitor algorithm outperforms the Greedy and Locally Dominant edge algorithms by one to two orders of magnitude on a serial processor. The b-Suitor algorithm has a high degree of concurrency, and it scales well up to 240 threads on a shared memory multiprocessor. The b-Suitor algorithm outperforms the Locally Dominant edge algorithm by a factor of fourteen on 16 cores of an Intel Xeon multiprocessor.

97 MATHEMATICS AND COMPUTING↗

Low latency memory access and synchronization

A low latency memory system access is provided in association with a weakly-ordered multiprocessor system. Bach processor in the multiprocessor shares resources, and each shared resource has an associated lock within a locking device that provides support for synchronization between the multiple processors in the multiprocessor and the orderly sharing of the resources. A processor only has permission to access a resource when it owns the lock associated with that resource, and an attempt by a processor to own a lock requires only a single load operation, rather than a traditional atomic load followed by store, such that the processor only performs a read operation and the hardware locking device performs a subsequent write operation rather than the processor. A simple prefetching for non-contiguous data structures is also disclosed. A memory line is redefined so that in addition to the normal physical memory data, every line includes a pointer that is large enough to point to any other line in the memory, wherein the pointers to determine which memory line to prefetch rather than some other predictive algorithm. This enables hardware to effectively prefetch memory access patterns that are non-contiguous, but repetitive.

Blumrich, Matthias A↗

Method for prefetching non-contiguous data structures

A low latency memory system access is provided in association with a weakly-ordered multiprocessor system. Each processor in the multiprocessor shares resources, and each shared resource has an associated lock within a locking device that provides support for synchronization between the multiple processors in the multiprocessor and the orderly sharing of the resources. A processor only has permission to access a resource when it owns the lock associated with that resource, and an attempt by a processor to own a lock requires only a single load operation, rather than a traditional atomic load followed by store, such that the processor only performs a read operation and the hardware locking device performs a subsequent write operation rather than the processor. A simple perfecting for non-contiguous data structures is also disclosed. A memory line is redefined so that in addition to the normal physical memory data, every line includes a pointer that is large enough to point to any other line in the memory, wherein the pointers to determine which memory line to prefect rather than some other predictive algorithm. This enables hardware to effectively prefect memory access patterns that are non-contiguous, but repetitive.

97 MATHEMATICS AND COMPUTING↗

Continuous-energy Monte Carlo neutron transport on GPUs in the Shift code

Here, a continuous-energy Monte Carlo neutron transport solver executing on GPUs has been developed within the Shift code. Several algorithmic approaches are considered, including both history-based and event-based implementations. Unlike in previous work involving multigroup Monte Carlo transport, it is demonstrated that event-based algorithms significantly outperform a history-based approach for continuous-energy transport as a result of increased device occupancy and reduced thread divergence. Numerical results are presented for detailed full-core models of a small modular reactor (SMR), including a model containing depleted fuel materials. These results demonstrate the substantial gains in performance that are possible with the latest-generation of GPUs. On the depleted SMR core configuration, an NVIDIA P100 GPU with 56 streaming multiprocessors provides performance equivalent to 90 CPU cores, and the latest V100 GPU with 80 multiprocessors offers the performance of more than 150 CPU cores.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Demonstrating GPU code portability and scalability for radiative heat transfer computations

High performance computing frameworks utilizing CPUs, Nvidia GPUs, and/or Intel Xeon Phis necessitate portable and scalable solutions for application developers. Nvidia GPUs in particular present numerous portability challenges with a different programming model, additional memory hierarchies, and partitioned execution units among streaming multiprocessors. This work presents modifications to the Uintah asynchronous many-task runtime and the Kokkos portability library to enable one single codebase for complex multiphysics applications to run across different architectures. Scalability and performance results are shown on multiple architectures for a globally coupled radiation heat transfer simulation, ranging from a single node to 16,384 Titan compute nodes.

Computer Science↗

Singular value decomposition utilizing parallel algorithms on graphical processors

One of the current challenges in underwater acoustic array signal processing is the detection of quiet targets in the presence of noise. In order to enable robust detection, one of the key processing steps requires data and replica whitening. This, in turn, involves the eigen-decomposition of the sample spectral matrix, Cx = 1/K xKX(k)XH(k) where X(k) denotes a single frequency snapshot with an element for each element of the array. By employing the singular value decomposition (SVD) method, the eigenvectors and eigenvalues can be determined directly from the data without computing the sample covariance matrix, reducing the computational requirements for a given level of accuracy (van Trees, Optimum Array Processing). (Recall that the SVD of a complex matrix A involves determining V, , and U such that A = U VH where U and V are orthonormal and is a positive, real, diagonal matrix containing the singular values of A. U and V are the eigenvectors of AAH and AHA, respectively, while the singular values are the square roots of the eigenvalues of AAH.) Because it is desirable to be able to compute these quantities in real time, an efficient technique for computing the SVD is vital. In addition, emerging multicore processors like graphical processing units (GPUs) are bringing parallel processing capabilities to an ever increasing number of users. Since the computational tasks involved in array signal processing are well suited for parallelization, it is expected that these computations will be implemented using GPUs as soon as users have the necessary computational tools available to them. Thus, it is important to have an SVD algorithm that is suitable for these processors. This work explores the effectiveness of two different parallel SVD implementations on an NVIDIA Tesla C2050 GPU (14 multiprocessors, 32 cores per multiprocessor, 1.15 GHz clock - peed). The first algorithm is based on a two-step algorithm which bidiagonalizes the matrix using Householder transformations, and then diagonalizes the intermediate bidiagonal matrix through implicit QR shifts. This is similar to that implemented for real matrices by Lahabar and Narayanan ("Singular Value Decomposition on GPU using CUDA", IEEE International Parallel Distributed Processing Symposium 2009). The implementation is done in a hybrid manner, with the bidiagonalization stage done using the GPU while the diagonalization stage is done using the CPU, with the GPU used to update the U and V matrices. The second algorithm is based on a one-sided Jacobi scheme utilizing a sequence of pair-wise column orthogonalizations such that A is replaced by AV until the resulting matrix is sufficiently orthogonal (that is, equal to U ). V is obtained from the sequence of orthogonalizations, while can be found from the square root of the diagonal elements of AH A and, once is known, U can be found from column scaling the resulting matrix. These implementations utilize CUDA Fortran and NVIDIA's CUB LAS library. The primary goal of this study is to quantify the comparative performance of these two techniques against themselves and other standard implementations (for example, MATLAB). Considering that there is significant overhead associated with transferring data to the GPU and with synchronization between the GPU and the host CPU, it is also important to understand when it is worthwhile to use the GPU in terms of the matrix size and number of concurrent SVDs to be calculated.

99 GENERAL AND MISCELLANEOUS↗

CUDA Computation of the Feynman Distribution

In 2006, NVIDIA Corporation introduced the Computer Unified Device Architecture (CUDA) for C, C++, and Fortran programming languages. The CUDA parallel computing platform uses the Graphics Processing Unit (GPU) of NVIDIA graphics card, and it is designed for parallel computations. In these computations, the same algorithms are executed on the data in parallel. The GPU contains many Streaming Multiprocessors (SM) that can independently execute blocks of threads. A thread is a sequence of serial programmed instructions. The SM executes threads in groups of 32 and a group of 32 threads is called warp. The threads within a block can be synchronized and have access to a shared memory that is visible only to them. Conversely, blocks of threads cannot be synchronized and have access only to the global memory of the graphics card. This memory is called device memory and is independent of the host memory controlled by the Central Processing Unit (CPU). A single thread has access to a local memory (register) that is not visible to any other thread. The part of the program that is executed on the graphics card is called kernel. Each thread of the kernel has a unique identification index that is accessible by the in-built variable threadIdx; the latter is a 3-component vector. A threads block can contain up to 1024 threads. Blocks of threads have a unique identification index that is accessible by the in-built variable blockIdx; the latter is a 3-component vector. The in-built variable blockDim contains the number of threads per block; this variable is a 3-component vector. Threads and blocks represent different levels of parallelism; a kernel is executed as a grid of blocks of threads. Figure 1 illustrates a grid of thread blocks for a GPU with 2 SM. A SM can execute multiple blocks of threads in parallel; a block of threads is executed on a single SM. A GPU can execute multiple kernels concurrently. This work has used the Quadro M5000 graphics card with the following technical specifications: - 2048 cores with 861 MHz frequency, - 16 streaming multiprocessors, - 5.2 trillion transistors, - 8.2 Gb device memory, - 64 kb register memory (per thread), - 48 kb shared memory (per block) - 211.6 GB/s memory bandwidth, and - 3.5 GFLOPS floating point performance.

97 MATHEMATICS AND COMPUTING↗

Neutronics Calculation of Fast Reactor Using Modern Computing Systems

The possibility of using the LUCKY-A three-dimensional multiprocessor transport code as a neutronics module for the COREMELT integrated non-stationary code for performing safety analysis, which is used for fast sodium reactor safety validation, is discussed. In this application, more stringent requirements are imposed on the computing speed of the LUCKY-A code. To test LUCKY-A possibilities, a typical computational problem is examined in a formulation characteristic for the proposed use of the code. The possibilities of the code for parallelized calculations are shown, and a comparison is made with calculations performed with the MMKKENO and TRIGEX codes. The efficacy of a parallel computational process is investigated as a function of the number of computing kernels used.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Real-Time Considerations for Rugged Embedded Systems

This chapter introduces the characterizing aspects of embedded systems, and discusses the specific features that a designer should address to an embedded system “rugged”, i.e., able to operate reliably in harsh environments. The chapter addresses both the hardware and the less obvious software aspect. After presenting a current list of certifications for ruggedization, the chapters present a case study that focuses on the interaction of the hardware and software layers in reactive real-time system. In particular, it shows how the use of fast FPGA prototyping could provide insights on unexpected factors that influence the performance and thus responsiveness to events of a scheduling algorithm for multiprocessor systems that manages both periodic, hard real-time task, and aperiodic tasks. The main lesson is that to make the system “rugged”, a designer should consider these issues by, for example, overprovisioning resources and/or computation capabilities.

FPGA↗

SYNTHETIC SPECTRA OF RADIO, MILLIMETER, SUB-MILLIMETER, AND INFRARED REGIMES WITH NON-LOCAL THERMODYNAMIC EQUILIBRIUM APPROXIMATION

We use a numerical code called PAKALMPI to compute synthetic spectra of the solar emission in quiet conditions at millimeter, sub-millimeter, and infrared wavelengths. PAKALMPI solves the radiative transfer equation, with non-local thermodynamic equilibrium (NLTE), in a three-dimensional geometry using a multiprocessor environment. The code is able to use three opacity functions: classical bremsstrahlung, H{sup -}, and inverse bremsstrahlung. In this work, we have computed and compared two synthetic spectra, one in the common way: using bremsstrahlung opacity function and considering a fully ionized atmosphere; and a new one considering bremsstrahlung, inverse bremsstrahlung, and H{sup -} opacity functions in NLTE. We analyzed in detail the local behavior of the low atmospheric emission at 17, 212, and 405 GHz (frequencies used by the Nobeyama Radio Heliograph and the Solar Submillimeter Telescope). We found that the H{sup -} is the major emission mechanism at low altitudes (below 500 km) and that at higher altitudes the classical bremsstrahlung becomes the major mechanism of emission. However, the brightness temperature remains unalterable. Finally, we found that the inverse bremsstrahlung process is not important for radio emission at these heights.

79 ASTRONOMY AND ASTROPHYSICS↗

SHELXT – Integrated space-group and crystal-structure determination

SHELXT automates routine small-molecule structure determination starting from single-crystal reflection data, the Laue group and a reasonable guess as to which elements might be present. The new computer program SHELXT employs a novel dual-space algorithm to solve the phase problem for single-crystal reflection data expanded to the space group P1. Missing data are taken into account and the resolution extended if necessary. All space groups in the specified Laue group are tested to find which are consistent with the P1 phases. After applying the resulting origin shifts and space-group symmetry, the solutions are subject to further dual-space recycling followed by a peak search and summation of the electron density around each peak. Elements are assigned to give the best fit to the integrated peak densities and if necessary additional elements are considered. An isotropic refinement is followed for non-centrosymmetric space groups by the calculation of a Flack parameter and, if appropriate, inversion of the structure. The structure is assembled to maximize its connectivity and centred optimally in the unit cell. SHELXT has already solved many thousand structures with a high success rate, and is optimized for multiprocessor computers. It is, however, unsuitable for severely disordered and twinned structures because it is based on the assumption that the structure consists of atoms.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

GPU-Centric Communication on NVIDIA GPU Clusters with InfiniBand: A Case Study with OpenSHMEM

GPUs have become an essential component for building compute clusters with high compute density and high performance per watt. As such clusters scale to have 1000s of GPUs, efficiently moving data between the GPUs becomes imperative to get maximum performance. NVSHMEM is an implementation of the OpenSHMEM standard for NVIDIA GPU clusters which allows communication to be issued from inside GPU kernels. In earlier work, we have shown how NVSHMEM can be used to achieve better application performance on GPUs connected through PCIe or NVLink. As part of this effort, we implement IB verbs for Mellanox InfiniBand adapters in CUDA. We evaluate different design alternatives, taking into consideration the relaxed memory model, automatic memory access coalescing and thread hierarchy on the GPU. We also consider correctness issues that arise in these designs. We take advantage of these designs transparently or through API extensions in NVSHMEM. With micro-benchmarks, we show that a Nvidia Pascal P100 GPU is able saturate the network bandwidth using only one or two of its 56 available streaming multiprocessors (SM). On a single GPU using a single IB EDR adapter, we achieve a throughput of around 90 million messages per second. In addition, we implement a 2dstencil application kernel using NVSHMEM and compare its performance with a CUDA-aware MPI-based implementation that uses GPUDirect RDMA. Speedups in the range of 23% to 42% are seen for input sizes large enough to fill the occupancy of Nvidia Pascal P100 GPUs on 2 to 4 nodes indicating that there are gains to be had by eliminating the CPU from the communication path when all computation runs on the GPU.

Potluri, Sreeram↗

Performance Assessment of Emerging Memories Through FPGA Emulation

Emerging memory technologies offer the prospect of large capacity, high bandwidth, and a range of access latencies ranging from DRAM-like to SSD-like. In this paper, we evaluate the performance of parallel applications on CPUs whose main memories sweep a wide range of latencies within a bandwidth cap. We use an FPGA emulator, the logic in memory emulator (LiME) to accelerate evaluation. The LiME framework uses a multiprocessor system on chip (MPSoC) combining multicore CPU, integrated memory controller, and FPGA fabric, and enables emulation orders of magnitude faster than software simulation. This work highlights the performance impact of higher latency on concurrent applications and identifies conditions under which future high latency memories can effectively be used as main memory.

97 MATHEMATICS AND COMPUTING↗

Composing Data Parallel Code for a SPARQL Graph Engine

Big data analytics process large amount of data to extract knowledge from them. Semantic databases are big data applications that adopt the Resource Description Framework (RDF) to structure metadata through a graph-based representation. The graph based representation provides several benefits, such as the possibility to perform in memory processing with large amounts of parallelism. SPARQL is a language used to perform queries on RDF-structured data through graph matching. In this paper we present a tool that automatically translates SPARQL queries to parallel graph crawling and graph matching operations. The tool also supports complex SPARQL constructs, which requires more than basic graph matching for their implementation. The tool generates parallel code annotated with OpenMP pragmas for x86 Shared-memory Multiprocessors (SMPs). With respect to commercial database systems such as Virtuoso, our approach reduces memory occupation due to join operations and provides higher performance. We show the scaling of the automatically generated graph-matching code on a 48-core SMP.

Big data↗

Select Advances in Computational Accelerator Physics

Computational accelerator physics has changed and broadened over the last decade or so. Part of the change is due to the advent of multiple ways of parallel computing. Another part comes from algorithmic developments. The multiple ways of parallel computing include distributed memory parallelism and on-chip parallelism, with the latter coming from architectures (CPU and GPU) having multiple processing elements (cores or streaming multiprocessors) and wide vector (SIMD) instruction units. The basics of these new architectures and their application to computational accelerator physics are briefly reviewed. Algorithmic advances in the select areas of spin tracking, cavity calculations, plasma acceleration, and electron cooling are also reviewed. In some cases the algorithms provide increased fidelity improving the overall accuracy, while in other cases, such as controlled dispersion, the algorithms provide increased fidelity by better modeling the essential physical interaction. Finally, the use of computational frameworks, which provide the basic computational infrastructure, while allowing the capability developer to concentrate on the math and physics, is reviewed in the context of the Vorpal application, which has found use across accelerator physics and many other fields.

Engineering↗

Fast and Accurate Simulation of the Cray XMT Multithreaded Supercomputer

Irregular applications, such as data mining and analysis or graph-based computations, show unpredictable memory/network access patterns and control structures. Highly multithreaded architectures with large processor counts, like the Cray MTA-1, MTA-2 and XMT, appear to address their requirements better than commodity clusters. However, the research on highly multithreaded systems is currently limited by the lack of adequate architectural simulation infrastructures due to issues such as size of the machines, memory footprint, simulation speed, accuracy and customization. At the same time, Shared-memory MultiProcessors (SMPs) with multi-core processors have become an attractive platform to simulate large scale machines. In this paper, we introduce a cycle-level simulator of the highly multithreaded Cray XMT supercomputer. The simulator runs unmodified XMT applications. We discuss how we tackled the challenges posed by its development, detailing the techniques introduced to make the simulation as fast as possible while maintaining a high accuracy. By mapping XMT processors (ThreadStorm with 128 hardware threads) to host computing cores, the simulation speed remains constant as the number of simulated processors increases, up to the number of available host cores. The simulator supports zero-overhead switching among different accuracy levels at run-time and includes a network model that takes into account contention. On a modern 48-core SMP host, our infrastructure simulates a large set of irregular applications 500 to 2000 times slower than real time when compared to a 128-processor XMT, while remaining within 10\% of accuracy. Emulation is only from 25 to 200 times slower than real time.

Cray XMT↗