Results for “parallel, scaling” — Page 9

DOE OSTI2018

Roofline scaling trajectories: A method for parallel application and architectural performance analysis

The end of Dennard scaling signaled a shift in HPC supercomputer architectures from systems built from single-core processor architectures to systems built from multicore and eventually manycore architectures. This transition substantially complicated performance optimization and analysis as new programming models were created, new scaling methodologies deployed, and on-chip contention became a bottleneck to performance. Existing distributed memory performance models like logP and logGP were unable to capture this contention. The Roofline model was created to address this contention and its interplay with locality. However, to date, the Roofline model has focused on full-node concurrency. In this paper, we extend the Roofline model to capture the effects of concurrency on data locality and on-chip contention. We demonstrate the value of this new technique by evaluating the NAS parallel benchmarks on both multicore and manycore architectures under both strong-And weak-scaling regimes. In order to quantify the interplay between programming model and locality, we evaluate scaling under both the OpenMP and flat MPI programming models.

Ibrahim, K↗

NASA NTRS1983

Scale lengths in quasi-parallel shocks

Examples of an interplanetary and the bow shock illustrate the small relative size of the electrostatic layer relative to the scale of the magnetic fluctuations in quasi-parallel shocks. While both examples are supercritical, the interplanetary example is marginally so, showing a thickness in absolute and convected ion larmor radii units that is thicker (approximately 13 U/omega sub ci) than at the bow shock (approximately omega sub ci). The fluid speed changes abruptly in the quasi-parallel shock on this shorter scale. The increase in electron and ion random energies also is clearly seen on this shorter scale. In the interplanetary example the scale of the electric layer is certainly less than 1/60th that of the up or downstreams magnetic fluctuations. The thickness of the earth's bow shock deceleration layer is dramatically narrower than any domain of upstream waves as controlled by reflected, intermediate, or diffuse ions.

Scudder, J. D.↗

NASA NTRS2000

Impact of Parallel Computing on Large Scale Aeroelastic Computations

Aeroelasticity is computationally one of the most intensive fields in aerospace engineering. Though over the last three decades the computational speed of supercomputers have substantially increased, they are still inadequate for large scale aeroelastic computations using high fidelity flow and structural equations. In addition to reaching a saturation in computational speed because of changes in economics, computer manufactures are stopping the manufacturing of mainframe type supercomputers. This has led computational aeroelasticians to face the gigantic task of finding alternate approaches for fulfilling their needs. The alternate path to over come speed and availability limitations of mainframe type supercomputers is to use parallel computers. During this decade several different architectures have evolved. In FY92 the US Government started the High Performance Computing and Communication (HPCC) program. As a participant in this program NASA developed several parallel computational tools for aeroelastic applications. This talk describes the impact of those application tools on high fidelity based multidisciplinary analysis.

Guruswamy, Guru P.↗

NASA NTRS1988

Application of the hypercube parallel processor to a large-scale moment method code

The applicability of a parallel computing architecture to the solution of a large-scale moment-method code is investigated. Specifically, the NEC (Numerical Electromagnetics Code) method-of-moments scattering program is implemented on a hypercube parallel processor. The accuracy and the increase in the speed of execution on this parallel architecture are demonstrated. The results show a very large reduction in execution time for large problems. The great potential of this parallel processor is shown for interactive solution of large NEC problems as well as other moment-method techniques such as the finite-element method.

Manshadi, Farzin↗

DOE OSTI2019

A Comparison of Classical and Aggregation-Based Algebraic Multigrid Preconditioners for High-Fidelity Simulation of Wind Turbine Incompressible Flows

This paper presents a comparison of parallel strong scaling performance of classical and aggregation algebraic multigrid (AMG) preconditioners in the context of wind turbine simulations. Fluid motion is governed by the incompressible Navier--Stokes equations, discretized in space with control-volume finite elements and in time with an inexact projection scheme using an implicit integrator. A discontinuous-Galerkin sliding-mesh algorithm captures rotor motion. The momentum equations are solved with iterative Krylov methods, preconditioned by symmetric Gauss--Seidel (SGS) in Trilinos and $\ell_1$ SGS in hypre. The mass-continuity equation is solved with GMRES preconditioned by AMG and can account for the majority of simulation time. Reducing this continuity solve time is crucial. Wind turbine simulations present two unique challenges for AMG preconditioned solvers: the computational meshes include strongly anisotropic elements, and mesh motion requires matrix reinitialization and computation of preconditioners at each time step. Detailed timing profiles are presented and analyzed, and best practices are discussed for both classical and aggregation-based AMG. Results are presented for simulations of two different wind turbines with up to 6 billion grid points on two different computer architectures. For moving mesh problems that require linear-system reinitialization, the well-established strategy of amortizing preconditioner setup costs over a large number of time steps to reduce the solve time is no longer valid. Instead, results show that faster time to solution is achieved by reducing preconditioner setup costs at the expense of linear-system solve costs. Standard smoothed aggregation with Chebyshev relaxation was found to perform poorly when compared with classical AMG in terms of solve time and robustness. However, plain aggregation was comparable to classical AMG.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

Large Scale Tensor Factorization via Parallel Sketches

Tensor factorization methods have recently gained increased popularity. A key feature that renders tensors attractive is the ability to directly model multi-relational data. In this work, we propose ParaSketch, a parallel tensor factorization algorithm that enables massive parallelism, to deal with large tensors. The idea is to compress the large tensor into multiple small tensors, decompose each small tensor in parallel, and combine the results to reconstruct the desired latent factors. Prior art in this direction entails potentially very high complexity in the (Gaussian) compression and final combining stages. Adopting sketching matrices for compression, the proposed method enjoys a dramatic reduction in compression complexity, and features a much lighter combining step. Moreover, theoretical analysis shows that the compressed tensors inherit latent identifiability under mild conditions, hence establishing correctness of the overall approach. Numerical experiments corroborate the theory and demonstrate the effectiveness of the proposed algorithm.

block term decomposition↗

DOE OSTI2024

OPER: Optimality-Guided Embedding Table Parallelization for Large-scale Recommendation Model

With the sharp increasing volume of user data, Deep Learning Recommendation Model (DLRM) becomes an indispensable infrastructure in large technology companies. However, large-scale DLRM on the multi-GPU platform is still inefficient due to unbalanced workload partitioning and intensive inter-GPU communication. To this end, we propose OPER, an OPtimality guided Embedding table placement for large-scale Recommendation model training and inference. OPER explores the potential of mitigating remote memory access latency in DLRM through fine-grained embedding table placement. Specifically, OPER proposes a theoretical modeling that builds up the relationship between EMT placement and the embedding communication latency in both training and inference. OPER proves the NP hardness of finding the optimal embedding table placement and proposes a heuristic algorithm that yields near optimal placement. OPER implements a SHMEM-based embedding table training system and a unified embedding index mapping to support fine-grained embedding table sharding and placement. Comprehensive experiments reveal that OPER achieves on average 3.4× and 5.1× speedup on training and inference respectively over state-of-the-art DLRM frameworks.

Wang, Zheng↗

NASA NTRS1991

Parallel Methods on Large-Scale Structural Analysis and Physics Applications; Symposium, Hampton, VA, Feb. 5, 6, 1991, Selected Papers

Recent advances in parallel methods and algorithms integrated into large-scale codes are presented. Consideration is given to problem decomposition (substructuring), efficient matrix solution algorithms for shared memory architectures, dynamic and transient analysis algorithms for shared memory architectures, and algorithms for distributed and massively parallel architectures. Particular attention is given to partitioning of unstructured problems for parallel processing, parallel-vector computation for linear-structural analysis and nonlinear unconstraint optimization problems, a parallel-vector equation solver for unsymmetric matrices on supercomputers, parallel nonlinear finite element dynamic response, multigrid algorithms for solving structural mechanics problems on supercomputers, structural analysis on massively parallel computers, explicit finite element methods with contact-impact on SIMD computers, and the impact of mapping and sparsity on parallelized finite element method modules.

Storaasli, Olaf O.↗

NASA NTRS1999

Parallel Visualization of Large-Scale Aerodynamics Calculations: A Case Study on the Cray T3E

This paper reports the performance of a parallel volume rendering algorithm for visualizing a large-scale, unstructured-grid dataset produced by a three-dimensional aerodynamics simulation. This dataset, containing over 18 million tetrahedra, allows us to extend our performance results to a problem which is more than 30 times larger than the one we examined previously. This high resolution dataset also allows us to see fine, three-dimensional features in the flow field. All our tests were performed on the Silicon Graphics Inc. (SGI)/Cray T3E operated by NASA's Goddard Space Flight Center. Using 511 processors, a rendering rate of almost 9 million tetrahedra/second was achieved with a parallel overhead of 26%.

Ma, Kwan-Liu↗

DOE OSTI2017

GPU-based parallel algorithm for generating massive scale-free networks using the preferential attachment model

A novel parallel algorithm is presented for generating random scale-free networks using the preferential-attachment model. The algorithm, named cuPPA, is custom-designed for single instruction multiple data (SIMD) style of parallel processing supported by modern processors such as graphical processing units (GPUs). To the best of our knowledge, our algorithm is the first to exploit GPUs, and also the fastest implementation available today, to generate scale-free networks using the preferential attachment model. A detailed performance study is presented to understand the scalability and runtime characteristics of the cuPPA algorithm. In one of the best cases, when executed on an NVidia GeForce 1080 GPU, cuPPA generates a scale-free network of two billion edges in less than 3 seconds.

Alam, Maksudul↗

DOE OSTI2020

Electromagnetic Shielding Design for 200 kW Stationary Wireless Charging of Light-Duty EV

Wireless power transfer (WPT) is a developing technology with the advantage of convenience and flexible charging. SAE recommended practice J2954 defines typical size and geometry with aluminum or ferrite plate shielding to limit leakage electromagnetic (EM) fields for WPT with power levels lower than 22 kVA from the input side. However, as the WPT power goes up to 100- or 200-kW level, EM safety surrounding the WPT becomes a critical concern. To address this oncoming safety challenge, a novel ferrite shielding design is proposed in this paper. Different misalignment scenarios in accordance with definitions in SAE J2954 are also taken into consideration to ensure EM safety under various operation scenarios. Simulation results, which are preliminarily verified by magnetic field measurements at 1.1 m from the center of the vehicle side coil under 100 kW operation, indicate that the magnetic field leakage can be maintained below the limits defined in SAE J2954 for 200 kW operation. A 3.3 kW scale-down test was also carried out and compared to parallel scale-down simulations. 26.8% field emission reduction is observed from the scale-down test, which supports the effectiveness of the proposed shielding design.

Zhang, Bo↗

DOE OSTI2020

Electromagnetic Shielding Design for 200 kW Stationary Wireless Charging of Light-Duty EV

Wireless power transfer (WPT) is a developing technology with the advantage of convenience and flexible charging. SAE recommended practice J2954 defines typical size and geometry with aluminum or ferrite plate shielding to limit leakage electromagnetic (EM) fields for WPT with power levels lower than 22 kVA from the input side. However, as the WPT power goes up to 100- or 200-kW level, EM safety surrounding the WPT becomes a critical concern. To address this oncoming safety challenge, a novel ferrite shielding design is proposed in this paper. Different misalignment scenarios in accordance with definitions in SAE J2954 are also taken into consideration to ensure EM safety under various operation scenarios. Simulation results, which are preliminarily verified by magnetic field measurements at 1.1 m from the center of the vehicle side coil under 100 kW operation, indicate that the magnetic field leakage can be maintained below the limits defined in SAE J2954 for 200 kW operation. A 3.3 kW scale-down test was also carried out and compared to parallel scale-down simulations. 26.8% field emission reduction is observed from the scale-down test, which supports the effectiveness of the proposed shielding design.

25 ENERGY STORAGE↗

DOE OSTI2021

Global Hybrid Simulations of Interaction Between Interplanetary Rotational Discontinuity and Bow Shock/Magnetosphere: Can Ion-Scale Magnetic Reconnection be Driven by Rotational Discontinuity Downstream of Quasi-Parallel Shock?

Ion-scale magnetic reconnection has been observed downstream of the terrestrial quasi-parallel (Q-∥) shock. Whether it is driven by interplanetary discontinuities or turbulent Q-∥ shock, however, is unclear. Using three-dimensional global hybrid simulation, we investigate the generation of magnetic reconnection downstream of the Q-∥ shock, while an interplanetary rotational discontinuity (RD) is launched to the bow shock. Cases with various solar wind Alfvén Mach numbers, M A = 3.0 to 8, and propagation directions of the RD are presented. The propagation direction n is assumed to be in the GSE xz plane and pointing earthward, with n = (-sin(θ 12 /2),0,-cos(θ 12 /2)), where θ 12 is the angle between the upstream (B1) and downstream (B2) magnetic fields across the transmitted RD. It is found that magnetic reconnection occurs inside the RD downstream of the Q-∥ shock, forming flux ropes extending along the dawn-dusk direction about tens of ion inertial lengths. Large-amplitude low-frequency waves originated from the Q-∥ shock lead to the bending and squeezing of the field lines around the RD, which play an important role in triggering reconnection inside the RD. As the RD impacts the dayside magnetopause, magnetopause reconnection takes place between the field lines behind the RD and geomagnetic field lines. Nevertheless, no reconnection is found downstream of the Q-∥ shock itself or outside the RD in the magnetosheath. The existent and structure of reconnection in the magnetosheath are found to strongly depend on the parameters M A and n. Our simulation shows that ion-scale magnetic reconnection is driven by an external driver in the form of the compression of an RD around the bow shock and in the magnetosheath, rather than caused by the turbulent Q-∥ shock alone.

79 ASTRONOMY AND ASTROPHYSICS↗

NASA NTRS2019

Real Time Photon-Counting Receiver for High Photon Efficiency Optical Communications

We present a scalable design for a photon-counting ground receiver based on superconducting nanowire single photon detectors (SNSPDs) and field programmable gate array (FPGA) real-time processing for applications to space-to-ground photon starved links, such as the Orion EM-2 Optical Communication Demonstration (O2O), and future deep space or low transmitter power missions. The receiver is designed to receive a serially concatenated pulse position modulation (SCPPM) waveform, which follows the Consultative Committee for Space Data Systems (CCSDS) Optical Communications Coding and Synchronization Red Book standard. The receiver design uses multiple individually fiber coupled, 80% detection efficiency commercial SNSPDs in parallel to scale to a required data rate, and is capable of achieving data rates up to 528 Mbps. For efficient fiber coupling from the telescope to the array of parallel detectors that can be scaled both to telescope aperture size and the number of detectors, we use either a single mode fiber (SMF) photonic lantern or a few-mode fiber (FMF) photonic lantern. In this paper we give an overview of the receiver system design, the characteristics of the photonic lanterns, the performance of the SNSPDs, and system level tests. We show that 40 Mbps can be received using a single SNSPD, and discuss aspects for scaling to higher data rates.

Vyhnalek, Brian E.↗

NASA NTRS2019

Real Time Photon-Counting Receiver for High Photon Efficiency Optical Communications

We present a scalable design for a photon-counting ground receiver based on superconducting nanowire single photon detectors (SNSPDs) and field programmable gate array (FPGA) real-time processing for applications to space-to-ground photon starved links, such as the Orion EM-2 Optical Communication Demonstration (O2O), and future deep space or low transmitter power missions. The receiver is designed to receive a serially concatenated pulse position modulation (SCPPM) waveform, which follows the Consultative Committee for Space Data Systems (CCSDS) Optical Communications Coding and Synchronization Red Book standard. The receiver design uses multiple individually fiber coupled, 80% detection efficiency commercial SNSPDs in parallel to scale to a required data rate, and is capable of achieving data rates up to 528 Mbps. For efficient fiber coupling from the telescope to the array of parallel detectors that can be scaled both to telescope aperture size and the number of detectors, we use either a single mode fiber (SMF) photonic lantern or a few-mode fiber (FMF) photonic lantern. In this paper we give an overview of the receiver system design, the characteristics of the photonic lanterns, the performance of the SNSPDs, and system level tests. We show that 40 Mbps can be received using a single SNSPD, and discuss aspects for scaling to higher data rates.

Vyhnalek, Brian E.↗

DOE OSTI2019

Fit Fly: A Case Study of Interconnect Innovation through Parallel Simulation

To meet the demand for exascale-level performance from high-performance computing (HPC) interconnects, many system architects are turning to simulation results for accurate and reliable predictions of the performance of prospective technologies. Testing full-scale networks with a variety of benchmarking tools, including synthetic workloads and application traces, can give crucial insight into what ideas are most promising without needing to physically construct a test network. While flexible, however, this approach is extremely compute time intensive. We address this time complexity challenge through the use of large-scale, optimistic parallel simulation that ultimately leads to faster HPC network architecture innovations. In this paper we demonstrate this innovation capability through a real-world network design case study. Specifically, we have simulated and compared four extreme-scale interconnects: Dragonfly, Megafly, Slim Fly, and a new dual-rail-dual-plane variation of the Slim Fly network topology. We present this new variant of Slim Fly, dubbed Fit Fly, to show how interconnect innovation and evaluation-beyond what is possible through analytic methods-can be achieved through parallel simulation. We validate and compare the model with various network designs using the CODES interconnect simulation framework. By running large-scale simulations in a parallel environment, we are able to quickly generate reliable performance results that can help network designers break ground on the next generation of high-performance network designs.

High Performance Computing↗

NASA NTRS1984

Scale lengths in quasi-parallel shocks

A review was carried out of ISEE and Voyager spacecraft magnetometer data to determine if quasi-parallel bow shocks are really broad, disordered regions. The key parameter was the deceleration scale (thickness, Lp) across which random energy would need to increase and a localized electrostatic field (E) would be present. Lp would define the breadth of the shock and be associated with a plasma deceleration. The ISEE 1 satellite collected data on the electron density, bulk speed, magnetic intensity, and electron temperature in November 1977 during five traverses of the bow shock. Similar data were gathered from an interplanetary shock wave in 1981. The evidence supported the concept of a plasma deceleration across a thin layer (Lp) in both types of shocks. The layers were about 50 times (interplanetary) and 20 times (earth) thinner than surrounding magnetic fluctuation regions. It is asserted that the regions of deceleration, although much thinner, are the actual shocks and not the entire regions of magnetic fluctuations.

Scudder, J. D.↗

NASA NTRS1995

Characterizing parallel file-access patterns on a large-scale multiprocessor

High-performance parallel file systems are needed to satisfy tremendous I/O requirements of parallel scientific applications. The design of such high-performance parallel file systems depends on a comprehensive understanding of the expected workload, but so far there have been very few usage studies of multiprocessor file systems. This paper is part of the CHARISMA project, which intends to fill this void by measuring real file-system workloads on various production parallel machines. In particular, we present results from the CM-5 at the National Center for Supercomputing Applications. Our results are unique because we collect information about nearly every individual I/O request from the mix of jobs running on the machine. Analysis of the traces leads to various recommendations for parallel file-system design.

Purakayastha, A.↗