Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Parallel Optimization Algorithms”

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 469 records · Page 26
All sourcesNASA NTRSDOE OSTI
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↗
DOE OSTI2024

Parallel interior-point solver for block-structured nonlinear programs on SIMD/GPU architectures

Here, we investigate how to port the standard interior-point method to new exascale architectures for block-structured nonlinear programs with state equations. Computationally, we decompose the interior-point algorithm into two successive operations: the evaluation of the derivatives and the solution of the associated Karush-Kuhn-Tucker (KKT) linear system. Our method accelerates both operations using two levels of parallelism. First, we distribute the computations on multiple processes using coarse parallelism. Second, each process uses SIMD/GPU accelerators locally to accelerate the operations using fine-grained parallelism. The KKT system is reduced by eliminating the inequalities and the state variables from the corresponding equations. We demonstrate our method's capability on the supercomputer Polaris, a testbed for the future exascale Aurora system. Each node is equipped with four GPUs, a setup amenable to our two-level approach. Our experiments on the stochastic optimal power flow problem show that the reduction method is 50x faster than the sparse linear solver HSL MA57 running in serial on the CPU, and 6x faster than Pardiso running in parallel on CPU on the same number of processes.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2017

TGE: Machine Learning Based Task Graph Embedding for Large-Scale Topology Mapping

Task mapping is an important problem in parallel and distributed computing. The goal in task mapping is to find an optimal layout of the processes of an application (or a task) onto a given network topology. We target this problem in the context of staging applications. A staging application consists of two or more parallel applications (also referred to as staging tasks) which run concurrently and exchange data over the course of computation. Task mapping becomes a more challenging problem in staging applications, because not only data is exchanged between the staging tasks, but also the processes of a staging task may exchange data with each other. We propose a novel method, called Task Graph Embedding (TGE), that harnesses the observable graph structures of parallel applications and network topologies. TGE employs a machine learning based algorithm to find the best representation of a graph, called an embedding, onto a space in which the task-to-processor mapping problem can be solved. We evaluate and demonstrate the effectiveness of TGE experimentally with the communication patterns extracted from runs of XGC, a large-scale fusion simulation code, on Titan.

Choi, Jong Youl↗
DOE OSTI2022

Optimization of Thermal Conductance at Interfaces Using Machine Learning Algorithms

We report optimization of thermal transport across the interface of two different materials is critical to micro-/nanoscale electronic, photonic, and phononic devices. Although several examples of compositional intermixing at the interfaces having a positive effect on interfacial thermal conductance (ITC) have been reported, an optimum arrangement has not yet been determined because of the large number of potential atomic configurations and the significant computational cost of evaluation. On the other hand, computation-driven materials design efforts are rising in popularity and importance. Yet, the scalability and transferability of machine learning models remain as challenges in creating a complete pipeline for the simulation and analysis of large molecular systems. In this work we present a scalable Bayesian optimization framework, which leverages dynamic spawning of jobs through the Message Passing Interface (MPI) to run multiple parallel molecular dynamics simulations within a parent MPI job to optimize heat transfer at the silicon and aluminum (Si/Al) interface. We found a maximum of 50% increase in the ITC when introducing a two-layer intermixed region that consists of a higher percentage of Si. Because of the random nature of the intermixing, the magnitude of increase in the ITC varies. We observed that both homogeneity/heterogeneity of the intermixing and the intrinsic stochastic nature of molecular dynamics simulations account for the variance in ITC.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2017

Extreme-Scale Algorithms & Software Resilience (EASIR) Architecture-Aware Algorithms for Scalable Performance and Resilience on Heterogeneous Architectures

This project addresses both communication-avoiding algorithms, and reproducible floating-point computation. Communication, i.e. moving data, either between levels of memory or processors over a network, is much more expensive per operation than arithmetic (measured in time or energy), so we seek algorithms that greatly reduce communication. We developed many new algorithms for both dense and sparse, and both direct and iterative linear algebra, attaining new communication lower bounds, and getting large speedups in many cases. We also extended this work in several ways: (1) We minimize writes separately from reads, since writes may be much more expensive than reads on emerging memory technologies, like Flash, sometimes doing asymptotically fewer writes than reads. (2) We extend the lower bounds and optimal algorithms to arbitrary algorithms that may be expressed as perfectly nested loops accessing arrays, where the array subscripts may be arbitrary affine functions of the loop indices (eg A(i), B(i,j+k, k+3*m-7, …) etc.). (3) We extend our communication-avoiding approach to some machine learning algorithms, such as support vector machines. This work has won a number of awards. We also address reproducible floating-point computation. We define reproducibility to mean getting bitwise identical results from multiple runs of the same program, perhaps with different hardware resources or other changes that should ideally not change the answer. Many users depend on reproducibility for debugging or correctness. However, dynamic scheduling of parallel computing resources, combined with nonassociativity of floating point addition, makes attaining reproducibility a challenge even for simple operations like summing a vector of numbers, or more complicated operations like the Basic Linear Algebra Subprograms (BLAS). We describe an algorithm that computes a reproducible sum of floating point numbers, independent of the order of summation. The algorithm depends only on a subset of the IEEE Floating Point Standard 754-2008, uses just 6 words to represent a “reproducible accumulator,” and requires just one read-only pass over the data, or one reduction in parallel. New instructions based on this work are being considered for inclusion in the future IEEE 754-2018 floating-point standard, and new reproducible BLAS are being considered for the next version of the BLAS standard.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Scalable branching on dual decomposition of stochastic mixed-integer programming problems

In this work, we present a scalable branching method for the dual decomposition of stochastic mixed-integer programming. Our new branching method is based on the branching method proposed by Caroe and Schultz that creates branching disjunctions on first-stage variables only. We propose improvements to the process for creating branching disjunctions, including (1) branching on the optimal solutions of the Dantzig-Wolfe reformulation of the restricted master problem and (2) using a more comprehensive (yet simple) measure for the dispersions associated with subproblem solution infeasibility. We prove that the proposed branching process leads to an algorithm that terminates finitely, and we provide conditions under which globally optimal solutions can be identified after termination. We have implemented our new branching method, as well as the Caroe-Schultz method and a branch-and-price method, in the open-source software package DSP. Using SIPLIB test instances, we present extensive numerical results to demonstrate that the proposed branching method significantly reduces the number of node subproblems and solution times.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2017

Aspen-based performance and energy modeling frameworks

Now with the anticipation of exascale architectures, energy consumption is becoming one of the critical design parameters, especially in light of the energy budget of 20-30 Megawatts set by the U.S. Department of Energy. Understanding an application’s execution pattern and its energy footprint is critical to improving application operation on a diverse heterogeneous architecture. Applying application-specific performance optimization can consequently improve energy consumption. However, this approach is only applicable to current systems. As we enter a new era of exascale architectures that is projected to contain more complex memory hierarchies, increased levels of parallelism, heterogeneity in hardware, and complex programming models and techniques, energy and performance management is getting more cumbersome. We therefore propose techniques that predict the energy consumption beforehand or at runtime to enable proactive tuning. Such energy prediction approaches must be generic and adapt themselves at runtime to changing application and hardware configurations. Most existing energy estimation and prediction approaches are empirical in nature and thus tied to current systems. To overcome this limitation, we propose two energy estimation techniques: ACEE (Algorithmic and Categorical Energy Estimation), which uses a combination of analytical and empirical modeling techniques; and AEEM (Aspen’s Embedded Energy Estimation), a system-level analytical energy estimation technique. Both of these models incorporate the Aspen domain specific language for performance modeling. We present the methodologies of these two models and test their accuracy using five proxy applications. We also describe three use cases.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2018

Exploiting Geometric Partitioning in Task Mapping for Parallel Computes

We present a new method for mapping applications' MPI tasks to cores of a parallel computer such that applications' communication time is reduced. We address the case of sparse node allocation, where the nodes assigned to a job are not necessarily located in a contiguous block nor within close proximity to each other in the network, although our methods generalize to contiguous allocations as well. The goal is to assign tasks to cores so that interdependent tasks are performed by "nearby' cores, thus lowering the distance messages must travel, the amount of congestion in the network, and the overall cost of communication. Our new method applies a geometric partitioning algorithm to both the tasks and the processors, and assigns task parts to the corresponding processor parts. We also present a number of algorithmic optimizations that exploit specific features of the network or application. We show that, for the structured finite difference mini-application MiniGhost, our mapping methods reduced communication time up to 75% relative to MiniGhost's default mapping on 128K cores of a Cray XK7 with sparse allocation. For the atmospheric modeling code E3SM/HOMME, our methods reduced communication time up to 31% on 32K cores of an IBM BlueGene/Q with contiguous allocation.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

MPMC and MCMD: Free High-Performance Simulation Software for Atomistic Systems

Advancements in parallel computing and hardware have allowed computational exploration of chemical systems of interest with unprecendented accuracy and efficiency. The typical development of molecular simulation software is initially inspired by a particular scientific inquiry. The softwares presented herein, MPMC (Massively Parallel Monte Carlo) and MCMD (Monte Carlo/Molecular Dynamics) were born out of a pursuit to simulate condensed phase physical and chemical interactions in porous materials. MPMC first began in 2005 and has been used for dozens of published results in the literature but has not yet been introduced in a standalone paper. MCMD is a more recent expansion and re-write with some published work, focused on adding molecular dynamics algorithms for transport and other time-dependent properties of chemical systems. Each software functions as a standalone with some unique and other overlapping features. Here, a driving aim of this work is to consider periodic, long-range polarization effects in classical simulation, and both codes are optimized to perform such calculations. Sample results will be presented which highlight methodically that inclusion of explicit polarization produces simulation results with predictive power which in general are in greater agreement with experiment than non-polarizable analogous simulations.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Multichannel Analysis of Surface Waves Accelerated (MASWAccelerated): Software for efficient surface wave inversion using MPI and GPUs

Multichannel Analysis of Surface Waves (MASW) is a technique frequently used in geotechnical engineering and engineering geophysics to infer 1D layered models of seismic shear wave velocities in the top tens to hundreds of meters of the subsurface. We aim to accelerate MASW calculations by capitalizing on modern computer hardware available in the workstations of most engineers: multiple cores and graphics processing units (GPUs). We propose new parallel and GPU accelerated algorithms for computing 1D MASW inversion, and provide software implementations in C using Message Passing Interface (MPI) and CUDA. These algorithms take advantage of sparsity that arises in the problem, and the work balance between processes considers typical data trends. We compare our methods to an existing open source Matlab MASW tool. Our serial C implementation achieves a 2x speedup over the Matlab software, and we continue to see improvements by parallelizing the problem with MPI. Here we see nearly perfect strong and weak scaling for uniform data, and improve strong scaling for realistic data by repartitioning the problem to process mapping. By utilizing GPUs available on most modern workstations, we observe an additional 1.3x speedup over the serial C implementation on the first use of the method. We typically repeatedly evaluate theoretical dispersion curves as part of an optimization procedure, and on the GPU the kernel can be cached for faster reuse on later runs. We observe a 3.2x speedup on the cached GPU runs compared to the serial C runs. This work is the first open-source parallel or GPU-accelerated software tool for MASW imaging, and should enable geotechnical engineers to fully utilize all computer hardware at their disposal.

58 GEOSCIENCES↗
DOE OSTI2019

The Continuous Readout Stream of the MicroBooNE Liquid Argon Time Projection Chamber for Detection of Supernova Neutrinos

The MicroBooNE Continuous Readout Stream is a parallel readout of the MicroBooNE liquid argon time projection chamber (LArTPC) which enables detection of non-beam events such as those from a supernova neutrino burst. The low energies of these neutrinos and the intense cosmic-ray flux due to the near-surface detector location makes triggering on these events very challenging. Instead, MicroBooNE has pioneered a novel approach for detecting supernova neutrinos relying on a delayed trigger generated by SNEWS (the Supernova Early Warning System). The continuous readout of the LArTPC is achieved through the use of real-time compression algorithms (zero suppression and Huffman compression) implemented in an FPGA (field-programmable gate array) in the readout electronics. We present the results of the optimization of the data reduction algorithms, and their operational performance. In order to demonstrate the capability of the continuous stream to detect low-energy electrons, a sample of Michel electrons from stopping cosmic-ray muons is reconstructed and compared to a similar sample from the lossless triggered readout stream.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2016

Development of a Stiffness-Based Chemistry Load Balancing Scheme, and Optimization of Input/Output and Communication, to Enable Massively Parallel High-Fidelity Internal Combustion Engine Simulations

A closed-cycle gasoline compression ignition engine simulation near top dead center (TDC) was used to profile the performance of a parallel commercial engine computational fluid dynamics code, as it was scaled on up to 4096 cores of an IBM Blue Gene/Q supercomputer. The test case has 9 million cells near TDC, with a fixed mesh size of 0.15 mm, and was run on configurations ranging from 128 to 4096 cores. Profiling was done for a small duration of 0.11 crank angle degrees near TDC during ignition. Optimization of input/output performance resulted in a significant speedup in reading restart files, and in an over 100-times speedup in writing restart files and files for post-processing. Improvements to communication resulted in a 1400-times speedup in the mesh load balancing operation during initialization, on 4096 cores. An improved, “stiffness-based” algorithm for load balancing chemical kinetics calculations was developed, which results in an over 3-times faster run-time near ignition on 4096 cores relative to the original load balancing scheme. With this improvement to load balancing, the code achieves over 78% scaling efficiency on 2048 cores, and over 65% scaling efficiency on 4096 cores, relative to 256 cores.

CFD, Mira, Supercomputing, High Performance Comput↗
DOE OSTI2015

GraphReduce: Processing Large-Scale Graphs on Accelerator-Based Systems

Recent work on real-world graph analytics has sought to leverage the massive amount of parallelism offered by GPU devices, but challenges remain due to the inherent irregularity of graph algorithms and limitations in GPU-resident memory for storing large graphs. We present GraphReduce, a highly efficient and scalable GPU-based framework that operates on graphs that exceed the device’s internal memory capacity. GraphReduce adopts a combination of edge- and vertex-centric implementations of the Gather-Apply-Scatter programming model and operates on multiple asynchronous GPU streams to fully exploit the high degrees of parallelism in GPUs with efficient graph data movement between the host and device.

Architecture optimization, reuse, performance, ene↗
DOE OSTI2016

iTOUGH2 v7.1

iTOUGH2 (inverse TOUGH2) provides inverse modeling capabilities for TOUGH2, a simulator for multi-dimensional, multi-phase, multi-component, non-isothermal flow and transport in fractured porous media. iTOUGH2 performs sensitivity analyses, data-worth analyses, parameter estimation, and uncertainty propagation analyses in geosciences and reservoir engineering and other application areas. iTOUGH2 supports a number of different combinations of fluids and components (equation-of-state (EOS) modules). In addition, the optimization routines implemented in iTOUGH2 can also be used for sensitivity analysis, automatic model calibration, and uncertainty quantification of any external code that uses text-based input and output files using the PEST protocol. iTOUGH2 solves the inverse problem by minimizing a non-linear objective function of the weighted differences between model output and the corresponding observations. Multiple minimization algorithms (derivative-free, gradient-based, and second-order; local and global) are available. iTOUGH2 also performs Latin Hypercube Monte Carlo simulations for uncertainty propagation analyses. A detailed residual and error analysis is provided. This upgrade includes (a) global sensitivity analysis methods, (b) dynamic memory allocation (c) additional input features and output analyses, (d) increased forward simulation capabilities, (e) parallel execution on multicore PCs and Linux clusters, and (f) bug fixes.

Finsterle, Stefan↗
DOE OSTI2021

The continuous readout stream of the MicroBooNE liquid argon time projection chamber for detection of supernova burst neutrinos

The MicroBooNE continuous readout stream is a parallel readout of the MicroBooNE liquid argon time projection chamber (LArTPC) which enables detection of non-beam events such as those from a supernova neutrino burst. The low energies of the supernova neutrinos and the intense cosmic-ray background flux due to the near-surface detector location makes triggering on these events very challenging. Instead, MicroBooNE relies on a delayed trigger generated by SNEWS (the Supernova Early Warning System) for detecting supernova neutrinos. The continuous readout of the LArTPC generates large data volumes, and requires the use of real-time compression algorithms (zero suppression and Huffman compression) implemented in an FPGA (field-programmable gate array) in the readout electronics. In this paper we present the results of the optimization of the data reduction algorithms, and their operational performance. To demonstrate the capability of the continuous stream to detect low-energy electrons, a sample of Michel electrons from stopping cosmic-ray muons is reconstructed and compared to a similar sample from the lossless triggered readout stream.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2015

GraphReduce: Large-Scale Graph Analytics on Accelerator-Based HPC Systems

Recent work on real-world graph analytics has sought to leverage the massive amount of parallelism offered by GPU devices, but challenges remain due to the inherent irregularity of graph algorithms and limitations in GPU-resident memory for storing large graphs. We present GraphReduce, a highly efficient and scalable GPU-based framework that operates on graphs that exceed the device’s internal memory capacity. GraphReduce adopts a combination of both edge- and vertex-centric implementations of the Gather-Apply-Scatter programming model and operates on multiple asynchronous GPU streams to fully exploit the high degrees of parallelism in GPUs with efficient graph data movement between the host and the device.

Architecture optimization, reuse, performance, ene↗
DOE OSTI2024

Porting Classical Approaches for Quantum Simulations to Quantum Computers

Simulating quantum many-body systems is one of the most promising problems in which we might anticipate that quantum computers should show quantum advantage. Unfortunately, there is still a gap between this promise and actual practice. New quantum algorithms need to be developed and the current quantum algorithms have various difficulties - e.g efficient state preparation - which must be overcome and improved upon. In many cases, classical approaches need to be ported over to quantum devices. In this project we have developed a suite of new quantum algorithms which makes progress in this regard. We developed a new optimization scheme for variational quantum eigensolvers, UBOS, which mitigates problems with local minimas and barren plateaus while improving convergence to the ground state by an order of magnitude. We developed a new way to utilize qubitization to find ground states of nearly frustration-free Hamiltonians faster than all previous methods. We developed a series of state preparation techniques which helps initialize parameterized quantum circuits into reasonable starting points on which quantum algorithms are then applied. In addition to the development of novel algorithms, it is critical to have classical simulation techniques for approximately simulating quantum circuits which can be used to benchmark and understand quantum algorithms. Toward that end, we developed a novel POVM formalism to simulate quantum circuits as well as exemplify the massive parallelization of tensor network methodologies. Finally, we developed physical understanding of entanglement phase transitions such as many-body localization and random tensor networks.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2018

Scaling Deep Learning workloads: NVIDIA DGX-1/Pascal and Intel Knights Landing

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. Here, this article 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. Our 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. We use sequentially equivalent implementations to maintain iso-accuracy between parallel and sequential DL models. 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.

97 MATHEMATICS AND COMPUTING↗