Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallel optimization”

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 757 records · Page 42

LAMMPS strong scaling performance optimization on Blue Gene/Q

LAMMPS "Large-scale Atomic/Molecular Massively Parallel Simulator" is an open-source molecular dynamics package from Sandia National Laboratories. Significant performance improvements in strong-scaling and time-to-solution for this application on IBM's Blue Gene/Q have been achieved through computational optimizations of the OpenMP versions of the short-range Lennard-Jones term of the CHARMM force field and the long-range Coulombic interaction implemented with the PPPM (particle-particle-particle mesh) algorithm, enhanced by runtime parameter settings controlling thread utilization. Additionally, MPI communication performance improvements were made to the PPPM calculation by re-engineering the parallel 3D FFT to use MPICH collectives instead of point-to-point. Performance testing was done using an 8.4-million atom simulation scaling up to 16 racks on the Mira system at Argonne Leadership Computing Facility (ALCF). Speedups resulting from this effort were in some cases over 2x.

Coffman, Paul↗

Distributed Parallel Processing and Dynamic Load Balancing Techniques for Multidisciplinary High Speed Aircraft Design

Multidisciplinary design optimization (MDO) for large-scale engineering problems poses many challenges (e.g., the design of an efficient concurrent paradigm for global optimization based on disciplinary analyses, expensive computations over vast data sets, etc.) This work focuses on the application of distributed schemes for massively parallel architectures to MDO problems, as a tool for reducing computation time and solving larger problems. The specific problem considered here is configuration optimization of a high speed civil transport (HSCT), and the efficient parallelization of the embedded paradigm for reasonable design space identification. Two distributed dynamic load balancing techniques (random polling and global round robin with message combining) and two necessary termination detection schemes (global task count and token passing) were implemented and evaluated in terms of effectiveness and scalability to large problem sizes and a thousand processors. The effect of certain parameters on execution time was also inspected. Empirical results demonstrated stable performance and effectiveness for all schemes, and the parametric study showed that the selected algorithmic parameters have a negligible effect on performance.

Krasteva, Denitza T.↗

A multi-platform evaluation of the randomized CX low-rank matrix factorization in Spark

We investigate the performance and scalability of the randomized CX low-rank matrix factorization and demonstrate its applicability through the analysis of a 1TB mass spectrometry imaging (MSI) dataset, using Apache Spark on an Amazon EC2 cluster, a Cray XC40 system, and an experimental Cray cluster. We implemented this factorization both as a parallelized C implementation with hand-tuned optimizations and in Scala using the Apache Spark high-level cluster computing framework. We obtained consistent performance across the three platforms: using Spark we were able to process the 1TB size dataset in under 30 minutes with 960 cores on all systems, with the fastest times obtained on the experimental Cray cluster. In comparison, the C implementation was 21X faster on the Amazon EC2 system, due to careful cache optimizations, bandwidth-friendly access of matrices and vector computation using SIMD units. We report these results and their implications on the hardware and software issues arising in supporting data-centric workloads in parallel and distributed environments.

Gittens, Alex↗

Parallel Memory-Independent Communication Bounds for SYRK

In this paper, we focus on the parallel communication cost of multiplying a matrix with its transpose, known as a symmetric rank-k update (SYRK). SYRK requires half the computation of general matrix multiplication because of the symmetry of the output matrix. Recent work (Beaumont et al., SPAA '22) has demonstrated that the sequential I/O complexity of SYRK is also a constant factor smaller than that of general matrix multiplication. Inspired by this progress, we establish memory-independent parallel communication lower bounds for SYRK with smaller constants than general matrix multiplication, and we show that these constants are tight by presenting communication-optimal algorithms. The crux of the lower bound proof relies on extending a key geometric inequality to symmetric computations and analytically solving a constrained nonlinear optimization problem. Here, the optimal algorithms use a triangular blocking scheme for parallel distribution of the symmetric output matrix and corresponding computation.

Communication costs↗

Reaching bandwidth saturation using transparent injection parallelization

Although logically available, applications may not exploit enough instantaneous communication concurrency to maximize network utilization on HPC systems. This is exacerbated in hybrid programming models that combine single program multiple data with OpenMP or CUDA. We present the design of a “multi-threaded” runtime able to transparently increase the instantaneous network concurrency and to provide near saturation bandwidth, independent of the application configuration and dynamic behavior. The runtime offloads communication requests from application level tasks to multiple communication servers. The servers use system specific performance models to attain network saturation. Our techniques alleviate the need for spatial and temporal application level message concurrency optimizations. Experimental results show improved message throughput and bandwidth by as much as 150% for 4 KB messages on InfiniBand and by as much as 120% for 4 KB messages on Cray Aries. For more complex operations such as all-to-all collectives, we observe as much as 30% speedup. This translates into 23% speedup on 12,288 cores for a NAS FT implemented using FFTW. We observe as much as 76% speedup on 1500 cores for an already optimized UPC+OpenMP geometric multigrid application using hybrid parallelism. For the geometric multigrid GPU implementation, we observe as much as 44% speedup on 512 GPUs.

Chaimov, Nicholas↗

Experience with approximations in the trust-region parallel direct search algorithm.

Recent years have seen growth in the number of algorithms designed to solve challenging simulation-based nonlinear optimization problems. One such algorithm is the Trust-Region Parallel Direct Search (TRPDS) method developed by Hough and Meza. In this paper, we take advantage of the theoretical properties of TRPDS to make use of approximation models in order to reduce the computational cost of simulation-based optimization. We describe the extension, which we call mTRPDS, and present the results of a case study for two earth penetrator design problems. In the case study, we conduct computational experiments with an array of approximations within the mTRPDS algorithm and compare the numerical results to the original TRPDS algorithm and a trust-region method implemented using the speculative gradient approach described by Byrd, Schnabel, and Shultz. The results suggest new ways to improve the algorithm.

97 MATHEMATICS AND COMPUTING↗

NEXTorch: A Design and Bayesian Optimization Toolkit for Chemical Sciences and Engineering

Automation and optimization of chemical systems require well-informed decisions on what experiments to run to reduce time, materials, and/or computations. Data-driven active learning algorithms have emerged as valuable tools to solve such tasks. Bayesian optimization, a sequential global optimization approach, is a popular active-learning framework. Past studies have demonstrated its efficiency in solving chemistry and engineering problems. Here we introduce NEXTorch, a library in Python/PyTorch, to facilitate laboratory or computational design using Bayesian optimization. NEXTorch offers fast predictive modeling, flexible optimization loops, visualization capabilities, easy interfacing with legacy software, and multiple types of parameters and data type conversions. It provides GPU acceleration, parallelization, and state-of-the-art Bayesian optimization algorithms and supports both automated an d human-in-the-loop optimization. The comprehensive online documentation introduces Bayesian optimization theory and several examples from catalyst synthesis, reaction condition optimization, parameter estimation, and reactor geometry optimization. NEXTorch is open-source and available on GitHub

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Parallelization of an Object-Oriented Unstructured Aeroacoustics Solver

A computational aeroacoustics code based on the discontinuous Galerkin method is ported to several parallel platforms using MPI. The discontinuous Galerkin method is a compact high-order method that retains its accuracy and robustness on non-smooth unstructured meshes. In its semi-discrete form, the discontinuous Galerkin method can be combined with explicit time marching methods making it well suited to time accurate computations. The compact nature of the discontinuous Galerkin method also makes it well suited for distributed memory parallel platforms. The original serial code was written using an object-oriented approach and was previously optimized for cache-based machines. The port to parallel platforms was achieved simply by treating partition boundaries as a type of boundary condition. Code modifications were minimal because boundary conditions were abstractions in the original program. Scalability results are presented for the SCI Origin, IBM SP2, and clusters of SGI and Sun workstations. Slightly superlinear speedup is achieved on a fixed-size problem on the Origin, due to cache effects.

Baggag, Abdelkader↗

Solving Unit Commitment Problems with Demand Responsive Loads

This work focuses on using variations of the Frank-Wolfe (FW) algorithm for solving unit commitment problems with high volumes of demand responsive loads on the power grid. We present a formulation of the unit commitment problem with demand responsive loads. We then show through reformulation and relaxations of the problem that variations of the Frank-Wolfe algorithm can be used to determine the time series decisions for the demand responsive loads. We show through computational experiments on the IEEE Reliability Test System that the timeseries of demand responsive load decisions obtained through our approach are near optimal and describe how large-scale parallel implementations of our approach can be highly computationally efficient.

demand response↗

Parallel simulated annealing with embedded machine learning and multifidelity models for reactor core design

This paper presents extensions to a penalty-free, parallel simulated annealing (SA) algorithm for multi-constrained combinatorial optimization with the aim of embedding multi-fidelity physics models into the annealing procedure. The method uses a low-fidelity, quickly executing model for rapid design space exploration and a high-fidelity model for detailed constraint resolution and on-the-fly bias correction. Machine learning models updated within the annealing procedure were used to bridge the gap between the multi-fidelity models, which led to accurate rapid exploration and efficient detailed constraint resolution. A software implementation of the new multi-fidelity optimization methods, called ML-PSA, was demonstrated on a continuous multi-fidelity optimization problem and a constrained combinatorial PWR lattice design problem. These problems demonstrate some of the features, parallel performance characteristics, and extensible nature of the multi-fidelity SA methods. This paper shows that the developed software and procedure are a general optimization tool that can be applied to a wide variety of scientific and engineering design optimization applications. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Selection of Thermal Worst-Case Orbits via Modified Efficient Global Optimization

Efficient Global Optimization (EGO) was used to select orbits with worst-case hot and cold thermal environments for the Stratospheric Aerosol and Gas Experiment (SAGE) III. The SAGE III system thermal model changed substantially since the previous selection of worst-case orbits (which did not use the EGO method), so the selections were revised to ensure the worst cases are being captured. The EGO method consists of first conducting an initial set of parametric runs, generated with a space-filling Design of Experiments (DoE) method, then fitting a surrogate model to the data and searching for points of maximum Expected Improvement (EI) to conduct additional runs. The general EGO method was modified by using a multi-start optimizer to identify multiple new test points at each iteration. This modification facilitates parallel computing and decreases the burden of user interaction when the optimizer code is not integrated with the model. Thermal worst-case orbits for SAGE III were successfully identified and shown by direct comparison to be more severe than those identified in the previous selection. The EGO method is a useful tool for this application and can result in computational savings if the initial Design of Experiments (DoE) is selected appropriately.

Moeller, Timothy M.↗

Synchronous and asynchronous optimized Schwarz methods for one-way subdivision of bounded domains

Convergence of both synchronous and asynchronous optimized Schwarz algorithms for the shifted Laplacian operator on a bounded rectangular domain, in a one-way subdivision of the computational domain, with overlap, is shown. Convergence results are obtained under very mild conditions on the size of the subdomains and on the amount of overlap. A couple of results are also given, relating the convergence rate of the asynchronous method to changes in the size of the domain. Numerical experiments illustrate the theoretical results.

97 MATHEMATICS AND COMPUTING↗

Augmented Lagrangian formulation of orbital-free density functional theory

We present an Augmented Lagrangian formulation and its real-space implementation for non-periodic Orbital-Free Density Functional Theory (OF-DFT) calculations. In particular, we rewrite the constrained minimization problem of OF-DFT as a sequence of minimization problems without any constraint, thereby making it amenable to powerful unconstrained optimization algorithms. Further, we develop a parallel implementation of this approach for the Thomas–Fermi–von Weizsacker (TFW) kinetic energy functional in the framework of higher-order finite-differences and the conjugate gradient method. With this implementation, we establish that the Augmented Lagrangian approach is highly competitive compared to the penalty and Lagrange multiplier methods. Additionally, we show that higher-order finite-differences represent a computationally efficient discretization for performing OF-DFT simulations. Overall, we demonstrate that the proposed formulation and implementation are both efficient and robust by studying selected examples, including systems consisting of thousands of atoms. We validate the accuracy of the computed energies and forces by comparing them with those obtained by existing plane-wave methods.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

High-throughput feedback-enabled optogenetic stimulation and spectroscopy in microwell plates

Abstract The ability to perform sophisticated, high-throughput optogenetic experiments has been greatly enhanced by recent open-source illumination devices that allow independent programming of light patterns in single wells of microwell plates. However, there is currently a lack of instrumentation to monitor such experiments in real time, necessitating repeated transfers of the samples to stand-alone analytical instruments, thus limiting the types of experiments that could be performed. Here we address this gap with the development of the optoPlateReader (oPR), an open-source, solid-state, compact device that allows automated optogenetic stimulation and spectroscopy in each well of a 96-well plate. The oPR integrates an optoPlate illumination module with a module called the optoReader, an array of 96 photodiodes and LEDs that allows 96 parallel light measurements. The oPR was optimized for stimulation with blue light and for measurements of optical density and fluorescence. After calibration of all device components, we used the oPR to measure growth and to induce and measure fluorescent protein expression in E. coli . We further demonstrated how the optical read/write capabilities of the oPR permit computer-in-the-loop feedback control, where the current state of the sample can be used to adjust the optical stimulation parameters of the sample according to pre-defined feedback algorithms. The oPR will thus help realize an untapped potential for optogenetic experiments by enabling automated reading, writing, and feedback in microwell plates through open-source hardware that is accessible, customizable, and inexpensive.

59 BASIC BIOLOGICAL SCIENCES↗

Power and Performance Trade-offs for Space Time Adaptive Processing

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 implementation 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.

Gawande, Nitin A.↗

Exploiting Vector and Multicore Parallelsim for Recursive, Data- and Task-Parallel Programs

Modern hardware contains parallel execution resources that are well-suited for data-parallelism-vector units-and task parallelism-multicores. However, most work on parallel scheduling focuses on one type of hardware or the other. In this work, we present a scheduling framework that allows for a unified treatment of task- and data-parallelism. Our key insight is an abstraction, task blocks, that uniformly handles data-parallel iterations and task-parallel tasks, allowing them to be scheduled on vector units or executed independently as multicores. Our framework allows us to define schedulers that can dynamically select between executing task- blocks on vector units or multicores. We show that these schedulers are asymptotically optimal, and deliver the maximum amount of parallelism available in computation trees. To evaluate our schedulers, we develop program transformations that can convert mixed data- and task-parallel pro- grams into task block-based programs. Using a prototype instantiation of our scheduling framework, we show that, on an 8-core system, we can simultaneously exploit vector and multicore parallelism to achieve 14×-108× speedup over sequential baselines.

Ren, Bin↗

LLNL/thicket

Thicket is a python-based toolkit for Exploratory Data Analysis (EDA) of parallel performance data that enables performance optimization and understanding of applications’ performance on supercomputers. It bridges the performance tool gap between being able to consider only a single instance of a simulation run (e.g., single platform, single measurement tool, or single scale) and finding actionable insights in multi-dimensional, multi-scale, multi-architecture, and multi-tool performance datasets.

Brink, Stephanie Labasan↗