Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “code 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 1,585 records · Page 88

Wind farm turbine type and placement optimization

The layout of turbines in a wind farm is already a challenging nonlinear, nonconvex, nonlinearly constrained continuous global optimization problem. Here we begin to address the next generation of wind farm optimization problems by adding the complexity that there is more than one turbine type to choose from. The optimization becomes a nonlinear constrained mixed integer problem, which is a very difficult class of problems to solve. Furthermore, this document briefly summarizes the algorithm and code we have developed, the code validation steps we have performed, and the initial results for multi-turbine type and placement optimization (TTP_OPT) we have run.

17 WIND ENERGY↗

Strong scaling of general-purpose molecular dynamics simulations on GPUs

We describe a highly optimized implementation of MPI domain decomposition in a GPU-enabled, general-purpose molecular dynamics code, HOOMD-blue (Anderson and Glotzer, 2013). Here, our approach is inspired by a traditional CPU-based code, LAMMPS (Plimpton, 1995), but is implemented within a code that was designed for execution on GPUs from the start (Anderson et al., 2008). The software supports short-ranged pair force and bond force fields and achieves optimal GPU performance using an autotuning algorithm. We are able to demonstrate equivalent or superior scaling on up to 3375 GPUs in Lennard-Jones and dissipative particle dynamics (DPD) simulations of up to 108 million particles. GPUDirect RDMA capabilities in recent GPU generations provide better performance in full double precision calculations. For a representative polymer physics application, HOOMD-blue 1.0 provides an effective GPU vs. CPU node speed-up of 12.5x.

97 MATHEMATICS AND COMPUTING↗

Gaussian processes with built-in dimensionality reduction: Applications to high-dimensional uncertainty propagation

Uncertainty quantification (UQ) tasks, such as model calibration, uncertainty propagation, and optimization under uncertainty, typically require several thousand evaluations of the underlying computer codes. To cope with the cost of simulations, one replaces the real response surface with a cheap surrogate based, e.g., on polynomial chaos expansions, neural networks, support vector machines, or Gaussian processes (GP). However, the number of simulations required to learn a generic multivariate response grows exponentially as the input dimension increases. This curse of dimensionality can only be addressed, if the response exhibits some special structure that can be discovered and exploited. A wide range of physical responses exhibit a special structure known as an active subspace (AS). An AS is a linear manifold of the stochastic space characterized by maximal response variation. The idea is that one should first identify this low dimensional manifold, project the high-dimensional input onto it, and then link the projection to the output. If the dimensionality of the AS is low enough, then learning the link function is a much easier problem than the original problem of learning a high-dimensional function. The classic approach to discovering the AS requires gradient information, a fact that severely limits its applicability. Furthermore, and partly because of its reliance to gradients, it is not able to handle noisy observations. The latter is an essential trait if one wants to be able to propagate uncertainty through stochastic simulators, e.g., through molecular dynamics codes. In this work, we develop a probabilistic version of AS which is gradient-free and robust to observational noise. Our approach relies on a novel Gaussian process regression with built-in dimensionality reduction. In particular, the AS is represented as an orthogonal projection matrix that serves as yet another covariance function hyper-parameter to be estimated from the data. To train the model, we design a two-step maximum likelihood optimization procedure that ensures the orthogonality of the projection matrix by exploiting recent results on the Stiefel manifold, i.e., the manifold of matrices with orthogonal columns. The additional benefit of our probabilistic formulation, is that it allows us to select the dimensionality of the AS via the Bayesian information criterion. We validate our approach by showing that it can discover the right AS in synthetic examples without gradient information using both noiseless and noisy observations. We demonstrate that our method is able to discover the same AS as the classical approach in a challenging one-hundred-dimensional problem involving an elliptic stochastic partial differential equation with random conductivity. Finally, we use our approach to study the effect of geometric and material uncertainties in the propagation of solitary waves in a one dimensional granular system.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Frame average optimization of cine-mode EPID images used for routine clinical in vivo patient dose verification of VMAT deliveries

Purpose: The in vivo 3D dose delivered to a patient during volumetric modulated arc therapy (VMAT) delivery can be calculated using electronic portal imaging device (EPID) images. These images must be acquired in cine-mode (i.e., “movie” mode) in order to capture the time-dependent delivery information. The angle subtended by each cine-mode EPID image during an arc can be changed via the frame averaging number selected within the image acquisition software. A large frame average number will decrease the EPID’s angular resolution and will result in a decrease in the accuracy of the dose information contained within each image. Alternatively, less EPID images acquired per delivery will decrease the overall 3D patient dose calculation time, which is appealing for large-scale clinical implementation. Therefore, the purpose of this study was to determine the optimal frame average value per EPID image, defined as the highest frame averaging that can be used without an appreciable loss in 3D dose reconstruction accuracy for VMAT treatments. Methods: Six different VMAT plans and six different SBRT-VMAT plans were delivered to an anthropomorphic phantom. Delivery was carried out on a Varian 2300ix model linear accelerator (Linac) equipped with an aS1000 EPID running at a frame acquisition rate of 7.5 Hz. An additional PC was set up at the Linac console area, equipped with specialized frame-grabber hardware and software packages allowing continuous acquisition of all EPID frames during delivery. Frames were averaged into “frame-averaged” EPID images using MATLAB. Each frame-averaged data set was used to calculate the in vivo dose to the patient and then compared to the single EPID frame in vivo dose calculation (the single frame calculation represents the highest possible angular resolution per EPID image). A mean percentage dose difference of low dose (<20% prescription dose) and high dose regions (>80% prescription dose) was calculated for each frame averaged scenario for each plan. The authors defined their unacceptable loss of accuracy as no more than a ±1% mean dose difference in the high dose region. Optimal frame average numbers were then determined as a function of the Linac’s average gantry speed and the dose per fraction. Results: The authors found that 9 and 11 frame averages were suitable for all VMAT and SBRT-VMAT treatments, respectively. This resulted in no more than a 1% loss to any of the dose region’s mean percentage difference when compared to the single frame reconstruction. The optimized number was dependent on the treatment’s dose per fraction and was determined to be as high as 14 for 12 Gy/fraction (fx), 15 for 8 Gy/fx, 11 for 6 Gy/fx, and 9 for 2 Gy/fx. Conclusions: The authors have determined an optimal EPID frame averaging number for multiple VMAT-type treatments. These are given as a function of the dose per fraction and average gantry speed. These optimized values are now used in the authors’ clinical, 3D, in vivo patient dosimetry program. This provides a reduction in calculation time while maintaining the authors’ required level of accuracy in the dose reconstruction.

61 RADIATION PROTECTION AND DOSIMETRY↗

ARPA-E Grid Optimization (GO) Competition Challenge 3

Synthetic Input Data and Team Results for the GO Competition Challenge 3 for Events 1 - 4 and the Sandbox, along with problem and format descriptions and code to validate data and solutions, are available here. Data for industry scenarios will not be made public. The Grid Optimization (GO) Competition Challenge 3 focused on the security-constrained optimal power flow (SCOPF) problem. It is part of a continuing effort begun with Challenges 1 and 2, to successfully discover, develop, and test innovative and disruptive software solutions for critical energy challenges and to overcome existing barriers. The broader goal of the of the GO Competition is to accelerate the development of transformational and disruptive methods for solving problems related to the electric power grid and to provide a transparent, fair, and comprehensive evaluation of new solution methods. Challenge 3 used multiperiod dynamic markets, including advisory models for extreme weather events, day-ahead markets, and the real-time markets with an extended look-ahead. In Event 4, whose submission window was August 31-September 4, 2023, 14 teams solved for the objective values of 669 scenarios (39 scenarios required solutions both with and without line switching being allowed). The 591 synthetic scenarios from 9 network models (3.6 GB) are available here. Ten teams were funded to participate and 7 won prizes totaling $2,400,000. The largest prize ($550,000) went to Mississippi State University. An additional $600,000 was awarded in Event 3 (6/15-16/2023). No prizes were awarded in Events 1 (1/25-27/2023) or 2 (4/13-14/2023). For more information on the competition and challenge see the "GO Competition Challenge 3 Information" resource below.

ACOPF↗

Advanced Code for Photocathode Design

The Phase I activity demonstrated that PhotoQE could be upgraded and modified to allow input using a graphical user interface. Specific calls to platform-dependent (e.g. IMSL) function calls were removed, and Fortran77 components were rewritten for Fortran95 compliance. The subroutines, specifically the common block structures and shared data parameters, were reworked to allow the GUI to update material parameter data, and the system was targeted for desktop personal computer operation. The new structures overcomes the previous rigid and unmodifiable library structures by implementing new, materials library data sets and repositioning the library values to external files. Material data may originate from published literature or experimental measurements. Further optimization and restructuring would allow custom and specific emission models for beam codes that rely on parameterized photoemission algorithms. These would be based on simplified and parametric representations updated and extended from previous versions (e.g., Modified Fowler-Dubridge, Modified Three-Step, etc.).

43 PARTICLE ACCELERATORS↗

CG-Kit: Code Generation Toolkit for performant and maintainable variants of source code applied to Flash-X hydrodynamics simulations

CG-Kit is a new Code Generation tool-Kit that we have developed as a part of the solution for portability and maintainability for multiphysics computing applications. The development of CG-Kit is rooted in the urgent need created by the shifting landscape of high-performance computing platforms and the algorithmic complexities of a particular large-scale multiphysics application: Flash-X. To efficiently use computing resources on a heterogeneous node, an application must have a map of computation to resources and a mechanism to move the data and computation to the resources according to the map. Most existing performance portability solutions are focussed on abstracting the expression of computations so that a unified source code can be specialized to run on different resources. However, such an approach is insufficient for a code like Flash-X, which has a multitude of code components that can be assembled in various permutations and combinations to form different instances of applications. Similar challenges apply to any code that has composability, where a single specified way of apportioning work among devices may not be optimal. Additionally, use cases arise where the optimal control flow of computation may differ for different devices while the underlying numerics remain identical. This combination leads to unique challenges including handling an existing large code base in Fortran and/or C/C++, subdivision of code into a great variety of units supporting a wide range of physics and numerical methods, different parallelization techniques for distributed and shared memory systems and accelerator devices, and heterogeneity of computing platforms requiring coexisting variants of parallel algorithms. All of these challenges demand that scientific software developers apply existing knowledge about domain applications, algorithms, and computing platforms to determine custom abstractions and granularity for code generation. There is a critical lack of tools to tackle those problems. CG-Kit is designed to fill this gap by providing a user with the ability to express their desired control flow and computation-to-resource map in the form a pseudocode-like recipe. It consists of standalone tools that can be combined into highly specific and, we argue, highly effective portability and maintainability toolchains. Here we present the design of our new tools: parametrized source trees, control flow graphs, and recipes. The tools are implemented in Python. They are agnostic to the programming language of the source code targeted for code generation. In conclusion, we demonstrate the capabilities of the toolkit with two examples, first, multithreaded variants of the basic AXPY operation, and second, variants of parallel algorithms within a hydrodynamics solver, called Spark, from Flash-X that operates on block-structured adaptive meshes.

Algorithmic portability↗

Asynchronous Iterative Solvers for Extreme-Scale Computing

The Asynchronous Iterative Solvers for Extreme-Scale Computing (AsyncIS) project aims to explore more efficient numerical algorithms by decreasing their overhead. AsyncIS does this by replacing the outer Krylov subspace solver with an asynchronous optimized Schwarz method, thereby removing the global synchronization and bulk synchronous operations typically used in numerical codes. AsyncIS—a U.S. Department of Energy (DOE)-funded collaboration between Georgia Tech, the University of Tennessee, Knoxville, Temple University, and Sandia National Laboratories—also focuses on the development and optimization of asynchronous preconditioners (i.e., preconditioners that are generated and/or applied in an asynchronous fashion). The novel preconditioning algorithms that provide fine-grained parallelism enable preconditioned Krylov solvers to run efficiently on large-scale distributed systems and manycore accelerators like GPUs.

97 MATHEMATICS AND COMPUTING↗

GX: a GPU-native gyrokinetic turbulence code for tokamak and stellarator design

GX is a code designed to solve the nonlinear gyrokinetic system for low-frequency turbulence in magnetized plasmas, particularly tokamaks and stellarators. In GX, our primary motivation and target is a fast gyrokinetic solver that can be used for fusion reactor design and optimization along with wide-ranging physics exploration. Here, this has led to several code and algorithm design decisions, specifically chosen to prioritize time to solution. First, we have used a discretization algorithm that is pseudospectral in the entire phase space, including a Laguerre–Hermite pseudospectral formulation of velocity space, which allows for smooth interpolation between coarse gyrofluid-like resolutions and finer conventional gyrokinetic resolutions and efficient evaluation of a model collision operator. Additionally, we have built GX to natively target graphics processors (GPUs), which are among the fastest computational platforms available today. Finally, we have taken advantage of the reactor-relevant limit of small $\rho _*$ by using the radially local flux-tube approach. In this paper we present details about the gyrokinetic system and the numerical algorithms used in GX to solve the system. We then present several numerical benchmarks against established gyrokinetic codes in both tokamak and stellarator magnetic geometries to verify that GX correctly simulates gyrokinetic turbulence in the small $\rho _*$. Moreover, we show that the convergence properties of the Laguerre–Hermite spectral velocity formulation are quite favourable for nonlinear problems of interest. Coupled with GPU acceleration, which we also investigate with scaling studies, this enables GX to be able to produce useful turbulence simulations in minutes on one (or a few) GPUs and higher fidelity results in a few hours using several GPUs. GX is open-source software that is ready for fusion reactor design studies.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Coding for Parallel Links to Maximize the Expected Value of Decodable Messages

When multiple parallel communication links are available, it is useful to consider link-utilization strategies that provide tradeoffs between reliability and throughput. Interesting cases arise when there are three or more available links. Under the model considered, the links have known probabilities of being in working order, and each link has a known capacity. The sender has a number of messages to send to the receiver. Each message has a size and a value (i.e., a worth or priority). Messages may be divided into pieces arbitrarily, and the value of each piece is proportional to its size. The goal is to choose combinations of messages to send on the links so that the expected value of the messages decodable by the receiver is maximized. There are three parts to the innovation: (1) Applying coding to parallel links under the model; (2) Linear programming formulation for finding the optimal combinations of messages to send on the links; and (3) Algorithms for assisting in finding feasible combinations of messages, as support for the linear programming formulation. There are similarities between this innovation and methods developed in the field of network coding. However, network coding has generally been concerned with either maximizing throughput in a fixed network, or robust communication of a fixed volume of data. In contrast, under this model, the throughput is expected to vary depending on the state of the network. Examples of error-correcting codes that are useful under this model but which are not needed under previous models have been found. This model can represent either a one-shot communication attempt, or a stream of communications. Under the one-shot model, message sizes and link capacities are quantities of information (e.g., measured in bits), while under the communications stream model, message sizes and link capacities are information rates (e.g., measured in bits/second). This work has the potential to increase the value of data returned from spacecraft under certain conditions.

Klimesh, Matthew A.↗

Accelerated application development: The ORNL Titan experience

The use of computational accelerators such as NVIDIA GPUs and Intel Xeon Phi processors is now widespread in the high performance computing community, with many applications delivering impressive performance gains. However, programming these systems for high performance, performance portability and software maintainability has been a challenge. In this paper we discuss experiences porting applications to the Titan system. Titan, which began planning in 2009 and was deployed for general use in 2013, was the first multi-petaflop system based on accelerator hardware. To ready applications for accelerated computing, a preparedness effort was undertaken prior to delivery of Titan. In this paper we report experiences and lessons learned from this process and describe how users are currently making use of computational accelerators on Titan.

97 MATHEMATICS AND COMPUTING↗

Thermal Conductivity of SiC/Si Composites – Porting PNNL EMTA Code for Fusion Analyses

An existing modeling method, the EMTA (Eshelby-Mori-Tanaka approach) modeling approach [1], is applied to the study of SiC/SiC 2D woven composites for fusion reactor applications for the first time, to the best of our knowledge, with excellent results. We compare EMTA model results to existing thermal conductivity data for these materials and suggest that in the future this approach can be beneficial by providing us with tools to further optimize these composite materials for fusion energy applications since the EMTA method and code can address both thermal and mechanical properties with the same framework.

Henager, Charles H.↗

Characterization of sub-picosecond laser-produced fast electrons by modeling angularly resolved bremsstrahlung measurements with 3D hybrid particle-in-cell code

Hard x-rays produced by intense laser-produced fast electrons interacting with solids are a vital source for producing radiographs of high-density objects and implosion cores for inertial confinement fusion. Accurate calculation of hard x-ray sources requires a three-dimensional (3D) simulation geometry that fully models the electron transport dynamics, including electron recirculation and the generation of absolute photon yields. To date, 3D simulations of laser-produced bremsstrahlung photons over tens of picoseconds and code benchmarking have not been performed definitively. In this study, we characterize sub-picosecond laser-produced fast electrons by modeling angularly resolved bremsstrahlung measurements for refluxing and non-refluxing targets using the 3D hybrid particle-in-cell (PIC), Large Scale Plasma code. Bremsstrahlung radiation and escaped electron data were obtained by focusing a 50-TW Leopard laser (15 J, 0.35 ps, 2 × 10 19 W/cm 2 ) on a 100-μm-thick Cu foil and a Cu with a large plastic backing (Cu–CH target). Data for both the Cu and Cu–CH targets were reproduced for simulations with a given set of electron parameters. Comparison of the simulations revealed that the hard x-ray emission from the Cu target was significantly longer in duration than that from the Cu–CH target. Here, the benchmarked hybrid PIC code could prove to be a powerful tool in the design and optimization of time- and angular-dependent bremsstrahlung sources for flash x-ray and gamma-ray radiography.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Fragment approach to constrained density functional theory calculations using Daubechies wavelets

In a recent paper, we presented a linear scaling Kohn-Sham density functional theory (DFT) code based on Daubechies wavelets, where a minimal set of localized support functions are optimized in situ and therefore adapted to the chemical properties of the molecular system. Thanks to the systematically controllable accuracy of the underlying basis set, this approach is able to provide an optimal contracted basis for a given system: accuracies for ground state energies and atomic forces are of the same quality as an uncontracted, cubic scaling approach. This basis set offers, by construction, a natural subset where the density matrix of the system can be projected. In this paper, we demonstrate the flexibility of this minimal basis formalism in providing a basis set that can be reused as-is, i.e., without reoptimization, for charge-constrained DFT calculations within a fragment approach. Support functions, represented in the underlying wavelet grid, of the template fragments are roto-translated with high numerical precision to the required positions and used as projectors for the charge weight function. We demonstrate the interest of this approach to express highly precise and efficient calculations for preparing diabatic states and for the computational setup of systems in complex environments.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Device for Detection of Explosives, Nuclear and Other Hazardous Materials in Luggage and Cargo Containers

Device for detection of explosives, radioactive and heavily shielded nuclear materials in luggage and cargo containers based on Nanosecond Neutron Analysis/Associated Particles Technique (NNA/APT) is under construction. Detection module consists of a small neutron generator with built-in position-sensitive detector of associated alpha-particles, and several scintillator-based gamma-ray detectors. Explosives and other hazardous chemicals are detected by analyzing secondary high-energy gamma-rays from reactions of fast neutrons with materials inside a container. The same gamma-ray detectors are used to detect unshielded radioactive and nuclear materials. An array of several neutron detectors is used to detect fast neutrons from induced fission of nuclear materials. Coincidence and timing analysis allows one to discriminate between fission neutrons and scattered probing neutrons. Mathematical modeling by MCNP5 and MCNP-PoliMi codes was used to estimate the sensitivity of the device and its optimal configuration. Comparison of the features of three gamma detector types--based on BGO, NaI and LaBr{sub 3} crystals is presented.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

An optimum PN code search technique for a given apriori signal location density

A method is presented for optimizing (minimizing) the acquisition time for a PN-type spread spectrum system when the a priori probability density function is not uniform. For an assumed a priori Gaussian density function, calculations show that the acquisition time, when the 0.5 probability acquisition time is used as a measure of acquisition time, is reduced by 40% for a cell detection probability of 0.25 and when three sweeps are used. For the same parameters and with four sweeps the acquisition time is reduced by 41%. When the acquisition time probability is set to 0.9 instead of 0.5, the reduction is 25% of the uniform sweep time.

Holmes, J. K.↗

Scalable molecular dynamics on CPU and GPU architectures with NAMD

NAMD is a molecular dynamics program designed for high-performance simulations of very large biological objects on CPU- and GPU-based architectures. NAMD offers scalable performance on petascale parallel supercomputers consisting of hundreds of thousands of cores, as well as on inexpensive commodity clusters commonly found in academic environments. It is written in C++ and leans on Charm++ parallel objects for optimal performance on low-latency architectures. NAMD is a versatile, multipurpose code that gathers state-of-the-art algorithms to carry out simulations in apt thermodynamic ensembles, using the widely popular CHARMM, AMBER, OPLS, and GROMOS biomolecular force fields. Here, we review the main features of NAMD that allow both equilibrium and enhanced-sampling molecular dynamics simulations with numerical efficiency. We describe the underlying concepts utilized by NAMD and their implementation, most notably for handling long-range electrostatics; controlling the temperature, pressure, and pH; applying external potentials on tailored grids; leveraging massively parallel resources in multiple-copy simulations; and hybrid quantum-mechanical/molecular-mechanical descriptions. We detail the variety of options offered by NAMD for enhanced-sampling simulations aimed at determining free-energy differences of either alchemical or geometrical transformations and outline their applicability to specific problems. Last, we discuss the roadmap for the development of NAMD and our current efforts toward achieving optimal performance on GPU-based architectures, for pushing back the limitations that have prevented biologically realistic billion-atom objects to be fruitfully simulated, and for making large-scale simulations less expensive and easier to set up, run, and analyze. NAMD is distributed free of charge with its source code at www.ks.uiuc.edu.

high-performance computing↗

JET ITER-Like Antenna Analysis Using the TOPICA Code

In this work, we carried out the analysis of the recently installed JET ITER-Like antenna with TOPICA code. Comparisons between TOPICA simulations and measurements taken during the actual experiment are instead part of the overview talk on the JET ITER-Like antenna [4]. As routinely done for all simulated antennas, TOPICA inputs are the technical drawings of the launcher and the accurate density and temperature profiles, which, in this case, have been provided by the JET team. The standard outputs are the input parameters of the antenna, namely the impedance matrix, the electric current distribution and the electric field pattern at the interface between the antenna region and the plasma column. This work, together with [4], provides an additional proof that the code can be adopted to predict the behavior of the JET ITER-Like antenna, and to confidently pass to the next step, i.e. using TOPICA for the challenging task of optimizing the complex design of the actual ITER antenna. More generally viewed, the possibility to reliably simulate the detailed geometry of an ICRF antenna, given a realistic plasma description, and to obtain the actual antenna input parameters, is of paramount importance to evaluate and predict the system performances, and to assist in system operation.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗