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 181 records · Page 10
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

A High Performance Sparse Tensor Algebra Compiler in MLIR

Sparse tensor algebra is widely used in many applications, including scientific computing, machine learning, and data analytics. The performance of sparse tensor algebra kernels strongly depends on the intrinsic characteristics of the input tensors, hence many storage formats are designed for tensors to achieve optimal performance for particular applications/architectures, which makes it challenging to implement and optimize every tensor operation of interest on a given architecture. We propose a tensor algebra domain-specific language (DSL) and compiler framework to automatically generate kernels for mixed sparse-dense tensor algebra operations. The proposed DSL provides high-level programming abstractions that resemble the familiar Einstein notation to represent tensor algebra operations. The compiler introduces a new Sparse Tensor Algebra dialect built on top of LLVM's extensible MLIR compiler infrastructure for efficient code generation while covering a wide range of tensor storage formats. Our compiler also leverages input-dependent code optimization to enhance data locality for better performance. Our results show that the performance of automatically generated kernels outperforms the state-of-the-art sparse tensor algebra compiler, with up to 20.92x, 6.39x, and 13.9x performance improvement over state-of-the-art tensor algebra compilers, for parallel SpMV, SpMM, and TTM, respectively.

Tian, Ruiqin↗
DOE OSTI2021

Geometry optimization speedup through a geodesic approach to internal coordinates

We present a new geodesic-based method for geometry optimization in a basis set of redundant internal coordinates. Overall, our method updates the molecular geometry by following the geodesic generated by a displacement vector on the internal coordinate manifold, which dramatically reduces the number of steps required to converge to a minimum. Our method can be implemented in any existing optimization code, requiring only implementation of derivatives of the Wilson B-matrix and the ability to numerically solve an ordinary differential equation.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

A regression algorithm for accelerated lattice QCD that exploits sparse inference on the D-Wave quantum annealer

We propose a regression algorithm that utilizes a learned dictionary optimized for sparse inference on a D-Wave quantum annealer. In this regression algorithm, we concatenate the independent and dependent variables as a combined vector, and encode the high-order correlations between them into a dictionary optimized for sparse reconstruction. On a test dataset, the dependent variable is initialized to its average value and then a sparse reconstruction of the combined vector is obtained in which the dependent variable is typically shifted closer to its true value, as in a standard inpainting or denoising task. Here, a quantum annealer, which can presumably exploit a fully entangled initial state to better explore the complex energy landscape, is used to solve the highly non-convex sparse coding optimization problem. The regression algorithm is demonstrated for a lattice quantum chromodynamics simulation data using a D-Wave 2000Q quantum annealer and good prediction performance is achieved. The regression test is performed using six different values for the number of fully connected logical qubits, between 20 and 64. The scaling results indicate that a larger number of qubits gives better prediction accuracy.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1999

Optimization of Aerospace Structure Subject to Damage Tolerance Criteria

The objective of this cooperative agreement was to seek computationally efficient ways to optimize aerospace structures subject to damage tolerance criteria. Optimization was to involve sizing as well as topology optimization. The work was done in collaboration with Steve Scotti, Chauncey Wu and Joanne Walsh at the NASA Langley Research Center. Computation of constraint sensitivity is normally the most time-consuming step of an optimization procedure. The cooperative work first focused on this issue and implemented the adjoint method of sensitivity computation in an optimization code (runstream) written in Engineering Analysis Language (EAL). The method was implemented both for bar and plate elements including buckling sensitivity for the latter. Lumping of constraints was investigated as a means to reduce the computational cost. Adjoint sensitivity computation was developed and implemented for lumped stress and buckling constraints. Cost of the direct method and the adjoint method was compared for various structures with and without lumping. The results were reported in two papers. It is desirable to optimize topology of an aerospace structure subject to a large number of damage scenarios so that a damage tolerant structure is obtained. Including damage scenarios in the design procedure is critical in order to avoid large mass penalties at later stages. A common method for topology optimization is that of compliance minimization which has not been used for damage tolerant design. In the present work, topology optimization is treated as a conventional problem aiming to minimize the weight subject to stress constraints. Multiple damage configurations (scenarios) are considered. Each configuration has its own structural stiffness matrix and, normally, requires factoring of the matrix and solution of the system of equations. Damage that is expected to be tolerated is local and represents a small change in the stiffness matrix compared to the baseline (undamaged) structure. The exact solution to a slightly modified set of equations can be obtained from the baseline solution economically without actually solving the modified system. Sherrnan-Morrison-Woodbury (SMW) formulas are matrix update formulas that allow this. SMW formulas were therefore used here to compute adjoint displacements for sensitivity computation and structural displacements in damaged configurations.

Akgun, Mehmet A.↗
NASA NTRS2000

Adjoint Techniques for Topology Optimization of Structures Under Damage Conditions

The objective of this cooperative agreement was to seek computationally efficient ways to optimize aerospace structures subject to damage tolerance criteria. Optimization was to involve sizing as well as topology optimization. The work was done in collaboration with Steve Scotti, Chauncey Wu and Joanne Walsh at the NASA Langley Research Center. Computation of constraint sensitivity is normally the most time-consuming step of an optimization procedure. The cooperative work first focused on this issue and implemented the adjoint method of sensitivity computation (Haftka and Gurdal, 1992) in an optimization code (runstream) written in Engineering Analysis Language (EAL). The method was implemented both for bar and plate elements including buckling sensitivity for the latter. Lumping of constraints was investigated as a means to reduce the computational cost. Adjoint sensitivity computation was developed and implemented for lumped stress and buckling constraints. Cost of the direct method and the adjoint method was compared for various structures with and without lumping. The results were reported in two papers (Akgun et al., 1998a and 1999). It is desirable to optimize topology of an aerospace structure subject to a large number of damage scenarios so that a damage tolerant structure is obtained. Including damage scenarios in the design procedure is critical in order to avoid large mass penalties at later stages (Haftka et al., 1983). A common method for topology optimization is that of compliance minimization (Bendsoe, 1995) which has not been used for damage tolerant design. In the present work, topology optimization is treated as a conventional problem aiming to minimize the weight subject to stress constraints. Multiple damage configurations (scenarios) are considered. Each configuration has its own structural stiffness matrix and, normally, requires factoring of the matrix and solution of the system of equations. Damage that is expected to be tolerated is local and represents a small change in the stiffness matrix compared to the baseline (undamaged) structure. The exact solution to a slightly modified set of equations can be obtained from the baseline solution economically without actually solving the modified system.. Shennan-Morrison-Woodbury (SMW) formulas are matrix update formulas that allow this (Akgun et al., 1998b). SMW formulas were therefore used here to compute adjoint displacements for sensitivity computation and structural displacements in damaged configurations.

Akgun, Mehmet A.↗
DOE OSTI2024

Accelerated Constrained Sparse Tensor Factorization on Massively Parallel Architectures

This study presents the first constrained sparse tensor factorization (cSTF) framework that optimizes and fully offloads computation to massively parallel GPU architectures, and the first performance characterization of cSTF on GPU architectures. In contrast to prior work on tensor factorization, where the matricized tensor times Khatri-Rao product (MTTKRP) is the primary performance bottleneck, our systematic analysis of the cSTF algorithm on GPUs reveals that adding constraints creates an additional bottleneck in the update operation for many real-world sparse tensors. While executing the update operation on the GPU brings significant speedup over its CPU counterpart, it remains a significant bottleneck. To further accelerate the update operation, we propose cuADMM, a new update algorithm that leverages algorithmic and code optimization strategies to minimize both computation and data movement on GPUs. As a result, our framework delivers significantly improved performance compared to prior state-of-the-art. On 10 real-world sparse tensors, our framework achieves geometric mean speedup of 5.1 × (max 41.59 ×) and 7.01 × (max 58.05 ×) on the NIVIDA A100 and H100 GPUs, respectively, over the state-of-the-art SPLATT library running on a 26-core Intel Ice Lake Xeon CPU.

Soh, Yongseok↗
DOE OSTI2019

Hierarchical Roofline analysis for GPUs: Accelerating performance optimization for the NERSC‐9 Perlmutter system

Summary The Roofline performance model provides an intuitive and insightful approach to identifying performance bottlenecks and guiding performance optimization. In preparation for the next‐generation supercomputer Perlmutter at NERSC, this paper presents a methodology to construct a hierarchical Roofline on NVIDIA GPUs and extends it to support reduced precision and Tensor Cores. The hierarchical Roofline incorporates L1, L2, device memory, and system memory bandwidths into one single figure, and it offers more profound insights into performance analysis than the traditional DRAM‐only Roofline. We use our Roofline methodology to analyze three proxy applications: GPP from BerkeleyGW, HPGMG from AMReX, and conv2d from TensorFlow. In doing so, we demonstrate the ability of our methodology to readily understand various aspects of performance and performance bottlenecks on NVIDIA GPUs and motivate code optimizations.

Yang, Charlene↗
DOE OSTI2024

Integrating quantum computing resources into scientific HPC ecosystems

Quantum Computing (QC) offers significant potential to enhance scientific discovery in fields such as quantum chemistry, optimization, and artificial intelligence. Yet QC faces challenges due to the noisy intermediate-scale quantum era’s inherent external noise issues. Here, this paper discusses the integration of QC as a computational accelerator within classical scientific high-performance computing (HPC) systems. By leveraging a broad spectrum of simulators and hardware technologies, we propose a hardware-agnostic framework for augmenting classical HPC with QC capabilities. Drawing on the HPC expertise of the Oak Ridge National Laboratory (ORNL) and the HPC lifecycle management of the Department of Energy (DOE), our approach focuses on the strategic incorporation of QC capabilities and acceleration into existing scientific HPC workflows. This includes detailed analyses, benchmarks, and code optimization driven by the needs of the DOE and ORNL missions. Our comprehensive framework integrates hardware, software, workflows, and user interfaces to foster a synergistic environment for quantum and classical computing research. This paper outlines plans to unlock new computational possibilities, driving forward scientific inquiry and innovation in a wide array of research domains.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
NASA NTRS2003

Megawatt Electromagnetic Plasma Propulsion

The NASA Glenn Research Center program in megawatt level electric propulsion is centered on electromagnetic acceleration of quasi-neutral plasmas. Specific concepts currently being examined are the Magnetoplasmadynamic (MPD) thruster and the Pulsed Inductive Thruster (PIT). In the case of the MPD thruster, a multifaceted approach of experiments, computational modeling, and systems-level models of self field MPD thrusters is underway. The MPD thruster experimental research consists of a 1-10 MWe, 2 ms pulse-forming-network, a vacuum chamber with two 32 diffusion pumps, and voltage, current, mass flow rate, and thrust stand diagnostics. Current focus is on obtaining repeatable thrust measurements of a Princeton Benchmark type self field thruster operating at 0.5-1 gls of argon. Operation with hydrogen is the ultimate goal to realize the increased efficiency anticipated using the lighter gas. Computational modeling is done using the MACH2 MHD code, which can include real gas effects for propellants of interest to MPD operation. The MACH2 code has been benchmarked against other MPD thruster data, and has been used to create a point design for a 3000 second specific impulse (Isp) MPD thruster. This design is awaiting testing in the experimental facility. For the PIT, a computational investigation using MACH2 has been initiated, with experiments awaiting further funding. Although the calculated results have been found to be sensitive to the initial ionization assumptions, recent results have agreed well with experimental data. Finally, a systems level self-field MPD thruster model has been developed that allows for a mission planner or system designer to input Isp and power level into the model equations and obtain values for efficiency, mass flow rate, and input current and voltage. This model emphasizes algebraic simplicity to allow its incorporation into larger trajectory or system optimization codes. The systems level approach will be extended to the pulsed inductive thruster and other electrodeless thrusters at a future date.

Gilland, James↗
DOE OSTI2021

Instruction Roofline: An insightful visual performance model for GPUs

The Roofline performance model provides an intuitive approach to identify performance bottlenecks and guide performance optimization. However, the classic FLOP-centric approach is inappropriate for the emerging applications that perform more integer operations than floating point operations. In this article, we reintroduce our Instruction Roofline Model on NVIDIA GPUs and expand our evaluation of it. The Instruction Roofline incorporates instructions and memory transactions across all memory hierarchies together, and provides more performance insights than the FLOP-oriented Roofline Model, that is, instruction throughput, stride memory access patterns, bank conflicts, and thread predication. We use our Instruction Roofline methodology to analyze eight proxy applications: HPGMG from AMReX, Matrix Transpose benchmarks, ADEPT from MetaHipMer's sequence alignment phase, EXTENSION from MetaHipMer's local assembly phase, CUSP, cuSPARSE, cudaTensorCoreGemm, and cuBLAS. We demonstrate the ability of our methodology to understand various aspects of performance and performance bottlenecks on NVIDIA GPUs and motivate code optimizations.

Ding, N↗
DOE OSTI2017

Nonlinear Force-free Coronal Magnetic Stereoscopy

Insights into the 3D structure of the solar coronal magnetic field have been obtained in the past by two completely different approaches. The first approach are nonlinear force-free field (NLFFF) extrapolations, which use photospheric vector magnetograms as boundary condition. The second approach uses stereoscopy of coronal magnetic loops observed in EUV coronal images from different vantage points. Both approaches have their strengths and weaknesses. Extrapolation methods are sensitive to noise and inconsistencies in the boundary data, and the accuracy of stereoscopy is affected by the ability of identifying the same structure in different images and by the separation angle between the view directions. As a consequence, for the same observational data, the 3D coronal magnetic fields computed with the two methods do not necessarily coincide. In an earlier work (Paper I) we extended our NLFFF optimization code by including stereoscopic constrains. The method was successfully tested with synthetic data, and within this work, we apply the newly developed code to a combined data set from SDO /HMI, SDO /AIA, and the two STEREO spacecraft. The extended method (called S-NLFFF) contains an additional term that monitors and minimizes the angle between the local magnetic field direction and the orientation of the 3D coronal loops reconstructed by stereoscopy. We find that when we prescribe the shape of the 3D stereoscopically reconstructed loops, the S-NLFFF method leads to a much better agreement between the modeled field and the stereoscopically reconstructed loops. We also find an appreciable decrease by a factor of two in the angle between the current and the magnetic field. This indicates the improved quality of the force-free solution obtained by S-NLFFF.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2012

MODELING MAGNETIC FIELD STRUCTURE OF A SOLAR ACTIVE REGION CORONA USING NONLINEAR FORCE-FREE FIELDS IN SPHERICAL GEOMETRY

We test a nonlinear force-free field (NLFFF) optimization code in spherical geometry using an analytical solution from Low and Lou. Several tests are run, ranging from idealized cases where exact vector field data are provided on all boundaries, to cases where noisy vector data are provided on only the lower boundary (approximating the solar problem). Analytical tests also show that the NLFFF code in the spherical geometry performs better than that in the Cartesian one when the field of view of the bottom boundary is large, say, 20 Degree-Sign Multiplication-Sign 20 Degree-Sign . Additionally, we apply the NLFFF model to an active region observed by the Helioseismic and Magnetic Imager on board the Solar Dynamics Observatory (SDO) both before and after an M8.7 flare. For each observation time, we initialize the models using potential field source surface (PFSS) extrapolations based on either a synoptic chart or a flux-dispersal model, and compare the resulting NLFFF models. The results show that NLFFF extrapolations using the flux-dispersal model as the boundary condition have slightly lower, therefore better, force-free, and divergence-free metrics, and contain larger free magnetic energy. By comparing the extrapolated magnetic field lines with the extreme ultraviolet (EUV) observations by the Atmospheric Imaging Assembly on board SDO, we find that the NLFFF performs better than the PFSS not only for the core field of the flare productive region, but also for large EUV loops higher than 50 Mm.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

Helium: lifting high-performance stencil kernels from stripped x86 binaries to halide DSL code

Highly optimized programs are prone to bit rot, where performance quickly becomes suboptimal in the face of new hardware and compiler techniques. In this paper we show how to automatically lift performance-critical stencil kernels from a stripped x86 binary and generate the corresponding code in the high-level domain-specific language Halide. Using Halide's state-of-the-art optimizations targeting current hardware, we show that new optimized versions of these kernels can replace the originals to rejuvenate the application for newer hardware. The original optimized code for kernels in stripped binaries is nearly impossible to analyze statically. Instead, we rely on dynamic traces to regenerate the kernels. We perform buffer structure reconstruction to identify input, intermediate and output buffer shapes. Here, we abstract from a forest of concrete dependency trees which contain absolute memory addresses to symbolic trees suitable for high-level code generation. This is done by canonicalizing trees, clustering them based on structure, inferring higher-dimensional buffer accesses and finally by solving a set of linear equations based on buffer accesses to lift them up to simple, high-level expressions. Helium can handle highly optimized, complex stencil kernels with input-dependent conditionals. We lift seven kernels from Adobe Photoshop giving a 75 % performance improvement, four kernels from Irfan View, leading to 4.97 x performance, and one stencil from the mini GMG multigrid benchmark netting a 4.25 x improvement in performance. We manually rejuvenated Photoshop by replacing eleven of Photoshop's filters with our lifted implementations, giving 1.12 x speedup without affecting the user experience.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

An efficient, partitioned ensemble algorithm for simulating ensembles of evolutionary MHD flows at low magnetic Reynolds number

Studying the propagation of uncertainties in a nonlinear dynamical system usually involves generating a set of samples in the stochastic parameter space and then repeated simulations with different sampled parameters. The main difficulty faced in the process is the excessive computational cost. In this paper, we present an efficient, partitioned ensemble algorithm to determine multiple realizations of a reduced Magnetohydrodynamics (MHD) system, which models MHD flows at low magnetic Reynolds number. The algorithm decouples the fully coupled problem into two smaller subphysics problems, which reduces the size of the linear systems that to be solved and allows the use of optimized codes for each subphysics problem. Moreover, the resulting coefficient matrices are the same for all realizations at each time step, which allows faster computation of all realizations and significant savings in computational cost. We prove this algorithm is first order accurate and long time stable under a time step condition. Numerical examples are provided to verify the theoretical results and demonstrate the efficiency of the algorithm.

Jiang, Nan↗
DOE OSTI2017

Experiences Evaluating Functionality and Performance of IBM POWER8+ Systems

In preparation for Summit, Oak Ridge National Laboratory’s next generation supercomputer, two IBM Power-based systems were deployed in late 2016 at the Oak Ridge Leadership Computing Facility (OLCF). This paper presents a detailed description of the acceptance of the first IBM Power-based early access systems installed at the OLCF. The two systems, Summitdev and Tundra, contain IBM POWER8+ processors with NVIDIA Pascal GPUs and were acquired to provide researchers with a platform to optimize codes for the Power architecture. In addition, this paper presents early functional and performance results obtained on Summitdev with the latest software stack available.

Vergara Larrea, Veronica G.↗
DOE OSTI2019

Finite element analyses of single particle crushing tests incorporating computed tomography imaging and damage mechanics

Fracture of individual grains within a granular assembly affect the strength and deformation behavior of granular materials under large stress states. Interparticle forces, leading to contact stresses and ultimately fracture initiation, are influenced by particle morphology. A numerical method that incorporates both particle fracture and morphology can provide a more accurate model of a granular material’s macro-scale response. This research investigates an approach to modeling fracture initiation and propagation within individual grains, utilizing high resolution X-ray computed tomography (CT) imaging to consider grain morphology and an explicit finite element code optimized for high performance computing which incorporates damage mechanics. Using precisely measured force-displacement curves from single particle crushing tests of manufactured quartz spheres, the method is validated, and the effects of fracture properties on grain fragmentation are determined. This approach is then used to simulate single particle crushing of Ottawa sand, with grain morphology incorporated through discretized CT images. Here the effect of grain orientation on the force and applied displacement at catastrophic splitting is also explored. The proposed discrete particle based finite element framework can be applied to modeling granular assemblies, considering the effect of individual particle shape and fracture on the assembly’s deformation response through well calibrated numerical simulations.

42 ENGINEERING↗
DOE OSTI2016

Direction Modulated Brachytherapy for Treatment of Cervical Cancer. II: Comparative Planning Study With Intracavitary and Intracavitary–Interstitial Techniques

Purpose: To perform a comprehensive comparative planning study evaluating the utility of the proposed direction modulated brachytherapy (DMBT) tandem applicator against standard applicators, in the setting of image guided adaptive brachytherapy of cervical cancer. Methods and Materials: A detailed conceptual article was published in 2014. The proposed DMBT tandem applicator has 6 peripheral grooves of 1.3-mm width, along a 5.4-mm-thick nonmagnetic tungsten alloy rod of density 18.0 g/cm{sup 3}, capable of generating directional dose profiles. We performed a comparative planning study with 45 cervical cancer patients enrolled consecutively in the prospective observational EMBRACE study. In all patients, MRI-based planning was performed while utilizing various tandem-ring (27 patients) and tandem-ring-needles (18 patients) applicators, in accordance with the Groupe Européen de Curiethérapie–European Society for Radiotherapy and Oncology recommendations. For unbiased comparisons, all cases were replanned with an in-house–developed inverse optimization code while enforcing a uniform set of constraints that are reflective of the clinical practice. All plans were normalized to the same high-risk clinical target volume D90 values achieved in the original clinical plans. Results: In general, if the standard tandem was replaced with the DMBT tandem while maintaining all other planning conditions the same, there was consistent improvement in the plan quality. For example, among the 18 tandem-ring-needles cases, the average D2cm{sup 3} reductions achieved were −2.48% ± 11.03%, −4.45% ± 5.24%, and −5.66% ± 6.43% for the bladder, rectum, and sigmoid, respectively. An opportunity may also exist in avoiding use of needles altogether for when the total number of needles required is small (approximately 2 to 3 needles or less), if DMBT tandem is used. Conclusions: Integrating the novel DMBT tandem onto both intracavitary and intracavitary–interstitial applicator assembly enabled consistent improvement in the sparing of the OARs, over a standard “single-channel” tandem, though individual variations in benefit were considerable. Although at an early stage of development, the DMBT concept design is demonstrated to be useful and pragmatic for potential clinical translation.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2019

Parallel accelerated Custom Correlation Coefficient calculations for genomics applications

The massive quantities of genomic data being made available through gene sequencing techniques are enabling breakthroughs in genomic science in many areas such as medical advances in the diagnosis and treatment of diseases. Analyzing this data, however, is a computational challenge insofar as the computational costs of the relevant algorithms can grow with quadratic, cubic or higher complexity—leading to the need for leadership scale computing. In this paper we describe a new approach to calculations of the Custom Correlation Coefficient (CCC) between Single Nucleotide Polymorphisms (SNPs) across a population, suitable for parallel systems equipped with graphics processing units (GPUs) or Intel Xeon Phi processors. We describe the mapping of the algorithms to accelerated processors, techniques used for eliminating redundant calculations due to symmetries, and strategies for efficient mapping of the calculations to many-node parallel systems. Results are presented demonstrating high per-node performance and near-ideal parallel scalability with rates of more than nine quadrillion (9 × 10 15 ) elementwise comparisons achieved per second with the latest optimized code on the ORNL Titan system, this being orders of magnitude faster than rates achieved using other codes and platforms as reported in the literature. Also it is estimated that as many as 90 quadrillion (90 × 10 15 ) comparisons per second may be achievable on the upcoming ORNL Summit system, an additional 10X performance increase. Finally, in a companion paper we describe corresponding techniques applied to calculations of the Proportional Similarity metric for comparative genomics applications.

97 MATHEMATICS AND COMPUTING↗