Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “mathematical libraries”

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 721 records · Page 40
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Do Molecular Fingerprints Identify Diverse Active Drugs in Large-Scale Virtual Screening? (No)

Computational approaches for small-molecule drug discovery now regularly scale to the consideration of libraries containing billions of candidate small molecules. One promising approach to increased the speed of evaluating billion-molecule libraries is to develop succinct representations of each molecule that enable the rapid identification of molecules with similar properties. Molecular fingerprints are thought to provide a mechanism for producing such representations. Here, we explore the utility of commonly used fingerprints in the context of predicting similar molecular activity. We show that fingerprint similarity provides little discriminative power between active and inactive molecules for a target protein based on a known active—while they may sometimes provide some enrichment for active molecules in a drug screen, a screened data set will still be dominated by inactive molecules. We also demonstrate that high-similarity actives appear to share a scaffold with the query active, meaning that they could more easily be identified by structural enumeration. Furthermore, even when limited to only active molecules, fingerprint similarity values do not correlate with compound potency. In sum, these results highlight the need for a new wave of molecular representations that will improve the capacity to detect biologically active molecules based on their similarity to other such molecules.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

DLSIA: Deep Learning for Scientific Image Analysis

DLSIA (Deep Learning for Scientific Image Analysis) is a Python-based machine learning library that empowers scientists and researchers across diverse scientific domains with a range of customizable convolutional neural network (CNN) architectures for a wide variety of tasks in image analysis to be used in downstream data processing. DLSIA features easy-to-use architectures, such as autoencoders, tunable U-Nets and parameter-lean mixed-scale dense networks (MSDNets). Additionally, this article introduces sparse mixed-scale networks (SMSNets), generated using random graphs, sparse connections and dilated convolutions connecting different length scales. For verification, several DLSIA-instantiated networks and training scripts are employed in multiple applications, including inpainting for X-ray scattering data using U-Nets and MSDNets, segmenting 3D fibers in X-ray tomographic reconstructions of concrete using an ensemble of SMSNets, and leveraging autoencoder latent spaces for data compression and clustering. As experimental data continue to grow in scale and complexity, DLSIA provides accessible CNN construction and abstracts CNN complexities, allowing scientists to tailor their machine learning approaches, accelerate discoveries, foster interdisciplinary collaboration and advance research in scientific image analysis.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Evaluating Retention Index Score Assumptions to Refine GC–MS Metabolite Identification

As metabolomics grows into a high-throughput and high demand research field, current metrics for the identification of small molecules in gas chromatography–mass spectrometry (GC–MS) still require manual verification. Though steps have been taken to improve scoring metrics by combining spectral similarity (SS) and retention index (RI), the problem persists. A large body of literature has analyzed and refined SS scores, but few studies have explicitly studied improvements to RI scores. Here, in this study, we examined whether uninvestigated assumptions of the RI score are valid and propose ways to improve them. Query RIs were matched to library RI with a generous window of ±35 to avoid unintentional removal of valid compound identifications. Each match was manually verified as a true positive (TP), true negative, or unknown. Metabolites with at least 30 TP identifications were included in downstream analyses, resulting in a total of 87 metabolites from samples of varying complexity and type (e.g., amino acid mixtures, human urine, fungal species, and so on.). Our results showed that the RI score assumptions of normality, consistent variance across metabolites, and a mean error centered at 0 are often violated. We demonstrated through a cross-validation analysis that modifying these underlying assumptions according to empirical metabolite-specific distributions improved the TP and negative rankings. Further, we statistically determined the minimum number of samples required to estimate distributional parameters for scoring metrics. Overall, this work proposes a robust statistical pipeline to reduce the time bottleneck of metabolite identification by improving RI scores and thus minimize the effort to complete manual verification.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Progress on Demonstration of a MOOSE-Based Coupled Capability for Hot Channel Factors in Fast Reactors

Hot channel factors (HCFs) are computed values that account for the impact on predicted peak fuel, cladding, and coolant temperatures due to uncertainties in the as-built reactor’s material properties and geometry as well as uncertainties due to modeling approximations. Reduction in computed HCF values via reduction or elimination of modeling approximations may translate to significant economic savings if the reactor power can be raised due to the extra temperature margin gained. While limited historical datasets exist for sodium-cooled fast reactors (SFRs), there are no available HCF data for lead-cooled fast reactors (LFRs) outside of work generated previously within NEAMS. The computation of HCFs involves insights from reactor physics, thermal fluids and heat conduction calculations to determine how the peak temperatures respond to various uncertainties in the design. Due to the significant advantages for multi-physics coupling offered by the MOOSE framework, Griffin (MOOSE-based reactor physics code), MOOSE Heat Conduction Module, and Cardinal (MOOSE-wrapped multi-physics application which includes the NekRS thermal fluids code) are being coupled together using the MOOSE MultiApp System to develop a highfidelity multi-physics modeling capability for HCF simulations. This high-fidelity coupling workflow may also be beneficial for other fast reactor applications in the future. In previous work, Griffin and NekRS were individually assessed to ensure the necessary capabilities were in place. This work describes initial efforts to couple the codes (including folding in the MOOSE Heat Conduction Module) and determining the workflow for the perturbed calculations which will leverage the Stochastic Tools Module (STM). To our knowledge, this is the first coupling of Griffin and NekRS as well as the first exploratory use of Stochastic Tools Module for Cardinal. In this report, the neutronics code Griffin, the heat conduction solver in MOOSE, and the MOOSE-wrapped application containing NekRS (Cardinal) are linked together to demonstrate the coupled capability. Griffin and Cardinal are linked dynamically by specifying shared libraries. Different coupling hierarchies are tested for selecting the most appropriate coupling strategy. A coupling scheme is selected based on the efficiency of calculation and ease of data communication. Multiple tests are performed to choose suitable mesh structure, model configurations, scheme setup and boundary conditions to avoid loss of energy due to data interpolation between different modules or weak imposition of fluxes in finite element codes. Computational experiments are performed to study the tolerance control of each type of iteration to avoid false convergence. The coupled capability is demonstrated in both single pin and 7-pin models based on LFR materials and geometry. The study finds that the use of too large a time step size in the heat conduction module can lead to temperature oscillation even though the heat conduction equation does not have a time-derivative kernel, but only the time-dependent boundary condition. A 7-pin model without duct region achieved good convergence in the coupled calculation while a 7 pin model with duct region experienced data communication issues which need to be resolved.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2021

Application of Machine Learning Algorithms to Identify Problematic Nuclear Data

In this work we aim to show that Machine learning algorithms are promising tools for the identification of nuclear data that contribute to increased errors in transport simulations. We demonstrate this through an application of a machine learning algorithm (Random Forest) to the Whisper/MCNP6 criticality validation library to identify nuclear data that are associated with an increase of the bias (simulated - experimental $k_{eff}$) in the calculations. Specifically, the $k_{eff}$ sensitivity profiles (w.r.t. nuclear data) of 233 U solution benchmarks are used to predict the bias and Shapley Additive Explanations (SHAP) are used to explain how the sensitivities are related to the predicted bias. The SHAP values can be interpreted as sensitivity coefficients of the machine learning model to the $k_{eff}$ sensitivities which are used to make predictions of bias. Using the SHAP values we can identify specific subsets of nuclear data which have the highest probability of influencing bias. We demonstrate the utility of this method by showing how SHAP values were used to identify an inconsistency in the 19 F inelastic scattering nuclear data. The methodology presented here is not limited to transport problems and can be applied to other simulations if there are experimental measurements to compare against, simulations of those experimental measurements, and the ability to calculate sensitivities of the model output with respect to the data inputs.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2025

hashin_shtrikman_mp: a package for the optimal design and discovery of multi-phase composite materials

hashin_shtrikman_mp is a tool for composites designers who have desired composite properties in mind, but who do not yet have an underlying formulation. The library utilizes the tightest theoretical bounds on the effective properties of composite materials with unspecified microstructure – the Hashin-Shtrikman bounds – to identify candidate theoretical materials, find real materials that are close to the candidates, and determine the optimal volume fractions for each of the constituents in the resulting composite. Its features include (i) leveraging of materials in the Materials Project database, (ii) integration with the Materials Project API, (iii) use of genetic machine-learning, (iv) agnosticism to underlying microstructure, and (v) ultimate engineering application, make it a tool with much broader applications than its predecessors.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Spiner vs EOSPAC6: capabilities, performance, and accuracy considerations

The Cross-Cutting Capabilities Project (XCAP) has significant interest in collecting and implementing a common set of libraries to be used across the various production hydro-codes at LANL. A common set of libraries will help facilitate comparisons between codes. In addition to minimizing variables for comparisons and physics validation, efforts toward optimization and porting of packages for future architectures will be more efficient. As part of this push Singularity is being considered as an inclusive materials interface library. One crucial part of this library is the importing, inverting, and interpolation of equation of state (EOS) data. Currently the Lagrangian Applications Project (LAP) and Safety Applications Project (SAP) are using EOSPAC6 for delivery EOS data. The Eulerian Applications Project (EAP) has historically used TEOS and more recently an implementation of Singularly via a package called Spiner for EOS interpolation. Singularity currently has the option of using several analytical EOS models, directly employing SESAME via Spiner, or directly using SESAME via EOSPAC6 (not optimized yet). Assuming Singularity moves forward as a common platform for the implementation of materials models, a decision will need to be made on which EOS interpolation package XCAP should move forward with, given resources and people are finite. We will attempt to address pros and cons of Spiner and EOSPAC6 and the trade-offs that should be considered during the decision making process. This report is intended as an ASC-PEM-EOS perspective on what is needed in an EOS interpolation package. Performance and Accuracy sections will mostly address data from comparison studies in LA-UR-22-22699.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2025

Toucan: A performance portable, scalable implementation of the DECA algorithm

In the field of additive manufacturing (AM), cellular automata (CA) is extensively used to simulate microstructural evolution during solidification. However, while traditional CA approaches are relatively fast, they still require a substantial number of time steps, are limited to moderate volumes, and are relatively difficult to improve through parallelism due to the highly localized nature of the solidification front. Here, to address these issues of time to solution and load balancing, we introduce Toucan, a parallel, performance-portable, and scalable code written in C++ with the Kokkos library that leverages the discrete event inspired cellular automata (DECA) algorithm to perform parallel-in-time (PinT) grain growth simulations. Toucan effectively mitigates load balancing issues by distributing the computational workload more evenly across processors, enhancing scalability and efficiency. We conduct both strong and weak scaling studies on up to 64 GPUs on the Frontier supercomputer, demonstrating that Toucan significantly outperforms the current state-of-the-art, time-stepped CA code, ExaCA, on both single and multi-GPU simulations. Even in AM-specific weak scaling scenarios, Toucan maintains near-ideal scaling, in contrast to the linear increase observed with ExaCA due to the moving laser raster pattern. This study highlights Toucan’s potential to transform microstructural simulations in AM by radically improving both efficiency and scalability over existing methods.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Computational Methods to Characterize Panel Loading Conditions for Accelerated Testing

Panel cracking and degradation due to wind loading are known to have a detrimental effect on power output. Recreating these damaging conditions in a controlled experimental setting requires an understanding of the loads being generated at the panel surface as a function of both wind speed, wind direction, and panel orientation. To better understand these relationships, a computational fluid dynamics (CFD) simulation package was constructed using an open-source Python library for solving partial differential equations with the finite element method. This CFD package allows the simulated wind speed and panel orientation of a multi-panel array to be easily changed and provides the capability to measure the traction forces at discrete points along the sun-facing and ground-facing surface of an interior panel residing in the wake of one or more upstream panels. A parameter exploration was performed in which the panel angle was varied in even increments from -70 deg to 70 deg and the wind speed was varied from 2 to 30 m/s. During post processing, the measured traction along the panel surface was averaged spatially and interpreted as a time-varying signal, where further processing of this signal yielded the root mean square amplitude and a characteristic frequency associated with the loading. This study found that higher wind speeds are generally associated with increased amplitude of loading and that the panel orientation angle can significantly exacerbate or mitigate this loading. These outcomes are presented along with current work on higher-fidelity verification simulations and recommendations for performing accelerated experimental testing.

loading↗
DOE OSTI2024

Differentiable Multiphysics Codes: A Breakthrough Technology for Simulation and Computing

This document summarizes the findings of a strategic planning exercise commissioned by the Weapons Simulation and Computing, Computational Physics (WSC/CP) program at the Lawrence Livermore National Laboratory (LLNL) in FY24. During the year, the committee met with multiple stakeholder communities to gather input, opinions, suggestions and concerns which have been incorporated throughout this document. The key findings from this exercise are summarized: • The development of multiphysics modelling and simulation (mod/sim) codes and software technologies, their deployment on exascale compute platforms, and their broad adoption across the NNSA is a major success of the Advanced Simulation and Computing (ASC) program and the Exascale Computing Project (ECP). Sustained investment in these core technologies is essential. • Today’s state of the art involves running ensembles of O(100K) simulations to perform uncertainty quantification (UQ) and design studies using multiple statistical methods such as Bayesian optimization to understand sensitivities of our models and explore parameterized design spaces. Even with exascale computing, we are practically limited to O(10) parameters in these studies since the number of simulations required to sample the space scales exponentially with the number of design parameters. • The data from these simulation ensembles is increasingly being used to train machine learned (ML) surrogates (or reduced order models, ROMs) which can then be used for optimization or real time design exploration. However, the trained surrogates are still limited in the number of parameters they can represent due to the sampling limitations previously noted. • Augmenting our suite of integrated multiphysics simulation codes, both current and emerging, with the ability to compute gradients (solution derivatives) of arbitrary simulation outputs with respect to (some or all) simulation inputs would be a breakthrough technology, opening the door to a new era of efficient and automated inverse design based on verified and validated mod/sim capabilities. • This capability, which we refer to as differentiable multiphysics codes (DMCs), would revolutionize both UQ and optimization studies by breaking the curse of dimensionality that presently limits our “gradient-free” ensemble based computing approach. A similar breakthrough occurred in the AI/ML community once the ability to compute gradients of arbitrary loss functions using back-propagation became commonplace. Gradient information from the multiphysics codes can also be used to dramatically improve the efficiency and scale of training of ML/ROM surrogates for rapid assessments. • Achieving this in our suite of codes will be a grand challenge, similar to the amount of effort that was required to transition from CPU to GPU computing. It will require buy-in from the entire WSC/CP program and beyond, including all integrated codes, physics and engineering models, third-party library dependencies and performance portability abstractions. It will also require investment in research and development of numerical methods for computing adjoints of coupled physics across multiple adaptively refined moving meshes and of stochastic (Monte Carlo) and mesh free (SPH) methods. • New software and numerical techniques, largely pioneered by the AI/ML community, make this feasible. Chief among these is automatic differentiation (AD), the ability to employ AD at point-wise locations in a physics calculation (instead of traditional black-box approaches) and the ability to perform “back-propagation in time” (or reverse mode AD) for non-linear partial differential equations (PDEs). Fundamentally, the conclusion of this strategic planning exercise is that the time is right to undertake a large scale effort in WSC, centered on the existing integrated codes, to continue the natural evolution of mod/sim in the age of AI/ML. Instead of attempting to replace mod/sim with purely data driven AI/ML models, we believe the key to success is to integrate AI/ML by building on top of the decades of hard-won knowledge and the verified/validated multiphysics modelling capability that is the hallmark of the ASC program.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

A method for crystallographic mapping of an alpha-beta titanium alloy with nanometre resolution using scanning precession electron diffraction and open-source software libraries

An approach for the crystallographic mapping of two-phase alloys on the nanoscale using a combination of scanned precession electron diffraction and open-source python libraries is introduced in this paper. This method is demonstrated using the example of a two-phase α/β titanium alloy. The data were recorded using a direct electron detector to collect the patterns, and recently developed algorithms to perform automated indexing and analyse the crystallography from the results. Very high-quality mapping is achieved at a 3 nm step size. The results show the expected Burgers orientation relationships between the α laths and β matrix, as well as the expected misorientations between α laths. A minor issue was found that one area was affected by 180° ambiguities in indexing occur due to this area being aligned too close to a zone axis of the α with twofold projection symmetry (not present in 3D) in the zero-order Laue Zone, and this should be avoided in data acquisition in the future. Nevertheless, this study demonstrates a good workflow for the analysis of nanocrystalline two- or multi-phase materials, which will be of widespread use in analysing two-phase titanium and other systems and how they evolve as a function of thermomechanical treatments.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Improved Gas Plume Identification Using Nearest Neighbor Methods for Background Estimation

Longwave infrared (LWIR) hyperspectral imaging (HSI) can be used for many tasks in remote sensing, including detecting and identifying effluent gases by LWIR sensors on airborne platforms. Identification is used after detection to increase confidence in weakly detected plumes, reduce false positives from detection, and distinguish between similar and confounding material signatures. Background estimation is an important step used to reveal the unique spectral characteristics of the detected gas, allowing the identification model to determine what the gas is specifically. The importance of proper background estimation increases when dealing with weak signals, large libraries of gases of interest, and uncommon or heterogeneous backgrounds. In this article, we propose two methods for background estimation: a novel k-nearest segments (KNS) algorithm and the standard k-nearest neighbors (KNN) algorithm. We test our methods and three existing background estimation methods for comparison against global background estimation to determine which performs best at estimating the true background radiance under a plume and for increasing identification confidence using a neural network classification model. We compare the different methods using 640 simulated weak plumes in an urban environment. For identification, our KNS algorithm improves median neural network identification confidence by 53.2%. For background radiance estimation, the KNN algorithm provides a median of 49 times less RMSE than global background estimation. Furthermore, KNN is the easiest method to tune for different plumes, making it an excellent “out of the box” background estimator.

47 OTHER INSTRUMENTATION↗
DOE OSTI2021

Sensitivity Analysis of MFiX-PIC Parameters Using Nodeworks, PSUADE, and DAKOTA

The study presented in this report was aimed to demonstrate UQ analysis performed not only with Nodeworks, but also two other well-established UQ software tools from the U.S. DOE’s National Laboratories (PSUADE from Lawrence Livermore National Laboratory and DAKOTA from Sandia National Laboratory). It is important to emphasize that the motivation of this study was not to determine the best UQ software, but to verify if the global sensitivity analyses from the end-to-end workflow in Nodeworks are consistent with the results of other two UQ software. The components of Nodeworks from Python’s ecosystem have been tested as standalone libraries. However, an assessment study for the complete workflow targeting a specific UQ analysis has not been performed for Nodeworks. Hence, this study is expected to serve as an equivalent of solution verification for Nodeworks using other established UQ tools as reference solution. For this purpose, three distinct flow configurations (i.e., settling bed, bubbling fluidized, and circulating fluidized bed) have been used as representative multiphase flow problems of interest. The results of the systematic simulation campaigns performed in an earlier study using the particle-in-cell (PIC) approach in the Multiphase Flow with Interphase eXchanges (MFIX) suite of solvers (i.e., MFiX-PIC) was utilized. The same set of tabulated results was provided as input to the different UQ software for global sensitivity analysis. Results for the three cases indicate that based on the Sobol’ Sensitivity Indices method the order of importance ranking determined by Nodeworks for the Sobol’ Total Sensitivity Indices is consistent with PSUADE and DAKOTA in each case for the five model parameters considered. The input files for Nodeworks for the three cases are also shared through NETL’s Gitlab repository for the reader interested in reproducibility and further analysis (See Section 1.2).

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Enhancing Docking Accuracy with PECAN2, a 3D Atomic Neural Network Trained without Co-Complex Crystal Structures

Decades of drug development research have explored a vast chemical space for highly active compounds. The exponential growth of virtual libraries enables easy access to billions of synthesizable molecules. Computational modeling, particularly molecular docking, utilizes physics-based calculations to prioritize molecules for synthesis and testing. Nevertheless, the molecular docking process often yields docking poses with favorable scores that prove to be inaccurate with experimental testing. To address these issues, several approaches using machine learning (ML) have been proposed to filter incorrect poses based on the crystal structures. However, most of the methods are limited by the availability of structure data. Here, we propose a new pose classification approach, PECAN2 (Pose Classification with 3D Atomic Network 2), without the need for crystal structures, based on a 3D atomic neural network with Point Cloud Network (PCN). The new approach uses the correlation between docking scores and experimental data to assign labels, instead of relying on the crystal structures. We validate the proposed classifier on multiple datasets including human mu, delta, and kappa opioid receptors and SARS-CoV-2 Mpro. Our results demonstrate that leveraging the correlation between docking scores and experimental data alone enhances molecular docking performance by filtering out false positives and false negatives.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Invertible neural networks for real-time control of extrusion additive manufacturing

Material extrusion additive manufacturing (AM) has enabled an elegant fabrication pathway for a vast material library. Nonetheless, each material requires optimization of printing parameters generally determined through significant trial-and-error testing. To eliminate arduous, iteration-based optimization approaches, many researchers have used machine learning (ML) algorithms which provide opportunities for automated process optimization. Here, in this work, we demonstrate the use of an ML-driven approach for real-time material extrusion print-parameter optimization through in-situ monitoring of printed line geometry. To do this, we use deep invertible neural networks (INNs) which can solve both forward and inverse, or optimization, problems using a single network. By combining in-situ computer vision and deep INNs, the printing parameters can be autonomously optimized to print a target line width in 1.2 s. Furthermore, defects that occur during printing can be rapidly identified and corrected autonomously. The methods developed and presented in this work eliminate user-intensive, time-consuming, and iterative parameter discovery approaches that currently limit accelerated implementation of extrusion-based AM processes. Furthermore, the presented approach can be generalized to provide real-time monitoring and optimization pathways for increasingly complex AM environments.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Application of SAMMY for Non-Destructive Characterization of Irradiated Samples Using Neutron Transmission Measurements at VENUS

This report presents the application of SAMMY (Larson 2008), an R-matrix Bayesian fitting system, for analyzing neutron transmission data to enable non-destructive isotopic identification. Transmission measurements were conducted at the VENUS (Bilheux et al. 2023) beamline of the Spallation Neutron Source (SNS) using samples with unknown or partially characterized compositions. Through the application of SAMMY’s R-matrix Bayesian fitting capabilities to the measured transmission spectra and comparison against evaluated nuclear data libraries, isotopic constituents can be quantitatively determined based on their characteristic neutron resonance signatures. This methodology demonstrates significant potential for applications in nuclear safeguards, forensics, and post-irradiation examination.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2022

From PeleC to PeleACC, to PeleC++

PeleC is an Exascale Computing Project application for simulating compressible combustion in complex geometries. It has been built on top of the popular AMReX library. In the beginning of the Exascale Computing Project, PeleC was focused on KNL. It uses a mixture of C++, C, and kernels written in Fortran to obtain performance by focusing on vectorization. Recently we have taken two approaches in deciding PeleC's future for obtaining performance on exascale GPU machines. In the first programming model, we decorated the Fortran kernels with OpenACC directives. This expedited our ability to run at large scales on Summit's GPUs, where we achieved a significant speedup over the CPUs on Summit. The second programming model involved rewriting the Fortran kernels in C++ and using AMReX's Kokkos-like lambda abstractions for running on the GPU. This resulted in similar speedups on Summit's GPUs over merely utilizing the CPUs. Both approaches involved AMReX's management of memory transfers between the device and host. In this work, we compare and contrast the benefits and pitfalls to both programming approaches regarding performance, performance portability, and productivity. We also discuss advantages we have found in taking the time to modernize our code and why have chosen a specific pathway to prepare our code for the future DOE exascale machines.

exascale computing↗
DOE OSTI2024

A graphics processing unit accelerated sparse direct solver and preconditioner with block low rank compression

We present the GPU implementation efforts and challenges of the sparse solver package STRUMPACK. The code is made publicly available on github with a permissive BSD license. STRUMPACK implements an approximate multifrontal solver, a sparse LU factorization which makes use of compression methods to accelerate time to solution and reduce memory usage. Multiple compression schemes based on rank-structured and hierarchical matrix approximations are supported, including hierarchically semi-separable, hierarchically off-diagonal butterfly, and block low rank. Here, in this paper, we present the GPU implementation of the block low rank (BLR) compression method within a multifrontal solver. Our GPU implementation relies on highly optimized vendor libraries such as cuBLAS and cuSOLVER for NVIDIA GPUs, rocBLAS and rocSOLVER for AMD GPUs and the Intel oneAPI Math Kernel Library (oneMKL) for Intel GPUs. Additionally, we rely on external open source libraries such as SLATE (Software for Linear Algebra Targeting Exascale), MAGMA (Matrix Algebra on GPU and Multi-core Architectures), and KBLAS (KAUST BLAS). SLATE is used as a GPU-capable ScaLAPACK replacement. From MAGMA we use variable sized batched dense linear algebra operations such as GEMM, TRSM and LU with partial pivoting. KBLAS provides efficient (batched) low rank matrix compression for NVIDIA GPUs using an adaptive randomized sampling scheme. The resulting sparse solver and preconditioner runs on NVIDIA, AMD and Intel GPUs. Interfaces are available from PETSc, Trilinos and MFEM, or the solver can be used directly in user code. We report results for a range of benchmark applications, using the Perlmutter system from NERSC, Frontier from ORNL, and Aurora from ALCF. For a high frequency wave equation on a regular mesh, using 32 Perlmutter compute nodes, the factorization phase of the exact GPU solver is about 6.5× faster compared to the CPU-only solver. The BLR-enabled GPU solver is about 13.8× faster than the CPU exact solver. For a collection of SuiteSparse matrices, the STRUMPACK exact factorization on a single GPU is on average 1.9× faster than NVIDIA’s cuDSS solver.

97 MATHEMATICS AND COMPUTING↗