Search NASA⌕ Search

SEARCH · Search NASA

Results for “PARALLEL PROCESSING”

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 1,711 records · Page 95

Digital Channel Simulator Developed and Tested

The Digital Channel Simulator (DCS) is a real-time test set developed in-house by the NASA Glenn Research Center at Lewis Field that simulates the characteristics of the modulator, demodulator, and transmission medium in a typical communications system to enable controlled laboratory testing of codec pairs. The DCS can support data rates up to 100 megasymbols per second (Msymbols/sec) with symbol sizes up to 10 bits and is compatible with both TTL (transistor transistor logic) and ECL (emitter coupled logic) interfaces. Because of its use of digital integrated circuits (IC's), the DCS offers the user accurate and repeatable testing while maintaining a simple reconfiguration of the modulation scheme and noise characteristics. The PC-based graphical user interface (GUI) assures user friendly operation for configuring, controlling, and monitoring the DCS and system during tests. In a typical communications system, the modulator places a symbol in constellation space and puts it on a carrier to be sent to the demodulator. Because of noise on the channel, the I and Q position in constellation space cannot be recovered exactly, and the received coordinates shift. To mimic this process in the laboratory, the DCS uses a mapper to place the symbol in constellation space. It simulates the shift in coordinates by digitally adding "noise" to the I and Q values. The mapper and noise source are implemented in lookup tables. Modulation schemes and noise characteristics are set by the values loaded in these tables. The mapper also has a pass-through mode to facilitate modulator testing, allowing noise to be added to 8-bit I and Q values of modulated data without a second mapping. To achieve high symbol rates, eight processing circuits are placed in parallel between an ECL demultiplexer and multiplexer. A graphical user interface was developed to calculate, load, and verify the values for the lookup tables. This interface can also be used to debug and verify proper operation of the channel simulator or to control an experiment. Operation of the DCS has been verified through three tests: a low-speed comprehensive system test, a high-speed (20 Msymbols/sec) test of the TTL interface, and a high-speed (100 Msymbols/sec) test of the ECL interface. The DCS is now ready for use by NASA and external customers.

Bizon, Thomas P.↗

Assessment of Near-Field Sonic Boom Simulation Tools

A recent study for the Supersonics Project, within the National Aeronautics and Space Administration, has been conducted to assess current in-house capabilities for the prediction of near-field sonic boom. Such capabilities are required to simulate the highly nonlinear flow near an aircraft, wherein a sonic-boom signature is generated. There are many available computational fluid dynamics codes that could be used to provide the near-field flow for a sonic boom calculation. However, such codes have typically been developed for applications involving aerodynamic configuration, for which an efficiently generated computational mesh is usually not optimum for a sonic boom prediction. Preliminary guidelines are suggested to characterize a state-of-the-art sonic boom prediction methodology. The available simulation tools that are best suited to incorporate into that methodology are identified; preliminary test cases are presented in support of the selection. During this phase of process definition and tool selection, parallel research was conducted in an attempt to establish criteria that link the properties of a computational mesh to the accuracy of a sonic boom prediction. Such properties include sufficient grid density near shocks and within the zone of influence, which are achieved by adaptation and mesh refinement strategies. Prediction accuracy is validated by comparison with wind tunnel data.

Casper, J. H.↗

Robotically Assembled Aerospace Structures: Digital Material Assembly using a Gantry-Type Assembler

This paper evaluates the development of automated assembly techniques for discrete lattice structures using a multi-axis gantry type CNC machine. These lattices are made of discrete components called digital materials. We present the development of a specialized end effector that works in conjunction with the CNC machine to assemble these lattices. With this configuration we are able to place voxels at a rate of 1.5 per minute. The scalability of digital material structures due to the incremental modular assembly is one of its key traits and an important metric of interest. We investigate the build times of a 5x5 beam structure on the scale of 1 meter (325 parts), 10 meters (3,250 parts), and 30 meters (9,750 parts). Utilizing the current configuration with a single end effector, performing serial assembly with a globally fixed feed station at the edge of the build volume, the build time increases according to a scaling law of n4, where n is the build scale. Build times can be reduced significantly by integrating feed systems into the gantry itself, resulting in a scaling law of n3. A completely serial assembly process will encounter time limitations as build scale increases. Automated assembly for digital materials can assemble high performance structures from discrete parts, and techniques such as built in feed systems, parallelization, and optimization of the fastening process will yield much higher throughput.

Mechanical Structure↗

Impacts of Spontaneous Hot Flow Anomalies on the Magnetosheath and Magnetopause

Spacecraft observations and global hybrid (kinetic ions and fluid electrons) simulations have demonstrated that ion dissipation processes at the quasi-parallel bow shock are associated with the formation of structures called spontaneous hot flow anomalies (SHFAs). Previous simulations and recent spacecraft observations have also established that SHFAs result in the formation of magnetosheath filamentary structures(MFS). In this paper we demonstrate that in addition to MFS, SHFAs also result in the formation of magnetos heath cavities that are associated with decreases in density, velocity, and magnetic field and enhancements in temperature. We use the results of a global MHD run to determine the change in the magnetosheath properties associated with cavities due to ion kinetic effects. The results also show the formation of regions of high flow speed called magnetosheath jets whose properties as a function of solar wind Mach number are described in this study. Comparing the properties of the simulated magnetosheath cavities and jets to past spacecraft observations provides good agreement in both cases. We also demonstrate that pressure variations associated with cavities and SHFAs in the sheath result in a continuous sunward and anti sunward magnetopause motion. This result is consistent with previous suggestions that SHFAs may be responsible for the generation of ion cyclotron waves and precipitation of ring current protons in the outer magnetosphere.

Omidi, N.↗

Robotically Assembled Aerospace Structures: Digital Material Assembly using a Gantry-Type Assembler

This paper evaluates the development of automated assembly techniques for discrete lattice structures using a multi-axis gantry type CNC machine. These lattices are made of discrete components called "digital materials." We present the development of a specialized end effector that works in conjunction with the CNC machine to assemble these lattices. With this configuration we are able to place voxels at a rate of 1.5 per minute. The scalability of digital material structures due to the incremental modular assembly is one of its key traits and an important metric of interest. We investigate the build times of a 5x5 beam structure on the scale of 1 meter (325 parts), 10 meters (3,250 parts), and 30 meters (9,750 parts). Utilizing the current configuration with a single end effector, performing serial assembly with a globally fixed feed station at the edge of the build volume, the build time increases according to a scaling law of n4, where n is the build scale. Build times can be reduced significantly by integrating feed systems into the gantry itself, resulting in a scaling law of n3. A completely serial assembly process will encounter time limitations as build scale increases. Automated assembly for digital materials can assemble high performance structures from discrete parts, and techniques such as built in feed systems, parallelization, and optimization of the fastening process will yield much higher throughput.

Manufacturing↗

Aladyn - Adaptive Neural Network Molecular Dynamics Simulation Code: Computational Materials Mini-Application

This report provides an overview and commands description of the Computational Materials mini-application, Aladyn. Aladyn is a simple molecular dynamics code written in FORTRAN 2008, which is designed to demonstrate the use of adaptive neural networks (ANNs) in atomistic simulations. The role of ANNs is to reproduce the very complex energy landscape resulting from the atomic interactions in materials with the accuracy of quantum mechanics-based energy calculations. The ANN is trained on a large set of atomic structures calculated using the density functional theory (DFT) method. The Aladyn code is being released to serve as a training testbed for students and professors in academia to explore possible optimization algorithms for parallel computing on multicore central processing unit (CPU) computers or computers utilizing many core architectures based on graphic processing units (GPUs). The effort is related to the High Performance Computing Incubator (HPCI) project at NASA Langley Research Center.

Yamakov, Vesselin I.↗

Investigation of Magmatic Activities on Early Mars Using Igneous Mineral Chemistry in Gale Crater, Mars

One objective of rover missions is exploring the geological context of the surroundings. Over the years, igneous petrology and sedimentology have been disconnected, the first investigating magmatic processes and volcanic activities, and the second seeking environmental conditions in the past and assessing the habitability of the planet. Although different, one is related to the other: igneous rocks are altered and broken down, leading to the formation of sedimentary rocks, which can in turn be used to back out the nature of their magmatic source. The Curiosity rover that landed in the 3.7 Gyr old impact crater Gale is traveling through sedimentary rocks. About fifty float rocks have been observed, and several of them with ambiguous texture and composition have been classified as igneous or sedimentary depending on studies such as Jake_M. The composition of several unambiguous igneous rocks has been analyzed [4- 6] but their heterogeneity at a larger instrumental (measurement size < 2 cm) scale prevents the measurement of a bulk composition as performed on Earth. An original approach avoiding these two last issues is to consider igneous mineral chemistry analyzed within igneous and sedimentary rocks to assess magmatic processes that could have formed them. Most Curiosity data are used to explore ancient environmental conditions, and a significant number of compositional analyses are under-explored for constraining magmatic activities. We will present how we can make use of sedimentary data for investigating igneous processes in the vicinity of Gale crater. Geological Context: We focus on the first 750 martian days, corresponding to measurements in a coherent lacustrine sedimentary unit called Bradbury, because all sedimentary rocks were sourced from the same watershed and appear to have a consistent source with minimal alteration [2-3]. Igneous detrital minerals including feldspar and pyroxene, are observed in sedimentary rocks. Monte Carlo models showed that minimal cation loss is observed based on the composition of all Bradbury rocks, implying negligible weathering [3]. Although clay minerals are detected in few rocks [7], chemical compositions of rocks can be explained by a mixture of primary igneous minerals [3]. Variation of composition within Bradbury rocks can be explained by mineral sorting and one distinct source component. While a common magmatic source is suggested, Bradbury sediments likely come from several volcanic eruptions from a single magmatic chamber [9- 10]. The occurrence of alkali minerals like sanidine and K-rich rocks throughout Bradbury supports the presence of a potassic component, likely trachytic, while plagioclase and a mafic composition suggest a basaltic component [8-9]. Instruments: Mineral chemistry can be estimated by three instruments onboard Curiosity. The CheMin instrument enables detection of mineral assemblages using X-ray diffraction (XRD). Using Rietveld refinement, each mineral is identified according to their 1D XRD pattern [11]. Note that distinction between pyroxene minerals is challenging with the CheMin instrument due to overlapping peaks on XRD patterns and low angular resolution of the instrument [12]. Then, using least square regression and optimization algorithms based on unit-cell parameters, mineral chemistry has been estimated by [11]. Plagioclase composition has been estimated using the NaAlSi3O8- CaAl2Si2O8 system and alkali feldspar is based on the NaAlSi3O8-KAlSi3O8 system (stars in Fig.1). Two mudstone samples (John Klein and Cumberland) and one sandstone sample (Windjana) were analyzed by CheMin at Yellowknife Bay and Kimberley, respectively. Figure 1. Ternary diagrams of feldspars (top) and pyroxene (bottom) quadrilateral. Stars correspond to CheMin composition and the gray patches to ChemCam composition. The colored dots are the composition of feldspar and pyroxene that crystallized during fractional crystallization at FMQ+1 of a melt extracted at distinct melting degree during the adiabatic ascent of a primitive mantle composition, without any water (left panels) and with 0.5 wt.% of water (right panels) at distinct pressure. The ChemCam instrument enables the analysis of the chemical compositions of rocks at hundreds of micrometer scale (350-550 μm) using laser induced breakdown spectroscopy (LIBS), which may provide the composition of minerals when they are larger than the beam spot (>550 μm) [13]. Within >5000 LIBS points, we performed a typical stoichiometric filtering allowing us to distinguish 56 feldspar and 10 pyroxene mineral compositions (grey patches in Fig. 1). Finally, the Alpha Particle X-ray Spectrometer (APXS) analyzes the composition of rocks with a 1.6 cm diameter spot size. Monte Carlo mass balance modeling allowed [3] to decipher a feldspar range varying between An30 and An40 (Fig. 1). Discussion: Although there could be a more complex history and other ways to form the whole compositional range of igneous minerals analyzed within the Bradbury formation, we are presenting here simple magmatic pathways commonly occurring on Earth using the thermodynamical softwares pMELTS and rhyoliteMELTS [14]. The objective is to find reasonable igneous processes that produce minerals that parallel the compositions of feldspar and pyroxene analyzed by the Curiosity rover. As commonly observed for mid-ocean ridge basalts, the adiabatic ascent of a primitive mantle composition [15] partially melting at 2 GPa has been modeled, followed by the extraction of a liquid at distinct degrees of partial melting, which undergoes fractional crystallization at an oxygen fugacity +1 log unit above the fayalite-magnetite-quartz (FMQ) buffer within the crust (0.02-0.4 GPa) with H2O = 0- 0.5 wt. %. These latter conditions correspond to those recorded within igneous clasts from the Noachian martian breccia NWA 7034 and paired and within Gale igneous rocks (colored dots in Fig. 1) [16-17]. To check the reliability of these 2-step models, we also tested fractional crystallization at similar conditions (FMQ+1; P=0.02-0.4 GPa; H2O = 0-0.5 wt. %) of starting compositions corresponding to that of magmas with distinct melting degrees obtained from isobaric experiments at 2 GPa [18]. Mineral compositions obtained from both models are similar. As shown on Fig. 1, the whole range of observed feldspar compositions cannot be reproduced by fractionation of one magma only. Indeed, while alkali feldspar and Na-plagioclase likely crystallized from fractional crystallization of a low-degree melt (here <15%), plagioclase and pyroxene can only be formed by fractional crystallization of a higher degree melt (here >19%). The corresponding liquid descent lines are broadly in agreement with compositions estimated by ChemCam corresponding to float igneous rocks (Fig. 2) [4-6]. Figure 2. Silica versus alkali content. Lines show the liquid lines of descent from magmas with distinct degrees of melting. Gray patches represents the composition of Gale igneous rocks [4-6]. Trachytic to rhyolitic magmas crystallize alkali feldspar, and andesite to dacite magmas likely form plagioclase. Therefore, at least two starting magmas at distinct melting degrees, which could easily come from a single mantle source, are necessary to explain the whole compositional range of feldspar and pyroxene analyzed within Bradbury rocks. Conclusion: Because rocks from the Bradbury formation are likely originating from the same magmatic source with minimal weathering as supported by several studies using different approaches, igneous mineral chemistry analyzed by CheMin and ChemCam allows us to back out reasonable magmatic pathways that could have crystallized them. Fractional crystallization of at least two starting magmas originating from distinct melting degrees of a single mantle source can explain the whole range of feldspar and pyroxene composition. Both alkaline and sub-alkaline liquids can be produced, with compositions corresponding to those of the igneous rocks analyzed by ChemCam within the Bradbury formation, highlighting the complexity of Mars magmatism.

Payre, V.↗

VIGILANTE: An Advanced Sensing/Processing Testbed for ATR Applications

VIGILANTE consists of two major components:1. the Viewing Imager/Gimballed Instrumentation Laboratory (VIGIL)- advanced infrared, visible, and ultraviolet sensors with appropriate optics and camera electronics & 2. the Analog Neural Three-dimensional processing Experiment (ANTE)-a massively parallel, neural network-based, high-speed processor.

automatic↗

Feasibility Study of Distributed Decision-Making on the Edge for Urban Air Mobility

The Concept of Operations for Urban Air Mobility (UAM) put forward by FAA, NASA, and several industry stakeholders acknowledges the diversity and complexity in UAM operations and, thereby, envisions a federated architecture for UAM management. In this architecture, the decision-making is distributed to a set of service providers who collectively manage the shared airspace usage by different stakeholders. This notionally brings autonomy closer to the UAM businesses and encourages to explore the feasibility of decision making on the very edge, which is the topic of the presented research. This paper reports research conducted on the hypothesis based on which the residual compute capability onboard smart unmanned aerial systems (UASs) is utilized to build situational awareness and resolve conflicts by passive and active coordination among multiple UASs, thereby implementing a layer of distributed autonomy in UAM. Key features of the edge-computing approach involve inter-UAS information exchange, independent assessment of own flight and environmental conditions, and estimation of other UASs’ flight preferences, incorporating machine learning techniques in the last two. Parallel computing on portable graphics processing unit (GPU) enables the machine learning workflow on the edge. A custom-built 3D simulator is used to evaluate the efficacy of the distributed decision-making on the edge. Each edge node, representing a smart UAS, connects to the simulator from a remote location and independently controls the behavior of the corresponding virtual asset in the simulator, analogous to participants in an online multi-player game. The presented edge-computing-based distributed decision-making framework is envisioned to pave the way for collective mobility of autonomous air vehicles in the future shared airspace, while allowing the inclusion of the business preferences of the UAS operators within allowed regulatory limits.

Edge computing↗

Feasibility Study of Distributed Decision-Making on the Edge for Urban Air Mobility

The Concept of Operations for Urban Air Mobility (UAM) put forward by FAA, NASA, and several industry stakeholders acknowledges the diversity and complexity in UAM operations and, thereby, envisions a federated architecture for UAM management. In this architecture, the decision-making is distributed to a set of service providers who collectively manage the shared airspace usage by different stakeholders. This notionally brings autonomy closer to the UAM businesses and encourages to explore the feasibility of decision making on the very edge, which is the topic of the presented research. This paper reports research conducted on the hypothesis based on which the residual compute capability onboard smart unmanned aerial systems (UASs) is utilized to build situational awareness and resolve conflicts by passive and active coordination among multiple UASs, thereby implementing a layer of distributed autonomy in UAM. Key features of the edge-computing approach involve inter-UAS information exchange, independent assessment of own flight and environmental conditions, and estimation of other UASs’ flight preferences, incorporating machine learning techniques in the last two. Parallel computing on portable graphics processing unit (GPU) enables the machine learning workflow on the edge. A custom-built 3D simulator is used to evaluate the efficacy of the distributed decision-making on the edge. Each edge node, representing a smart UAS, connects to the simulator from a remote location and independently controls the behavior of the corresponding virtual asset in the simulator, analogous to participants in an online multi-player game. The presented edge-computing-based distributed decision-making framework is envisioned to pave the way for collective mobility of autonomous air vehicles in the future shared airspace, while allowing the inclusion of the business preferences of the UAS operators within allowed regulatory limits.

Edge computing↗

System and method for continuous solids slurry depressurization

A continuous slag processing system includes a rotating parallel disc pump, coupled to a motor and a brake. The rotating parallel disc pump includes opposing discs coupled to a shaft, an outlet configured to continuously receive a fluid at a first pressure, and an inlet configured to continuously discharge the fluid at a second pressure less than the first pressure. The rotating parallel disc pump is configurable in a reverse-acting pump mode and a letdown turbine mode. The motor is configured to drive the opposing discs about the shaft and against a flow of the fluid to control a difference between the first pressure and the second pressure in the reverse-acting pump mode. The brake is configured to resist rotation of the opposing discs about the shaft to control the difference between the first pressure and the second pressure in the letdown turbine mode.

36 MATERIALS SCIENCE↗

Systems and methods for powder bed additive manufacturing anomaly detection

Detection and classification of anomalies for powder bed metal additive manufacturing. Anomalies, such as recoater blade impacts, binder deposition issues, spatter generation, and some porosities, are surface-visible at each layer of the building process. A multi-scaled parallel dynamic segmentation convolutional neural network architecture provides additive manufacturing machine and imaging system agnostic pixel-wise semantic segmentation of layer-wise powder bed image data. Learned knowledge is easily transferrable between different additive manufacturing machines. The anomaly detection can be conducted in real-time and provides accurate and generalizable results.

Scime, Luke R.↗

Enabling the Broader Use of MOOSE for Nuclear Energy and Other Simulation

This Final Scientific and Technical Report summarizes work performed under the Phase IIA SBIR project “Enabling the Broader Use of MOOSE for Nuclear Energy and Other Simulation” (DE-SC0020906) from August 2023 through August 2025. The objective of the Phase IIA effort was to mature and harden capabilities developed during Phase II, with the goal of enabling practical interoperability between Coreform’s isogeometric analysis (IGA) technologies and the Multiphysics Object-Oriented Simulation Environment (MOOSE), while improving robustness, performance, and scalability for complex, nuclear-relevant geometries. Over the course of Phase IIA, the project established and validated an extraction-based interoperability pathway between Coreform tools and MOOSE. A combined mesh and matrix format was defined collaboratively with MOOSE developers and integrated into the solver, enabling standard MOOSE workflows to operate on data exported from Coreform’s IGA and Flex Representation Method (FRM) pipelines. Early demonstrations validated architectural compatibility using linear solid mechanics problems, while later efforts focused on benchmark testing and external use. By the end of the project period, engineers at BWXT were able to independently set up and execute a simulation using the Coreform–MOOSE workflow and provide direct feedback that informed further refinement. In parallel, substantial effort was devoted to improving the robustness of trimmed U-spline construction for complex CAD geometries. A growing test suite of nuclear-relevant models was compiled through collaboration with multiple stakeholders and used to drive extensive bug fixing and reliability improvements. These efforts resulted in improved robustness and performance, including the addition of fallback capabilities that enhance reliability when the underlying commercial CAD kernel fails. Performance-oriented work progressed later in the project, with the development and demonstration of methods to decompose complex geometries into structured subregions and updated data representations to support more efficient solver processing. Additionally, extensive enhancements to threadsafe parallel data structures and trimming operations established a foundation for scalable processing of large assemblies. Collaboration with Sandia National Laboratories on the SGM geometric modeling kernel advanced to a functioning interface test case, positioning the workflow for future kernel integration. Overall, the Phase IIA effort successfully transitioned the project from architectural proof-of-concept to externally exercised, solver-integrated capability, while clarifying remaining technical challenges related to standardization, performance optimization, and kernel integration.

42 ENGINEERING↗

Computing Flow Transition On Parallel Processors

Parallel algorithm developed on multiple-microprocessor computer. Program initiated to develop computer codes capable of directly simulating and mathematically modeling transition process at mach numbers ranging from subsonic to hypersonic. Parallel computers potentially offer reduction of processing time; processing time inversely proportional to number of available processors.

Bokhari, S.↗

Leveraging Pre-Built Catalogs and Object-Level Scheduling to Eliminate I/O Bottlenecks in HPC Environments

Modern High-Performance Computing (HPC) environments face mounting challenges due to the shift from large to small file datasets, along with an increasing number of users and parallelized applications. As HPC systems rely on Parallel File Systems (PFS), such as Lustre for data processing, performance bottlenecks stemming from Object Storage Target (OST) contention have become a significant concern. Existing solutions, such as LADS with its object-level scheduling approach, fall short in large-scale HPC environments due to their inability to effectively address metadata I/O bottlenecks and the growing number of I/O processes. This study highlights the pressing need for a comprehensive solution that tackles both OST contention and metadata I/O challenges in diverse HPC workloads. To address these challenges, we propose SwiftLoad, an object-level I/O scheduling framework that leverages a metadata catalog to enhance the performance and efficiency of parallel HPC utilities. The adoption of the metadata catalog mitigates the metadata I/O bottlenecks that commonly occur in HPC utilities, a challenge that is particularly pronounced in object-level I/O scheduling. SwiftLoad addresses OST contention and the uneven distribution of I/O processes across different OSTs through mathematical modeling and incorporates a Loader Configuration Module to regulate the number of I/O processes. Evaluated with two representative utilities—data deduplication profiling and data augmentation—SwiftLoad achieved performance improvements of up to 5.63x and 11.0x, respectively, on a production supercomputer.

HPC↗

Ion reflection and downstream thermalization at the quasi-parallel bow shock

Using the results of ISEE 2 plasma and magnetic field measurements, two features related to the ion thermalization process at high-Mach-number (M above 2) quasi-parallel collisionless shocks are discussed. These are the presence of a coherent secondary beam of ions within the shock layer which is considered to be produced by reflection, and downstream ion distributions which contain both a relatively cold core of directly transmitted ions and a hotter 'shell' of ions, which appear to result from the disruption and scattering of ions initially reflected at the shock. Evidence is presented that coherent ion reflection is an important element of the ion energy dissipation process at high-Mach-number quasi-parallel shocks.

Gosling, J. T.↗

Extreme Scaling of Production Visualization Software on Diverse Architectures

We present the results of a series of experiments studying how visualization software scales to massive data sets. Although several paradigms exist for processing large data, we focus on pure parallelism, the dominant approach for production software. These experiments utilized multiple visualization algorithms and were run on multiple architectures. Two types of experiments were performed. For the first, we examined performance at massive scale: 16,000 or more cores and one trillion or more cells. For the second, we studied weak scaling performance. These experiments were performed on the largest data set sizes published to date in visualization literature, and the findings on scaling characteristics and bottlenecks contribute to understanding of how pure parallelism will perform at high levels of concurrency and with very large data sets.

97 MATHEMATICS AND COMPUTING↗

PAGANI: a parallel adaptive GPU algorithm for numerical integration

We present a new adaptive parallel algorithm for the challenging problem of multi-dimensional numerical integration on massively parallel architectures. Adaptive algorithms have demonstrated the best performance, but efficient many-core utilization is difficult to achieve because the adaptive work-load can vary greatly across the integration space and is impossible to predict a priori. Existing parallel algorithms utilize sequential computations on independent processors, which results in bottlenecks due to the need for data redistribution and processor synchronization. Our algorithm employs a high-throughput approach in which all existing sub-regions are processed and sub-divided in parallel. Repeated sub-region classification and filtering improves upon a brute-force approach and allows the algorithm to make efficient use of computation and memory resources. A CUDA implementation shows orders of magnitude speedup over the fastest open-source CPU method and extends the achievable accuracy for difficult integrands. Our algorithm typically outperforms other existing deterministic parallel methods.

Sakiotis, Ioannis↗