Search NASA⌕ Search

SEARCH · Search NASA

Results for “Dynamic load balancing”

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 505 records · Page 28

Dynamic Force Reconstruction of Transient Flap Control Force Experiments in a Hypersonic Wind Tunnel

This work details the design of a sliced-cone model with a flap embedded within the slice. The flap is controlled by a fast acting servo to simulate a control force of a hypersonic vehicle. Design considerations such as sensor placement, expected loading, servo arm selection, and cavity temperatures are detailed in this work. Such a design imparts a dynamic response of the test article and as such, dynamic force measurement methodologies are required to reconstruct the control forcing. Three methods are of particular interest: the Sum of Weighted Accelerations Technique, Time Domain Deconvolution Method, and Frequency Domain Inverse Method. Novel versions of each method have been applied to this dataset to improve the reconstruction accuracy.

dynamic balance↗

Packaging strategies for printed circuit board components. Volume I, materials & thermal stresses.

Decisions on material selections for electronics packaging can be quite complicated by the need to balance the criteria to withstand severe impacts yet survive deep thermal cycles intact. Many times, material choices are based on historical precedence perhaps ignorant of whether those initial choices were carefully investigated or whether the requirements on the new component match those of previous units. The goal of this program focuses on developing both increased intuition for generic packaging guidelines and computational methodologies for optimizing packaging in specific components. Initial efforts centered on characterization of classes of materials common to packaging strategies and computational analyses of stresses generated during thermal cycling to identify strengths and weaknesses of various material choices. Future studies will analyze the same example problems incorporating the effects of curing stresses as needed and analyzing dynamic loadings to compare trends with the quasi-static conclusions.

36 MATERIALS SCIENCE↗

Efficient Parallelization of Irregular Applications on GPU Architectures

With the enlarging computation capacity of general Graphics Processing Units (GPUs), leveraging GPUs to accelerate parallel applications has become a critical topic in academia and industry. However, a wide range of irregular applications with the computation-/memory-intensive nature cannot easily achieve high GPU utilization. The challenges mainly involve the following aspects: first, data dependence leads to coarse-grained kernel and inefficient parallelism; second, heavy GPU memory usage may cause frequent memory evictions and extra overhead of I/O; third, specific computation patterns produce memory redundancies; last, workload balance and data reusability conjunctly benefit the overall performance, but there may exist a dynamic trade-off between them. Targeting these challenges, this dissertation proposes multiple optimizations to accelerate two real-world applications: many-body correlation functions to simulate nuclear physics in a large-scale scientific system; the other is the eALS-based matrix factorization recommendation system. To accelerate the calculations of many-body correlation functions, this dissertation presents three frameworks in GPU memory management and multi-GPU scheduling. Firstly, an optimized systematic GPU memory management framework, MemHC, utilizes a series of new memory reduction designs in GPU memory allocation, CPU/GPU communications, and GPU memory oversubscription. Secondly, an enhanced multi-GPU scheduling framework, MICCO, particularly by taking both data dimension (e.g., data reuse and data eviction) and computation dimension into account. MICCO designs a heuristic scheduling algorithm and a machine learning-based regression model to generate the optimal settings of a proposed new concept to manage the trade-off. Thirdly, a locality-aware multi-GPU scheduling framework. This scheduler leverages pipeline batch generation with a looking-ahead strategy by building local dependency graphs for memory transfer reduction and better data reuse, achieving up to 79.92% memory cost reduction and 1.67x speedup. To parallelize the eALS-based recommendation system, this dissertation proposes an efficient CPU/GPU heterogeneous recommendation system, HEALS. HEALS employs newly designed architecture-adaptive data formats to achieve load balance and good data locality on CPU and GPU. To mitigate the data dependence, HEALS presents a CPU/GPU collaboration model for both task parallelism and data parallelism with multiple kernel computation optimizations. In summary, this dissertation efficiently accelerates two typical irregular applications on GPUs by building four frameworks, including CPU/GPU collaboration, GPU memory management, and multi-GPU scheduling.

Wang, Qihan↗

Modeling, Analysis, and Control of Demand Response Resources

While the traditional goal of an electric power system has been to control supply to fulfill demand, the demand-side can plan an active role in power systems via Demand Response (DR), defined by the Department of Energy (DOE) as “a tariff or program established to motivate changes in electric use by end-use customers in response to changes in the price of electricity over time, or to give incentive payments designed to induce lower electricity use at times of high market prices or when grid reliability is jeopardized” [29]. DR can provide a variety of benefits including reducing peak electric loads when the power system is stressed and fast timescale energy balancing. Therefore, DR can improve grid reliability and reduce wholesale energy prices and their volatility. This dissertation focuses on analyzing both recent and emerging DR paradigms. Recent DR programs have focused on peak load reduction in commercial buildings and industrial facilities (C&I facilities). We present methods for using 15-minute-interval electric load data, commonly available from C&I facilities, to help building managers understand building energy consumption and ‘ask the right questions’ to discover opportunities for DR. Additionally, we present a regression-based model of whole building electric load, i.e., a baseline model, which allows us to quantify DR performance. We use this baseline model to understand the performance of 38 C&I facilities participating in an automated dynamic pricing DR program in California. In this program, facilities are expected to exhibit the same response each DR event. We find that baseline model error makes it difficult to precisely quantify changes in electricity consumption and understand if C&I facilities exhibit event-to-event variability in their response to DR signals. Therefore, we present a method to compute baseline model error and a metric to determine how much observed DR variability results from baseline model error rather than real variability in response. We find that, in general, baseline model error is large. Though some facilities exhibit real DR variability, most observed variability results from baseline model error. In some cases, however, aggregations of C&I facilities exhibit real DR variability, which could create challenges for power system operation. These results have implications for DR program design and deployment. Emerging DR paradigms focus on faster timescale DR. Here, we investigate methods to coordinate aggregations of residential thermostatically controlled loads (TCLs), including air conditioners and refrigerators, to manage frequency and energy imbalances in power systems. We focus on opportunities to centrally control loads with high accuracy but low requirements for sensing and communications infrastructure. Specifically, we compare cases when measured load state information (e.g., power consumption and temperature) is 1) available in real time; 2) available, but not in real time; and 3) not available. We develop Markov Chain models to describe the temperature state evolution of heterogeneous populations of TCLs, and use Kalman filtering for both state and joint parameter/state estimation. We present a look-ahead proportional controller to broadcast control signals to all TCLs, which always remain in their temperature dead-band. Simulations indicate that it is possible to achieve power tracking RMS errors in the range of 0.26–9.3% of steady state aggregated power consumption. Results depend upon the information available for system identification, state estimation, and control. We find that, depending upon the performance required, TCLs may not need to provide state information to the central controller in real time or at all. We also estimate the size of the TCL potential resource; potential revenue from participation in markets; and break-even costs associated with deploying DR-enabling technologies. We find that current TCL energy storage capacity in California is 8–11 GWh, with refrigerators contributing the most. Annual revenues from participation in regulation vary from $10 to $220 per TCL per year depending upon the type of TCL and climate zone, while load following and arbitrage revenues are more modest at $2 to $35 per TCL per year. These results lead to a number of policy recommendations that will make it easier to engage residential loads in fast timescale DR.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Challenges of Monte Carlo Transport

These are slides from a presentation for Parallel Summer School at Los Alamos National Laboratory. Solving discretized partial differential equations (PDEs) of interest can require a large number of computations. We can identify concurrency to allow parallel solution of discrete PDEs. Simulated particles histories can be used to solve the Boltzmann transport equation. Particle histories are independent in neutral particle transport, making them amenable to parallel computation. Physical parameters and method type determine the data dependencies of particle histories. Data requirements shape parallel algorithms for Monte Carlo. Then, Parallel Computational Physics and Parallel Monte Carlo are discussed and, finally, the results are given. The mesh passing method greatly simplifies the IMC implementation and allows simple load-balancing. Using MPI windows and passive, one-sided RMA further simplifies the implementation by removing target synchronization. The author is very interested in implementations of PGAS that may allow further optimization for one-sided, read-only memory access (e.g. Open SHMEM). The MPICH_RMA_OVER_DMAPP option and library is required to make one-sided messaging scale on Trinitite - Moonlight scales poorly. Interconnect specific libraries or functions are likely necessary to ensure performance. BRANSON has been used to directly compare the current standard method to a proposed method on idealized problems. The mesh passing algorithm performs well on problems that are designed to show the scalability of the particle passing method. BRANSON can now run load-imbalanced, dynamic problems. Potential avenues of improvement in the mesh passing algorithm will be implemented and explored. A suite of test problems that stress DD methods will elucidate a possible path forward for production codes.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Hybrid methods for rotordynamic analysis

Effective procedures are presented for the response analysis of the Space Shuttle Main Engine turbopumps under transient loading conditions. Of particular concern is the determination of the nonlinear response of the systems to rotor imbalance in presence of bearing clearances. The proposed procedures take advantage of the nonlinearities involved being localized at only a few rotor/housing coupling joints. The methods include those based on integral formulations for the incremental solutions involving the transition matrices of the rotor and housing. Alternatively, a convolutional representation of the housing displacements at the coupling points is proposed which would allow performing the transient analysis on a reduced model of the housing. The integral approach is applied to small dynamical models to demonstrate the efficiency of the approach. For purposes of assessing the numerical integration results for the nonlinear rotor/housing systems, a numerical harmonic balance procedure is developed to enable determining all possible harmonic, subharmonic, and nonperiodic solutions of the systems. A brief account of the Fourier approach is presented as applied to a two degree of freedon rotor-support system.

Noah, Sherif T.↗

Multi-Domain Routing in Delay Tolerant Networks

The goal of Delay Tolerant Networking (DTN) is to provide the missing ingredient for the ever-growing collection of communicating nodes in our solar system to become a Solar System Internet (SSI). Great strides have been made in modeling particular types of DTNs, such as schedule- or discovery-based. Now, analogously to the Internet, these smaller DTNs can be considered routing domains which must be stitched together to form the overall SSI. In this paper, we propose a framework for cross-domain routing in DTNs as well as methodologies for detecting these sub-domains. Example time-varying networks are given to demonstrate the techniques proposed. A basic component is the mathematical theory of sheaves, which unifies the underlying model of DTN routing algorithms, by giving rise to routing sheaves – these can be defined for the dynamic and scheduled networks as noted above, and can also be used to define the interfaces between these domains in order to route across them. An immediate application would be routing across discovery-based networks connected by scheduled networks. These DTN subdomains remain elusive, however, and need to become well-defined and properly sized for tractable computability. In particular, a balance must be determined between areas that are too large (i.e. large matrix computations) versus areas that are too small (i.e. “many” single-noded domains). Moreover, the connections between the domains should, at least locally, be chosen to optimize data flow and connectivity: we address this in three ways. First, tools from persistent homology are given to understand underlying structures, reminiscent of hierarchies in the Internet Protocol (IP) addressing. Second, we construct a notion of temporal graph curvature based on network geometry to analyze flows induced by dynamical processes on these networks. Finally, Schrodinger Bridges, a tool arising from statistical physics, are proposed as a method of constructing flows on time-evolving networks with desirable properties such as speed, robustness, and load sensitivity. We construct an approach to temporal hypergraphs to simultaneously model unicast, multicast, and broadcast, using the language of scheme theory, and then consider DTN network coding as a way to achieve network-level computation and organization. The paper concludes with a discussion and ideas for future work.

Alan Hylton↗

A Comparison of Tandem Walk Performance Between Bed Rest Subjects and Astronauts

Astronauts experience a microgravity environment during spaceflight, which results in a central reinterpretation of both vestibular and body axial-loading information by the sensorimotor system. Subjects in bed rest studies lie at 6deg head-down in strict bed rest to simulate the fluid shift and gravity-unloading of the microgravity environment. However, bed rest subjects still sense gravity in the vestibular organs. Therefore, bed rest isolates the axial-unloading component, thus allowing for the direct study of its effects. The Tandem Walk is a standard sensorimotor test of dynamic postural stability. In a previous abstract, we compared performance on a Tandem Walk test between bed rest control subjects, and short- and long-duration astronauts both before and after flight/bed rest using a composite index of performance, called the Tandem Walk Parameter (TWP), that takes into account speed, accuracy, and balance control. This new study extends the previous data set to include bed rest subjects who performed exercise countermeasures. The purpose of this study was to compare performance during the Tandem Walk test between bed rest subjects (with and without exercise), short-duration (Space Shuttle) crewmembers, and long-duration International Space Station (ISS) crewmembers at various time points during their recovery from bed rest or spaceflight.

Miller, Chris↗

MSC/NASTRAN DMAP Alter Used for Closed-Form Static Analysis With Inertia Relief and Displacement-Dependent Loads

Solving for the displacements of free-free coupled systems acted upon by static loads is a common task in the aerospace industry. Often, these problems are solved by static analysis with inertia relief. This technique allows for a free-free static analysis by balancing the applied loads with the inertia loads generated by the applied loads. For some engineering applications, the displacements of the free-free coupled system induce additional static loads. Hence, the applied loads are equal to the original loads plus the displacement-dependent loads. A launch vehicle being acted upon by an aerodynamic loading can have such applied loads. The final displacements of such systems are commonly determined with iterative solution techniques. Unfortunately, these techniques can be time consuming and labor intensive. Because the coupled system equations for free-free systems with displacement-dependent loads can be written in closed form, it is advantageous to solve for the displacements in this manner. Implementing closed-form equations in static analysis with inertia relief is analogous to implementing transfer functions in dynamic analysis. An MSC/NASTRAN (MacNeal-Schwendler Corporation/NASA Structural Analysis) DMAP (Direct Matrix Abstraction Program) Alter was used to include displacement-dependent loads in static analysis with inertia relief. It efficiently solved a common aerospace problem that typically has been solved with an iterative technique.

Source record↗

Stability and control of power systems with high penetrations of inverter-based resources: An accessible review of current knowledge and open questions

As power system renewable energy penetrations increase, the ways in which key renewable technologies such as wind and solar photovoltaics (PV) differ from thermal generators become more apparent. Many studies have examined the variability and uncertainty of such generators and described how generation and load can be balanced for a wide variety of annual energy penetrations, at timescales from seconds to years. Another important characteristic of these resources is asynchronicity, the result of using inverters to interface the prime energy source with the power system as opposed to synchronous generators. Unlike synchronous generators, whose frequency of alternating current (AC) injection is physically coupled to the rotation of the machine itself, inverter based asynchronous generators do not share the same physical coupling with the generated frequency. These subtle differences impact the operations of power systems developed around the characteristics of synchronous generators. In this paper we review current knowledge and open research questions concerning the interplay between asynchronous inverter-based resources (IBRs) and cycle- to second-scale power system dynamics, with a focus on how stability and control may be impacted or need to be achieved differently when there are high instantaneous penetrations of IBRs across an interconnection. This work does not seek to provide a comprehensive review of the latest developments, but is instead intended to be accessible to any reader with an engineering background and an interest in power systems and renewable energy. As such, the paper includes basic material on power electronics, control schemes for IBRs, and power system stability; and uses this background material to describe potential impacts of IBRs on power system stability, operational challenges associated with large amounts of distributed IBR generation, and modern power system simulation trends driven by IBR characteristics.

14 SOLAR ENERGY↗

Power flows and Mechanical Intensities in structural finite element analysis

The identification of power flow paths in dynamically loaded structures is an important, but currently unavailable, capability for the finite element analyst. For this reason, methods for calculating power flows and mechanical intensities in finite element models are developed here. Formulations for calculating input and output powers, power flows, mechanical intensities, and power dissipations for beam, plate, and solid element types are derived. NASTRAN is used to calculate the required velocity, force, and stress results of an analysis, which a post-processor then uses to calculate power flow quantities. The SDRC I-deas Supertab module is used to view the final results. Test models include a simple truss and a beam-stiffened cantilever plate. Both test cases showed reasonable power flow fields over low to medium frequencies, with accurate power balances. Future work will include testing with more complex models, developing an interactive graphics program to view easily and efficiently the analysis results, applying shape optimization methods to the problem with power flow variables as design constraints, and adding the power flow capability to NASTRAN.

Hambric, Stephen A.↗

Time-Accurate Computational Fluid Dynamics Analysis of a Supersonic Turbine to Assess Structural Mode Excitation

Gas turbines can experience significant unsteady loading during operation which can cause damage such as blade cracking or bearing wear. While computational fluid dynamics (CFD) modeling of full three-dimensional gas turbine flow to verify meanline design performance is standard practice within the industry, the use of time-accurate CFD makes it possible to assess the unsteady loading which can be especially problematic at operating conditions where unsteady fluid loading aligns with structural modes. The Fluid Dynamics branch (ER42) at NASA Marshall Space Flight Center has developed a time-accurate CFD analysis methodology using the unstructured, density-based, Loci-CHEM solver which is capable of simulating the relative motion required for turbomachinery applications through sliding interface meshes. The highly-parallel unstructured solver allows for full three-dimensional, time-accurate simulation of turbines including upstream and downstream components such as manifolds, nozzles, and stators, any of which could excite a turbine structural mode. Recently, Loci-CHEM was used to analyze the FASTRAC supersonic turbine to assess the loading on the rotor. Overall performance and time-averaged loading on the rotor was computed, including the axial force over the entire rotor disk which is required for force balance calculations on the turbopump. The FASTRAC turbine has supersonic nozzles upstream of the rotor that result in significant unsteady loading on the rotor blades at multiples of the nozzle pass frequency. The unsteady blade pressure loading was decomposed into the frequency domain for direct use in structural forced response analysis. Utilizing this type of CFD analysis coupled with structural analysis has proven to be a valuable tool in identifying structural modes that are being excited by unsteady fluid loading.

M.V. Fischels↗

Time-Accurate Computational Fluid Dynamics Analysis of a Supersonic Turbine to Assess Structural Mode Excitation

Gas turbines can experience significant unsteady loading during operation which can cause damage such as blade cracking or bearing wear. While computational fluid dynamics (CFD) modeling of full three-dimensional gas turbine flow to verify meanline design performance is standard practice within the industry, the use of time-accurate CFD makes it possible to assess the unsteady loading which can be especially problematic at operating conditions where unsteady fluid loading aligns with structural modes. The Fluid Dynamics branch (ER42) at NASA Marshall Space Flight Center has developed a time-accurate CFD analysis methodology using the unstructured, density-based, Loci-CHEM solver which is capable of simulating the relative motion required for turbomachinery applications through sliding interface meshes. The highly-parallel unstructured solver allows for full three-dimensional, time-accurate simulation of turbines including upstream and downstream components such as manifolds, nozzles, and stators, any of which could excite a turbine structural mode. Recently, Loci-CHEM was used to analyze the FASTRAC supersonic turbine to assess the loading on the rotor. Overall performance and time-averaged loading on the rotor was computed, including the axial force over the entire rotor disk which is required for force balance calculations on the turbopump. The FASTRAC turbine has supersonic nozzles upstream of the rotor that result in significant unsteady loading on the rotor blades at multiples of the nozzle pass frequency. The unsteady blade pressure loading was decomposed into the frequency domain for direct use in structural forced response analysis. Utilizing this type of CFD analysis coupled with structural analysis has proven to be a valuable tool in identifying structural modes that are being excited by unsteady fluid loading.

M V Fischels↗

Using hob offset to balance dynamic strength in spur gears

This paper presents an analytical study on the effect of hob offset on the dynamic tooth strength of spur gears. The study was limited to equal and opposite offset values applied to the pinion and gear to maintain the standard operating center distance. The analysis presented in this paper was performed using a new version of the NASA gear dynamics code DANST. The operating speed of the transmission has a significant influence on the amount of hob offset required to equalize the dynamic stresses in the pinion and gear. In the transmission studied, at low speeds, the optimum hob offset value was found to fluctuate. At higher speeds, the optimum value was constrained by the minimum allowed thickness at the tip of the pinion tooth. For gears that must operate over a range of speeds, an average offset value may be used. Spur gears designed with the procedure presented here can have significant improvements in load capacity.

Liou, Chuen-Huei↗

SOFIA 2 model telescope wind tunnel test report

This document outlines the tests performed to make aerodynamic force and torque measurements on the SOFIA wind tunnel model telescope. These tests were performed during the SOFIA 2 wind tunnel test in the 14 ft wind tunnel during the months of June through August 1994. The test was designed to measure the dynamic cross elevation moment acting on the SOFIA model telescope due to aerodynamic loading. The measurements were taken with the telescope mounted in an open cavity in the tail section of the SOFIA model 747. The purpose of the test was to obtain an estimate of the full scale aerodynamic disturbance spectrum, by scaling up the wind tunnel results (taking into account differences in sail area, air density, cavity dimension, etc.). An estimate of the full scale cross elevation moment spectrum was needed to help determine the impact this disturbance would have on the telescope positioning system requirements. A model of the telescope structure, made of a light weight composite material, was mounted in the open cavity of the SOFIA wind tunnel model. This model was mounted via a force balance to the cavity bulkhead. Despite efforts to use a 'stiff' balance, and a lightweight model, the balance/telescope system had a very low resonant frequency (37 Hz) compared to the desired measurement bandwidth (1000 Hz). Due to this mechanical resonance of the balance/telescope system, the balance alone could not provide an accurate measure of applied aerodynamic force at the high frequencies desired. A method of measurement was developed that incorporated accelerometers in addition to the balance signal, to calculate the aerodynamic force.

Keas, Paul↗

Advanced Stirling Radioisotope Generator (ASRG) Thermal Power Model in MATLAB

This paper presents a one-dimensional steady-state mathematical thermal power model of the ASRG. It aims to provide a guideline of understanding how the ASRG works and what can change its performance. The thermal dynamics and energy balance of the generator is explained using the thermal circuit of the ASRG. The Stirling convertor performance map is used to represent the convertor. How the convertor performance map is coupled in the thermal circuit is explained. The ASRG performance characteristics under i) different sink temperatures and ii) over the years of mission (YOM) are predicted using the one-dimensional model. Two Stirling converter control strategies, i) fixing the hot-end of temperature of the convertor by adjusting piston amplitude and ii) fixing the piston amplitude, were tested in the model. Numerical results show that the first control strategy can result in a higher system efficiency than the second control strategy when the ambient gets warmer or the general-purpose heat source (GPHS) fuel load decays over the YOM. The ASRG performance data presented in this paper doesn't pertain to the ASRG flight unit. Some data of the ASRG engineering unit (EU) and flight unit that are available in public domain are used in this paper for the purpose of numerical studies.

Wang, Xiao-Yen, J.↗

Arbitrary Order Virtual Element Methods for High‐Order Phase‐Field Modeling of Dynamic Fracture

ABSTRACT Accurate modeling of fracture nucleation and propagation in brittle and ductile materials subjected to dynamic loading is important in predicting material damage and failure under extreme conditions. Phase‐field fracture models have garnered a lot of attention in recent years due to their success in representing damage and fracture processes in a wide class of materials and under a variety of loading conditions. Second‐order phase‐field fracture models are by far the most popular among researchers (and increasingly, among practitioners), but fourth‐order models have started to gain broader acceptance since their more recent introduction. The exact solution corresponding to these high‐order phase‐field fracture models has higher regularity. Thus, numerical solutions of the model equations can achieve improved accuracy and higher spatial convergence rates. In this work, we develop a virtual element framework for the high‐order phase‐field model of dynamic fracture. The virtual element method (VEM) can be regarded as a generalization of the classical finite element method. In addition to many other desirable characteristics, the VEM allows computing on polytopal meshes. Here, we use ‐conforming virtual elements and the generalized‐ time integration method for the momentum balance equation, and adopt ‐conforming virtual elements for the high‐order phase‐field equation. We verify our virtual element framework using classical quasi‐static benchmark problems and demonstrate its capabilities with the aid of numerical simulations of dynamic fracture in brittle materials.

42 ENGINEERING↗

Performance Enhancement Strategies for Multi-Block Overset Grid CFD Applications

The overset grid methodology has significantly reduced time-to-solution of highfidelity computational fluid dynamics (CFD) simulations about complex aerospace configurations. The solution process resolves the geometrical complexity of the problem domain by using separately generated but overlapping structured discretization grids that periodically exchange information through interpolation. However, high performance computations of such large-scale realistic applications must be handled efficiently on state-of-the-art parallel supercomputers. This paper analyzes the effects of various performance enhancement strategies on the parallel efficiency of an overset grid Navier-Stokes CFD application running on an SGI Origin2000 machinc. Specifically, the role of asynchronous communication, grid splitting, and grid grouping strategies are presented and discussed. Details of a sophisticated graph partitioning technique for grid grouping are also provided. Results indicate that performance depends critically on the level of latency hiding and the quality of load balancing across the processors.

Djomehri, M. Jahed↗