Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallel Performance Data”

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 289 records · Page 16

TSE computers - A means for massively parallel computations

A description is presented of hardware concepts for building a massively parallel processing system for two-dimensional data. The processing system is to use logic arrays of 128 x 128 elements which perform over 16 thousand operations simultaneously. Attention is given to image data, logic arrays, basic image logic functions, a prototype negator, an interleaver device, image logic circuits, and an image memory circuit.

Strong, J. P., III↗

Implementation and validation of a wake model for vortex-surface interactions in low speed forward flight

The wake model was implemented using a VAX 750 and a Microvax II workstation. Online graphics capability using a DISSPLA graphics package. The rotor model used by Beddoes was significantly extended to include azimuthal variations due to forward flight and a simplified scheme for locating critical points where vortex elements are placed. A test case was obtained for validation of the predictions of induced velocity. Comparison of the results indicates that the code requires some more features before satisfactory predictions can be made over the whole rotor disk. Specifically, shed vorticity due to the azimuthal variation of blade loading must be incorporated into the model. Interactions between vortices shed from the four blades of the model rotor must be included. The Scully code for calculating the velocity field is being modified in parallel with these efforts to enable comparison with experimental data. To date, some comparisons with flow visualization data obtained at Georgia Tech were performed and show good agreement for the isolated rotor case. Comparison of time-resolved velocity data obtained at Georgia Tech also shows good agreement. Modifications are being implemented to enable generation of time-averaged results for comparison with NASA data.

Komerath, Narayanan M.↗

File concepts for parallel I/O

The subject of input/output (I/O) was often been neglected in the design of parallel computer systems, although for many problems I/O rates will limit the speedup attainable. The I/O problem is addressed by considering the role of files in parallel systems. The notion of parallel files is introduced. Parallel files provide for concurrent access by multiple processes, and utilize parallelism in the I/O system to improve performance. Parallel files can also be used conventionally by sequential programs. A set of standard parallel file organizations is proposed, based on common data partitioning techniques. Implementation strategies for the proposed organizations are suggested, using multiple storage devices. Problem areas are also identified and discussed.

Crockett, Thomas W.↗

Breaking down barriers in cooperative fault management: Temporal and functional information displays

At the highest level, the fundamental question addressed by this research is how to aid human operators engaged in dynamic fault management. In dynamic fault management there is some underlying dynamic process (an engineered or physiological process referred to as the monitored process - MP) whose state changes over time and whose behavior must be monitored and controlled. In these types of applications (dynamic, real-time systems), a vast array of sensor data is available to provide information on the state of the MP. Faults disturb the MP and diagnosis must be performed in parallel with responses to maintain process integrity and to correct the underlying problem. These situations frequently involve time pressure, multiple interacting goals, high consequences of failure, and multiple interleaved tasks.

Potter, Scott S.↗

Real-time software receiver

A real-time software receiver that executes on a general purpose processor. The software receiver includes data acquisition and correlator modules that perform, in place of hardware correlation, baseband mixing and PRN code correlation using bit-wise parallelism.

Ledvina, Brent M.↗

Real-time software receiver

A real-time software receiver that executes on a general purpose processor. The software receiver includes data acquisition and correlator modules that perform, in place of hardware correlation, baseband mixing and PRN code correlation using bit-wise parallelism.

Ledvina, Brent M.↗

Computational Investigation of Retropropulsion Operating Environments with a GPU-Enabled Detached Eddy Simulation Approach

Human exploration of the surface of Mars will require an extended powered descent phase of flight, during which aerodynamic-propulsive interference effects can be significant. Characterization of these environments to enable implementation of this technology into a flight vehicle will rely heavily on computational simulation. This work advances the understanding of retropropulsion aerodynamics through application of a massively parallel detached eddy simulation approach on a GPU-accelerated computational framework, yielding data that are largely unachievable with conventional high-performance computing resources. This work includes time-dependent and time-averaged forces and moments on a conceptual, full-scale vehicle in environments and operating conditions relevant to human Mars exploration. Conditions are examined where the engine exhaust flow transitions between over-expanded and under-expanded flow structures, and flight operation will require the ability to maintain control of the vehicle during such a transition. These transitions occur as the vehicle decelerates, and as such, this investigation includes supersonic, transonic, and subsonic flight conditions. Options for vehicle control during powered flight include differential throttling of the engines. This paper provides an overview of the computational campaign, approach, and discussion of results in characterizing the resulting aerodynamics for differential throttling with retropropulsion in atmospheric environments.

EDL↗

I/O in Machine Learning Applications on HPC Systems: A 360-degree Survey

Growing interest in Artificial Intelligence (AI) has resulted in a surge in demand for faster methods of Machine Learning (ML) model training and inference. This demand for speed has prompted the use of high performance computing (HPC) systems that excel in managing distributed workloads. Because data is the main fuel for AI applications, the performance of the storage and I/O subsystem of HPC systems is critical. In the past, HPC applications accessed large portions of data written by simulations or experiments or ingested data for visualizations or analysis tasks. ML workloads perform small reads spread across a large number of random files. This shift of I/O access patterns poses several challenges to modern parallel storage systems. In this paper, we survey I/O in ML applications on HPC systems, and target literature within a 6-year time window from 2019 to 2024. We define the scope of the survey, provide an overview of the common phases of ML, review available profilers and benchmarks, examine the I/O patterns encountered during offline data preparation, training, and inference, and explore I/O optimizations utilized in modern ML frameworks and proposed in recent literature. Lastly, we seek to expose research gaps that could spawn further R&D.

97 MATHEMATICS AND COMPUTING↗

A portable MPI-based parallel vector template library

This paper discusses the design and implementation of a polymorphic collection library for distributed address-space parallel computers. The library provides a data-parallel programming model for C++ by providing three main components: a single generic collection class, generic algorithms over collections, and generic algebraic combining functions. Collection elements are the fourth component of a program written using the library and may be either of the built-in types of C or of user-defined types. Many ideas are borrowed from the Standard Template Library (STL) of C++, although a restricted programming model is proposed because of the distributed address-space memory model assumed. Whereas the STL provides standard collections and implementations of algorithms for uniprocessors, this paper advocates standardizing interfaces that may be customized for different parallel computers. Just as the STL attempts to increase programmer productivity through code reuse, a similar standard for parallel computers could provide programmers with a standard set of algorithms portable across many different architectures. The efficacy of this approach is verified by examining performance data collected from an initial implementation of the library running on an IBM SP-2 and an Intel Paragon.

Sheffler, Thomas J.↗

A Portable MPI-Based Parallel Vector Template Library

This paper discusses the design and implementation of a polymorphic collection library for distributed address-space parallel computers. The library provides a data-parallel programming model for C + + by providing three main components: a single generic collection class, generic algorithms over collections, and generic algebraic combining functions. Collection elements are the fourth component of a program written using the library and may be either of the built-in types of c or of user-defined types. Many ideas are borrowed from the Standard Template Library (STL) of C++, although a restricted programming model is proposed because of the distributed address-space memory model assumed. Whereas the STL provides standard collections and implementations of algorithms for uniprocessors, this paper advocates standardizing interfaces that may be customized for different parallel computers. Just as the STL attempts to increase programmer productivity through code reuse, a similar standard for parallel computers could provide programmers with a standard set of algorithms portable across many different architectures. The efficacy of this approach is verified by examining performance data collected from an initial implementation of the library running on an IBM SP-2 and an Intel Paragon.

Sheffler, Thomas J.↗

Performance evaluation of the JPL interim digital SAR processor

The performance of the Interim Digital SAR Processor (IDP) was evaluated. The IDP processor was originally developed for experimental processing of digital SEASAT SAR data. One phase of the system upgrade which features parallel processing in three peripheral array processors, automated estimation for Doppler parameters, and unsupervised image pixel location determination and registration was executed. The method to compensate for the target range curvature effect was improved. A four point interpolation scheme is implemented to replace the nearest neighbor scheme used in the original IDP. The processor still maintains its fast throughput speed. The current performance and capability of the processing modes now available on the IDP system are updated.

Wu, C.↗

Simulation and dynamic performance of a 20-kHz spacecraft power system

A candidate 20-kHz spacecraft power system which includes a series-parallel combination of four Mapham inverters connected to several types of loads is described. A computer simulation of the power system is used to illustrate its steady-state and dynamic performance on an end-to-end basis. Comparisons with measured data are made. It is shown that significant distortion of the 20-kHz bus voltage can occur due to the switching of the load converters. This distortion can be reduced by including a shunt-connected parallel resonant filter on the 20-kHz side of the load converter. It is also shown that the distortion can be reduced by using a pulse-density-modulated switching strategy.

Wasynczuk, O.↗

ATM solar array in-flight performance analysis

The physical and electrical characteristics of the Apollo Telescope Mount (ATM) solar array are described and in-flight performance data are analyzed and compared with predicted results. Two solar cell module configurations were used. Type I module consists of 228 2 x 6 cm solar cells with two cells in parallel and 114 cells in series. Type II modules contain 684 2 x 2 cm cells with six cells in parallel and 114 cells in series. A different interconnection scheme was used for each type. Panels using type II modules with mesh interconnect system performed marginally better than those using type I module with loop interconnect system. The average degradation rate for the ATM array was 8.2% for a 271-day mission.

Thornton, J. P.↗

The paradigm compiler: Mapping a functional language for the connection machine

The Paradigm Compiler implements a new approach to compiling programs written in high level languages for execution on highly parallel computers. The general approach is to identify the principal data structures constructed by the program and to map these structures onto the processing elements of the target machine. The mapping is chosen to maximize performance as determined through compile time global analysis of the source program. The source language is Sisal, a functional language designed for scientific computations, and the target language is Paris, the published low level interface to the Connection Machine. The data structures considered are multidimensional arrays whose dimensions are known at compile time. Computations that build such arrays usually offer opportunities for highly parallel execution; they are data parallel. The Connection Machine is an attractive target for these computations, and the parallel for construct of the Sisal language is a convenient high level notation for data parallel algorithms. The principles and organization of the Paradigm Compiler are discussed.

Dennis, Jack B.↗

Pressure-Fed LOX/LCH4 Reaction Control System for Spacecraft: Transient Modeling and Thermal Vacuum Hotfire Test Results

An integrated cryogenic liquid oxygen, liquid methane (LOX/LCH4) reaction control system (RCS) was tested at NASA Glenn Research Center's Plum Brook Station in the Spacecraft Propulsion Research Facility (B-2) under vacuum and thermal vacuum conditions. The RCS is a subsystem of the Integrated Cryogenic Propulsion Test Article (ICPTA), a pressure-fed LOX/LCH4 propulsion system composed of a single 2,800 lbf main engine, two 28 lbf RCS engines, and two 7 lbf RCS engines. Propellants are stored in four 48 inch diameter 5083 aluminum tanks that feed both the main engine and RCS engines in parallel. Helium stored cryogenically in a composite overwrapped pressure vessel (COPV) flows through a heat exchanger on the main engine before being used to pressurize the propellant tanks to a design operating pressure of 325 psi. The ICPTA is capable of simultaneous main engine and RCS operation. The RCS engines utilize a coil-on-plug (COP) ignition system designed for operation in a vacuum environment, eliminating corona discharge issues associated with a high voltage lead. There are two RCS pods on the ICPTA, with two engines on each pod. One of these two engines is a heritage flight engine from Project Morpheus. Its sea level nozzle was removed and replaced by an 85:1 nozzle machined using Inconel 718, resulting in a maximum thrust of 28 lbf under altitude conditions. The other engine is a scaled down version of the 28 lbf engine, designed to match the core and overall mixture ratios as well as other injector characteristics. This engine can produce a maximum thrust of 7 lbf with an 85:1 nozzle that was additively manufactured using Inconel 718. Both engines are film-cooled and capable of limited duration gas-gas and gas-liquid operation, as well as steady-state liquid-liquid operation. Each pod contains one of each version, such that two engines of the same thrust level can be fired as a couple on opposite pods. The RCS feed system is composed of symmetrical 3/8 inch lines that tap off of the main propellant manifold to send LOX and LCH4 outboard to the RCS pods. A Thermodynamic Vent System (TVS) is used to condition propellants at each pod by venting through an orifice and then routing the cold expansion products back through tubing that is welded along a large portion of the main RCS feed lines. Prior to final installation on the ICPTA, the RCS engines were tested in a small vacuum chamber at the Johnson Space Center (JSC) Energy Systems Test Area (ESTA) to verify functionality of the new COP ignition system and check out operation of the vacuum nozzles. After engine-level testing, the RCS engines were installed on the vehicle and a series of integrated hot-fire tests were performed at JSC consisting of various pulsing and steady-state firings as well as integrated main engine/RCS operation. The ICPTA was then integrated into the Plum Brook B-2 facility for vacuum and thermal/vacuum testing. Testing in the B-2 facility was composed of multiple thermal and pressure environments. The first set of tests were performed under ambient temperature and altitude pressure conditions. These tests consisted of a range of minimum impulse bit (MIB) pulsing sequences with low duty cycle, analogous to a coast phase in which the RCS is primarily used for station keeping. The primary goal of this sequence is to understand how propellant conditions were effected without an active TVS. In this scenario, consistent gas-gas operation is desirable since it results in a smaller MIB and more efficient propellant consumption. Multiple skin thermocouples are mounted on the feedlines, in addition to a submerged thermocouple on each commodity, in order to gather thermal data on the system. Higher duty cycle pulsing tests were then performed, analogous to an ascent or landing mission phase. The primary goal of this sequence was to examine how well the engines self-conditioned without active TVS when starting from a quiescent state. The TVS was then activated during some tests to demonstrate the capability to quickly condition the engines for higher pulsing demand scenarios. A thermocouple at the TVS outlet allows for the calculation of energy absorbed by the vented propellant. Lastly, tests with longer pulses and multiple engines firing either in sequence or simultaneously were run in order to gather transient system response data on waterhammer. Six total high-speed pressure transducers are installed on the RCS system, one sensor at the end of each propellant manifold line on the pods, and one at the tap-off location for each commodity. This will allow for the accurate characterization of waterhammer in the system under various propellant conditions and firing sequences. Other instrumentation for this test series includes nozzle throat thermocouples, chamber pressure measurement, heat soakback measurement, and tank wall plume impingement temperature measurement. The next set of tests were performed to demonstrate simultaneous main engine and RCS operation. Data from this test will be used to examine if there is any change to nominal operation of the RCS as a result of feed system interaction or other phenomenon. Some of these tests began under high vacuum conditions (target ambient pressure less than 1x10(exp -3) torr) and others began at altitude conditions. The last set of tests were performed with the B-2 cold wall active. Under these tests, many of the same low duty cycle MIB tests were repeated in order to characterize how propellant conditions changed with the lower heat leak. In this scenario the RCS manifold experiences much less heat leak, resulting in a change to how well the engines self-condition. As a result, an increase in maximum waterhammer pressures and a change in natural frequency of the system was expected due to higher density propellants. The lower heat leak should also result in a change to the MIB pulse profile, and data will be examined to understand how MIB repeatability is affected in the different operating environments. Parallel to the test efforts, a set of transient model development efforts were made to predict RCS performance. The primary effort was aimed at producing a SINDA/FLUINT model to predict propellant conditioning up to the engine inlet as a function of different environmental and operating parameters, with the goal of predicting chamber pressure, TVS performance, and propellant consumption over time. Preliminary results for this effort will be presented in comparison with test data. Additional modeling efforts were made using SINDA/FLUINT to predict waterhammer in the system since the software is capable of handling multiphase transient fluid dynamics. These results will be compared with the high-speed pressure transducer test data for validation purposes.

Atwell, Matthew J.↗

ORBIT: Oak Ridge Base Foundation Model for Earth System Predictability

Earth system predictability is challenged by the complexity of environmental dynamics and the multitude of variables involved. Current AI foundation models, although advanced by leveraging large and heterogeneous data, are often constrained by their size and data integration, limiting their effectiveness in addressing the full range of Earth system prediction challenges. To overcome these limitations, we introduce the Oak Ridge Base Foundation Model for Earth System Predictability (ORBIT), an advanced vision transformer model that scales up to 113 billion parameters using a novel hybrid tensor-data orthogonal parallelism technique. As the largest model of its kind, ORBIT surpasses the current climate AI foundation model size by a thousandfold. Performance scaling tests conducted on the Frontier supercomputer have demonstrated that ORBIT achieves 684 petaFLOPS to 1.6 exaFLOPS sustained throughput, with scaling efficiency maintained at 41% to 85% across 49,152 AMD GPUs. These breakthroughs establish new advances in AI-driven climate modeling and demonstrate promise to significantly improve the Earth system predictability.

Wang, Xiao↗

Performance Modeling and Measurement of Parallelized Code for Distributed Shared Memory Multiprocessors

This paper presents a model to evaluate the performance and overhead of parallelizing sequential code using compiler directives for multiprocessing on distributed shared memory (DSM) systems. With increasing popularity of shared address space architectures, it is essential to understand their performance impact on programs that benefit from shared memory multiprocessing. We present a simple model to characterize the performance of programs that are parallelized using compiler directives for shared memory multiprocessing. We parallelized the sequential implementation of NAS benchmarks using native Fortran77 compiler directives for an Origin2000, which is a DSM system based on a cache-coherent Non Uniform Memory Access (ccNUMA) architecture. We report measurement based performance of these parallelized benchmarks from four perspectives: efficacy of parallelization process; scalability; parallelization overhead; and comparison with hand-parallelized and -optimized version of the same benchmarks. Our results indicate that sequential programs can conveniently be parallelized for DSM systems using compiler directives but realizing performance gains as predicted by the performance model depends primarily on minimizing architecture-specific data locality overhead.

Waheed, Abdul↗

Single- and Multiple-Objective Optimization with Differential Evolution and Neural Networks

Genetic and evolutionary algorithms have been applied to solve numerous problems in engineering design where they have been used primarily as optimization procedures. These methods have an advantage over conventional gradient-based search procedures became they are capable of finding global optima of multi-modal functions and searching design spaces with disjoint feasible regions. They are also robust in the presence of noisy data. Another desirable feature of these methods is that they can efficiently use distributed and parallel computing resources since multiple function evaluations (flow simulations in aerodynamics design) can be performed simultaneously and independently on ultiple processors. For these reasons genetic and evolutionary algorithms are being used more frequently in design optimization. Examples include airfoil and wing design and compressor and turbine airfoil design. They are also finding increasing use in multiple-objective and multidisciplinary optimization. This lecture will focus on an evolutionary method that is a relatively new member to the general class of evolutionary methods called differential evolution (DE). This method is easy to use and program and it requires relatively few user-specified constants. These constants are easily determined for a wide class of problems. Fine-tuning the constants will off course yield the solution to the optimization problem at hand more rapidly. DE can be efficiently implemented on parallel computers and can be used for continuous, discrete and mixed discrete/continuous optimization problems. It does not require the objective function to be continuous and is noise tolerant. DE and applications to single and multiple-objective optimization will be included in the presentation and lecture notes. A method for aerodynamic design optimization that is based on neural networks will also be included as a part of this lecture. The method offers advantages over traditional optimization methods. It is more flexible than other methods in dealing with design in the context of both steady and unsteady flows, partial and complete data sets, combined experimental and numerical data, inclusion of various constraints and rules of thumb, and other issues that characterize the aerodynamic design process. Neural networks provide a natural framework within which a succession of numerical solutions of increasing fidelity, incorporating more realistic flow physics, can be represented and utilized for optimization. Neural networks also offer an excellent framework for multiple-objective and multi-disciplinary design optimization. Simulation tools from various disciplines can be integrated within this framework and rapid trade-off studies involving one or many disciplines can be performed. The prospect of combining neural network based optimization methods and evolutionary algorithms to obtain a hybrid method with the best properties of both methods will be included in this presentation. Achieving solution diversity and accurate convergence to the exact Pareto front in multiple objective optimization usually requires a significant computational effort with evolutionary algorithms. In this lecture we will also explore the possibility of using neural networks to obtain estimates of the Pareto optimal front using non-dominated solutions generated by DE as training data. Neural network estimators have the potential advantage of reducing the number of function evaluations required to obtain solution accuracy and diversity, thus reducing cost to design.

Rai, Man Mohan↗