Search NASA⌕ Search

SEARCH · Search NASA

Results for “memory efficiency”

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 1,261 records · Page 70

Detecting Spin-Bath Polarization with Quantum Quench Phase Shifts of Single Spins in Diamond

Single-qubit sensing protocols can be used to measure qubit-bath coupling parameters. However, for sufficiently large coupling, the sensing protocol itself perturbs the bath, which is predicted to result in a characteristic response in the sensing measurements. Here, we observe this bath perturbation, also known as a quantum quench, by preparing the nuclear spin bath of a nitrogen-vacancy (NV) center in polarized initial states and performing phase-resolved spin-echo measurements on the NV electron spin. These measurements reveal a time-dependent phase determined by the initial state of the bath. We derive the relationship between the sensor phase and the Gaussian spin-bath polarization and apply it to reconstruct both the axial and transverse polarization components. Using this insight, we optimize the transfer efficiency of our dynamic nuclear polarization sequence. This technique for directly measuring bath polarization may assist in preparing high-fidelity quantum memory states, improving nanoscale NMR methods, and investigating non-Gaussian quantum baths.

36 MATERIALS SCIENCE↗

ENSIGN

ENSIGN is a data analytics software package offering a modern unsupervised machine learning solution for scalable discovery in Big Data. The analytics in ENSIGN are based on an advanced mathematical tool called tensor decomposition and they are optimized to run efficiently on a range of computing platforms (from small multicore Desktop platforms to large Supercomputing clusters and novel high-end memory-driven computing platforms such as HPE Superdome Flex). ENSIGN enables the user to extract deep insights from the entirety of massive-scale (100s of Gigabytes or Terabytes scale) multidimensional data. ENSIGN uncovers latent patterns in data without the user having to specify or describe what the patterns are; the user, in the first place, may not even know such patterns existed and that they have to look for such patterns. The insights gained from ENSIGN could be trailheads that can be used as starting points for deeper forensic investigation.

Baskaran, Muthu↗

ENSIGN

ENSIGN is a data analytics software package offering a modern unsupervised machine learning solution for scalable discovery in Big Data. The analytics in ENSIGN are based on an advanced mathematical tool called tensor decomposition and they are optimized to run efficiently on a range of computing platforms (from small multicore Desktop platforms to large Supercomputing clusters and novel high-end memory-driven computing platforms such as HPE Superdome Flex). ENSIGN enables the user to extract deep insights from the entirety of massive-scale (100s of Gigabytes or Terabytes scale) multidimensional data. ENSIGN uncovers latent patterns in data without the user having to specify or describe what the patterns are; the user, in the first place, may not even know such patterns existed and that they have to look for such patterns. The insights gained from ENSIGN could be trailheads that can be used as starting points for deeper forensic investigation.

Baskaran, Muthu↗

Main memory unit

The development of a main memory unit (MMU) for the space ultrareliable module computer (SUMC) model HTC is discussed. The design, fabrication, and test of basic memory modules (BMM) which were to be used in the design and construction of the MMU are described. The BMM was designed from state-of-the-art technologies which included large scale integration devices mounted and interconnected on a substrate to form a functional module to be utilized in the MMU development. A SUMC memory system design study is discussed which addressed itself to the BMM design and analysis to be conducted to determine the most efficient organization of the BMM in order to establish such modularity features as: word length expandability without redesign, high reliability, and fault tolerance. One MMU was designed, fabricated, tested, and delivered which will be electrical and mechanically compatible with the hybrid technology computer (HTC) model of the SUMC family of computers. The MMU will contain a storage capacity of 8196 36 bit words which includes a parity bit for each 8 bit byte of data.

Source record↗

Efficiently modeling neural networks on massively parallel computers

Neural networks are a very useful tool for analyzing and modeling complex real world systems. Applying neural network simulations to real world problems generally involves large amounts of data and massive amounts of computation. To efficiently handle the computational requirements of large problems, we have implemented at Los Alamos a highly efficient neural network compiler for serial computers, vector computers, vector parallel computers, and fine grain SIMD computers such as the CM-2 connection machine. This paper describes the mapping used by the compiler to implement feed-forward backpropagation neural networks for a SIMD (Single Instruction Multiple Data) architecture parallel computer. Thinking Machines Corporation has benchmarked our code at 1.3 billion interconnects per second (approximately 3 gigaflops) on a 64,000 processor CM-2 connection machine (Singer 1990). This mapping is applicable to other SIMD computers and can be implemented on MIMD computers such as the CM-5 connection machine. Our mapping has virtually no communications overhead with the exception of the communications required for a global summation across the processors (which has a sub-linear runtime growth on the order of O(log(number of processors)). We can efficiently model very large neural networks which have many neurons and interconnects and our mapping can extend to arbitrarily large networks (within memory limitations) by merging the memory space of separate processors with fast adjacent processor interprocessor communications. This paper will consider the simulation of only feed forward neural network although this method is extendable to recurrent networks.

Farber, Robert M.↗

Monitoring cognitive function and need with the automated neuropsychological assessment metrics in Decompression Sickness (DCS) research

Hypobaric decompression sickness (DCS) research presents the medical monitor with the difficult task of assessing the onset and progression of DCS largely on the basis of subjective symptoms. Even with the introduction of precordial Doppler ultrasound techniques for the detection of venous gas emboli (VGE), correct prediction of DCS can be made only about 65 percent of the time according to data from the Armstrong Laboratory's (AL's) hypobaric DCS database. An AL research protocol concerned with exercise and its effects on denitrogenation efficiency includes implementation of a performance assessment test battery to evaluate cognitive functioning during a 4-h simulated 30,000 ft (9144 m) exposure. Information gained from such a test battery may assist the medical monitor in identifying early signs of DCS and subtle neurologic dysfunction related to cases of asymptomatic, but advanced, DCS. This presentation concerns the selection and integration of a test battery and the timely graphic display of subject test results for the principal investigator and medical monitor. A subset of the Automated Neuropsychological Assessment Metrics (ANAM) developed through the Office of Military Performance Assessment Technology (OMPAT) was selected. The ANAM software provides a library of simple tests designed for precise measurement of processing efficiency in a variety of cognitive domains. For our application and time constraints, two tests requiring high levels of cognitive processing and memory were chosen along with one test requiring fine psychomotor performance. Accuracy, speed, and processing throughout variables as well as RMS error were collected. An automated mood survey provided 'state' information on six scales including anger, happiness, fear, depression, activity, and fatigue. An integrated and interactive LOTUS 1-2-3 macro was developed to import and display past and present task performance and mood-change information.

Nesthus, Thomas E.↗

Parallel adaptive mesh refinement within the PUMAA3D Project

To enable the solution of large-scale applications on distributed memory architectures, we are designing and implementing parallel algorithms for the fundamental tasks of unstructured mesh computation. In this paper, we discuss efficient algorithms developed for two of these tasks: parallel adaptive mesh refinement and mesh partitioning. The algorithms are discussed in the context of two-dimensional finite element solution on triangular meshes, but are suitable for use with a variety of element types and with h- or p-refinement. Results demonstrating the scalability and efficiency of the refinement algorithm and the quality of the mesh partitioning are presented for several test problems on the Intel DELTA.

Freitag, Lori↗

Runtime Detection of C-Style Errors in UPC Code

Unified Parallel C (UPC) extends the C programming language (ISO C 99) with explicit parallel programming support for the partitioned global address space (PGAS), which provides a global memory space with localized partitions to each thread. Like its ancestor C, UPC is a low-level language that emphasizes code efficiency over safety. The absence of dynamic (and static) safety checks allows programmer oversights and software flaws that can be hard to spot. In this paper, we present an extension of a dynamic analysis tool, ROSE-Code Instrumentation and Runtime Monitor (ROSECIRM), for UPC to help programmers find C-style errors involving the global address space. Built on top of the ROSE source-to-source compiler infrastructure, the tool instruments source files with code that monitors operations and keeps track of changes to the system state. The resulting code is linked to a runtime monitor that observes the program execution and finds software defects. We describe the extensions to ROSE-CIRM that were necessary to support UPC. We discuss complications that arise from parallel code and our solutions. We test ROSE-CIRM against a runtime error detection test suite, and present performance results obtained from running error-free codes. ROSE-CIRM is released as part of the ROSE compiler under a BSD-style open source license.

97 MATHEMATICS AND COMPUTING↗

Toward Performance Portable Programming for Heterogeneous System-on-Chips: Case Study with Qualcomm Snapdragon SoC

Future heterogeneous Domain-Specific System-on-Chips (DSSoC) will be extraordinarily complex in terms of processors, memory hierarchies, and interconnection networks.To manage this complexity, architects, system software designers, and application developers need programming technologies that are flexible, accurate, efficient, and productive. These technologies will need to be as independent of any one specific architecture as is practical, because the sheer dimensionality and scale of the complexity will not allow porting and optimizing applications foreach given DSSoC. To address these issues, we are developing Cosmic Castle, a performance portable programming toolchain for streaming applications on heterogeneous architectures. The primary focus of Cosmic Castle is on enabling efficient and performant code generation through the smart compiler and intelligent runtime system. This paper presents the preliminary evaluation of our ongoing work toward Cosmic Castle. Specifically, we detail our code porting efforts and evaluate various benchmarks on the Qualcomm Snapdragon SoC using tools developed through Cosmic Castle.

Cabrera, Anthony↗

Evaluation of Design Concepts for Collapsible Cryogenic Storage Vessels

Future long-duration missions to Mars using in situ resource production to obtain oxygen from the Martian atmosphere for use as a propellant or for life support will require long term oxygen storage facilities. This report describes preliminary analysis of design concepts for lightweight, collapsible liquid oxygen storage tanks to be used on the surface of Mars. With storage at relatively low pressures, an inflatable tank concept in which the cryogen is stored within a fiber-reinforced Teflon FEP bladder is an efficient approach. The technology required for such a tank is well-developed through similar previous applications in positive expulsion bladders for zero-g liquid fuel rocket tanks and inflatable space habitat technology, though the liquid oxygen environment presents unique challenges. The weight of the proposed structure is largely dominated by the support structure needed to hold the tank off the ground and permit a vacuum insulation space to be maintained around the tank. In addition to the inflatable tank concept, telescoping tank concepts are studied. For a telescoping tank, the greatest difficulty is in making effective joints and seals. The use of shape memory alloy to produce a passive clamping ring is evaluated. Although the telescoping tank concepts are a viable option, it appears that inflatable tank concepts will be more efficient and are recommended.

Fleming, David C.↗

A new era of ferroelectric thin films for nonvolatile memories

Ferroelectric films have potential applications in nonvolatile memory devices. In addition to the well-established perovskite-structure oxide ferroelectrics, hafnia- and wurtzite-based ferroelectrics have recently attracted considerable attention because of the improved scalability of the ferroelectric response (down to nanometer thicknesses), their compatibility with silicon fabrication processes, and the availability of deposition methods that realize three-dimensional structures. Furthermore, due to their high compatibility with silicon processes, these are also expected to be used in emerging energy-efficient applications, such as neuromorphic computing and reservoir computing. In this article, we review the fundamentals of hafnia- and wurtzite-based ferroelectrics and their advantages and issues for developing in nonvolatile memory devices. Then, possible emerging applications are discussed.

36 MATERIALS SCIENCE↗

Parallel Grid Manipulations in Earth Science Calculations

The National Aeronautics and Space Administration (NASA) Data Assimilation Office (DAO) at the Goddard Space Flight Center is moving its data assimilation system to massively parallel computing platforms. This parallel implementation of GEOS DAS will be used in the DAO's normal activities, which include reanalysis of data, and operational support for flight missions. Key components of GEOS DAS, including the gridpoint-based general circulation model and a data analysis system, are currently being parallelized. The parallelization of GEOS DAS is also one of the HPCC Grand Challenge Projects. The GEOS-DAS software employs several distinct grids. Some examples are: an observation grid- an unstructured grid of points at which observed or measured physical quantities from instruments or satellites are associated- a highly-structured latitude-longitude grid of points spanning the earth at given latitude-longitude coordinates at which prognostic quantities are determined, and a computational lat-lon grid in which the pole has been moved to a different location to avoid computational instabilities. Each of these grids has a different structure and number of constituent points. In spite of that, there are numerous interactions between the grids, e.g., values on one grid must be interpolated to another, or, in other cases, grids need to be redistributed on the underlying parallel platform. The DAO has designed a parallel integrated library for grid manipulations (PILGRIM) to support the needed grid interactions with maximum efficiency. It offers a flexible interface to generate new grids, define transformations between grids and apply them. Basic communication is currently MPI, however the interfaces defined here could conceivably be implemented with other message-passing libraries, e.g., Cray SHMEM, or with shared-memory constructs. The library is written in Fortran 90. First performance results indicate that even difficult problems, such as above-mentioned pole rotation- a sparse interpolation with little data locality between the physical lat-lon grid and a pole rotated computational grid- can be solved efficiently and at the GFlop/s rates needed to solve tomorrow's high resolution earth science models. In the subsequent presentation we will discuss the design and implementation of PILGRIM as well as a number of the problems it is required to solve. Some conclusions will be drawn about the potential performance of the overall earth science models on the supercomputer platforms foreseen for these problems.

Sawyer, W.↗

Improving Block-level Efficiency with scsi-mq

Current generation solid-state storage devices are exposing a new bottlenecks in the SCSI and block layers of the Linux kernel, where IO throughput is limited by lock contention, inefficient interrupt handling, and poor memory locality. To address these limitations, the Linux kernel block layer underwent a major rewrite with the blk-mq project to move from a single request queue to a multi-queue model. The Linux SCSI subsystem rework to make use of this new model, known as scsi-mq, has been merged into the Linux kernel and work is underway for dm-multipath support in the upcoming Linux 4.0 kernel. These pieces were necessary to make use of the multi-queue block layer in a Lustre parallel filesystem with high availability requirements. We undertook adding support of the 3.18 kernel to Lustre with scsi-mq and dm-multipath patches to evaluate the potential of these efficiency improvements. In this paper we evaluate the block-level performance of scsi-mq with backing storage hardware representative of a HPC-targerted Lustre filesystem. Our findings show that SCSI write request latency is reduced by as much as 13.6%. Additionally, when profiling the CPU usage of our prototype Lustre filesystem, we found that CPU idle time increased by a factor of 7 with Linux 3.18 and blk-mq as compared to a standard 2.6.32 Linux kernel. Our findings demonstrate increased efficiency of the multi-queue block layer even with disk-based caching storage arrays used in existing parallel filesystems.

Cluster Computing↗

Antiferroelectric Ceramics for Energy–Efficient Capacitors by Theory–Guided Discovery

Antiferroelectric ceramics, via the electric-field-induced antiferroelectric (AFE)–ferroelectric (FE) phase transitions, show great promise for high-energy-density capacitors. Yet, currently, only 70–80% energy release is found during a charge–discharge cycle. Here, for PbZrO 3 -based oxides, geometric nonlinear theory of martensitic phase transitions is applied (first used to guide supercompatible shape-memory alloys) to predict the reversibility of the AFE–FE transition by using density-functional theory to assess AFE/FE interfacial lattice-mismatch strain that assures ultralow electric hysteresis and extended fatigue lifetime. A good correlation of mismatch strain with electric hysteresis, hence, with energy efficiency of AFE capacitors is observed. Here, guided by theory, high-throughput material search is conducted and AFE compositions with a near-perfect charge–discharge energy efficiency (98.2%), i.e., near-zero hysteresis are discovered. And the fatigue life of the capacitor reaches 79.5 million charge–discharge cycles, a factor of 80 enhancement over AFE ceramics with large electric hysteresis.

36 MATERIALS SCIENCE↗

Probabilistic Neural Computing with Stochastic Devices

Abstract The brain has effectively proven a powerful inspiration for the development of computing architectures in which processing is tightly integrated with memory, communication is event‐driven, and analog computation can be performed at scale. These neuromorphic systems increasingly show an ability to improve the efficiency and speed of scientific computing and artificial intelligence applications. Herein, it is proposed that the brain's ubiquitous stochasticity represents an additional source of inspiration for expanding the reach of neuromorphic computing to probabilistic applications. To date, many efforts exploring probabilistic computing have focused primarily on one scale of the microelectronics stack, such as implementing probabilistic algorithms on deterministic hardware or developing probabilistic devices and circuits with the expectation that they will be leveraged by eventual probabilistic architectures. A co‐design vision is described by which large numbers of devices, such as magnetic tunnel junctions and tunnel diodes, can be operated in a stochastic regime and incorporated into a scalable neuromorphic architecture that can impact a number of probabilistic computing applications, such as Monte Carlo simulations and Bayesian neural networks. Finally, a framework is presented to categorize increasingly advanced hardware‐based probabilistic computing technologies.

Misra, Shashank↗

A deep learning-based Bayesian framework for high-resolution calibration of building energy models

Calibrating building energy models (BEMs), i.e., closing discrepancy between modeling and field measurements, is of significance to support its applications in building sustainability and resilience analysis. However, as being widely used in practice, current Bayesian calibration is mostly performed in low-resolution (annual or monthly), instead of high-resolution (hourly or sub-hourly), which is crucial to support emerging BEM applications, such as building-renewable energy integration (demand response) and smart control. This is attributable to the gaps in current Bayesian calibration process, including (1) difficulty in supporting reliable high-resolution calibration with over-parameterization and multi-solution issues, (2) inadequacy of meta-model to capture temporal building dynamics in high-resolution, and (3) excessive computational burdens of covariance matrix calculation in Bayesian inference. Therefore, to close these gaps, this research proposes a novel deep learning-based Bayesian calibration framework, involving pre-calibration mechanism, Long Short-Term Memory as surrogate models, and simplified covariance matrix calculation, to calibrate BEMs in high temporal resolution (i.e., hourly) with enhanced accuracy and computational efficiency. Finally, the case study demonstrates its effectiveness to match modeling outcomes with measurements and realize CV-RMSE of < 30 % and NMBE of < 6 % in hourly resolution, as well as a significant reduction of calibration time (by > 99 %, from > 600 h to ~ 1.5 h).

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Scaling Deep Learning workloads: NVIDIA DGX-1/Pascal and Intel Knights Landing

Deep Learning (DL) algorithms have become ubiquitous in data analytics. As a result, major computing vendors—including NVIDIA, Intel, AMD, and IBM—have architectural road maps influenced by DL workloads. Furthermore, several vendors have recently advertised new computing products as accelerating large DL workloads. Unfortunately, it is difficult for data scientists to quantify the potential of these different products. Here, this article provides a performance and power analysis of important DL workloads on two major parallel architectures: NVIDIA DGX-1 (eight Pascal P100 GPUs interconnected with NVLink) and Intel Knights Landing (KNL) CPUs interconnected with Intel Omni-Path or Cray Aries. Our evaluation consists of a cross section of convolutional neural net workloads: CifarNet, AlexNet, GoogLeNet, and ResNet50 topologies using the Cifar10 and ImageNet datasets. The workloads are vendor-optimized for each architecture. We use sequentially equivalent implementations to maintain iso-accuracy between parallel and sequential DL models. Our analysis indicates that although GPUs provide the highest overall performance, the gap can close for some convolutional networks; and the KNL can be competitive in performance/watt. We find that NVLink facilitates scaling efficiency on GPUs. However, its importance is heavily dependent on neural network architecture. Furthermore, for weak-scaling—sometimes encouraged by restricted GPU memory—NVLink is less important.

97 MATHEMATICS AND COMPUTING↗

A fast band–Krylov eigensolver for macromolecular functional motion simulation on multicore architectures and graphics processors

We introduce a new iterative Krylov subspace-based eigensolver for the simulation of macromolecular motions on desktop multithreaded platforms equipped with multicore processors and, possibly, a graphics accelerator (GPU). The method consists of two stages, with the original problem first reduced into a simpler band-structured form by means of a high-performance compute-intensive procedure. This is followed by a memory-intensive but low-cost Krylov iteration, which is off-loaded to be computed on the GPU by means of an efficient data-parallel kernel. The experimental results reveal the performance of the new eigensolver. Concretely, when applied to the simulation of macromolecules with a few thousands degrees of freedom and the number of eigenpairs to be computed is small to moderate, the new solver outperforms other methods implemented as part of high-performance numerical linear algebra packages for multithreaded architectures.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗