Search NASA⌕ Search

SEARCH · Search NASA

Results for “hardware accelerators”

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 487 records · Page 27

Program Development Tools and Infrastructures: Activities by ASC CSSE/CCE and Partners

Exascale class machines will exhibit a new level of complexity: they will feature an unprecedented number of cores and threads, will most likely be heterogeneous and deeply hierarchical, and offer a range of new hardware techniques (such as speculative threading, transactional memory, programmable prefetching, and programmable accelerators), which all have to be utilized for an application to realize the full potential of the machine. Additionally, users will be faced with less memory per core, fixed total power budgets, and sharply reduced MTBFs. At the same time, it is expected that the complexity of applications will rise sharply for exascale systems, both to implement new science possible at exascale and to exploit the new hardware features necessary to achieve exascale performance. This is particularly true for many of the NNSA codes, which are large and often highly complex integrated simulation codes that push the limits of everything in the system including language features. To overcome these limitations and to enable users to reach exascale performance, users will expect a new generation of tools that address the bottlenecks of exascale machines, that work seamlessly with the (set of) programming models on the target machines, that scale with the machine, that provide automatic analysis capabilities, and that are flexible and modular enough to overcome the complexities and changing demands of the exascale architectures. Further, any tool must be robust enough to handle the complexity of large integrated codes while keeping the user's learning curve low. With the ASC program, in particular the CSSE (Computational Systems and Software Engineering) and CCE (Common Compute Environment) projects, we are working towards a new generation of tools that fulfill these requirements and that provide our users as well as the larger HPC community with the necessary tools, techniques, and methodologies required to make exascale performance a reality.

43 PARTICLE ACCELERATORS↗

Data-driven analysis to understand GPU hardware resource usage of optimizations

With heterogeneous systems, the number of GPUs per chip increases to provide computational capabilities for solving science at a nanoscopic scale. However, low utilization for single GPUs defies the need to invest more money in expensive accelerators. Although related work develops optimizations to improve application performance, none studies how these optimizations impact hardware resource usage or average GPU utilization. Here, this paper takes a data-driven analysis approach in addressing this gap by (1) characterizing how hardware resource usage affects device utilization, execution time, or both, (2) presenting a multiobjective metric to identify important application-device interactions that can be optimized to improve device utilization and application performance jointly, (3) studying hardware resource usage behaviors of several optimizations for a benchmark application, and finally (4) identifying optimization opportunities for several scientific proxy applications based on their hardware resource usage behaviors. Furthermore, we demonstrate the applicability of our methodology by applying the identified optimizations to a proxy application, which improves the execution time, device utilization, and power consumption by up to 29.6%, 5.3% and 26.5% respectively.

Computer science↗

Integration of Ag-CBRAM crossbars and Mott ReLU neurons for efficient implementation of deep neural networks in hardware

In-memory computing with emerging non-volatile memory devices (eNVMs) has shown promising results in accelerating matrix-vector multiplications. However, activation function calculations are still being implemented with general processors or large and complex neuron peripheral circuits. Here, we present the integration of Ag-based conductive bridge random access memory (Ag-CBRAM) crossbar arrays with Mott rectified linear unit (ReLU) activation neurons for scalable, energy and area-efficient hardware (HW) implementation of deep neural networks. We develop Ag-CBRAM devices that can achieve a high ON/OFF ratio and multi-level programmability. Compact and energy-efficient Mott ReLU neuron devices implementing ReLU activation function are directly connected to the columns of Ag-CBRAM crossbars to compute the output from the weighted sum current. We implement convolution filters and activations for VGG-16 using our integrated HW and demonstrate the successful generation of feature maps for CIFAR-10 images in HW. Our approach paves a new way toward building a highly compact and energy-efficient eNVMs-based in-memory computing system.

Mott insulators↗

Optimizing the Fast Fourier Transform Using Mixed Precision on Tensor Core Hardware

The Fast Fourier Transform is a fundamental tool in scientific and technical computation. The highly parallelizable nature of the algorithm makes it a suitable candidate for GPU acceleration. This paper focuses on exploiting the speedup due to using the half precision multiplication capability of the latest GPUs' tensor core hardware without significantly degrading the precision of the Fourier Transform result. We develop an algorithm that dynamically splits the input single precision dataset into two half precision sets at the lowest level, uses half precision multiplication, and recombines the result at a later step. This work paves the way for using tensor cores for high precision inputs.

Sorna, Anumeena↗

Computational Offload with BlueField Smart NICs

The recent introduction of a new generation of "smart NICs" have provided new accelerator platforms that include CPU cores or reconfigurable fabric in addition to traditional networking hardware and packet offloading capabilities. While there are currently several proposals for using these smartNICs for low-latency, in-line packet processing operations, there remains a gap in knowledge as to how they might be used as computational accelerators for traditional high-performance applications. This work aims to look at benchmarks and mini-applications to evaluate possible benefits of using a smartNIC as a compute accelerator for HPC applications. We investigate NVIDIA's current-generation BlueField-2 card, which includes eight Arm CPUs along with a small amount of storage, and we test the networking and data movement performance of these cards compared to a standard Intel server host. We then detail how two different applications, YASK and miniMD can be modified to make more efficient use of the BlueField-2 device with a focus on overlapping computation and communication for operations like neighbor building and halo exchanges. Our results show that while the overall compute performance of these devices is limited, using them with a modified miniMD algorithm allows for potential speedups of 5 to 20% over the host CPU baseline with no loss in simulation accuracy.

97 MATHEMATICS AND COMPUTING↗

AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing

The recent development of deep learning has been mostly focusing on Euclidean data, such as images, videos, audios, etc. However, most real-world information and relation are often expressed as graphs. To efficiently learn from graph data, graph convolutional networks (GCNs) emerge as a promising approach, showing advantages in several practical applications such as social network analysis, knowledge discovery, 3D modeling, motion capturing, etc. Real-world graphs are usually extremely large and imbalanced, posting significant performance demand and design challenges on the hardware dedicated for GCN inference. In this paper, we propose an architecture design called UW-GCN to accelerate graph convolutional network inference. To tackle the major performance bottleneck from workload imbalance, we propose dynamic neighborhood stealing and remote chunk shuffling techniques, relying on hardware flexibility to achieve hardware auto-tuning under negligible area or delay overhead. Specifically, UW-GCN is able to smartly profile the sparse graph pattern while continuously adjusting the workload distribution via routing reconfiguration among parallel processing elements (PEs). The ideal configuration is then reused in the remaining iterations. To the best of our knowledge, this is the first accelerator design particularly for GCN and the first work relying on hardware auto-tuning, which is normally based on software, to achieve near-optimal workload balance in processing sparse structures.

Geng, Tong↗

Investigation and Development of Control Laws for the NASA Langley Research Center Cockpit Motion Facility

The ability to develop highly advanced simulators is a critical need that has the ability to significantly impact the aerospace industry. The aerospace industry is advancing at an ever increasing pace and flight simulators must match this development with ever increasing urgency. In order to address both current problems and potential advancements with flight simulator techniques, several aspects of current control law technology of the National Aeronautics and Space Administration (NASA) Langley Research Center's Cockpit Motion Facility (CMF) motion base simulator were examined. Preliminary investigation of linear models based upon hardware data were examined to ensure that the most accurate models are used. This research identified both system improvements in the bandwidth and more reliable linear models. Advancements in the compensator design were developed and verified through multiple techniques. The position error rate feedback, the acceleration feedback and the force feedback were all analyzed in the heave direction using the nonlinear model of the hardware. Improvements were made using the position error rate feedback technique. The acceleration feedback compensator also provided noteworthy improvement, while attempts at implementing a force feedback compensator proved unsuccessful.

Coon, Craig R.↗

Development of intelligent instruments with embedded HTTP servers for control and data acquisition in a cryogenic setup—The hardware, firmware, and software implementation

The power of Ethernet for control and automation technology is being largely understood by the automation industry in recent times. Ethernet with HTTP (Hypertext Transfer Protocol) is one of the most widely accepted communication standards today. Ethernet is best known for being able to control through internet from anywhere in the globe. The Ethernet interface with built-in on-chip embedded servers ensures global connections for crate-less model of control and data acquisition systems which have several advantages over traditional crate-based control architectures for slow applications. This architecture will completely eliminate the use of any extra PLC (Programmable Logic Controller) or similar control hardware in any automation network as the control functions are firmware coded inside intelligent meters itself. Here, we describe the indigenously built project of a cryogenic control system built for linear accelerator at Inter University Accelerator Centre, known as “CADS,” which stands for “Complete Automation of Distribution System.” CADS deals with complete hardware, firmware, and software implementation of the automated linac cryogenic distribution system using many Ethernet based embedded cryogenic instruments developed in-house. Each instrument works as an intelligent meter called device-server which has the control functions and control loops built inside the firmware itself. Dedicated meters with built-in servers were designed out of ARM (Acorn RISC (Reduced Instruction Set Computer) Machine) and ATMEL processors and COTS (Commercially Off-the-Shelf) SMD (Surface Mount Devices) components, with analog sensor front-end and a digital back-end web server implementing remote procedure call over HTTP for digital control and readout functions. At present, 24 instruments which run 58 embedded servers inside, each specific to a particular type of sensor-actuator combination for closed loop operations, are now deployed and distributed across control LAN (Local Area Network). A group of six categories of such instruments have been identified for all cryogenic applications required for linac operation which were designed to build this medium-scale cryogenic automation setup. These devices have special features like remote rebooters, daughter boards for PIDs (Proportional Integral Derivative), etc., to operate them remotely in radiation areas and also have emergency switches by which each device can be taken to emergency mode temporarily. Finally, all the data are monitored, logged, controlled, and analyzed online at a central control room which has a user-friendly control interface developed using LabVIEW{sup ®}. This paper discusses the overall hardware, firmware, software design, and implementation for the cryogenics setup.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

XACC - eXtreme-scale Accelerator Programming Framework

Hybrid programming models for beyond-CMOS technologies will prove critical for integrating new computing technologies alongside our existing infrastructure. Unfortunately the software infrastructure required to enable this is lacking or not available. XACC is a programming framework for extreme-scale, post-exascale accelerator architectures that integrates alongside existing conventional applications. It is a pluggable framework for programming languages developed for next-gen computing hardware architectures like quantum and neuromorphic computing. It lets computational scientists efficiently off-load classically intractable work to attached accelerators through user-friendly Kernel definitions. XACC makes post-exascale hybrid programming approachable for domain computational scientists.

McCaskey, AlexanderJ.↗

XACC - eXtreme-scale Accelerator Programming Framework

Hybrid programming models for beyond-CMOS technologies will prove critical for integrating new computing technologies alongside our existing infrastructure. Unfortunately the software infrastructure required to enable this is lacking or not available. XACC is a programming framework for extreme-scale, post-exascale accelerator architectures that integrates alongside existing conventional applications. It is a pluggable framework for programming languages developed for next-gen computing hardware architectures like quantum and neuromorphic computing. It lets computational scientists efficiently off-load classically intractable work to attached accelerators through user-friendly Kernel definitions. XACC makes post-exascale hybrid programming approachable for domain computational scientists.

McCaskey, Alexander J.↗

Center of Gravity Quasi-Static Load Estimate Using Random Vibration Responses

The quasi-static load (QSL) is often levied on flight hardware structural designers as a requirement. This requirement is verified either by test or analysis with engineers using different approaches to meet this requirement. A commonly used approach is to consider the rms acceleration Center of Gravity (CG) responses of test articles undergoing random vibration tests. The differences between structural engineers and/or test conductors within the space industry exist on how the data obtained from flight hardware random vibration tests are used to estimate QSLs. In this paper, a method to accurately estimate the CG grms from the acceleration power spectral densities with an appropriate sigma value to estimate peak CG accelerations is discussed. Recommendations are made to ensure that the community follows consistent QSL estimation from random vibration test data. The method of estimating the CG acceleration responses may also be used to response limit if the flight hardware responses exceed flight limit loads with margins.

Kolaini, Ali↗

The EMMA Non-Scaling FFAG

The Electron Model for Many Applications (EMMA) will be the World's first non-scaling FFAG and is under construction at the STFC Daresbury Laboratory in the UK. Construction is due for completion in Summer 2010 and will be followed by commissioning with beam and a detailed experimental programme to study the functioning of this type of accelerator. This paper will give an overview of the motivation for the project and describe the EMMA design and hardware. The first results from commissioning will be presented in a separate paper.

43 PARTICLE ACCELERATORS↗

I ntera C hem : Exploring Excited States in Virtual Reality with Ab Initio Interactive Molecular Dynamics

InteraChem is an ab initio interactive molecular dynamics (AI-IMD) visualizer that leverages recent advances in virtual reality hardware and software, as well as the graphical processing unit (GPU)-accelerated TeraChem electronic structure package, in order to render quantum chemistry in real time. We introduce the exploration of electronically excited states via AI-IMD using the floating occupation molecular orbital-complete active space configuration interaction method. The optimization tools in InteraChem enable identification of excited state minima as well as minimum energy conical intersections for further characterization of excited state chemistry in small- to medium-sized systems. We demonstrate that finite-temperature Hartree–Fock theory is an efficient method to perform ground state AI-IMD. InteraChem allows users to track electronic properties such as molecular orbitals and bond order in real time, resulting in an interactive visualization tool that aids in the interpretation of excited state chemistry data and makes quantum chemistry more accessible for both research and educational purposes.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Multi-frequency recirculating planar magnetrons

The multi-frequency recirculating planar magnetron (MFRPM) is the first magnetron capable of simultaneous generation of significantly different output frequencies (1 and 2 GHz) in a single operating pulse. Design and simulation of a prototype MFRPM were followed by hardware fabrication and experimental verification using the Michigan Electron Long Beam Accelerator with a Ceramic insulator at −300 kV, 1–5 kA, and 0.14–0.23 T axial magnetic field. Preliminary results demonstrated simultaneous generation of microwave pulses near 1 GHz and 2 GHz at powers up to 44 MW and 21 MW, respectively, with peak total efficiencies up to 9%.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

A Case For Intra-rack Resource Disaggregation in HPC

The expected halt of traditional technology scaling is motivating increased heterogeneity in high-performance computing (HPC) systems with the emergence of numerous specialized accelerators. As heterogeneity increases, so does the risk of underutilizing expensive hardware resources if we preserve today’s rigid node configuration and reservation strategies. This has sparked interest in resource disaggregation to enable finer-grain allocation of hardware resources to applications. However, there is currently no data-driven study of what range of disaggregation is appropriate in HPC. To that end, we perform a detailed analysis of key metrics sampled in NERSC’s Cori, a production HPC system that executes a diverse open-science HPC workload. In addition, we profile a variety of deep-learning applications to represent an emerging workload. We show that for a rack (cabinet) configuration and applications similar to Cori, a central processing unit with intra-rack disaggregation has a 99.5% probability to find all resources it requires inside its rack. In addition, ideal intra-rack resource disaggregation in Cori could reduce memory and NIC resources by 5.36% to 69.01% and still satisfy the worst-case average rack utilization.

97 MATHEMATICS AND COMPUTING↗

Calibration and Measurement techniques in the LLRF systems of the Fermilab PIP-II Linac

he PIP-II Accelerator is an 800 MeV superconducting Linac in the injection chain of the Fermilab accelerator complex. The LLRF systems are a based on two different hardware platforms controlling a variety of cavity types and resonance control systems including temperature, pneumatic and piezzo tuners. The various calibrations required prior to beam operation include, signal power, gradient, amplifier characterization, cavity Q measurement and detune constants. Measurements such as piezzo capacitance, cavity piezo transfer function help in determining tuner health and in devising microphonics control strategies. These measurement and calibration methods of the PIP-II LLRF system are discussed here.

Varghese, P.↗

Computational fluid dynamics analysis of SSME phase 2 and phase 2+ preburner injector element hydrogen flow paths

Phase 2+ Space Shuttle Main Engine powerheads, E0209 and E0215 degraded their main combustion chamber (MCC) liners at a faster rate than is normal for phase 2 powerheads. One possible cause of the accelerated degradation was a reduction of coolant flow through the MCC. Hardware changes were made to the preburner fuel leg which may have reduced the resistance and, therefore, pulled some of the hydrogen from the MCC coolant leg. A computational fluid dynamics (CFD) analysis was performed to determine hydrogen flow path resistances of the phase 2+ fuel preburner injector elements relative to the phase 2 element. FDNS was implemented on axisymmetric grids with the hydrogen assumed to be incompressible. The analysis was performed in two steps: the first isolated the effect of the different inlet areas and the second modeled the entire injector element hydrogen flow path.

Ruf, Joseph H.↗