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 433 records · Page 24

Quantum Circuit Partitioning for Scalable Noise-Aware Quantum Circuit Re-Synthesis

Re-synthesis techniques are utilized to optimize the quantum circuit. To enable scalable re-synthesis a divide-and-conquer approach is adopted that partitions the circuit into smaller blocks, which are optimized independently. Several algorithms have been proposed to minimize the block number while maximizing the gate count of each block. However, they vary in their performance and may not yield the highest output fidelity. We propose a reinforcement learning-based quantum circuit partitioning framework that incorporates the physical properties of the quantum hardware to maximize the output fidelity post-quantum circuit optimization. To accelerate the training, we also propose a noise injection method that enables on-the-fly optimization in the reinforcement learning environment, independent of the adopted optimization/re-synthesis method at the block level. We evaluate our approach compared to different partitioning techniques using various quantum benchmarks executed on IBM Q Hanoi quantum computer.

Charrwi, Mohammad Walid↗

Rotary Valve & Beamline Highlights for Fiscal Year 2017

This Fiscal Year (FY) work was divided between continued testing and characterization work of the Rotary Valve (RV) and mechanical engineering support for the beamline hardware stands. This configuration is more like the final setup with the accelerator firing deuterons down the evacuated beamline toward the RV for interaction with the deuterium and neutron production. The beamline cells were part of an experiment to reduce the impact that RV gas would have on the beamline vacuum. This work will be reported separately from this report. Previous testing had been with the beamline at atmospheric pressure and now the goal was to get test results of the RV with it connected to a beamline that’s running at some level of vacuum.

42 ENGINEERING↗

A NEW DIFFERENTIAL AND ERRANT BEAM CURRENT MONITOR FOR THE SNS* ACCELERATOR

A new Differential and errant Beam Current Monitor (DBCM) is being implemented for both the Spallation Neutron Source's Medium Energy Beam Transport (MEBT) and the Super Conducting Linac (SCL) accelerator sections. These new current monitors will abort the beam when the difference between two toroidal pickups exceeds a threshold. The MEBT DBCM will protect the MEBT chopper target, while the SCL DBCM will abort beam to minimize fast beam losses in the SCL cavities. The new DBCM will also record instances of errant beam, such as beam dropouts, to assist in further optimization of the SNS Accelerator. A software Errant Beam Monitor was implemented on the regular BCM hardware to study errant beam pulses. The new system will take over this functionality and will also be able to abort beam on pulse-to-pulse variations. Because the system is based on the FlexRIO hardware and programmed in LabVIEW FPGA, it will be able to abort beam in about 5 us. This paper describes the development, implementation, and initial test results of the DBCM, as well as errant beam examples.

beam current monitor↗

Accelerating Lossy and Lossless Compression on Emerging BlueField DPU Architectures

Data compression has become a crucial technique in addressing performance bottlenecks caused by increasing data volumes in High-Performance Computing (HPC), Big Data, and Deep Learning (DL). Despite its potential to boost system performance, recent studies have identified significant challenges with existing compression methods, mainly due to their high computational demands amidst continuously growing data sizes. Concurrently, the advent of Data Processing Units (DPUs), equipped with programmable System-on-Chip (SoC) and specialized compression accelerators, offers a promising opportunity to alter the landscape of data compression. This paper explores the complexities and potential of leveraging NVIDIA BlueField DPUs to accelerate lossy and lossless compression. Towards this, we introduce PEDAL, an innovative library that leverages the hardware capabilities of DPUs to unify and optimize data compression designs. Moreover, we seamlessly co-design PEDAL with the popular MPICH MPI library, demonstrating up to 101x speedup in compression time and 88x decrease in communication latency. Drawing on these achievements, we share our experience with various research communities about accelerating data compression on DPUs in communication-oriented HPC scenarios.

Li, Yuke↗

wa-hls4ml and lui-gnn: A benchmark and GNN-based surrogate model for hls4ml resource and latency estimation

As machine learning (ML) increasingly serves as a tool for addressing real-time challenges in scientific applications, the development of advanced tooling has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as model synthesis, are now becoming limiting factors in the rapid iteration of designs. To reduce these emerging constraints, multiple efforts are being launched toward designing an ML-based surrogate model that estimates resource usage of synthesized accelerator architectures. This model would reduce the design iteration time, especially when designing within a set of given hardware constraints. This approach shows considerable potential, but as it stands, the effort is early and would benefit from coordination and standardization to assist future work as it emerges. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of more than 100,000 fully connected neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. In addition to the resource utilization and latency data provided, the dataset includes generated artifacts and log files for many of the synthesized neural networks, in order to support future research in ML-based code generation. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, as well as the average performance across a subset of the dataset. We measure the performance of a given predictor model through multiple metrics, including $R^2$ score and SMAPE on regression tasks, as well as inference time to further characterize the estimator under test. Additionally, we introduce the latency/utilization inference graph neural network (lui-gnn), a surrogate model that uses a graph neural network to represent input architectures in the form of a directed graph. This graph representation allows for a diverse set of model architectures to all be effectively handled by a surrogate model. We present the architecture and performance of the model, as evaluated by the new proposed benchmark, including SMAPE, $R^2$ score, and inference times, and find that lui-gnn generally predicts latency and utilization for the 75\% quantile within several percent of the synthesized resources on the synthetic test dataset, indicating that this approach of estimating resource and latency via a surrogate models has promise and warrants further research.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

3D-ReG: A 3D ReRAM-based Heterogeneous Architecture for Training Deep Neural Networks

Deep neural network (DNN) models are being expanded to a broader range of applications. The computational capability of traditional hardware platforms cannot accommodate the growth of model complexity. Among recent technologies to accelerate DNN, resistive memory (ReRAM)-based processing-in-memory (PIM) emerged as a promising solution for DNN inference due to its high efficiency for matrix-based computation. We face two major technical challenges in extending the use of ReRAM-based accelerators for training: (1) full-precision data is essential in back-propagation; (2) the need to support both feed-forward and back-propagation aggravates the data-movement burden. We propose a heterogeneous architecture named as 3D-ReG, which leverages full-precision GPU to ensure training accuracy and low-overhead 3D integration to provide low-cost data movements. Moreover, we introduce conservative and aggressive task-mapping schemes, which partition the computation phases in different ways to balance execution efficiency and training accuracy. We evaluate 3D-ReG implemented with two 3D integration technologies, through-silicon vias (TSVs) and monolithic inter-tier vias (MIVs), and compare them with GPU-only and PIM-only counterparts. Various GPU-only platforms using two main-memory technologies (DRAM, ReRAM) and three interconnect technologies (2D, TSV, MIV) are evaluated as well. Experimental results show that 3D-ReG can achieve on average 5.64× training speedup and 3.56× higher energy efficiency compared with the GPU with DRAM as main memory, at the cost of 0.05%–3.39% accuracy drop. We define a new metric, gain-loss ratio (GLR), which quantitatively evaluates the capability of a DNN training hardware in terms of the model accuracy and hardware efficiency. The results of our comparison show that the aggressive task-mapping scheme on MIV-based 3D-ReG outperforms the other methods.

Computer Science↗

Python for Development of OpenMP and CUDA Kernels for Multidimensional Data

Design of data structures for high performance computing (HPC) is one of the principal challenges facing researchers looking to utilize heterogeneous computing machinery. Heterogeneous systems derive cost, power, and speed efficiency by being composed of the appropriate hardware for the task. Yet, each type of processor requires a specific organization of the application state in order to achieve peak performance. Discovering this and refactoring the code can be a challenging and time-consuming task for the researcher, as the data structures and the computational model must be co-designed. We present a methodology that uses Python as the environment for which to explore tradeoffs in both the data structure design as well as the code executing on the computation accelerator. Our method enables multi-dimensional arrays to be used effectively in any target environment. We have chosen to focus on OpenMP and CUDA environments, thus exploring the development of optimized kernels for the two most common classes of computing hardware available today: multi-core CPU and GPU. Python s large palette of file and network access routines, its associative indexing syntax and support for common HPC environments makes it relevant for diverse hardware ranging from laptops through computing clusters to the highest performance supercomputers. Our work enables researchers to accelerate the development of their codes on the computing hardware of their choice.

43 PARTICLE ACCELERATORS↗

GronOR: Massively Parallel and GPU-Accelerated Non-Orthogonal Configuration Interaction for Large Molecular Systems

GronOR is a program package for non-orthogonal configuration interaction calculations for an electronic wave function built in terms of anti-symmetrized products of multi-configuration molecular fragment wave functions. The two-electron integrals that have to be processed may be expressed in terms of atomic orbitals or in terms of an orbital basis determined from the molecular orbitals of the fragments. The code has been specifically designed for execution on distributed memory massively parallel and Graphics Processing Unit (GPU)-accelerated computer architectures, using an MPI+OpenACC/OpenMP programming approach. The task-based execution model used in the implementation allows for linear scaling with the number of nodes on the largest pre-exascale architectures available, provides hardware fault resiliency, and enables effective execution on systems with distinct central processing unit-only and GPU-accelerated partitions. The code interfaces with existing multi-configuration electronic structure codes that provide optimized molecular fragment orbitals, configuration interaction coefficients, and the required integrals. Algorithm and implementation details, parallel and accelerated performance benchmarks, and an analysis of the sensitivity of the accuracy of results and computational performance to thresholds used in the calculations are presented.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Upgrading the Data Acquisition and Control System of the LANSCE LINAC

Los Alamos National Laboratory LANL is in the process of upgrading the control system for the Los Alamos Neutron Science Center (LANSCE) linear accelerator. The 38 year-old data acquisition and control equipment is being replaced with COTS hardware. An overview of the current system requirements and how the National Instruments cRIO system meets these requirements will be given, as well as an update on the installation and operation of a prototype system in the LANSCE LINAC.

Accelerator Design, Technology, and Operations↗

GPU COMPUTING FOR PARTICLE TRACKING

This is a feasibility study of using a modern Graphics Processing Unit (GPU) to parallelize the accelerator particle tracking code. To demonstrate the massive parallelization features provided by GPU computing, a simplified TracyGPU program is developed for dynamic aperture calculation. Performances, issues, and challenges from introducing GPU are also discussed. General purpose Computation on Graphics Processing Units (GPGPU) bring massive parallel computing capabilities to numerical calculation. However, the unique architecture of GPU requires a comprehensive understanding of the hardware and programming model to be able to well optimize existing applications. In the field of accelerator physics, the dynamic aperture calculation of a storage ring, which is often the most time consuming part of the accelerator modeling and simulation, can benefit from GPU due to its embarrassingly parallel feature, which fits well with the GPU programming model. In this paper, we use the Tesla C2050 GPU which consists of 14 multi-processois (MP) with 32 cores on each MP, therefore a total of 448 cores, to host thousands ot threads dynamically. Thread is a logical execution unit of the program on GPU. In the GPU programming model, threads are grouped into a collection of blocks Within each block, multiple threads share the same code, and up to 48 KB of shared memory. Multiple thread blocks form a grid, which is executed as a GPU kernel. A simplified code that is a subset of Tracy++ [2] is developed to demonstrate the possibility of using GPU to speed up the dynamic aperture calculation by having each thread track a particle.

42 ENGINEERING↗

Navier: Dataflow Architecture for Computation Chemistry

Navier’s objectives were two evaluate the use of emerging technologies, especially dataflow accelerators, for high-performance computing (HPC) applications, specifically in the domain of chemistry, and to develop a prototype software stack to support such applications. Navier builds on capabilities previously developed by synergistic projects, such as PNNL Data Model Convergence (DMC) LDRD Hardware Advanced Workflows (HAW) and DuOMO, as well as DOE ARIAA. Throughout its 18 months, the Navier team developed new capabilities and artifacts at all levels of the HW/SW stack, provided a seamless way to integrate novel computing architectures (Sambanova SN10 and Xilinx Versal AI) into an existing software stack, developed chemistry workflows, data analytics tools, and HPC molecular dynamics workflows that leverage the developed stack and PNNL institutional investments in emerging architectures. Navier also explored the use of active learning to accelerate a computational chemistry workflow for organic molecules on PNNL Junction cluster (in collaboration with AMD/Xilinx). Navier developed tools, methodologies, and studies for hardware software co-design and (sparse) dataflow accelerators that are composable and can be used together or separately. These methodologies are now used in other projects, such as DOE AMAIS and HPDA. This report describes Navier’s achievement, the developed tools and methodologies, and the research findings and conclusions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

High-power test of a C-band linear accelerating structure with an RFSoC-based LLRF system

Normal conducting linear particle accelerators consist of multiple rf stations with accelerating structure cavities. Low-level rf (LLRF) systems are employed to set the phase and amplitude of the field in the accelerating structure and to compensate for the pulse-to-pulse fluctuation of the rf field in the accelerating structures with a feedback loop. The LLRF systems are typically implemented with analog rf mixers, heterodyne-based architectures, and discrete data converters. There are multiple rf signals from each of the rf stations, so the number of rf channels required increases rapidly with multiple rf stations. With a large number of rf channels, the footprint, component cost, and system complexity of the LLRF hardware will increase significantly. To meet the design goals of being compact and affordable for future accelerators, we have designed the next-generation LLRF (NG-LLRF) with a higher integration level based on RFSoC technology. The NG-LLRF system samples rf signals directly and performs rf mixing digitally. Further, the NG-LLRF has been characterized in loopback mode to evaluate the performance of the system and has also been tested with a standing-wave accelerating structure, a prototype for the Cool Copper Collider (C 3 ) with a peak rf power level up to 16.45 MW. The loopback test demonstrated amplitude fluctuation below 0.15% and phase fluctuation below 0.15°, which are considerably better than the requirements of C 3 . The rf signals from the different stages of the accelerating structure at different power levels are measured by the NG-LLRF, which will be critical references for the control algorithm designs. The NG-LLRF also offers flexibility in waveform modulation, so we have used rf pulses with various modulation schemes, which could be useful for controlling some of the rf stations in accelerators. In this paper, the high-power test results at different stages of the test setup will be summarized, analyzed, and discussed.

47 OTHER INSTRUMENTATION↗

An extensible operating system design for large-scale parallel machines.

Running untrusted user-level code inside an operating system kernel has been studied in the 1990's but has not really caught on. We believe the time has come to resurrect kernel extensions for operating systems that run on highly-parallel clusters and supercomputers. The reason is that the usage model for these machines differs significantly from a desktop machine or a server. In addition, vendors are starting to add features, such as floating-point accelerators, multicore processors, and reconfigurable compute elements. An operating system for such machines must be adaptable to the requirements of specific applications and provide abstractions to access next-generation hardware features, without sacrificing performance or scalability.

43 PARTICLE ACCELERATORS↗

JEMMRLA - Electron Model of a Muon RLA with Multi-pass Arcs

We propose a demonstration experiment for a new concept of a 'dogbone' RLA with multi-pass return arcs -- JEMMRLA (Jlab Electron Model of Muon RLA). Such an RLA with linear-field multi-pass arcs was introduced for rapid acceleration of muons for the next generation of Muon Facilities. It allows for efficient use of expensive RF while the multi-pass arc design based on linear combined-function magnets exhibits a number of advantages over separate-arc or pulsed-arc designs. Here we describe a test of this concept by scaling a GeV scale muon design for electrons. Scaling muon momenta by the muon-to-electron mass ratio leads to a scheme, in which a 4.5 MeV electron beam is injected in the middle of a 3 MeV/pass linac with two double-pass return arcs and is accelerated to 18 MeV in 4.5 passes. All spatial dimensions including the orbit distortion are scaled by a factor of 7.5, which arises from scaling the 200 MHz muon RF to a readily available 1.5 GHz. The hardware requirements are not very demanding making it straightforward to implement. Such an RLA may have applications going beyond muon acceleration: in medical isotope production, radiation cancer therapy and homeland security.

43 PARTICLE ACCELERATORS↗

FPGAs as a Service to Accelerate Machine Learning Inference [PowerPoint]

Large-scale particle physics experiments face challenging demands for high-throughput computing resources both now and in the future. New heterogeneous computing paradigms on dedicated hardware with increased parallelization, such as Field Programmable Gate Arrays (FPGAs), offer exciting solutions with large potential gains. The growing applications of machine learning algorithms in particle physics for simulation, reconstruction, and analysis are naturally deployed on such platforms. We demonstrate that the acceleration of machine learning inference as a web service represents a heterogeneous computing solution for particle physics experiments that requires minimal modification to the current computing model. As examples, we retrain the ResNet-50 convolutional neural network to demonstrate state-of-the-art performance for top quark jet tagging at the LHC, and apply a ResNet-50 model with transfer learning for neutrino event classification. Using Project Brainwave b y Microsoft to accelerate the ResNet-50 image classification model, we achieve average inference times of 60 (10) milliseconds with our experimental physics software framework using Brainwave as a cloud (edge or on-premises) service, representing an improvement by a factor of approximately 30 (175) in model inference latency over traditional CPU inference in current experimental hardware. A single FPGA service accessed by many CPUs achieves a throughput of 600--700 inferences per second using an image batch of one, comparable to large batch-size GPU throughput and significantly better than small batch-size GPU throughput. Deployed as an edge or cloud service for the particle physics computing model, coprocessor accelerators can have a higher duty cycle and are potentially much more cost-effective.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A Hierarchical Task Scheduler for Heterogeneous Computing

Heterogeneous computing is one of the future directions of HPC. Task scheduling in heterogeneous computing must balance the challenge of optimizing the application performance and the need for an intuitive interface with the programming run-time to maintain programming portability. The challenge is further compounded by the varying data communication time between tasks. This paper proposes RANGER, a hardware-assisted task-scheduling framework. By integrating RISC-V cores with accelerators, the RANGER scheduling framework divides scheduling into global and local levels. At the local level, RANGER further partitions each task into fine-grained subtasks to reduce the overall makespan. At the global level, RANGER maintains the coarse granularity of the task specification, thereby maintaining programming portability. The extensive experimental results demonstrate that RANGER achieves a 12.7× performance improvement on average, while only requires 2.7% of area overhead.

Miniskar, Narasinga Rao↗

Piecewise empirical model (PEM) of resistive memory for pulsed analog and neuromorphic applications

With the end of Dennard scaling and the ever-increasing need for more efficient, faster computation, resistive switching devices (ReRAM), often referred to as memristors, are a promising candidate for next generation computer hardware. These devices show particular promise for use in an analog neuromorphic computing accelerator as they can be tuned to multiple states and be updated like the weights in neuromorphic algorithms. Modeling a ReRAM-based neuromorphic computing accelerator requires a compact model capable of correctly simulating the small weight update behavior associated with neuromorphic training. These small updates have a nonlinear dependence on the initial state, which has a significant impact on neural network training. Consequently, we propose an empirically derived general purpose model that can accurately capture the nonlinearity of an arbitrary two-terminal device to match pulse measurements important for neuromorphic computing applications. By defining the state of the device to be proportional to its current, the model parameters can be extracted from a series of voltages pulses that mimic the behavior of a device in an analog neuromorphic computing accelerator. This allows for a general, accurate, and intuitive compact circuit model that is applicable to different resistance switching device technologies. In this work we explain the details of the model, implement the model in the circuit simulator Xyce, and give an example of its usage to model a specific Ta/TaO x device.

97 MATHEMATICS AND COMPUTING↗