Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Bit manipulation”

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 19 records

BitGNN: Unlocking the Performance Potential of Binary Graph Neural Networks on GPUs

Graph Neural Networks (GNNs) have shown compelling results in many graph-based learning tasks. They are, however, time-consuming. Recent work has shown a promising direction in improving GNN speed and shrinking the size — network binarization, which binarizes network values and operations. Prior work, however, mainly focused on algorithm designs, leaving it open on how to fully materialize the performance potential. This work fills the gap by proposing techniques to best map binary GNNs and their computations to fit the nature of bit manipulations, optimizations and algorithms to maximize BSpMM kernel efficiency, and solutions to other factors influencing the end-to-end time on GPUs. Results on real-world graphs show that the proposed techniques outperform state of-the-art binary GNN implementations by 21-67× with little accuracy loss.

Chen, Jou-An↗

BSTC: A Novel Binarized-Soft-Tensor-Core Design for Accelerating Bit-Based Approximated Neural Nets

Binarized neural network (or BNN) promises tremendous performance improvement over traditional DNNs through simplified bit-level computation and significantly reduced memory access/storage cost. In addition, it has other advantages of low-cost, low-energy and high-robustness, showing great utilization potential in resources-constrained, volatile and latency-critical applications, which are critical for future HPC execution. However, the promised significant performance gain of BNN inference has never been fully demonstrated on general-purpose processors, particularly on GPUs, due to: (i) the challenge to extract and leverage sufficient fine-grained bit-level-parallelism to saturate GPU cores when batch is small; (ii) the fundamental design conflict between bit-based BNN algorithm and underlying word-based architecture; (iii) architecture \& performance unfriendly BNN network design. To address (i) and (ii), we propose binarized-soft-tensor-core as a software-hardware codesign approach to construct bit-manipulation capability for modern GPUs to effectively harvest the emerging bit-level-parallelism. To tackle (iii), we propose intra- and inter-layer fusion techniques so that the entire BNN inference can be packed into a single GPU kernel, to avoid high-cost frequent launching. Experiments demonstrate that our design can achieve over 1000x speedup for raw inference latency and 10x for inference throughput over the state-of-the-art full-precision simulated BNN inference for AlexNet on ImageNet.

Li, Ang↗

SystemBurn

SystemBurn is a tool for creating a synthetic computational load for the purpose of measuring how much power a computer will draw under that type of load. The loads include fundamental library function calls like matrix multiply, memory copies, fourier transforms, bit manipulation, I/O, network packet transfers, and some code contrived to cause the processor to dray more or less power. The code produces some diagnostic and progress output, but the actual measurements would be recorded from the power panels within the computer room.

Josh Lothian, JeffKuehn↗

SystemBurn

SystemBurn is a tool for creating a synthetic computational load for the purpose of measuring how much power a computer will draw under that type of load. The loads include fundamental library function calls like matrix multiply, memory copies, fourier transforms, bit manipulation, I/O, network packet transfers, and some code contrived to cause the processor to dray more or less power. The code produces some diagnostic and progress output, but the actual measurements would be recorded from the power panels within the computer room.

Josh Lothian, Jeff Kuehn↗

The Tiny Median Filter: A Small Size, Flexible Arbitrary Percentile Finder Scheme Suitable for FPGA Implementation

This document reports the design, implementation and testing of a small silicon resource usage, very flexible arbitrary percentile finding scheme called the Tiny Median Filter. It can be used not only as a median filter in image processing with square filtering windows, but also for applications of any percentile filter or maximum or minimum finder with any size of data set as long as the number of bits of the data is finite. It opens possibilities for image processing tasks with non-square or irregular filter windows. In this scheme, data swapping or data bit manipulating are avoided and high functional efficiency of the logic components is applied to save silicon resources. Some logic functions are absorbed into other functions to further reduce the complexity. The combinational logic paths are designed to be sufficiently short so that the firmware can be compiled to the maximum operating frequency allowed by the block memories of the FPGA devices. The Tiny Median Filter receives, processes and output data in non-stop manner with no irregular timing which helps to simplify design of surrounding stages.

Wu, Jinyuan [Fermilab] (ORCID:0000000344329521)↗

Accelerating matrix-centric graph processing on GPUs through bit-level optimizations

Even though it is well known that binary values are common in graph applications (e.g., adjacency matrix), how to leverage the phenomenon for efficiency has not yet been adequately explored. This paper presents a systematic study on how to unlock the potential of the bit-level optimizations of graph computations that involve binary values. It proposes a two-level representation named Bit-Block Compressed Sparse Row (B2SR) and presents a series of optimizations to the graph operations on B2SR by the intrinsics of modern GPUs. It additionally introduces Deep Reinforcement Learning (DRL) as an efficient way to best configure the bit-level optimizations on the fly. Additionally, the DQN-based adaptive tile size selector with dedicated model training can reach 68% prediction accuracy. Evaluations on NVIDIA Pascal and Volta GPUs show that the optimizations bring up to 40× and 6555× for essential GraphBLAS kernels SpMV and SpGEMM, respectively, making GraphBLAS-based BFS accelerate up to 433×, SSSP, PR, and CC up to 35×, and TC up to 52×.

79 ASTRONOMY AND ASTROPHYSICS↗

Coherent coupling between a quantum dot and a donor in silicon

Individual donors in silicon chips are used as quantum bits with extremely low error rates. However, physical realizations have been limited to one donor because their atomic size causes fabrication challenges. Quantum dot qubits, in contrast, are highly adjustable using electrical gate voltages. This adjustability could be leveraged to deterministically couple donors to quantum dots in arrays of qubits. In this work, we demonstrate the coherent interaction of a 31P donor electron with the electron of a metal-oxide-semiconductor quantum dot. We form a logical qubit encoded in the spin singlet and triplet states of the two-electron system. We show that the donor nuclear spin drives coherent rotations between the electronic qubit states through the contact hyperfine interaction. This provides every key element for compact two-electron spin qubits requiring only a single dot and no additional magnetic field gradients, as well as a means to interact with the nuclear spin qubit.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Real-time multi-mode neutron multiplicity counter

Embodiments are directed to a digital data acquisition method that collects data regarding nuclear fission at high rates and performs real-time preprocessing of large volumes of data into directly useable forms for use in a system that performs non-destructive assaying of nuclear material and assemblies for mass and multiplication of special nuclear material (SNM). Pulses from a multi-detector array are fed in parallel to individual inputs that are tied to individual bits in a digital word. Data is collected by loading a word at the individual bit level in parallel, to reduce the latency associated with current shift-register systems. The word is read at regular intervals, all bits simultaneously, with no manipulation. The word is passed to a number of storage locations for subsequent processing, thereby removing the front-end problem of pulse pileup. The word is used simultaneously in several internal processing schemes that assemble the data in a number of more directly useable forms. The detector includes a multi-mode counter that executes a number of different count algorithms in parallel to determine different attributes of the count data.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Evaluation of hole-spin superposition in GaAs/AlGaAs quantum wells through time-resolved photoluminescence measurements

Relaxation dynamics of the up–down superposition of hole spins in undoped GaAs/AlGaAs quantum wells were investigated by means of time-resolved photoluminescence (PL) measurements with a view toward achieving quantum bits using hole spins. Linearly polarized PL signals reflecting the hole-spin superposition were observed using linearly polarized excitation light under quasi-resonant excitation conditions. In contrast to circular-polarization experiments conducted under nonresonant excitation conditions, the decay time of the degree of linear polarization obtained from the time-resolved PL signal increased slightly with increasing quantum confinement energy. These experimental results demonstrate that the creation and observation of hole-spin superposition under resonant conditions are crucial for manipulating hole-spin quantum bits.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Lithographically defined few-electron silicon quantum dots based on a silicon-on-insulator substrate

Silicon quantum dot (QD) devices with a proximal single-electron transistor (SET) charge sensor have been fabricated in a metal-oxide-semiconductor structure based on a silicon-on-insulator substrate. The charge state of the QDs was clearly read out using the charge sensor via the SET current. The lithographically defined small QDs enabled clear observation of the few-electron regime of a single QD and a double QD by charge sensing. Tunnel coupling on tunnel barriers of the QDs can be controlled by tuning the top-gate voltages, which can be used for manipulation of the spin quantum bit via exchange interaction between tunnel-coupled QDs. The lithographically defined silicon QD device reported here is technologically simple and does not require electrical gates to create QD confinement potentials, which is advantageous for the integration of complicated constructs such as multiple QD structures with SET charge sensors for the purpose of spin-based quantum computing.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Improving Signal to Noise Ratios in Ion Mobility Spectrometry and Structures for Lossless Ion Manipulations (SLIM) using a High Dynamic Range Analog-to-Digital Converter

Signal digitization is a commonly overlooked part of ion mobility-mass spectrometry (IMS-MS) workflows, yet it is a significant contributor for determining signal-to-noise ratios and MS resolution. Here we report on the integration of a 2 GS/s, 14-bit ADC with a structures for lossless ion manipulations (SLIM)-IMS-MS and compare the performance to a commonly used 8-bit ADC. The 14-bit ADC provided an effective reduction in digitized noise by factor of ~6, owing largely to the use of smaller bit sizes. The low baseline allowed the threshold voltage levels to be set very close to the MCP baseline voltage, allowing for as much signal to be acquired as possible without causing overloading or excessive digitization of MCP baseline noise. Analyses of Agilent tuning mixture ions and a complex mixture of heavy labeled phosphopeptides showed that the 14-bit ADC (compared to the 8-bit ADC) provided a modest signal-to-noise increase (~1.5 to 2-fold) for high intensity ions, such as the Agilent tuning mixture ions and the 2+ and 3+ charge states of many phosphopeptide constituents. However, signal enhancements were as much as 10-fold for low intensity ions, and the 14-bit ADC enabled discernable signal intensities otherwise lost using an 8-bit digitizer. Additionally, the 14-bit ADC required ~14-fold fewer mass spectra to be averaged to produce a mass spectrum with similar S/N as the 8-bit ADC under identical conditions, potentially providing an order of magnitude higher measurement throughput. The high resolution, low baseline, and fast speed of the new 14-bit ADC enables high performance digitization of MS, IMS-MS, and SLIM-IMS-MS spectra, and allows a much fuller picture of analyte profiles in complex mixtures to be acquired.

data acquisition, digitization, ion mobility spect↗

Semiconductor Nanotechnology: Novel Materials and Devices for Electronics, Photonics, and Renewable Energy Applications

Electronic and photonic information technology and renewable energy alternatives, such as solar energy, fuel cells and batteries, have now reached an advanced stage in their development. Cost-effective improvements to current technological approaches have made great progress, but certain challenges remain. As feature sizes of the latest generations of electronic devices are approaching atomic dimensions, circuit speeds are now being limited by interconnect bottlenecks. This has prompted innovations such as the introduction of new materials into microelectronics manufacturing at an unprecedented rate and alternative technologies to silicon CMOS architectures. Despite the environmental impact of conventional fossil fuel consumption, the low cost of these energy sources has been a long-standing economic barrier to the development of alternative and more efficient renewable energy sources, fuel cells and batteries. In the face of mounting environmental concerns, interest in such alternative energy sources has grown. It is now widely accepted that nanotechnology offers potential solutions for securing future progress in information and energy technologies. The Canadian Semiconductor Technology Conference (CSTC) forum was established 25 years ago in Ottawa as an important symbol of the intrinsic strength of the Canadian semiconductor research and development community, and the Canadian semiconductor industry as a whole. In 2007, the 13th CSTC was held in Montreal, moving for the first time outside the national capital region. The first three meetings in the series of Nano and Giga Challenges in Electronics and Photonics NGCM2002 in Moscow, NGCM2004 in Krakow, and NGC2007 in Phoenix were focused on interdisciplinary research from the fundamentals of materials science to the development of new system architectures. In 2009 NGC2009 and the 14th Canadian Semiconductor Technology Conference (CSTC2009) were held as a joint event, hosted by McMaster University (10 14 August, Hamilton, Ontario, Canada) and the scope was expanded to include renewable energy research and development. This special issue of Nanotechnology is devoted to a better understanding of the function and design of semiconductor devices that are relevant to information technology (both electronics and photonics based) and renewable energy applications. The papers contained in this special issue are selected from the NGC/CSTC2009 symposium. Among them is a report by Ray LaPierre from McMaster University and colleagues at the University of Waterloo in Canada on the ability to manipulate single spins in nanowire quantum bits. The paper also reports the development of a testbed of a few qubits for general quantum information processing tasks [1]. Lower cost and greater energy conversion efficiency compared with thin film devices have led to a high level of activity in nanowire research related to photovoltaic applications. This special issue also contains results from an impedance spectroscopy study of core shell GaAs nanowires to throw light on the transport and recombination mechanisms relevant to solar cell research [2]. Information technology research and renewable energy sources are research areas of enormous public interest. This special issue addresses both theoretical and experimental achievements and provides a stimulating outlook for technological developments in these highly topical fields of research. References [1] Caram J, Sandoval C, Tirado M, Comedi D, Czaban J, Thompson D A and LaPierre R R 2010 Electrical characteristics of core shell p-n GaAs nanowire structures with Te as the n-dopant Nanotechnology 21 134007 [2] Baugh J, Fung J S and LaPierre R R 2010 Building a spin quantum bit register using semiconductor nanowires Nanotechnology 21 134018

14 SOLAR ENERGY↗

Usefulness of multiqubit W-type states in quantum information processing

We analyze the efficiency of multiqubit W-type states as resources for quantum information. For this, we identify and generalize four-qubit W-type states. Our results show that these states can be used as resources for deterministic quantum information processing. The utility of results, however, is limited by the availability of experimental setups to perform and distinguish multiqubit measurements. We therefore emphasize protocols where two users want to establish an optimal bipartite entanglement using the partially entangled W-type states. We find that for such practical purposes, four-qubit W-type states can be a better resource in comparison to three-qubit W-type states. For a dense coding protocol, our states can be used deterministically to send two bits of classical message by locally manipulating a single qubit. In addition, we also propose a realistic experimental method to prepare the four-qubit W-type states using standard unitary operations and weak measurements.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

High-resolution tunnelling spectroscopy of fractional quantum Hall states

Strong interactions between electrons in two-dimensional systems in the presence of a high magnetic field give rise to fractional quantum Hall states that host quasiparticles with a fractional charge and fractional exchange statistics. Here, in this work, we demonstrate high-resolution scanning tunnelling microscopy and spectroscopy of fractional quantum Hall states in ultra-clean Bernal-stacked bilayer graphene devices. Spectroscopy measurements show sharp excitations that have been predicted to emerge when electrons fractionalize into bound states of quasiparticles. We found energy gaps for candidate non-abelian fractional states that are larger by a factor of five than those in other related systems, for example, semiconductor heterostructures, and this suggests that bilayer graphene is an ideal platform for manipulating these quasiparticles and for creating topological quantum bits. We also found previously unobserved fractional states in our very clean graphene samples.

quantum Hall↗

TTHRESH: Tensor Compression for Multidimensional Visual Data

Memory and network bandwidth are decisive bottlenecks when handling high-resolution multidimensional data sets in visualization applications, and they increasingly demand suitable data compression strategies. In this paper, we introduce a novel lossy compression algorithm for multidimensional data over regular grids. It leverages the higher-order singular value decomposition (HOSVD), a generalization of the SVD to three dimensions and higher, together with bit-plane, run-length and arithmetic coding to compress the HOSVD transform coefficients. Our scheme degrades the data particularly smoothly and achieves lower mean squared error than other state-of-the-art algorithms at low-to-medium bit rates, as it is required in data archiving and management for visualization purposes. Further advantages of the proposed algorithm include very fine bit rate selection granularity and the ability to manipulate data at very small cost in the compression domain, for example to reconstruct filtered and/or subsampled versions of all (or selected parts) of the data set.

97 MATHEMATICS AND COMPUTING↗

Nuclear spin qubits in a trapped-ion quantum computer

Physical systems must fulfill a number of conditions to qualify as useful quantum bits (qubits) for quantum-information processing, including ease of manipulation, long decoherence times, and high fidelity readout operations. Since these conditions are hard to satisfy with a single system, it may be necessary to combine different degrees of freedom. Here we discuss a possible system based on electronic and nuclear spin degrees of freedom in trapped ions. The nuclear spin yields long decoherence times, while the electronic spin, in a magnetic field gradient, provides efficient manipulation, and the optical transitions of the ions assure a selective and efficient initialization and readout.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Quantum holographic encoding in a two-dimensional electron gas

The advent of bottom-up atomic manipulation heralded a new horizon for attainable information density, as it allowed a bit of information to be represented by a single atom. The discrete spacing between atoms in condensed matter has thus set a rigid limit on the maximum possible information density. While modern technologies are still far from this scale, all theoretical downscaling of devices terminates at this spatial limit. Here, however, we break this barrier with electronic quantum encoding scaled to subatomic densities. We use atomic manipulation to first construct open nanostructures - 'molecular holograms' - which in turn concentrate information into a medium free of lattice constraints: the quantum states of a two-dimensional degenerate Fermi gas of electrons. The information embedded in the holograms is transcoded at even smaller length scales into an atomically uniform area of a copper surface, where it is densely projected into both two spatial degrees of freedom and a third holographic dimension mapped to energy. In analogy to optical volume holography, this requires precise amplitude and phase engineering of electron wavefunctions to assemble pages of information volumetrically. This data is read out by mapping the energy-resolved electron density of states with a scanning tunnelling microscope. As the projection and readout are both extremely near-field, and because we use native quantum states rather than an external beam, we are not limited by lensing or collimation and can create electronically projected objects with features as small as {approx}0.3 nm. These techniques reach unprecedented densities exceeding 20 bits/nm{sup 2} and place tens of bits into a single fermionic state.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Molecular spins for quantum computation

Spins in solids or in molecules possess discrete energy levels, and the associated quantum states can be tuned and coherently manipulated by means of external electromagnetic fields. Spins therefore provide one of the simplest platforms to encode a quantum bit (qubit), the elementary unit of future quantum computers. Notably, performing any useful computation demands much more than realizing a robust qubit—one also needs a large number of qubits and a reliable manner with which to integrate them into a complex circuitry that can store and process information and implement quantum algorithms. This ‘scalability’ is arguably one of the challenges for which a chemistry-based bottom-up approach is best-suited. Molecules, being much more versatile than atoms, and yet microscopic, are the quantum objects with the highest capacity to form non-trivial ordered states at the nanoscale and to be replicated in large numbers using chemical tools.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗