Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “mixed state compiling”

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

Quantum mixed state compiling

The task of learning a quantum circuit to prepare a given mixed state is a fundamental quantum subroutine. We present a variational quantum algorithm (VQA) to learn mixed states which is suitable for near-term hardware. Our algorithm represents a generalization of previous VQAs that aimed at learning preparation circuits for pure states. We consider two different ansätze for compiling the target state; the first is based on learning a purification of the state and the second on representing it as a convex combination of pure states. In both cases, the resources required to store and manipulate the compiled state grow with the rank of the approximation. Thus, by learning a lower rank approximation of the target state, our algorithm provides a means of compressing a state for more efficient processing. As a byproduct of our algorithm, one effectively learns the principal components of the target state, and hence our algorithm further provides a new method for principal component analysis. We investigate the efficacy of our algorithm through extensive numerical implementations, showing that typical random states and thermal states of many body systems may be learnt this way. Additionally, we demonstrate on quantum hardware how our algorithm can be used to study hardware noise-induced states.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Mixing effects on spectroscopy and partonic observables of heavy mesons with logarithmic confining potential in a light-front quark model

Using the variational principle, we systematically investigate the mass spectra and wave functions of both 1⁢𝑆 and 2⁢𝑆 state heavy pseudoscalar (𝑃) and vector (𝑉) mesons within the light-front quark model. This approach incorporates a Coulomb plus logarithmic confinement potential to accurately describe the constituent quark and antiquark dynamics. Additionally, spin hyperfine interactions are introduced perturbatively to compute the masses of pseudoscalar and vector mesons. The present analyses of the 1⁢𝑆 and 2⁢𝑆 states require the consideration of mixing between them to account for empirical constraints. These constraints include the mass gap Δ⁢𝑀 𝑃 >Δ⁢𝑀 𝑉 , where Δ⁢𝑀 𝑃⁡(𝑉) =𝑀$^{2⁢𝑆}_{𝑃⁡(𝑉)}$−𝑀$^{1⁢𝑆}_{𝑃⁡(𝑉)}$ and the hierarchy of the decay constants 𝑓 1⁢𝑆 >𝑓 2⁢𝑆 . We find the optimal value of the mixing angle to be 𝜃 =1⁢8°, significantly enhancing the consistency between our spectroscopic predictions and the experimental data compiled by the Particle Data Group. Furthermore, based on the predicted mass, the newly observed resonance 𝐵 𝐽⁡ (5840) could be assigned as a 2 1⁢ 𝑆 0 state in the 𝐵 meson family. The study also reports various pertinent observables, including twist-two distribution amplitudes, electromagnetic form factors, charge radii, 𝜉 moments, and transition form factors that are found to be consistent with both available lattice simulations and experimental data. In addition, our predicted branching ratios for the channels of 𝐵 + →𝜏 + ⁢𝜈 𝜏 as well as rare decays of 𝐵 0 and 𝐵$^0_𝑠$ appear in accordance with experimental data.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Residual waste detection in HLW tanks - 15701

This research uses commercial sonar technology to monitor residual waste in the U.S. Department of Energy's (DOE) Hanford Site high-level-waste staging tanks, with primary focus on the detection and imaging of the settled solids at specified areas of interest along the tank surface within a limited amount of time. High-level waste (HLW) is generated at the US-DOE Hanford Site as a by-product from the processing of nuclear materials. Waste is currently stored in 177 underground carbon-steel waste tanks. Pulse jet mixers (PJM) are often used to break loose and suspend the solids that have settled at the bottom of the tanks. A typical tank can have 6-8 pulse tubes. The Solid Liquid Interface Monitor (SLIM) must be capable of measuring dynamic movements in Region 1 zone between each PJM cycle. SLIM was originally developed to map the sludge/supernate interface within the U.S. DOE Hanford Site's HLW tanks. SLIM is a dual-axis mechanically rotating sonar and consists of an acoustic transducer, rotation motor with gearbox, tilt motor, position reference sensor, and pressure balancing mechanisms and its associated deployment platform. The sonar gathers a swath of data in the horizontal plane and rotates the transducer through a programmed angle around the vertical axis. Oil-filled device. No user serviceable parts. No semiconductors. Hard anodized aluminum and polyurethane delrin. May be deployed in areas of significant radioactivity. Kaolin will be used during mixing operations in order to simulate the physical properties of the expected slurry. In order to reduce the amount of time needed to image and object, all settings were decreased to their minimal state. The low resolution settings provide insufficient data for the manufacturer's 3D profiler to compile a visible point cloud and much less to allow the meshing algorithm to work properly. Several filtering algorithms have been applied to each scan in order to remove erroneous spikes and ghosting effects. Data has been collected for the mixing process of kaolin. This data consists of scans executed while the pump was active and inactive. The imaging algorithm has also been used with data derived from scans of 3% or less of kaolin. Future work: Mixing of kaolin within an experimental tank will be conducted in order to simulate the settling time of the HLW within the tank. The same filtering algorithms will be applied to each of these scans to evaluate the SLIM's efficiency to image solids at the bottom of a tank through increasing volume of kaolin.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Qompress: Efficient Compilation for Ququarts Exploiting Partial and Mixed Radix Operations for Communication Reduction

Quantum computing is in an era of limited resources. Current hardware lacks high fidelity gates, long coherence times, and the number of computational units required to perform meaningful computation. Contemporary quantum devices typically use a binary system, where each qubit exists in a superposition of the 0 and 1 states. Furthermore, it is often possible to access the 2 or even 3 states in the same physical unit by manipulating the system in different ways. In this work, we consider automatically encoding two qubits into one four-state ququart via a compression scheme. We use quantum optimal control to design efficient proof-of-concept gates that fully replicate standard qubit computation on these encoded qubits.

compilation↗

A SQL to C compiler in 500 lines of code

Abstract We present the design and implementation of a SQL query processor that outperforms existing database systems and is written in just about 500 lines of Scala code – a convincing case study that high-level functional programming can handily beat C for systems-level programming where the last drop of performance matters. The key enabler is a shift in perspective toward generative programming. The core of the query engine is an interpreter for relational-algebra operations, written in Scala. Using the open-source lightweight modular staging framework, we turn this interpreter into a query compiler with very low effort. To do so, we capitalize on an old and widely known result from partial evaluation: the first Futamura projection, which states that a process that can specialize an interpreter to any given input program is equivalent to a compiler. In this context, we discuss lightweight modular staging programming patterns such as mixed-stage data structures (e.g., data records with static schema and dynamic field components) and techniques to generate low-level C code, including specialized data structures and data loading primitives.

Computer Science↗

Development of a Conceptual Process for Selective CO 2 Capture from Fuel Gas Streams Using [hmim][Tf 2 N] Ionic Liquid as a Physical Solvent

The Ionic Liquid (IL) [hmim][Tf 2 N] was used as a physical solvent in an Aspen Plus simulation, employing the Peng-Robinson Equation of State (P-R EOS) with Boston-Mathias (BM) alpha function and standard mixing rules, to develop a conceptual process for CO 2 capture from a shifted warm fuel gas stream produced from Pittsburgh # 8 coal for a 400 MWe power plant. The physical properties of the IL, including density, viscosity, surface tension, vapor pressure and heat capacity were obtained from literature and modeled as a function of temperature. Also, available experimental solubility values for CO 2 , H 2 , H 2 S, CO, and CH 4 in this IL were compiled and their binary interaction parameters (Δ ij and l ij ) were optimized and correlated as functions of temperature. The Span-Wager Equation-of-State EOS was also employed to generate CO 2 solubilities in [hmim][Tf 2 N] at high pressures (up to 10 MPa) and temperatures (up to 510 K). The conceptual process developed consisted of 4 adiabatic absorbers (2.4 m ID, 30 m high) arranged in parallel and packed with Plastic Pall Rings of 0.025 m for CO 2 capture; 3 flash drums arranged in series for solvent (IL) regeneration with the pressure-swing option; and a pressure-intercooling system for separating and pumping CO 2 up to 153 bar to the sequestration sites. The compositions of all process streams, CO 2 capture efficiency, and net power were calculated using Aspen Plus simulator. The results showed that, based on the composition of the inlet gas stream to the absorbers, 95.67 mol% of CO 2 was captured and sent to sequestration sites; 99.5 mol% of H 2 was separated and sent to turbines; the solvent exhibited a minimum loss of 0.31 mol%; and the net power balance of the entire system was 30.81 MW. These results indicated that [hmim][Tf 2 N] IL could be used as a physical solvent for CO 2 capture from warm shifted fuel gas streams with high efficiency.

20 FOSSIL-FUELED POWER PLANTS↗

Toward closure between predicted and observed particle viscosity over a wide range of temperatures and relative humidity

Abstract. Atmospheric aerosols can exist in amorphous semi-solid or glassy phase states whose viscosity varies with atmospheric temperature and relative humidity. The temperature and humidity dependence of viscosity has been hypothesized to be predictable from the combination of a water–organic binary mixing rule of the glass transition temperature, a glass-transition-temperature-scaled viscosity fragility parameterization, and a water uptake parameterization. This work presents a closure study between predicted and observed viscosity for sucrose and citric acid. Viscosity and glass transition temperature as a function of water content are compiled from literature data and used to constrain the fragility parameterization. New measurements characterizing viscosity of sub-100 nm particles using the dimer relaxation method are presented. These measurements extend the available data of temperature- and humidity-dependent viscosity to −28 ∘C. Predicted relationships agree well with observations at room temperature and with measured isopleths of constant viscosity at ∼107 Pa s at temperatures warmer than −28 ∘C. Discrepancies at colder temperatures are observed for sucrose particles. Simulations with the kinetic multi-layer model of gas–particle interactions suggest that the observed deviations at colder temperature for sucrose can be attributed to kinetic limitations associated with water uptake at the timescales of the dimer relaxation experiments. Using the available information, updated equilibrium phase-state diagrams (-80∘C

54 ENVIRONMENTAL SCIENCES↗

VUV Photoionization Dynamics and Reactivity of Heterogeneous Water Clusters

This feature focuses on bridging isolated water and bulk water studies. It assembles experiments and theory on water clusters, mainly probed by vacuum ultraviolet (VUV) radiation, and summarizes what the effects are of intermolecular interactions on both the spectroscopy and the VUV-induced processes in water. In particular, it highlights studies of heterogeneous water clusters─those incorporating other molecular species such as naphthalene, methane, formic acid, and glycerol─which serve as model systems to investigate the fundamental roles of water in hydrogen bonding networks, proton transfer, and astrochemical processes. These mixed clusters provide a platform to revisit the stability of protonated water clusters and compile observations of ionization-induced structural rearrangements and fragmentation, especially in systems involving hydroxyl-rich cosolvents. This feature then explores energy transfer mechanisms in molecular clusters, following VUV photoexcitation. It concludes with potential future directions. First, we investigate excited-state dynamics in molecular clusters through direct probing, complementing the synchrotron studies discussed here. Second, exploring gas-phase molecule evaporation from confined spaces and interfaces using advanced spectroscopic techniques sheds light on these ubiquitous, yet currently debated, molecular processes.

Cluster chemistry↗

Integrated Analyses of Multiple Worldwide Aerosol Mass Spectrometer Datasets for Improved Understanding of Aerosol Sources and Processes and for Comparison with Global Models

The AMS is the only current instrument that provides real-time, quantitative, and size-resolved data on submicron non-refractory aerosol species with a time resolution of a few minutes or better. The AMS field data are multidimensional and massive, containing extremely rich information on aerosol chemistry, microphysics and dynamics—basic information that is required to evaluate and quantify the radiative climate forcing of atmospheric aerosols. The high time resolution of the AMS data also reveals details of aerosol dynamic variations that are vital to understanding the physico-chemical processes of atmospheric aerosols that govern aerosol properties relevant to the climate. There are two primary objectives of this 3-year project. Our first objective is to perform highly integrated analysis of dozens of AMS datasets acquired from various urban, forested, coastal, marine, mountain peak, and rural/remote locations around the world and synthesize and inter-compare results with a focus on the sources and the physico-chemical processes that govern aerosol properties relevant to aerosol climate forcing. Our second objective is to support our collaboration with global aerosol modelers, in which we will supply the size-resolved aerosol composition and temporal variation data (via a public web interface) and our analysis results for use in model testing and validation and for translation of the rich AMS database into model constraints that can improve climate forcing simulations. Several prominent global aerosol modelers have expressed enthusiastic support for this collaboration. The specific tasks that we propose to accomplish include 1) to develop, validate, and apply multivariate analysis techniques for improved characterization and source apportionment of organic aerosols; 2) to evaluate aerosol source regions and relative contributions based on back-trajectory integration (PSCF method); 3) to summarize and synthesize submicron aerosol information, including composition, concentration, size distribution and (inferred) shape and mixing state in various environments and their regional and seasonal variations within the context of regional and global modeling; and 4) to quantitatively evaluate important processes in various atmospheric environments and during different seasons, focusing on acid-catalyzed SOA formation, new particle growth, and photochemical processes of atmospheric organic aerosols (i.e., SOA production and POA oxidation). We will also examine the correlations and compile the ratios between important pairs of aerosol and gas phase species using region-specific and season-specific correlations and as a function of photochemical age and compare them with the ratios produced by various models. To enable our collaborations with the modelers, we will supply (via a public web interface) AMS data and our analysis results for use in model testing and validation and facilitate the use of the AMS information to constrain calculations of radiative forcing. Model output and AMS measurements and derived parameters will be compared with a focus on regional variability of model/measurement discrepancies and their causes. Finally we will share results, insights and data mining algorithms through peer-reviewed publications, presentations/tutorials at conferences/workshops, and web dissemination of analysis results and in-house developed software packages.

58 GEOSCIENCES↗

Efficient DNA sequence compression with neural networks

Abstract Background The increasing production of genomic data has led to an intensified need for models that can cope efficiently with the lossless compression of DNA sequences. Important applications include long-term storage and compression-based data analysis. In the literature, only a few recent articles propose the use of neural networks for DNA sequence compression. However, they fall short when compared with specific DNA compression tools, such as GeCo2. This limitation is due to the absence of models specifically designed for DNA sequences. In this work, we combine the power of neural networks with specific DNA models. For this purpose, we created GeCo3, a new genomic sequence compressor that uses neural networks for mixing multiple context and substitution-tolerant context models. Findings We benchmark GeCo3 as a reference-free DNA compressor in 5 datasets, including a balanced and comprehensive dataset of DNA sequences, the Y-chromosome and human mitogenome, 2 compilations of archaeal and virus genomes, 4 whole genomes, and 2 collections of FASTQ data of a human virome and ancient DNA. GeCo3 achieves a solid improvement in compression over the previous version (GeCo2) of $2.4\%$, $7.1\%$, $6.1\%$, $5.8\%$, and $6.0\%$, respectively. To test its performance as a reference-based DNA compressor, we benchmark GeCo3 in 4 datasets constituted by the pairwise compression of the chromosomes of the genomes of several primates. GeCo3 improves the compression in $12.4\%$, $11.7\%$, $10.8\%$, and $10.1\%$ over the state of the art. The cost of this compression improvement is some additional computational time (1.7–3 times slower than GeCo2). The RAM use is constant, and the tool scales efficiently, independently of the sequence size. Overall, these values outperform the state of the art. Conclusions GeCo3 is a genomic sequence compressor with a neural network mixing approach that provides additional gains over top specific genomic compressors. The proposed mixing method is portable, requiring only the probabilities of the models as inputs, providing easy adaptation to other data compressors or compression-based data analysis tools. GeCo3 is released under GPLv3 and is available for free download at https://github.com/cobilab/geco3.

Silva, Milton↗

Unique Facilities, Innovative Treatment Processes, and a Talented Workforce Allow Fluor Idaho to Successfully Disposition Transuranic Waste - 18388

Fluor Idaho is utilizing state-of-the-art equipment, treatment technologies, and employee expertise at the Department of Energy's (DOE) Idaho Site to treat contact-handled transuranic (CH-TRU) waste, as well as extremely challenging remote-handled transuranic (RH-TRU) waste forms, to prepare the material for shipment and permanent disposal at the Department's Waste Isolation Pilot Plant (WIPP). Nowhere else in the DOE Complex - perhaps the world - will one find the compilation of tools and an experienced workforce with an unparalleled history of safely retrieving, characterizing, treating, repackaging, certifying, and shipping TRU waste. Since 2003, the Advanced Mixed Waste Treatment Project (AMWTP) has used both conventional and unique retrieval concepts, high-tech characterization equipment - including real-time radiography and radioassay - robotics, and automated treatment processes to retrieve and treat 65,000 cubic meters of above-ground CH-TRU waste. Additionally, engineered controls and specialized personal protective equipment have allowed employees to make thousands of entries into high-radiation areas without a skin or clothing contamination incident - evidence of the employees' commitment to safety. AMWTP has encountered and found solutions for many challenging waste forms, including polychlorinated biphenyl-contaminated wastes, TRU waste containers with up to 1,000 fissile gram equivalent plutonium-239, oversized TRU boxes, and partially absorbed organic sludge waste. By utilizing additional facilities and equipment at the Accelerated Retrieval Project V and VII, Fluor Idaho has developed new processing lines and capabilities to ensure worker safety and milestone completion. These two soft-sided buildings, located at a facility adjacent to AMWTP, were originally constructed to exhume Comprehensive Environmental Response, Compensation and Liability Act (CERCLA) TRU wastes from a section of a 97-acre landfill. After a total of 1,150 cubic meters of targeted waste were removed, personnel, working in collaboration with state regulators, successfully implemented a unique regulatory strategy that allowed these CERCLA facilities to be modified and permitted to treat Resource Conservation and Recovery Act (RCRA) wastes from the AMWTP - a savings of tens of millions of dollars over the cost of constructing new buildings. Adding to the AMWTP's technological uniqueness is its ability to treat wastes from other facilities in the DOE Complex. Two prior decision documents and an agreement with the state of Idaho have allowed AMWTP to treat wastes from 15 other DOE sites. It is the only project in the complex to have done so. At another project located next to AMWTP, targeted waste exhumation within the Subsurface Disposal Area is approximately two years ahead of schedule. Crews have removed more than 8,600 cubic meters of CH-TRU and hazardous wastes from eight exhumation areas that equal a combined footprint of 4.75 acres. Less than one acre remains to be exhumed to satisfy the 2008 Record of Decision waste exhumation end state between the DOE, Environmental Protection Agency, and state of Idaho. Construction of the final building where targeted waste exhumation will continue was completed during the summer of 2017. Exhumation will be completed within the next few years. Following exhumation completion, an evapotranspiration barrier (or cap) will be constructed over the entire landfill. Fluor Idaho continues to make progress on a more challenging project involving an inventory of RH-TRU waste that also contains the reactive metal sodium. Crews are utilizing hot cells, once used for spent nuclear fuel reprocessing and high-level waste treatment at the Idaho Nuclear Technology and Engineering Center, to render this challenging waste form safe for disposal in a national geologic waste repository. The treatment methods being employed on this RH-TRU waste are unique in the DOE Complex. Minute concentrations of water are sprayed onto the sodium-contaminated material in a process called 'spritzing' to chemically change the sodium from a reactive metal to a non-reactive salt. In other cases, sodium-contaminated waste is dipped in water to initiate a reaction in a process called 'immersion.' Lastly, for more challenging waste shapes that could not be successfully treated with the two prior methods, crews lower the sodium-contaminated RH-TRU waste into a one-of-a-kind distillation unit that uses high temperature and vacuum to volatilize the sodium and recapture it in a condenser unit for disposal. The RH-TRU waste treated with this technology is then safe for permanent disposal. Fluor Idaho is proud to operate DOE's CH- and RH-TRU waste treatment facilities, as well as use innovative processes and its talented workforce to complete this important work-scope. Such capabilities allow the contractor to meet the commitments made by the DOE to the state of Idaho. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Technology inventory: energy efficiency measures for residential, commercial, and industrial sectors in the United States

Energy efficiency is expected to continue to be one of main levers for governments to address the energy and climate crisis. While a diverse mix of energy efficiency programs has been regarded as an important part of governmental actions, an energy efficiency policy or program needs to identify and assess eligible savings measures, i.e., specific technologies or practices, for designing key features, e.g., target type, target stringency, covered load, and cost recovery. This report briefly discusses U.S. energy efficiency policies and programs, explores publicly available resources of energy efficiency measures for residential, commercial, and industrial sectors in the U.S., and describes how to compile key measures and characteristics into an integrated spreadsheet database. The database in the initial version offers, for more than 1,200 measures, efficiency or improvement potential, technical notes, costs at a given year, where applicable. Energy efficiency measures compiled here could play an important role in determining savings targets and target stringency under a potential energy efficiency policy such as an energy efficiency resource standard in Korea, and offer significant opportunities for energy efficiency improvement.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Dataset_for_Conserved_macromolecular_architecture_of_Poplar_secondary_cell_walls_revealed_by_ssNMR_and_atomistic_modeling

This dataset contains solid-state 13C NMR data and atomistic molecular dynamics simulation files supporting the study of nanoscale secondary cell wall architecture across 13 genetically diverse Populus trichocarpa genotypes grown under uniform greenhouse conditions in 13C-enriched CO2 atmospheres (~89% 13C enrichment).The dataset contains two collections of solid-state 13C NMR data. (1) 200 MHz data (Bruker Avance III HD, 4 mm HX probe, 10 kHz MAS): raw Bruker TopSpin experiment folders and DMFIT-exported ascii spectra for selective and non-selective 1D 13C-13C spin diffusion experiments (3000 ms mixing) used to quantify inter-polymer spatial proximities, and short-mixing (1 ms) reference spectra used for polymeric abundance quantification by spectral deconvolution. (2) 600 MHz data (Bruker Avance III, 1.6 mm PhoenixNMR HXY probe, 30 kHz MAS): raw Bruker TopSpin experiment folders containing 2D CORD, 2D CP-INADEQUATE, and 13C/1H relaxation (T1, T1rho) experiments for all 13 genotypes, with processed Excel workbooks per experiment type. Molecular dynamics simulation code, coordinate files, and analysis scripts (NAMD/CHARMM/Python) for six atomistic cell wall models are included. Summarized ssNMR data are compiled into a single excel file and subjected to statistical analysis. Multivariate analysis code (PCA, Pearson correlation) and summary data are provided as excel worksheets and Jupyter notebooks (Python 3).

09 BIOMASS FUELS↗

A High Performance Sparse Tensor Algebra Compiler in MLIR

Sparse tensor algebra is widely used in many applications, including scientific computing, machine learning, and data analytics. The performance of sparse tensor algebra kernels strongly depends on the intrinsic characteristics of the input tensors, hence many storage formats are designed for tensors to achieve optimal performance for particular applications/architectures, which makes it challenging to implement and optimize every tensor operation of interest on a given architecture. We propose a tensor algebra domain-specific language (DSL) and compiler framework to automatically generate kernels for mixed sparse-dense tensor algebra operations. The proposed DSL provides high-level programming abstractions that resemble the familiar Einstein notation to represent tensor algebra operations. The compiler introduces a new Sparse Tensor Algebra dialect built on top of LLVM's extensible MLIR compiler infrastructure for efficient code generation while covering a wide range of tensor storage formats. Our compiler also leverages input-dependent code optimization to enhance data locality for better performance. Our results show that the performance of automatically generated kernels outperforms the state-of-the-art sparse tensor algebra compiler, with up to 20.92x, 6.39x, and 13.9x performance improvement over state-of-the-art tensor algebra compilers, for parallel SpMV, SpMM, and TTM, respectively.

Tian, Ruiqin↗

Singularity-EOS: Performance Portable Equations of State and Mixed Cell Closures

We present Singularity-EOS, a new performance-portable library for equations of state and related capabilities. Singularity-EOS provides a large set of analytic equations of state, such as the Gruneisen equation of state, and tabulated equation of state data under a unified interface. It also provides support capabilities around these equations of state, such as Python wrappers, solvers for finding pressure-temperature equilibrium between multiple equations of state, and a unique modifier framework, allowing the user to transform a base equation of state, for example by shifting or scaling the specific internal energy. All capabilities are performance portable, meaning they compile and run on both CPU and GPU for a wide variety of architectures.

97 MATHEMATICS AND COMPUTING↗

FLASH: FPGA-Accelerated Smart Switches with GCN Case Study

Some communication switches, e.g., the Mellanox SHArP and those in the IBM BlueGene clusters, are augmented to process packets at the application level with fixed-function collectives. This approach, however, lacks flexibility, which limits their applicability in diverse and dynamic workloads. Recently, a new type of programmable packet processor, which uses high-level languages, e.g., P4, has emerged as possible candidates. P4-based switches, however, fall short in certain applications, including machine learning, where capabilities not currently supported by P4 are needed. These include more complex calculation, such as sparse computation and fused multiply-accumulate, data-intensive floating point operations, data reuse, and significant memory. The problem addressed here is that such a switch augmentation needs to support: a large amount of state, significant flexible compute capability, and ease of programming, all while maintaining full functionality, including ensuring high throughput, and demonstrating utility. In this work, we propose a programmable look-aside-type accelerator that can be embedded into, or attached to, existing communication switch pipelines and that is capable of processing packets at line-rate. The proposed in-switch accelerator is based on mixing an ISA (subset of RISC-V instructions) with dataflow graphs (found in CGRAs). To augment performance, vector instructions are also supported. To facilitate usability, we have developed a complete toolchain to compile user-provided C/C++ codes to appropriate back-end instructions for configuring the accelerator. While this approach is flexible enough to support various workloads, in this paper, we consider Graph Convolutional Networks (GCNs) as a case study. Experimental results show that this approach considerably improves the performance of distributed GCN applications.

Haghi, Pouya↗

Mapping Spiking Neural Networks to Heterogeneous Crossbar Architectures using Integer Linear Programming

Advances in novel hardware devices and architectures allow Spiking Neural Network (SNN) evaluation using ultra-low power, mixed-signal, memristor crossbar arrays. As individual network sizes quickly scale beyond the dimensional capabilities of single crossbars, networks must be mapped onto multiple crossbars. Crossbar sizes within modern Memristor Crossbar Architectures (MCAs) are determined predominately not by device technology but by network topology; more, smaller crossbars consume less area thanks to the high structural sparsity found in larger, brain-inspired SNNs. Motivated by continuing increases in SNN sparsity due to improvements in training methods, we propose utilizing heterogeneous crossbar sizes to further reduce area consumption. This approach was previously unachievable as prior compiler studies only explored solutions targeting homogeneous MCAs. Our work improves on the state-of-the-art by providing Integer Linear Programming (ILP) formulations supporting arbitrarily heterogeneous architectures. By modeling axonal interactions between neurons, our methods produce better mappings while removing inhibitive a priori knowledge requirements. We first show a 16.7-27.6% reduction in area consumption for square-crossbar homogeneous architectures. Then, we demonstrate 66.9-72.7% further reduction when using a reasonable configuration of heterogeneous crossbar dimensions. Next, we present a new optimization formulation capable of minimizing the number of inter-crossbar routes. When applied to solutions already near-optimal in area, an 11.9-26.4% routing reduction is observed without impacting area consumption. Finally, we present a profile-guided optimization capable of minimizing the number of runtime spikes between crossbars. Compared to the best-area-then-route optimized solutions, we observe a further 0.5-14.8% inter-crossbar spike reduction while requiring 1–3 orders of magnitude less solver time.

Pohl, Devin [ORNL] (ORCID:0009000040149027)↗

Coulomb excitation of 102 Ru with 12 C and 16 O

Here, the Coulomb excitation of 102 Ru was performed with beams of 12 C and 16 O ions. The beam particles scattered at forward angles were momentum analyzed with a magnetic spectrograph. The resolution achieved enabled the populations of the $2^{+}_{1}$ state, the unresolved $2^{+}_{2}$/ $4^{+}_{1}$, and $2^{+}_{4}$/$3^{-}_{1}$, doublets of states, and the $3^{-}_{2}$ state to be determined as a function of the scattering angle. These populations are compared with gosia calculations, yielding B(E2; $2^{+}_{1}$→$0^{+}_{1}$) = 41.5 ± 2.3 W.u., B(E2; $2^{+}_{2}$→$0^{+}_{1}$) = 1.75 ± 0.11 W.u., B(E3; $3^{-}_{1}$→$0^{+}_{1}$) = 31.5 ± 3.5 W.u., and B(E3; $3^{-}_{2}$→$0^{+}_{1}$) = 6.8 ± 0.5 W.u. The B(E3; $3^{-}_{1}$→$0^{+}_{1}$) value is significantly larger than previously measured. The weakly populated $2^{+}_{3}$ state, presumed to be a member of the band built on the $0^{+}_{2}$ state, was observed clearly for a single angle only, and a fit to its population results in B(E2; $2^{+}_{3}$→$0^{+}_{1}$) = 0.053 ± 0.011 W.u. Using the known γ -ray branching ratios for the $2^{+}_{3}$ level, the B(E2; $2^{+}_{3}$→$0^{+}_{2}$) value is calculated to be 18 ± 4 W.u., substantially less than the B(E2; $2^{+}_{1}$→$0^{+}_{1}$) . This suggests that the deformation of the $0^{+}_{2}$ state is lower than that of the $0^{+}_{1}$ state. The results are compared with beyond-mean-field calculations with the Gogny-D1S interaction using the symmetry-conserving configuration-mixing method.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗