Search

AI workloads, particularly those driven by deep learning, are introducing novel usage patterns to high-performance computing (HPC) systems that are not comprehensively captured by standard HPC benchmarks. As one of the largest academic research centers dedicated to deep learning, Mila identified the need to develop a custom benchmarking suite to address the diverse requirements of its community, which consists of over 1,000 researchers. This report introduces Milabench, the resulting benchmarking suite. Its design was informed by an extensive literature review encompassing 867 papers, as well as surveys conducted with Mila researchers. This rigorous process led to the selection of 26 primary benchmarks tailored for procurement evaluations, alongside 16 optional benchmarks for in-depth analysis. We detail the design methodology, the structure of the benchmarking suite, and provide performance evaluations using GPUs from NVIDIA, AMD, and Intel. The Milabench suite is open source and can be accessed at github.com/mila-iqia/milabench.

For a hypergraph $\mathbb{H}$ on $[n]$, the hypergraphic poset $P_\mathbb{H}$ is the transitive closure of the oriented skeleton of the hypergraphic polytope $\triangle_\mathbb{H}$ (the Minkowski sum of the standard simplices $\triangle_H$ for all $H \in \mathbb{H}$). Hypergraphic posets include the weak order for the permutahedron (when $\mathbb{H}$ is the complete graph on $[n]$) and the Tamari lattice for the associahedron (when $\mathbb{H}$ is the set of all intervals of $[n]$), which motivates the study of lattice properties of hypergraphic posets. In this paper, we focus on interval hypergraphs, where all hyperedges are intervals of $[n]$. We characterize the interval hypergraphs $\mathbb{I}$ for which $P_\mathbb{I}$ is a lattice, a distributive lattice, a semidistributive lattice, and a lattice quotient of the weak order., Comment: 30 pages, 8 figures

The fermionic many-body problem in the strong correlation regime is notoriously difficult to tackle. In a previous work (Phys. Rev. B 101, 045109 (2020)), we have proposed to extend the single-reference coupled-cluster (SRCC) method to the strong correlation regime using low-rank tensor decompositions (LRTD) to express the cluster operator, without truncating it with respect to the number of excitations. For that purpose, we have proposed a new type of LRTD called ``superpositions of tree-tensor networks'' (STTN), which use the same set of building blocs to define all the tensors involved in the CC equations, and combine different ``channels'', i.e. different types of pairing among excited particles and holes, in the decomposition of a given tensor. Those two principles are aimed at globally minimizing the total number of free parameters required to accurately represent the ground state. In this work, we show that STTN can indeed be compact and accurate representations of strongly correlated ground states by using them to express the CC cluster operator amplitudes and wave function coefficients of exact grounds states of small two-dimensional Hubbard clusters, at half-filling, up to three particle-hole excitations. We show the compactness of STTN by using a number of free parameters smaller than the number of equations in the CCSD approximation, i.e. much smaller than the number of fitted tensor elements. We find that, for the systems considered, the STTN are more accurate as the size of the system increases and that combining different channels in the decompositions of the most strongly correlated tensors is crucial to obtain good accuracy., Comment: 12 pages, 12 figures

In this study, the Method of Characteristics (MOC) for Delayed Neutron Precursors (DNPs) is used to solve the precursors balance equation with turbulent diffusion. The diffusivity of DNPs, significantly higher than molecular diffusivity, emerges in turbulent flows from the time-averaging of the DNPs mass balance equation. To integrate this effect within the MOC framework, the advection-reaction component of the DNPs balance equation is solved using the MOC, while the diffusive source is computed from the concentration of the previous iteration. The method is validated on a 2D recirculating pipe reactor with high Reynolds number flow, comparing the MOC with diffusion to a standard finite volume (FV) discretization of the fission products balance equation. Additionally, the impact of the diffusivity term on DNP distributions and reactor reactivity is quantified as a function of the turbulent Schmidt number.

We demonstrate a method for radionuclide assay that is spectroscopic with 100 % counting efficiency for alpha decay. Advancing both cryogenic decay energy spectrometry (DES) and drop-on-demand inkjet metrology, a solution of Am-241 was assayed for massic activity (Bq/g) with a relative combined standard uncertainty less than 1 %. We implement live-timed counting, spectroscopic analysis, validation by liquid scintillation (LS) counting, and confirmation of quantitative solution transfer. Experimental DES spectra are well modeled with a Monte Carlo simulation. The model was further used to simulate Pu-238 and Pu-240 impurities, calculate detection limits, and demonstrate the potential for tracer-free multi-nuclide analysis, which will be valuable for new cancer therapeutics based on decay chains, Standard Reference Materials (SRMs) containing impurities, and more widely in nuclear energy, environmental monitoring, security, and forensics.

There is a tremendous amount of interest in AI/ML technologies due to the proliferation of generative AI applications such as ChatGPT. This trend has significantly increased demand on GPUs, which are the workhorses for training AI models. Due to the high costs of GPUs and lacking supply, it has become of interest to optimize GPU usage in HPC centers. MIT Lincoln Laboratory Supercomputing Center (LLSC) has developed an easy-to-use GPU sharing feature supported by LLSC-developed tools including LLsub and LLMapReduce. This approach overcomes some of the limitations with the existing methods for GPU sharing. This allows users to apply GPU sharing whenever possible while they are developing their AI/ML models and/or doing parametric study on their AI models or executing other GPU applications. Based on our initial experimental results with GPU sharing, GPU sharing with triples mode is easy to use and achieved significant improvement in GPU usage and throughput performance for certain types of AI applications.

The increasing use and cost of high performance computing (HPC) requires new easy-to-use tools to enable HPC users and HPC systems engineers to transparently understand the utilization of resources. The MIT Lincoln Laboratory Supercomputing Center (LLSC) has developed a simple command, LLload, to monitor and characterize HPC workloads. LLload plays an important role in identifying opportunities for better utilization of compute resources. LLload can be used to monitor jobs both programmatically and interactively. LLload can characterize users' jobs using various LLload options to achieve better efficiency. This information can be used to inform the user to optimize HPC workloads and improve both CPU and GPU utilization. This includes improvements using judicious oversubscription of the computing resources. Preliminary results suggest significant improvement in GPU utilization and overall throughput performance with GPU overloading in some cases. By enabling users to observe and fix incorrect job submission and/or inappropriate execution setups, LLload can increase the resource usage and improve the overall throughput performance. LLload is a light-weight, easy-to-use tool for both HPC users and HPC systems engineers to monitor HPC workloads to improve system utilization and efficiency.

In nuclear reactors, Delayed Neutron Precursors (DNPs) are important for reactor safety and operation. In liquid nuclear fuels, DNPs are transported by the flow, and an advection-reaction balance equation for their concentration must be solved in addition to the Neutron Balance Equation. This research paper applies the method of characteristics to solve the DNPs equation, using techniques previously developed in petroleum engineering and groundwater analysis. The calculation strategy incorporates a pathline generation algorithm covering all the mesh cells of the geometry for concentration calculations, and pathlines are reconstructed for both step and piecewise linear velocity fields. The analytical solutions are used on a Cartesian mesh to compute the DNPs concentrations assuming a step source of DNPs. The results obtained on the steady state phases of a benchmark problem illustrate this new method's capability to solve advection-dominated problems in liquid-fueled reactors. The method developed in this work is suited to smooth velocity fields (e.g., laminar or RANS flows) where pathlines can be tracked, and where DNP diffusivity is negligible.

GD1400AB was one of the first known white dwarf$+$brown dwarf binaries, and is the only one of these systems where the white dwarf is a ZZ Ceti pulsator. Here we present both radial velocity measurements and time series photometry, analysing both the white dwarf pulsations and the effects of irradiation on the brown dwarf. We find the brightness temperatures of 1760$/pm$10 K for the night side and 1860$/pm$10 K for the day side indicate the brown dwarf is hotter than spectra have previously suggested, although brightness temperatures calculated using a larger radius for the brown dwarf are consistent with previously determined spectral types. We also discuss the likely evolutionary pathway of this binary, and put its common envelope phase into context with the other known systems., Comment: 10 pages, 9 figures, accepted for publication in MNRAS: https://doi.org/10.1093/mnras/stae2301

Real-time supercomputing performance analysis is a critical aspect of evaluating and optimizing computational systems in a dynamic user environment. The operation of supercomputers produce vast quantities of analytic data from multiple sources and of varying types so compiling this data in an efficient matter is critical to the process. MIT Lincoln Laboratory Supercomputing Center has been utilizing the Unity 3D game engine to create a Digital Twin of our supercomputing systems for several years to perform system monitoring. Unity offers robust visualization capabilities making it ideal for creating a sophisticated representation of the computational processes. As we scale the systems to include a diversity of resources such as accelerators and the addition of more users, we need to implement new analysis tools for the monitoring system. The workloads in research continuously change, as does the capability of Unity, and this allows us to adapt our monitoring tools to scale and incorporate features enabling efficient replay of system wide events, user isolation, and machine level granularity. Our system fully takes advantage of the modern capabilities of the Unity Engine in a way that intuitively represents the real time workload performed on a supercomputer. It allows HPC system engineers to quickly diagnose usage related errors with its responsive user interface which scales efficiently with large data sets.

HPC systems used for research run a wide variety of software and workflows. This software is often written or modified by users to meet the needs of their research projects, and rarely is built with security in mind. In this paper we explore several of the key techniques that MIT Lincoln Laboratory Supercomputing Center has deployed on its systems to manage the security implications of these workflows by providing enforced separation for processes, filesystem access, network traffic, and accelerators to make every user feel like they are running on a personal HPC.

The MIT/IEEE/Amazon GraphChallenge encourages community approaches to developing new solutions for analyzing graphs and sparse data derived from social media, sensor feeds, and scientific data to discover relationships between events as they unfold in the field. The anonymized network sensing Graph Challenge seeks to enable large, open, community-based approaches to protecting networks. Many large-scale networking problems can only be solved with community access to very broad data sets with the highest regard for privacy and strong community buy-in. Such approaches often require community-based data sharing. In the broader networking community (commercial, federal, and academia) anonymized source-to-destination traffic matrices with standard data sharing agreements have emerged as a data product that can meet many of these requirements. This challenge provides an opportunity to highlight novel approaches for optimizing the construction and analysis of anonymized traffic matrices using over 100 billion network packets derived from the largest Internet telescope in the world (CAIDA). This challenge specifies the anonymization, construction, and analysis of these traffic matrices. A GraphBLAS reference implementation is provided, but the use of GraphBLAS is not required in this Graph Challenge. As with prior Graph Challenges the goal is to provide a well-defined context for demonstrating innovation. Graph Challenge participants are free to select (with accompanying explanation) the Graph Challenge elements that are appropriate for highlighting their innovations., Comment: Accepted to IEEE HPEC 2024

We present a study of the white dwarf (WD) cooling sequence (CS) in the globular cluster (GC) Omega Centauri, the primary goal of a dedicated Hubble Space Telescope (HST) programme. Our analysis has revealed that the peak at the termination of the WD CS is located at $m_{\rm F606W}$=30.1$\pm$0.2 (equivalent to $V$$\sim$31). The brighter part of Omega Centauri's WD CS is consistent with the presence of massive He-core WDs, in agreement with previous HST analyses with ultraviolet and blue filters. Comparative analyses of the WD luminosity function (LF) with theoretical counterparts have shown that a single-age population for the cluster is compatible with the data. However, an analysis of just the WD LF cannot entirely exclude the possibility of an age range, due to uncertainties in the present-day WD mass function, with a star formation history potentially spanning up to 5 billion years, predominantly comprising stars about 13 Gyr old, and with just a minority potentially as young as 8 Gyr. This underscores the need for global spectroscopic and photometric investigations that include simultaneously the WD populations together with the previous evolutionary phases to fully understand the cluster's diverse chemical compositions and ages., Comment: 10 pages, 9 figures. Accepted for publication in Astronomy & Astrophysics on Sep 5, 2024

Understanding what is normal is a key aspect of protecting a domain. Other domains invest heavily in observational science to develop models of normal behavior to better detect anomalies. Recent advances in high performance graph libraries, such as the GraphBLAS, coupled with supercomputers enables processing of the trillions of observations required. We leverage this approach to synthesize low-parameter observational models of anonymized Internet traffic with a high regard for privacy., Comment: Accepted to IEEE HPEC, 7 pages, 6 figures, 1 table, 41 references

This work develops a zero-shot mechanism for an agent to satisfy a Linear Temporal Logic (LTL) specification given existing task primitives. Oftentimes, autonomous robots need to satisfy spatial and temporal goals that are unknown until run time. Prior research addresses the problem by learning policies that are capable of executing a high-level task specified using LTL, but they incorporate the specification into the learning process; therefore, any change to the specification requires retraining the policy. Other related research addresses the problem by creating skill-machines which, given a specification change, do not require full policy retraining but require fine-tuning on the skill-machine to guarantee satisfaction. We present a more a flexible approach -- to learn a set of minimum-violation (MV) task primitive policies that can be used to satisfy arbitrary LTL specifications without retraining or fine-tuning. Task primitives can be learned offline using reinforcement learning (RL) methods and combined using Boolean composition at deployment. This work focuses on creating and pruning a transition system (TS) representation of the environment in order to solve for deterministic, non-ambiguous, and feasible solutions to LTL specifications given an environment and a set of MV task primitive policies. We show that our pruned TS is deterministic, contains no unrealizable transitions, and is sound. Through simulation, we show that our approach is executable and we verify our MV policies produce the expected symbols., Comment: 16 pages, 11 figures

We present follow-up spectroscopy and a detailed model atmosphere analysis of 29 wide double white dwarfs, including eight systems with a crystallized C/O core member. We use state-of-the-art evolutionary models to constrain the physical parameters of each star, including the total age. Assuming that the members of wide binaries are coeval, any age difference between the binary members can be used to test the cooling physics for white dwarf stars, including potential delays due to crystallization and $^{22}$Ne distillation. We use our control sample of 14 wide binaries with non-crystallized members to show that this method works well; the control sample shows an age difference of only $\Delta$Age = $-0.03 \pm$ 0.15 Gyr between its members. For the eight crystallized C/O core systems we find a cooling anomaly of $\Delta$Age= 1.13$^{+1.20}_{-1.07}$ Gyr. Even though our results are consistent with a small additional cooling delay ($\sim1$ Gyr) from $^{22}$Ne distillation and other neutron-rich impurities, the large uncertainties make this result not statistically significant. Nevertheless, we rule out cooling delays longer than 3.6 Gyr at the 99.7% ($3\sigma$) confidence level for 0.6-0.9 $M_{\odot}$ white dwarfs. Further progress requires larger samples of wide binaries with crystallized massive white dwarf members. We provide a list of subgiant + white dwarf binaries that could be used for this purpose in the future., Comment: 19 pages, including 6 figures and 4 tables, accepted for publication in ApJ

Wide-angle fisheye images are becoming increasingly common for perception tasks in applications such as robotics, security, and mobility (e.g. drones, avionics). However, current models often either ignore the distortions in wide-angle images or are not suitable to perform pixel-level tasks. In this paper, we present an encoder-decoder model based on a radial transformer architecture that adapts to distortions in wide-angle lenses by leveraging the physical characteristics defined by the radial distortion profile. In contrast to the original model, which only performs classification tasks, we introduce a U-Net architecture, DarSwin-Unet, designed for pixel level tasks. Furthermore, we propose a novel strategy that minimizes sparsity when sampling the image for creating its input tokens. Our approach enhances the model capability to handle pixel-level tasks in wide-angle fisheye images, making it more effective for real-world applications. Compared to other baselines, DarSwin-Unet achieves the best results across different datasets, with significant gains when trained on bounded levels of distortions (very low, low, medium, and high) and tested on all, including out-of-distribution distortions. We demonstrate its performance on depth estimation and show through extensive experiments that DarSwin-Unet can perform zero-shot adaptation to unseen distortions of different wide-angle lenses.

We present a detailed model atmosphere analysis of massive white dwarfs with $M > 0.9~M_\odot$ and $T_{\rm eff}\geq11,000$ K in the Montreal White Dwarf Database 100 pc sample and the Pan-STARRS footprint. We obtained follow-up optical spectroscopy of 109 objects with no previous spectral classification in the literature. Our spectroscopic follow-up is now complete for all 204 objects in the sample. We find 118 normal DA white dwarfs, including 45 massive DAs near the ZZ Ceti instability strip. There are no normal massive DBs: the six DBs in the sample are strongly magnetic and/or rapidly rotating. There are 20 massive DQ white dwarfs in our sample, and all are found in the crystallization sequence. In addition, 66 targets are magnetic (32% of the sample). We use magnetic white dwarf atmosphere models to constrain the field strength and geometry using offset dipole models. We also use magnetism, kinematics, and rotation measurements to constrain the fraction of merger remnant candidates among this population. The merger fraction of this sample increases from 25% for 0.9-$1~M_{\odot}$ white dwarfs to 49% for 1.2-$1.3~M_{\odot}$. However, this fraction is as high as $78_{-7}^{+4}$% for 1.1-$1.2~M_{\odot}$ white dwarfs. Previous works have demonstrated that 5-9% of high-mass white dwarfs stop cooling for $\sim8$ Gyr due to the $^{22}$Ne distillation process, which leads to an overdensity of Q-branch stars in the solar neighborhood. We demonstrate that the over-abundance of the merger remnant candidates in our sample is likely due to the same process., Comment: 34 pages, 16 figures, accepted for publication in ApJ

One of the more complex tasks for researchers using HPC systems is performance monitoring and tuning of their applications. Developing a practice of continuous performance improvement, both for speed-up and efficient use of resources is essential to the long term success of both the HPC practitioner and the research project. Profiling tools provide a nice view of the performance of an application but often have a steep learning curve and rarely provide an easy to interpret view of resource utilization. Lower level tools such as top and htop provide a view of resource utilization for those familiar and comfortable with Linux but a barrier for newer HPC practitioners. To expand the existing profiling and job monitoring options, the MIT Lincoln Laboratory Supercomputing Center created LLoad, a tool that captures a snapshot of the resources being used by a job on a per user basis. LLload is a tool built from standard HPC tools that provides an easy way for a researcher to track resource usage of active jobs. We explain how the tool was designed and implemented and provide insight into how it is used to aid new researchers in developing their performance monitoring skills as well as guide researchers in their resource requests.

In this paper, we give a rule for the multiplication of a Schubert class by a tautological class in the (small) quantum cohomology ring of the flag manifold. As an intermediate step, we establish a formula for the multiplication of a Schubert class by a quantum Schur polynomial indexed by a hook partition. This entails a detailed analysis of chains and intervals in the quantum Bruhat order. This analysis allows us to use results of Leung--Li and of Postnikov to reduce quantum products by hook Schur polynomials to the (known) classical product., Comment: 34 pages, 16 color pictures, some text and graphic in color. Full paper with all proofs

Commercially impactful quantum algorithms such as quantum chemistry and Shor's algorithm require a number of qubits and gates far beyond the capacity of any existing quantum processor. Distributed architectures, which scale horizontally by networking modules, provide a route to commercial utility and will eventually surpass the capability of any single quantum computing module. Such processors consume remote entanglement distributed between modules to realize distributed quantum logic. Networked quantum computers will therefore require the capability to rapidly distribute high fidelity entanglement between modules. Here we present preliminary demonstrations of some key distributed quantum computing protocols on silicon T centres in isotopically-enriched silicon. We demonstrate the distribution of entanglement between modules and consume it to apply a teleported gate sequence, establishing a proof-of-concept for T centres as a distributed quantum computing and networking platform., Comment: 14 pages, 13 figures

Companies are facing a persistent and growing skills gap, which has been exacerbated by the tight labor market. Employers use industry certifications, college degrees, boot camps, and badges to measure job skills, but until recently it was impossible to compare certifications or degree attainment. This paper describes credential transparency and its benefits--how it provides specific details on a credential program's length, cost, competencies, skills provided, level of mastery targeted, earnings potential, and employment outcomes, and allows individuals to compare credentials to see which is more likely to result in a preferred career or higher wages. The federal government has taken steps to promote the use of credential transparency and currently 26 states are engaged in important work around its use, allowing them to measure return on investment, align credential offerings with economic needs, and support the creation of digital transcripts and learning and employment records.

We report on transport characteristics of field effect two-dimensional electron gases (2DEG) in 24 nm wide indium arsenide surface quantum wells. High quality single-subband magnetotransport with clear quantized integer quantum Hall plateaus are observed to filling factor $\nu=2$ in magnetic fields of up to B = 18 T, at electron densities up to 8$\times 10^{11}$ /cm$^2$. Peak mobility is 11,000 cm$^2$/Vs at 2$\times 10^{12}$ /cm$^2$. Large Rashba spin-orbit coefficients up to 124 meV$\cdot$\r{A} are obtained through weak anti-localization (WAL) measurements. Proximitized superconductivity is demonstrated in Nb-based superconductor-normal-superconductor (SNS) junctions, yielding 78$-$99% interface transparencies from superconducting contacts fabricated ex-situ (post-growth), using two commonly-used experimental techniques for measuring transparencies. These transparencies are on a par with those reported for epitaxially-grown superconductors. These SNS junctions show characteristic voltages $I_c R_{\text{N}}$ up to 870 $\mu$V and critical current densities up to 9.6 $\mu$A/$\mu$m, among the largest values reported for Nb-InAs SNS devices., Comment: main text 8 pages with 3 figures; supplementary material 17 pages with 11 figures

We propose a new method to investigate signatures of a quantum gravity phase in the primordial state of cosmological perturbations. We formulate and study a quantum model of a perturbed Friedmann-Lemaitre-Robertson-Walker universe beyond a tensor-product Born-Oppenheimer-like factorization, that is, without restricting the wave function of the universe to the product of the background and perturbation wave functions. We show that the quantum dynamics generically does not preserve the product form of the universe's wave function, which spontaneously evolves into a more general entangled state. Upon expanding this state in a suitable basis of background wave functions and setting Gaussian initial conditions for the perturbations, we numerically find that each of these wave functions becomes associated with a non-Gaussian state of an inhomogeneous perturbation., Comment: 10 pages, 7 figures

We present James Webb Space Telescope observations of the globular cluster NGC 6397 and use them to extend to infrared wavelengths the characterization of the cluster's entire white dwarf (WD) cooling sequence (CS). The data allows us to probe fundamental astrophysical WD properties and to search for evidence in their colors for (or against) the existence of ancient planetary systems. The existing archival Hubble Space Telescope imaging data obtained ~18 years ago reach ultra-deep optical magnitudes (V~31) and allow us to derive a near-perfect separation between field and cluster members. We detect an apparent split in the lower part of the WD CS of NGC 6397. The red part of the WD CS, containing about 25% of the total, exhibits significant IR-excess of up to Delta m_F322W2 ~ 0.5 mag. These infrared excesses require both theoretical and observational follow-ups to confirm their veracity and to ascertain their true nature., Comment: 16 pages, 10+A2 figures. Accepted for publication on Astronomische Nachrichten on 2nd May 2024

Four years after the discovery of a unique DAQ white dwarf with a hydrogen-dominated and carbon-rich atmosphere, we report the discovery of four new DAQ white dwarfs, including two that were not recognized properly in the literature. We find all five DAQs in a relatively narrow mass and temperature range of $M=1.14-1.19~M_{\odot}$ and $T_{\rm eff}=13,000-17,000$ K. In addition, at least two show photometric variations due to rapid rotation with $\approx10$ min periods. All five are also kinematically old, but appear photometrically young with estimated cooling ages of about 1 Gyr based on standard cooling tracks, and their masses are roughly twice the mass of the most common white dwarfs in the solar neighborhood. These characteristics are smoking gun signatures of white dwarf merger remnants. Comparing the DAQ sample with warm DQ white dwarfs, we demonstrate that there is a range of hydrogen abundances among the warm DQ population, and the distinction between DAQ and warm DQ white dwarfs is superficial. We discuss the potential evolutionary channels for the emergence of the DAQ subclass, and suggest that DAQ white dwarfs are trapped on the crystallization sequence, and may remain there for a significant fraction of the Hubble time., Comment: ApJ, in press

Among superconductor/semiconductor hybrid structures, in-situ aluminum (Al) grown on InGaAs/InAs is widely pursued for the experimental realization of Majorana Zero Mode quasiparticles. This is due to the high carrier mobility, low effective mass, and large Land\'e g-factor of InAs, coupled with the relatively high value of the in-plane critical magnetic field in thin Al films. However, growing a thin, continuous Al layer using the Molecular Beam Epitaxy (MBE) is challenging due to aluminum's high surface mobility and tendency for 3D nucleation on semiconductor surfaces. A study of epitaxial Al thin film growth on In0.75Ga0.25As with MBE is presented, focusing on the effects of the Al growth rate and substrate temperature on the nucleation of Al layers. We find that for low deposition rates, 0.1 {\AA}/s and 0.5 {\AA}/s, the growth continues in 3D mode during the deposition of the nominal 100 {\AA} of Al, resulting in isolated Al islands. However, for growth rates of 1.5 {\AA}/s and above, the 3D growth mode quickly transitions into island coalescence, leading to a uniform 2D Al layer. Moreover, this transition is very abrupt, happening over an Al flux increase of less than 1%. We discuss the growth mechanisms explaining these observations. The results give new insights into the kinetics of Al deposition and show that with sufficiently high Al flux, a 2D growth on substrates at close to room temperature can be achieved already within the first few Al monolayers. This eliminates the need for complex cryogenic substrate cooling and paves the way for the development of high-quality superconductor-semiconductor interfaces in standard MBE systems., Comment: The following article has been published in Journal of Vacuum Science & Technology A. DOI: https://doi.org/10.1116/6.0003459

Hot Jupiters were many of the first exoplanets discovered in the 1990s, but in the decades since their discovery, the mysteries surrounding their origins remain. Here, we present nine new hot Jupiters (TOI-1855 b, TOI-2107 b, TOI-2368 b, TOI-3321 b, TOI-3894 b, TOI-3919 b, TOI-4153 b, TOI-5232 b, and TOI-5301 b) discovered by NASA's TESS mission and confirmed using ground-based imaging and spectroscopy. These discoveries are the first in a series of papers named the Migration and Evolution of giant ExoPlanets (MEEP) survey and are part of an ongoing effort to build a complete sample of hot Jupiters orbiting FGK stars, with a limiting Gaia $G$-band magnitude of 12.5. This effort aims to use homogeneous detection and analysis techniques to generate a set of precisely measured stellar and planetary properties that is ripe for statistical analysis. The nine planets presented in this work occupy a range of masses (0.55 Jupiter masses (M$_{\rm{J}}$) $<$ M$_{\rm{P}}$ $<$ 3.88 M$_{\rm{J}}$) and sizes (0.967 Jupiter radii (R$_{\rm{J}}$) $<$ R$_{\rm{P}}$ $<$ 1.438 R$_{\rm{J}}$) and orbit stars that range in temperature from 5360 K $<$ Teff $<$ 6860 K with Gaia $G$-band magnitudes ranging from 11.1 to 12.7. Two of the planets in our sample have detectable orbital eccentricity: TOI-3919 b ($e = 0.259^{+0.033}_{-0.036}$) and TOI-5301 b ($e = 0.33^{+0.11}_{-0.10}$). These eccentric planets join a growing sample of eccentric hot Jupiters that are consistent with high-eccentricity tidal migration, one of the three most prominent theories explaining hot Jupiter formation and evolution., Comment: 35 pages, 7 tables, and 14 figures. Submitted to AAS Journals on 2023 Dec 28

We report the discovery of spectroscopic variations in the magnetic DBA white dwarf SDSS J091016.43+210554.2. Follow-up time-resolved spectroscopy at the Apache Point Observatory (APO) and the MMT show significant variations in the H absorption lines over a rotation period of 7.7 or 11.3 h. Unlike recent targets that show similar discrepancies in their H and He line profiles, such as GD 323 and Janus (ZTF J203349.8+322901.1), SDSS J091016.43+210554.2 is confirmed to be magnetic, with a field strength derived from Zeeman-split H and He lines of B ~ 0.5 MG. Model fits using a H and He atmosphere with a constant abundance ratio across the surface fail to match our time-resolved spectra. On the other hand, we obtain excellent fits using magnetic atmosphere models with varying H/He surface abundance ratios. We use the oblique rotator model to fit the system geometry. The observed spectroscopic variations can be explained by a magnetic inhomogeneous atmosphere where the magnetic axis is offset from the rotation axis by beta = 52 degrees, and the inclination angle between the line of sight and the rotation axis is i = 13 - 16 degrees. This magnetic white dwarf offers a unique opportunity to study the effect of the magnetic field on surface abundances. We propose a model where H is brought to the surface from the deep interior more efficiently along the magnetic field lines, thus producing H polar caps, Comment: 12 pages, 13 figures, accepted for publication in MNRAS

We propose a conjecture extending the classical construction of elliptic units to complex cubic number fields $K$. The conjecture concerns special values of the elliptic gamma function, a holomorphic function of three complex variables arising in mathematical physics whose transformation properties under $\mathrm{SL}_3(\mathbf{Z})$ were studied by Felder and Varchenko in the early 2000s. Using this function we construct complex numbers that we conjecture to be units in narrow ray class fields of $K$. We also propose a reciprocity law for the action of the Galois group on these units in the style of Shimura. To support our conjecture we offer numerical evidence and also prove a new type of Kronecker limit formula relating the logarithm of the modulus of these complex numbers to the derivatives at $s = 0$ of partial zeta functions of $K$. Our constructions unveil the role played by the elliptic gamma function in Hilbert's twelfth problem for complex cubic fields., Comment: 44 pages, 1 figure

We have discovered a class of dynamically stable coherent states for motion on the half-line. The regularization of the half-line boundary and the consequent quantum motion are expounded within the framework of covariant affine quantization, although alternative approaches are also feasible. The former approach is rooted in affine coherent states and offers a consistent semiclassical representation of quantum motion. However, this method has been known to possess two shortcomings: (a) the dependence of affine coherent states on the choice of a vector, denoted as 'fiducial vector' (which remains unspecified), introduces significant arbitrariness in boundary regularization, and (b) regardless of the choice of 'fiducial vector,' affine coherent states fail to evolve parametrically under the Schr\"odinger equation, thus limiting the accuracy of the semiclassical description. This limitation, in particular, hampers their suitability for approximating the evolution of compound observables. We demonstrate that a distinct and more refined definition of affine coherent states can simultaneously address both of these issues. In other words, these new affine coherent states exhibit parametric evolution only when the 'fiducial vector,' denoted as $|\psi_0>$, possesses a highly specific character, such as being an eigenstate of a well-defined Hamiltonian. Our discovery holds significant relevance in the field of quantum cosmology, particularly in scenarios where the positive variable is the scale factor of the universe, and its regularized motion plays a crucial role in avoiding the big-bang singularity., Comment: 15 pages, 3 figures

Low-power small form factor data processing units (DPUs) enable offloading and acceleration of a broad range of networking and security services. DPUs have accelerated the transition to programmable networking by enabling the replacement of FPGAs/ASICs in a wide range of network oriented devices. The GraphBLAS sparse matrix graph open standard math library is well-suited for constructing anonymized hypersparse traffic matrices of network traffic which can enable a wide range of network analytics. This paper measures the performance of the GraphBLAS on an ARM based NVIDIA DPU (BlueField 2) and, to the best of our knowledge, represents the first reported GraphBLAS results on a DPU and/or ARM based system. Anonymized hypersparse traffic matrices were constructed at a rate of over 18 million packets per second.

Large language models (LLMs) have exploded in popularity due to their new generative capabilities that go far beyond prior state-of-the-art. These technologies are increasingly being leveraged in various domains such as law, finance, and medicine. However, these models carry significant computational challenges, especially the compute and energy costs required for inference. Inference energy costs already receive less attention than the energy costs of training LLMs -- despite how often these large models are called on to conduct inference in reality (e.g., ChatGPT). As these state-of-the-art LLMs see increasing usage and deployment in various domains, a better understanding of their resource utilization is crucial for cost-savings, scaling performance, efficient hardware usage, and optimal inference strategies. In this paper, we describe experiments conducted to study the computational and energy utilization of inference with LLMs. We benchmark and conduct a preliminary analysis of the inference performance and inference energy costs of different sizes of LLaMA -- a recent state-of-the-art LLM -- developed by Meta AI on two generations of popular GPUs (NVIDIA V100 \& A100) and two datasets (Alpaca and GSM8K) to reflect the diverse set of tasks/benchmarks for LLMs in research and practice. We present the results of multi-node, multi-GPU inference using model sharding across up to 32 GPUs. To our knowledge, our work is the one of the first to study LLM inference performance from the perspective of computational and energy resources at this scale.

Expanding the scientific tools available to protect computer networks can be aided by a deeper understanding of the underlying statistical distributions of network traffic and their potential geometric interpretations. Analyses of large scale network observations provide a unique window into studying those underlying statistics. Newly developed GraphBLAS hypersparse matrices and D4M associative array technologies enable the efficient anonymized analysis of network traffic on the scale of trillions of events. This work analyzes over 100,000,000,000 anonymized packets from the largest Internet telescope (CAIDA) and over 10,000,000 anonymized sources from the largest commercial honeyfarm (GreyNoise). Neither CAIDA nor GreyNoise actively emit Internet traffic and provide distinct observations of unsolicited Internet traffic (primarily botnets and scanners). Analysis of these observations confirms the previously observed Cauchy-like distributions describing temporal correlations between Internet sources. The Gull lighthouse problem is a well-known geometric characterization of the standard Cauchy distribution and motivates a potential geometric interpretation for Internet observations. This work generalizes the Gull lighthouse problem to accommodate larger classes of coastlines, deriving a closed-form solution for the resulting probability distributions, stating and examining the inverse problem of identifying an appropriate coastline given a continuous probability distribution, identifying a geometric heuristic for solving this problem computationally, and applying that heuristic to examine the temporal geometry of different subsets of network observations. Application of this method to the CAIDA and GreyNoise data reveals a several orders of magnitude difference between known benign and other traffic which can lead to potentially novel ways to protect networks., Comment: 9 pages, 7 figures, IEEE HPEC 2023 (accepted)

The efficient frontier (EF) is a fundamental resource allocation problem where one has to find an optimal portfolio maximizing a reward at a given level of risk. This optimal solution is traditionally found by solving a convex optimization problem. In this paper, we introduce NeuralEF: a fast neural approximation framework that robustly forecasts the result of the EF convex optimization problem with respect to heterogeneous linear constraints and variable number of optimization inputs. By reformulating an optimization problem as a sequence to sequence problem, we show that NeuralEF is a viable solution to accelerate large-scale simulation while handling discontinuous behavior., Comment: Accepted at the Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS 2023)

pPython seeks to provide a parallel capability that provides good speed-up without sacrificing the ease of programming in Python by implementing partitioned global array semantics (PGAS) on top of a simple file-based messaging library (PythonMPI) in pure Python. pPython follows a SPMD (single program multiple data) model of computation. pPython runs on a single-node (e.g., a laptop) running Windows, Linux, or MacOS operating systems or on any combination of heterogeneous systems that support Python, including on a cluster through a Slurm scheduler interface so that pPython can be executed in a massively parallel computing environment. It is interesting to see what performance pPython can achieve compared to the traditional socket-based MPI communication because of its unique file-based messaging implementation. In this paper, we present the point-to-point and collective communication performances of pPython and compare them with those obtained by using mpi4py with OpenMPI. For large messages, pPython demonstrates comparable performance as compared to mpi4py., Comment: arXiv admin note: substantial text overlap with arXiv:2208.14908

Matrix/array analysis of networks can provide significant insight into their behavior and aid in their operation and protection. Prior work has demonstrated the analytic, performance, and compression capabilities of GraphBLAS (graphblas.org) hypersparse matrices and D4M (d4m.mit.edu) associative arrays (a mathematical superset of matrices). Obtaining the benefits of these capabilities requires integrating them into operational systems, which comes with its own unique challenges. This paper describes two examples of real-time operational implementations. First, is an operational GraphBLAS implementation that constructs anonymized hypersparse matrices on a high-bandwidth network tap. Second, is an operational D4M implementation that analyzes daily cloud gateway logs. The architectures of these implementations are presented. Detailed measurements of the resources and the performance are collected and analyzed. The implementations are capable of meeting their operational requirements using modest computational resources (a couple of processing cores). GraphBLAS is well-suited for low-level analysis of high-bandwidth connections with relatively structured network data. D4M is well-suited for higher-level analysis of more unstructured data. This work demonstrates that these technologies can be implemented in operational settings., Comment: Accepted to IEEE HPEC, 8 pages, 8 figures, 1 table, 69 references. arXiv admin note: text overlap with arXiv:2203.13934. text overlap with arXiv:2309.01806

Defending community-owned cyber space requires community-based efforts. Large-scale network observations that uphold the highest regard for privacy are key to protecting our shared cyberspace. Deployment of the necessary network sensors requires careful sensor placement, focusing, and calibration with significant volumes of network observations. This paper demonstrates novel focusing and calibration procedures on a multi-billion packet dataset using high-performance GraphBLAS anonymized hypersparse matrices. The run-time performance on a real-world data set confirms previously observed real-time processing rates for high-bandwidth links while achieving significant data compression. The output of the analysis demonstrates the effectiveness of these procedures at focusing the traffic matrix and revealing the underlying stable heavy-tail statistical distributions that are necessary for anomaly detection. A simple model of the corresponding probability of detection ($p_{\rm d}$) and probability of false alarm ($p_{\rm fa}$) for these distributions highlights the criticality of network sensor focusing and calibration. Once a sensor is properly focused and calibrated it is then in a position to carry out two of the central tenets of good cybersecurity: (1) continuous observation of the network and (2) minimizing unbrokered network connections., Comment: Accepted to IEEE HPEC, 9 pages, 12 figures, 1 table, 63 references, 2 appendices