arXiv subjects

Ying Xiong

Publications and source records attributed to Ying Xiong.

At least 55 records · Page 3Linked to original sources

arXiv2023

TSoR: TCP Socket over RDMA Container Network for Cloud Native Computing

Cloud-native containerized applications constantly seek high-performance and easy-to-operate container network solutions. RDMA network is a potential enabler with higher throughput and lower latency than the standard TCP/IP network stack. However, several challenges remain in equipping containerized applications with RDMA network: 1) How to deliver transparent improvements without modifying application code; 2) How to integrate RDMA-based network solutions with container orchestration systems; 3) How to efficiently utilize RDMA for container networks. In this paper, we present an RDMA-based container network solution, TCP Socket over RDMA (TSoR), which addresses all the above challenges. To transparently accelerate applications using POSIX socket interfaces without modifications, we integrate TSoR with a container runtime that can intercept system calls for socket interfaces. To be compatible with orchestration systems like Kubernetes, TSoR implements a container network following the Kubernetes network model and satisfies all requirements of the model. To leverage RDMA benefits, TSoR designs a high-performance network stack that efficiently transfers TCP traffic using RDMA network. Thus, TSoR provides a turn-key solution for existing Kubernetes clusters to adopt the high-performance RDMA network with minimal effort. Our evaluation results show that TSoR provides up to 2.3x higher throughput and 64\% lower latency for existing containerized applications, such as Redis key-value store and Node.js web server, with no code changes. TSoR code will be open-sourced.

cs.NI↗

arXiv2023

Hibernate Container: A Deflated Container Mode for Fast Startup and High-density Deployment in Serverless Computing

Serverless computing is a popular cloud computing paradigm, which requires low response latency to handle on-demand user requests. There are two prominent techniques employed for reducing the response latency: keep fully initialized containers alive (Warm Container) or reduce the new container startup (cold start) latency. This paper presents the 3rd container startup mode: Hibernate Container, which starts faster than the cold start container mode and consumes less memory than the Warm Container mode. Hibernate Container is essentially a "deflated" Warm Container. Its application memory is swapped out to disk, the freed memory is reclaimed and file based mmap memory is cleaned-up. The Hibernate Container's deflated memory is inflated in response to user requests. As Hibernate Container's application is fully initialized, its response latency is less than the cold start mode; and as the application memory is deflated, its memory consumption is less than the Warm Container mode. Additionally, when a Hibernate Container is "woken up" to process a request, the Woken-up Container has similar response latency to Warm Container but less memory consumption because not all the deflated memory needs to be inflated. We implemented the Hibernate technique as part of the open source Quark secure container runtime project and our test demonstrated that Hibernate Container consumes about 7\% to 25\% of the Warm Container memory. All of this results in a higher deployment density, lower latency and appreciable improvements in the overall system performance.

cs.DC↗

arXiv2023

Polariton-drag enabled quantum geometric photocurrents in high symmetry materials

Lowered symmetry enables access to a wide set of responses not typically accessible in high symmetry materials. Prime examples are time-reversal forbidden quantum geometric photocurrent responses (e.g., linear injection and circular shift photocurrents) that are thought to vanish in non-magnetic materials. Here we argue that polariton-drag processes enable to unblock such quantum geometric photocurrents even in non-magnetic and centrosymmetric materials. Strikingly, we uncover how a cooperative effect between finite q irradiation and the Fermi surface position leads to a polariton selective photoexcitation (PSP). PSP enables to directly address carriers within tight momentum resolved windows of the Fermi surface to yield giant enhancements of quantum geometric photocurrents. This selectivity enables to directly track momentum resolved quantum geometric quantities along the Fermi surface providing a new tool to interrogate the quantum geometry of high symmetry materials.

cond-mat.mes-hall↗

arXiv2023

Layer photovoltaic effect in van der Waals heterostructures

We argue that the layer electric polarization of noncentrosymmetric layered heterostructures can be generically controlled by light yielding a layer photovoltaic effect (LPE). The LPE possesses a rich phenomenology and can arise from myriad distinct mechanisms displaying strong sensitivity to symmetry (e.g., point group and time-reversal) as well as the presence/absence of a Fermi surface. We systematically classify these and unveil how LPE manifests for a range of light polarizations and even for unpolarized light. These unusual layer photoresponses can be realized in a range of layered heterostructures such as bilayer graphene aligned on hexagonal Boron Nitride and manifest sizeable layer polarization susceptibilities in the terahertz frequency range that can be used as novel means of bulk photodetection.

cond-mat.mes-hall↗

arXiv2022

A separation property for iterated function systems of similitudes

Let $E$ be the attractor of an iterated function system $\{ϕ_i(x)=ρR_ix+a_i\}_{i=1}^N$ on $\Bbb R^d$, where $0<ρ<1$, $a_i\in \Bbb R^d$ and $R_i$ are orthogonal transformations on $\Bbb R^d$. Suppose that $\{ϕ_i\}_{i=1}^N$ satisfies the open set condition, but not the strong separation condition. We show that $E$ can not be generated by any iterated function system of similitudes satisfying the strong separation condition. This gives a partial answer to a folklore question about the separation conditions on the generating iterated function systems of self-similar sets.

math.DS↗

arXiv2022

LightSeq2: Accelerated Training for Transformer-based Models on GPUs

Transformer-based neural models are used in many AI applications. Training these models is expensive, as it takes huge GPU resources and long duration. It is challenging because typical data like sentences have variable lengths, and Transformer's computation patterns are more complex than convolutional neural networks. Existing systems either only focus on model inference or optimization for only BERT-like encoder models. In this paper, we present LightSeq2, a system to accelerate training for a general family of Transformer models on GPUs. We propose a series of GPU optimization techniques tailored to the specific computation flow and memory access patterns of Transformer models. LightSeq2 supports many model architectures, including BERT (encoder-only), GPT (decoder-only), Transformer (encoder-decoder), and vision Transformer. Our experiments for a variety of models and benchmarks show that LightSeq2 is consistently faster (1.4-3.5x) than previous systems on different GPUs. In particular, it gains 308% training speedup compared with existing systems on a large public machine translation benchmark (WMT14 English-German).

cs.CL↗

arXiv2021

Sets of Exact Approximation Order by Complex rational numbers

For a nonincreasing function $ψ$, let $\textrm{Exact}(ψ)$ be the set of complex numbers that are approximable by complex rational numbers to order $ψ$ but to no better order. In this paper, we obtain the Hausdorff dimension and packing dimension of $\textrm{Exact}(ψ)$ when $ψ(x)=o(x^{-2})$. We also prove that the lower bound of the Hausdorff dimension is greater than $2-τ/(1-2τ)$ when $τ=\limsup_{x\to\infty}ψ(x)x^2$ small enough.

math.NT↗

arXiv2021

The difference between the Hurwitz continued fraction expansions of a complex number and its rational approximations

For regular continued fraction, if a real number $x$ and its rational approximation $p/q$ satisfying $|x-p/q|<1/q^2$, then, after deleting the last integer of the partial quotients of $p/q$, the sequence of the remaining partial quotients is a prefix of that of $x$. In this paper, we show that the situation is completely different if we consider the Hurwitz continued fraction expansions of a complex number and its rational approximations. More specifically, we consider the set $E(ψ)$ of complex numbers which are well approximated with the given bound $ψ$ and have quite different Hurwitz continued fraction expansions from that of their rational approximations. The Hausdorff and packing dimensions of such set are determined. It turns out that its packing dimension is always full for any given approximation bound $ψ$ and its Hausdorff dimension is equal to that of the $ψ$-approximable set $W(ψ)$ of complex numbers. As a consequence, we also obtain an analogue of the classical Jarník Theorem in real case.

math.NT↗

arXiv2021

Atomic configuration controlled photocurrent in van der Waals homostructures

Conventional photocurrents at a p-n junction depend on macroscopic built-in fields and are typically insensitive to the microscopic details of a crystal's atomic configuration. Here we demonstrate how atomic configuration can control photocurrent in van der Waals (vdW) materials. In particular, we find bulk shift photocurrents (SPC) can display a rich (atomic) configuration dependent phenomenology that range from contrasting SPC currents for different stacking arrangements in a vdW homostructure (e.g., AB vs BA stacking) to a strong light polarization dependence for SPC that align with crystallographic axes. Strikingly, we find that SPC in vdW homostructures can be directed by modest strain, yielding sizeable photocurrent magnitudes under unpolarized light irradiation and manifesting even in the absence of p-n junctions. These demonstrate that SPC are intimately linked to how the Bloch wavefunctions are embedded in real space, and enables a new macroscopic transport probe (photocurrent) of lattice-scale registration in vdW materials.

cond-mat.mes-hall↗

arXiv2021

LightSeq: A High Performance Inference Library for Transformers

Transformer, BERT and their variants have achieved great success in natural language processing. Since Transformer models are huge in size, serving these models is a challenge for real industrial applications. In this paper, we propose LightSeq, a highly efficient inference library for models in the Transformer family. LightSeq includes a series of GPU optimization techniques to to streamline the computation of neural layers and to reduce memory footprint. LightSeq can easily import models trained using PyTorch and Tensorflow. Experimental results on machine translation benchmarks show that LightSeq achieves up to 14x speedup compared with TensorFlow and 1.4x compared with FasterTransformer, a concurrent CUDA implementation. The code is available at https://github.com/bytedance/lightseq.

cs.MS↗

arXiv2020

Multistable excitonic Stark effect

The optical Stark effect is a tell-tale signature of coherent light-matter interaction in excitonic systems, wherein an irradiating light beam tunes exciton transition frequencies. Here we show that, when excitons are placed in a nanophotonic cavity, the excitonic Stark effect can become highly nonlinear, exhibiting multi-valued and hysteretic Stark shifts that depend on the history of the irradiating light. This multistable Stark effect (MSE) arises from feedback between the cavity mode occupation and excitonic population, mediated by the Stark-induced mutual tuning of the cavity and excitonic resonances. Strikingly, the MSE manifests even for very dilute exciton concentrations and can yield discontinuous Stark shift jumps of order meV. We expect that the MSE can be realized in readily available transition metal dichalcogenide excitonic systems placed in planar photonic cavities, at modest pump intensities. This phenomenon can provide new means to engineer coupled states of light and matter that can persist even in the single exciton limit.

cond-mat.mes-hall↗

arXiv2019

On the Separation Structure of Lalley-Gatzouras Fractals

We obtain a necessary and sufficient condition for Lalley-Gatzouras sets to be uniform disconnected. This enable us to find all Lalley-Gatzouras sets which are quasisymmetrically equivalent to the Cantor ternary set. As another application, we also study the limit behavior of the gap sequence of Lalley-Gatzouras sets.

math.MG↗

arXiv2018

On Assouad dimension and arithmetic progressions in sets defined by digit restrictions

We show that the set defined by digit restrictions contains arbitrarily long arithmetic progressions if and only if its Assouad dimension is one. Moreover, we show that for any $0\le s\le 1$, there exists some set on $\mathbb{R}$ with Hausdorff dimension $s$ whose Fourier dimension is zero and it contains arbitrarily long arithmetic progressions.

math.CA↗

arXiv2018

Every component of a fractal square is a Peano continuum

This paper concerns the local connectedness of components of self-similar sets. Given an equal partition of the unit square into n*n small squares, we may choose arbitrarily two or more of them and form an iterated function system. The attractor F resulted from this IFS is called a fractal square. We prove that every component of F is locally connected. The same result for three-dimensional analogues of F does not hold.

math.GN↗

arXiv2016

Affine embeddings of Cantor sets and dimension of $αβ$-sets

Let $E, F\subset {\Bbb R}^d$ be two self-similar sets, and suppose that $F$ can be affinely embedded into $E$. Under the assumption that $E$ is dust-like and has a small Hausdorff dimension, we prove the logarithmic commensurability between the contraction ratios of $E$ and $F$. This gives a partial affirmative answer to Conjecture 1.2 in \cite{FHR14}. The proof is based on our study of the box-counting dimension of a class of multi-rotation invariant sets on the unit circle, including the $αβ$-sets initially studied by Engelking and Katznelson.

math.CA↗

arXiv2015

Low-level Vision by Consensus in a Spatial Hierarchy of Regions

We introduce a multi-scale framework for low-level vision, where the goal is estimating physical scene values from image data---such as depth from stereo image pairs. The framework uses a dense, overlapping set of image regions at multiple scales and a "local model," such as a slanted-plane model for stereo disparity, that is expected to be valid piecewise across the visual field. Estimation is cast as optimization over a dichotomous mixture of variables, simultaneously determining which regions are inliers with respect to the local model (binary variables) and the correct co-ordinates in the local model space for each inlying region (continuous variables). When the regions are organized into a multi-scale hierarchy, optimization can occur in an efficient and parallel architecture, where distributed computational units iteratively perform calculations and share information through sparse connections between parents and children. The framework performs well on a standard benchmark for binocular stereo, and it produces a distributional scene representation that is appropriate for combining with higher-level reasoning and other low-level cues.

cs.CV↗

arXiv2014

Modeling Radiometric Uncertainty for Vision with Tone-mapped Color Images

To produce images that are suitable for display, tone-mapping is widely used in digital cameras to map linear color measurements into narrow gamuts with limited dynamic range. This introduces non-linear distortion that must be undone, through a radiometric calibration process, before computer vision systems can analyze such photographs radiometrically. This paper considers the inherent uncertainty of undoing the effects of tone-mapping. We observe that this uncertainty varies substantially across color space, making some pixels more reliable than others. We introduce a model for this uncertainty and a method for fitting it to a given camera or imaging pipeline. Once fit, the model provides for each pixel in a tone-mapped digital photograph a probability distribution over linear scene colors that could have induced it. We demonstrate how these distributions can be useful for visual inference by incorporating them into estimation algorithms for a representative set of vision tasks.

cs.CV↗

arXiv2014

From Shading to Local Shape

We develop a framework for extracting a concise representation of the shape information available from diffuse shading in a small image patch. This produces a mid-level scene descriptor, comprised of local shape distributions that are inferred separately at every image patch across multiple scales. The framework is based on a quadratic representation of local shape that, in the absence of noise, has guarantees on recovering accurate local shape and lighting. And when noise is present, the inferred local shape distributions provide useful shape information without over-committing to any particular image explanation. These local shape distributions naturally encode the fact that some smooth diffuse regions are more informative than others, and they enable efficient and robust reconstruction of object-scale shape. Experimental results show that this approach to surface reconstruction compares well against the state-of-art on both synthetic images and captured photographs.

cs.CV↗