Search

Today's large-scale scientific applications running on high-performance computing (HPC) systems generate vast data volumes. Thus, data compression is becoming a critical technique to mitigate the storage burden and data-movement cost. However, existing lossy compressors for scientific data cannot achieve a high compression ratio and throughput simultaneously, hindering their adoption in many applications requiring fast compression, such as in-memory compression. To this end, in this work, we develop a fast and high-ratio error-bounded lossy compressor on GPUs for scientific data (called FZ-GPU). Specifically, we first design a new compression pipeline that consists of fully parallelized quantization, bitshuffle, and our newly designed fast encoding. Then, we propose a series of deep architectural optimizations for each kernel in the pipeline to take full advantage of CUDA architectures. We propose a warp-level optimization to avoid data conflicts for bit-wise operations in bitshuffle, maximize shared memory utilization, and eliminate unnecessary data movements by fusing different compression kernels. Finally, we evaluate FZ-GPU on two NVIDIA GPUs (i.e., A100 and RTX A4000) using six representative scientific datasets from SDRBench. Results on the A100 GPU show that FZ-GPU achieves an average speedup of 4.2X over cuSZ and an average speedup of 37.0X over a multi-threaded CPU implementation of our algorithm under the same error bound. FZ-GPU also achieves an average speedup of 2.3X and an average compression ratio improvement of 2.0X over cuZFP under the same data distortion., Comment: 14 pages, 12 figures, accepted by ACM HPDC '23

Existing error-bounded lossy compression techniques control the pointwise error during compression to guarantee the integrity of the decompressed data. However, they typically do not explicitly preserve the topological features in data. When performing post hoc analysis with decompressed data using topological methods, preserving topology in the compression process to obtain topologically consistent and correct scientific insights is desirable. In this paper, we introduce TopoSZ, an error-bounded lossy compression method that preserves the topological features in 2D and 3D scalar fields. Specifically, we aim to preserve the types and locations of local extrema as well as the level set relations among critical points captured by contour trees in the decompressed data. The main idea is to derive topological constraints from contour-tree-induced segmentation from the data domain, and incorporate such constraints with a customized error-controlled quantization strategy from the classic SZ compressor.Our method allows users to control the pointwise error and the loss of topological features during the compression process with a global error bound and a persistence threshold.

This paper is concerned with establishing a trace minimization principle for two Hermitian matrix pairs. Specifically, we will answer the question: when is $\inf_X\operatorname{tr}(\widehat AX^{\rm H}AX)$ subject to $\widehat BX^{\rm H}BX=I$ (the identity matrix of apt size) finite? Sufficient and necessary conditions are obtained and, when the infimum is finite, an explicit formula for it is presented in terms of the finite eigenvalues of the matrix pairs. Our results extend Fan's trace minimization principle (1949) for a Hermitian matrix, a minimization principle of Kova\v{c}-Striko and Veseli\'c (1995) for a Hermitian matrix pair, and most recent ones by the authors and their collaborators for a Hermitian matrix pair and a Hermitian matrix., Comment: 26 pages

One-sided dense matrix decompositions (e.g., Cholesky, LU, and QR) are the key components in scientific computing in many different fields. Although their design has been highly optimized for modern processors, they still consume a considerable amount of energy. As CPU-GPU heterogeneous systems are commonly used for matrix decompositions, in this work, we aim to further improve the energy saving of one-sided matrix decompositions on CPU-GPU heterogeneous systems. We first build an Algorithm-Based Fault Tolerance protected overclocking technique (ABFT-OC) to enable us to exploit reliable overclocking for key matrix decomposition operations. Then, we design an energy-saving matrix decomposition framework, Bi-directional Slack Reclamation(BSR), that can intelligently combine the capability provided by ABFT-OC and DVFS to maximize energy saving and maintain performance and reliability. Experiments show that BSR is able to save up to 11.7% more energy compared with the current best energy saving optimization approach with no performance degradation and up to 14.1% Energy * Delay^2 reduction. Also, BSR enables the Pareto efficient performance-energy trade-off, which is able to provide up to 1.43x performance improvement without costing extra energy.

Most existing semi-supervised graph-based clustering methods exploit the supervisory information by either refining the affinity matrix or directly constraining the low-dimensional representations of data points. The affinity matrix represents the graph structure and is vital to the performance of semi-supervised graph-based clustering. However, existing methods adopt a static affinity matrix to learn the low-dimensional representations of data points and do not optimize the affinity matrix during the learning process. In this paper, we propose a novel dynamic graph structure learning method for semi-supervised clustering. In this method, we simultaneously optimize the affinity matrix and the low-dimensional representations of data points by leveraging the given pairwise constraints. Moreover, we propose an alternating minimization approach with proven convergence to solve the proposed nonconvex model. During the iteration process, our method cyclically updates the low-dimensional representations of data points and refines the affinity matrix, leading to a dynamic affinity matrix (graph structure). Specifically, for the update of the affinity matrix, we enforce the data points with remarkably different low-dimensional representations to have an affinity value of 0. Furthermore, we construct the initial affinity matrix by integrating the local distance and global self-representation among data points. Experimental results on eight benchmark datasets under different settings show the advantages of the proposed approach.

Quantum semiconductor structures are commonly achieved by bandgap engineering that relies on the ability to switch from one semiconductor to another during their growth. Growth of a superlattice is typically demanding technologically. In contrast, accumulated evidence points to a tendency among a certain class of multiple-cation binary oxides to self-assemble spontaneously as superlattice structures. This class has been dubbed the homologous superlattices. For a famous example, when a mixture of indium and zinc is oxidized, the phases of In-O and ZnO separate in an orderly periodic manner, along the ZnO polar axis, with polarity inversion taking place between consecutive ZnO sections. As we review here, the same structure has been observed when the indium was replaced with other metals, and perhaps even in ZnO alone. This peculiar self-assembled structure has been attracting research over the past decade. The purpose of this study is to gain understanding of the physics underlying the formation of this unique structure. Here, we first provide an extensive review of the accumulated literature on these spontaneously-formed structures and then propose an explanation for the long-standing mystery of this intriguing self-assembly in the form of an electrostatic growth phenomenon and test the proposed model on experimental data.

Stochastic algebraic Riccati equations, also known as rational algebraic Riccati equations, arising in linear-quadratic optimal control for stochastic linear time-invariant systems, were considered to be not easy to solve. The-state-of-art numerical methods most rely on differentiability or continuity, such as Newton-type method, LMI method, or homotopy method. In this paper, we will build a novel theoretical framework and reveal the intrinsic algebraic structure appearing in this kind of algebraic Riccati equations. This structure guarantees that to solve them is almost as easy as to solve deterministic/classical ones, which will shed light on the theoretical analysis and numerical algorithm design for this topic., Comment: 17 pages

Deep learning has been widely applied for the channel state information (CSI) feedback in frequency division duplexing (FDD) massive multiple-input multiple-output (MIMO) system. For the typical supervised training of the feedback model, the requirements of large amounts of task-specific labeled data can hardly be satisfied, and the huge training costs and storage usage of the model in multiple scenarios are hindrance for model application. In this letter, a multi-task learning-based approach is proposed to improve the feasibility of the feedback network. An encoder-shared feedback architecture and the corresponding training scheme are further proposed to facilitate the implementation of the multi-task learning approach. The experimental results indicate that the proposed multi-task learning approach can achieve comprehensive feedback performance with considerable reduction of training cost and storage usage of the feedback model., Comment: 5 pages, 2 figures

This paper studies the optimal state estimation for a dynamic system, whose transfer function can be nonlinear and the input noise can be of arbitrary distribution. Our algorithm differs from the conventional extended Kalman filter (EKF) and the particle filter (PF) in that it estimates not only the state vector but also the Cramer-Rao bound (CRB), which serves as an accuracy indicator. Combining the state estimation, the CRB, and the incoming new measurement, the algorithm updates the state estimation according to the maximum likelihood (ML) criterion. To illustrate the effectiveness of the proposed method for autonomous driving, we apply it to estimate the position and velocity of a vehicle based on the noisy measurements of distance and Doppler offset. Simulation results show that the proposed algorithm can achieve estimation significantly more accurate than the standard EKF and the PF., Comment: The algorithm description in section III is incomplete

Due to the short decohorence time of qubits available in the NISQ-era, it is essential to pack (minimize the size and or the depth of) a logical quantum circuit as efficiently as possible given a sparsely coupled physical architecture. In this work we introduce a locality-aware qubit routing algorithm based on a graph theoretic framework. Our algorithm is designed for the grid and certain "grid-like" architectures. We experimentally show the competitiveness of algorithm by comparing it against the approximate token swapping algorithm, which is used as a primitive in many state-of-the-art quantum transpilers. Our algorithm produces circuits of comparable depth (better on random permutations) while being an order of magnitude faster than a typical implementation of the approximate token swapping algorithm., Comment: 7 pages, 4 figure, IPDPS-QC3 2022