Orders Of Magnitude Speedup Research Articles

ZoomOut

We present a simple and efficient method for refining maps or correspondences by iterative upsampling in the spectral domain that can be implemented in a few lines of code. Our main observation is that high quality maps can be obtained even if the input correspondences are noisy or are encoded by a small number of coefficients in a spectral basis. We show how this approach can be used in conjunction with existing initialization techniques across a range of application scenarios, including symmetry detection, map refinement across complete shapes, non-rigid partial shape matching and function transfer. In each application we demonstrate an improvement with respect to both the quality of the results and the computational speed compared to the best competing methods, with up to two orders of magnitude speed-up in some applications. We also demonstrate that our method is both robust to noisy input and is scalable with respect to shape complexity. Finally, we present a theoretical justification for our approach, shedding light on structural properties of functional maps.

Effective Resistance of Two-Dimensional Truncated Infinite Mesh Structures

Determining the effective resistance of a mesh is a powerful tool for simplifying the analysis of complex electrical structures, ranging from transistors to power delivery systems. A common structure in science and engineering is a two-dimensional resistive grid. Applications of this structure include IR drop analysis and decoupling capacitor allocation in on-chip power and ground networks in VLSI systems, and the analysis of electrical and thermal conductive media, such as a semiconductor substrate. In the case where the effective resistance is evaluated far from the grid edges, an infinite resistive lattice can be used to simplify the analysis of a grid. If however the target nodes are located close to the grid edges, the assumption of infinite dimensions becomes invalid, producing inaccurate results. To bridge this gap, a resistive mesh truncated along a single or two dimensions is discussed here. Integral expressions for the effective resistance within a truncated infinite mesh structure are provided, as well as a closed-form approximation, which exhibits good agreement with the exact integral equation, exhibiting an average error of 0.27% and a maximum error of 4.77%. These expressions significantly improve the accuracy of the effective resistance estimation near the edges and corner of a resistive mesh, providing a tenfold reduction in error. In case studies, the expressions provide four to six orders of magnitude speedup in IR analysis of a $10^{4} \times 10^{4}$ grid, while providing accuracy comparable to nodal analysis.

An efficient hybrid-grid crossflow equilibrium model for field-scale fractured reservoir simulation

Multiphase flow simulation in fractured reservoirs at field scale is a significant challenge. Despite recent advances and a wide range of applications in both hydrology and hydrocarbon reservoir engineering, discussing efficient methods that cover computational complexity, accuracy, and flexibility aspects is still of paramount importance for a better understanding of these complex media. In this work, we present a new method that handles both the topological and computational complexities of fractures, taking into consideration advantages of various existing approaches, which include hybrid-grid and crossflow equilibrium models. The hybrid-grid (HG) model consists of representing fractures as lower-dimensional objects that still are represented as control volumes in a computational grid. The HG model is equivalent to a single-porosity model with a practical solution for the small control volumes at the intersection between fractures; however, the overall simulation run time is still dominated by the remaining fracture small control volumes. To overcome single-porosity computational challenges, a crossflow equilibrium (CFE) concept between discrete fractures and a small neighborhood in the matrix blocks can be employed. The CFE model consists of combining fractures with a small fraction of the neighborhood matrix blocks on either side in larger elements to achieve a better computational efficiency than conventional single-porosity models. The implementation of a CFE model at field scale is not practical because of the fracture topological challenges associated with the construction of an accurate computational grid for the CFE elements. In this work, we propose a method based on a combination of HG and CFE models to overcome the challenges associated with the HG fracture small control volumes and field-scale CFE computational grid construction. First, we assess the performance of the existing CFE model, and we propose an improved model. In addition, we suggest an input data handling method that is sufficient to account for fractional flow inside the CFE elements for flow in homogeneous fractured reservoirs without the need of any change in the simulator. Second, we describe the uniqueness of the proposed method, and we discuss different numerical examples to assess both the accuracy and computational efficiency. The results obtained are very accurate, and, computationally, one to two orders of magnitude speedup can be achieved. The improved CFE results are superior over the traditional CFE model. Combined with the HG model, the results are significantly improved while retaining a very good performance.

PrivFT: Private and Fast Text Classification With Homomorphic Encryption

The need for privacy-preserving analytics is higher than ever due to the severity of privacy risks and to comply with new privacy regulations leading to an amplified interest in privacy-preserving techniques that try to balance between privacy and utility. In this work, we present an efficient method for Text Classification while preserving the privacy of the content using Fully Homomorphic Encryption (FHE). Our system (named \textbf{Priv}ate \textbf{F}ast \textbf{T}ext (PrivFT)) performs two tasks: 1) making inference of encrypted user inputs using a plaintext model and 2) training an effective model using an encrypted dataset. For inference, we train a supervised model and outline a system for homomorphic inference on encrypted user inputs with zero loss to prediction accuracy. In the second part, we show how to train a model using fully encrypted data to generate an encrypted model. We provide a GPU implementation of the Cheon-Kim-Kim-Song (CKKS) FHE scheme and compare it with existing CPU implementations to achieve 1 to 2 orders of magnitude speedup at various parameter settings. We implement PrivFT in GPUs to achieve a run time per inference of less than 0.66 seconds. Training on a relatively large encrypted dataset is more computationally intensive requiring 5.04 days.

Efficient Identification of Approximate Best Configuration of Training in Large Datasets

A configuration of training refers to the combinations of feature engineering, learner, and its associated hyperparameters. Given a set of configurations and a large dataset randomly split into training and testing set, we study how to efficiently identify the best configuration with approximately the highest testing accuracy when trained from the training set. To guarantee small accuracy loss, we develop a solution using confidence interval (CI)-based progressive sampling and pruning strategy. Compared to using full data to find the exact best configuration, our solution achieves more than two orders of magnitude speedup, while the returned top configuration has identical or close test accuracy.

DAEGEN: A Modular Compiler for Exploring Decoupled Spatial Accelerators

Specialized hardware accelerators, particularly those that are programmable and flexible to target multiple problems in their domain, have proven to provide orders of magnitude speedup and energy efficiency. However, their design requires extensive manual effort, due to the need for hardware-software codesign to balance the degree and forms of specialization to the domains or program behaviors of interest. This article provides the first steps towards one approach for automating much of these processes. The insight behind our work is to recognize that decoupled spatial architectures both define a rich design space with many tradeoffs for different kinds of applications, and also can be composed out of a simple set of well-defined primitives. Therefore, we propose a modular accelerator design framework, DAEGEN, a.k.a. Decoupled Access Excution Accelerator Generator. This article defines an initial compiler and architecture primitives, and we discuss key challenges.

Structure Property Analysis of the Solution and Solid-State Properties of Bistable Photochromic Hydrazones.

The development of new photochromic compounds and the optimization of their photophysical and switching properties are prerequisites for accessing new functions and opportunities that are not possible with currently available systems. To this end we recently developed a new bistable hydrazone switch that undergoes efficient photoswitching and emission ON/OFF toggling in both solution and solid-state. Here, we present a systematic structure-property analysis using a family of hydrazones and show how their properties, including activation wavelengths, photostationary states (PSSs), photoisomerization quantum yields, thermal half-lives (τ1/2), and solution/solid-state fluorescence characteristics vary as a function of electron donating (EDG) and/or withdrawing (EWG) substituents. These studies resulted in the red-shifting of the absorption profiles of the Z and E isomers of the switches, while maintaining excellent PSSs in almost all of the compounds. The introduction of para-NMe2 and/or para-NO2 groups improved the photoisomerization quantum yields, and the extremely long thermal half-lives (tens to thousands of years) were maintained in most cases, even in a push-pull system, which can be activated solely with visible light. Hydrazones bearing EDGs at the stator phenyl group are an exception and show up to 6 orders of magnitude acceleration in τ1/2 (i.e., days) because of a change in the isomerization mechanism. Moreover, we discovered that a para-NMe2 group is required to have reasonable fluorescence quantum yields in solution and that rigidification enhances the emission in the solid-state. Finally, X-ray crystallography analysis showed that the switching process is more efficient in the solid-state when the hydrazone is loosely packed.

Darwin: A Genomics Coprocessor

Long read sequencing is promising as it provides knowledge of a full spectrum of mutations in the human genome and generates more contiguous de novo assemblies. But high error rate in long reads imposes a computational barrier for genome assembly. Darwin, a specialized coprocessor, which provides orders of magnitude speedup over conventional processors in long read assembly, can eliminate this barrier.

Model-based analysis and quality-by-design framework for high aspect ratio crystals in crystallizer-wet mill systems using GPU acceleration enabled optimization

The simultaneous control of crystal size and shape is particularly important in pharmaceutical crystallization processes. These two quantities not only influence the dissolution rate and bioavailability of final drug products, but also contribute to the manufacturability and efficiency of downstream operations. The manipulation of crystal shape, however, is difficult since it requires decoupled growth rate control of individual crystal faces. The aim of this work is to analyze and evaluate the effectiveness of wet milling for control of bivariate size distribution of high-aspect ratio crystals during solution crystallization. Three configurations are compared: the pure batch crystallizer, the crystallizer with internal wet mill (immersion mill) and crystallizer with external wet mill. Population balance models (PBMs) are employed to describe bivariate size density function dynamics, with stirring rate dependent breakage to enable the direct optimization of wet mill operation. The generated system of hyperbolic partial differential and ordinary differential equations is solved by a fully discretized high-resolution finite volume method (HR-FVM), involving graphical processing unit (GPU) acceleration, which brings up to two orders of magnitude speed-up by compared to the serial C implementation. The simulation studies revealed that wet milling significantly extends the achievable crystal size and shape domain, but it also broadens the product size distribution. The analysis of the optimized operating profiles, which include temperature, wet mill rotation (RPM) and recirculation rate, enabled the generalization of optimal operation, which can efficiently guide a QbD approach by considerably reducing the design space without compromising the process performance.

Efficient Algorithms for the Line-SIAC Filter

Visualizing high-order finite element simulation data using current visualization tools has many challenges: discontinuities at element boundaries, interpolating artifacts, and evaluating derived quantities. These challenges have been addressed by postprocessing the simulation data using the L-SIAC filter. However, the time required to postprocess using this filter needs to be addressed to enable using it on large datasets. In this work, we introduce an efficient technique to speed-up the L-SIAC filter and alternate ways to gain further speed-up at the cost of accuracy. This method is also ideal to postprocess at regularly spaced locations, which would be suitable for standard visualization software. Our results show that our method can achieve up to two orders of magnitude speed-up as compared to our interpretation of the technique presented in Docampo-Sanchez (SIAM J Sci Comput 39(5):A2179–A2200, 2017).

LoCool: Fighting Hot Spots Locally for Improving System Energy Efficiency

Elevated on-chip temperatures significantly degrade performance, energy-efficiency, and lifetime of processors. The cooling system for a chip is typically designed to remove the worst-case heat generated per unit area. Cooling demand, however, spatially and temporally varies across a chip as hot spots occur on different locations with different intensities. Thus, designing a homogeneous cooling system for a chip can be inefficient. Recently, hybrid cooling strategies, such as integrating thermoelectric coolers (TECs) with microchannel liquid cooling, have been explored for hot spot mitigation. The efficiency of such a cooling system strongly depends on the operating point of each cooling method, as well as the locations and intensities of the hot spots. To this end, we first devise a compact thermal modeling method for the design and evaluation of hybrid cooling systems in a fast and accurate way. The proposed model provides up to four orders of magnitude speedup in simulation time compared to COMSOL multiphysics simulations with less than 2.9 °C average temperature error. Leveraging our fast model, we develop LoCool, a hybrid cooling optimization method, which jointly determines the most energy-efficient cooling settings for a given chip power distribution and temperature constraint. LoCool determines the liquid flow rate and the input current for each TEC depending on the cooling requirements for individual hot spots as well as for the background heat. Experimental evaluation shows up to 40% cooling energy savings compared to designing homogeneous cooling systems under the same thermal constraints.

ProXray: Protocol Model Learning and Guided Firmware Analysis

The number of Internet of Things (IoT) has reached 7 billion globally in early 2018 and are nearly ubiquitous in daily life. Knowing whether or not these devices are safe and secure to use is becoming critical. IoT devices usually implement communication protocols such as USB and Bluetooth within firmware to allow a wide range of functionality. Thus analyzing firmware using domain knowledge from these protocols is vital to understand device behavior, detect implementation bugs, and identify malicious components. Unfortunately, due to the complexity of these protocols, there is usually no formal specification available that can help automate the firmware analysis; as a result significant manual effort is currently required to study these protocols and to reverse engineer the device firmware. In this paper, we propose a new firmware analysis methodology using symbolic execution called ProXray, which can learn a protocol model from known firmware, and apply the model to recognize the protocol relevant fields and detect functionality within unknown firmware automatically. After the training phase, ProXray can fully automate the firmware analysis process while supporting user’s queries in the form of protocol relevant constraints. We have applied ProXray to the USB and the Bluetooth protocols by learning protocol constraint models from firmware that implement these protocols. We are then able to map protocol fields and identify USB functionality automatically within all 6 unknown USB firmware while achieving more than an order of magnitude speedup in reaching protocol relevant targets in unknown Bluetooth firmware. Our model achieved high coverage of the USB and Bluetooth specifications for several important protocol fields. ProXray provides a new method to apply domain knowledge to firmware analysis automatically.

SWIFT

Event streams generated by smart devices common in modern Internet of Things applications must be continuously mined to monitor the behavior of the underlying system. In this work, we propose a stream pattern mining system for supporting online IoT applications. First, to solve the pattern explosion problem of existing stream pattern mining strategies, we now design pattern semantics that continuously produce a compact set of patterns that max-imumly compresses the dynamic data streams, called MDL-based Representative Patterns (MRP). We then design a one-pass SWIFT approach that continuously mines the up-to-date MRP pattern set for each stream window upon the arrival or expiration of individual events. We show that SWIFT is guaranteed to select the update operation for each individual incoming event that leads to the most compact encoding of the sequence in the current window. We further enhance SWIFT to support batch updates, called B-SWIFT. B-SWIFT adopts alazy updatestrategy that guarantees that only the minimal number of operations are conducted to process an incoming event batch for MRP pattern mining. Evaluation by our industry lighting lab collaborator demonstrates that SWIFT successfully solves their use cases and finds more representative patterns than the alternative approaches adapting the state-of-the-art static representative pattern mining methods. Our experimental study confirms that SWIFT outperforms the best existing method up to 50% in the compactness of produced pattern encodings, while providing a 4 orders of magnitude speedup.

Automated Phase-Noise-Aware Design of RF Clock Distribution Circuits

This paper presents a methodology for phase-noise-aware design of RF clock distribution networks. The methodology is embedded within a commercial digital toolchain. It consists of a method for phase noise analysis based on lookup tables, a buffer insertion algorithm providing an optimal buffering for large interconnect trees and a gate sizing algorithm that offers fast power minimization, subject to noise constraints. The phase noise estimate produced by the proposed method on average deviates by less than 3.45% from SPICE simulation results; however, it can be computed three to five orders of magnitude faster. The proposed buffer insertion methodology can easily handle very large industrial examples. It completes within seconds, compared to days for methodologies used in industry today. The gate sizing optimization offers three to four orders of magnitude speedup, compared to commercial products, while producing comparable results. The proposed design methodology reduces the design time of RF clock distribution networks from weeks to a few hours by eliminating the need for manual circuit design, analog simulations, and manual layout. To the extent of our knowledge, this is the first paper on the design automation of RF clock distribution networks.

Combining Metadynamics and Integrated Tempering Sampling.

The simulation of rare events is one of the key problems in atomistic simulations. Toward its solution, a plethora of methods have been proposed. Here we combine two such methods: metadynamics and integrated tempering sampling. In metadynamics, the fluctuations of a carefully chosen collective variable are amplified, while in integrated tempering sampling the system is pushed to visit an approximately uniform interval of energies and allows exploring a range of temperatures in a single run. We describe our approach and apply it to the two prototypical systems a SN2 chemical reaction and to the freezing of silica. The combination of metadynamics and integrated tempering sampling leads to a powerful method. In particular in the case of silica we have measured more than 1 order of magnitude acceleration.

Profit: <underline>Pr</underline>iority and P<underline>o</underline>wer/Per<underline>f</underline>ormance Opt<underline>i</underline>miza<underline>t</underline>ion for Many-Core Systems

As power density emerges as the main constraint for many-core systems, controlling power consumption under the thermal design power while maximizing the performance becomes increasingly critical. To dynamically save power, dynamic voltage frequency scaling techniques have proved to be effective and are widely available commercially. Meanwhile, systems have certain performance constraints that the applications should satisfy to ensure quality of service. In this paper, we present an online distributed reinforcement learning (OD-RL)-based DVFS control algorithm for many-core system performance improvement under both power and performance constraints. At the finer grain, a per-core RL method is used to learn the optimal control policy of the voltage/frequency (VF) levels in a model-free manner. At the coarser grain, an efficient global power budget reallocation algorithm is used to maximize the overall performance. The experiments show that compared to the state-of-the-art algorithms: 1) OD-RL produces up to 98% less budget overshoot; 2) up to 23% higher energy efficiency; and 3) two orders of magnitude speedup over state-of-the-art techniques for systems with hundreds of cores. Furthermore, priority-aware OD-RL can better satisfy performance constraints than OD-RL with: 1) $17.8\boldsymbol {\times }$ more epochs satisfying the performance constraints; 2) $5.6\boldsymbol {\times }$ better performance gain; and 3) $20.0\boldsymbol {\times }$ better performance-power tradeoffs under similar efficiency and scalability.

Superpixel Hierarchy.

Superpixel segmentation has been one of the most important tasks in computer vision. In practice, an object can be represented by a number of segments at finer levels with consistent details or included in a surrounding region at coarser levels. Thus, a superpixel segmentation hierarchy is of great importance for applications that require different levels of image details. However, there is no method that can generate all scales of superpixels accurately in real time. In this paper, we propose the superhierarchy algorithm which is able to generate multi-scale superpixels as accurately as the state-of-the-art methods but with one to two orders of magnitude speed-up. The proposed algorithm can be directly integrated with recent efficient edge detectors to significantly outperform the state-of-the-art methods in terms of segmentation accuracy. Quantitative and qualitative evaluations on a number of applications demonstrate that the proposed algorithm is accurate and efficient in generating a hierarchy of superpixels.

Efficient Representative Subset Selection over Sliding Windows

Representative subset selection (RSS) is an important tool for users to draw insights from massive datasets. Existing literature models RSS as the submodular maximization problem to capture the "diminishing returns" property of the representativeness of selected subsets, but often only has a single constraint (e.g., cardinality), which limits its applications in many real-world problems. To capture the data recency issue and support different types of constraints, we formulate dynamic RSS in data streams as maximizing submodular functions subject to general $d$-knapsack constraints (SMDK) over sliding windows. We propose a \textsc{KnapWindow} framework (KW) for SMDK. KW utilizes the \textsc{KnapStream} algorithm (KS) for SMDK in append-only streams as a subroutine. It maintains a sequence of checkpoints and KS instances over the sliding window. Theoretically, KW is $\frac{1-\varepsilon}{1+d}$-approximate for SMDK. Furthermore, we propose a \textsc{KnapWindowPlus} framework (KW$^{+}$) to improve upon KW. KW$^{+}$ builds an index \textsc{SubKnapChk} to manage the checkpoints and KS instances. \textsc{SubKnapChk} deletes a checkpoint whenever it can be approximated by its successors. By keeping much fewer checkpoints, KW$^{+}$ achieves higher efficiency than KW while still guaranteeing a $\frac{1-\varepsilon'}{2+2d}$-approximate solution for SMDK. Finally, we evaluate the efficiency and solution quality of KW and KW$^{+}$ in real-world datasets. The experimental results demonstrate that KW achieves more than two orders of magnitude speedups over the batch baseline and preserves high-quality solutions for SMDK over sliding windows. KW$^{+}$ further runs 5-10 times faster than KW while providing solutions with equivalent or even better utilities.

Path planning with divergence-based distance functions

Distance functions between points in a domain can be used to automatically plan a gradient-descent path towards a given target point in the domain, avoiding obstacles that may be present. A key requirement from such distance functions is the absence of spurious local minima, and this has led to the common use of harmonic potential functions. This choice guarantees the absence of spurious minima, but is well known to be slow to numerically compute and prone to numerical precision issues. To alleviate the first of these problems, we propose a family of novel divergence distances. These are based on f-divergence of the Poisson kernel of the domain. Using the concept of conformal invariance, we show that divergence distances are equivalent to the harmonic potential function on simply-connected domains, namely generate paths which are identical to those generated by the potential function. We then discuss how to compute two special cases of divergence distances, one based on the Kullback–Leibler, the other on the total variation divergence, in practice by discretizing the domain with a triangle mesh and using Finite Elements (FEM) computation. We show that using divergence distances instead of the potential function and other distances has a significant computational advantage, as, following a pre-processing stage, they may be computed online in a multi-query scenario up to an order of magnitude faster than the others when taking advantage of certain sparsity properties of the Poisson kernel. Furthermore, the computation is “embarrassingly parallel”, so may be implemented on a GPU with up to three orders of magnitude speedup.

Quantifying Topological Uncertainty in Fractured Systems using Graph Theory and Machine Learning

Fractured systems are ubiquitous in natural and engineered applications as diverse as hydraulic fracturing, underground nuclear test detection, corrosive damage in materials and brittle failure of metals and ceramics. Microstructural information (fracture size, orientation, etc.) plays a key role in governing the dominant physics for these systems but can only be known statistically. Current models either ignore or idealize microscale information at these larger scales because we lack a framework that efficiently utilizes it in its entirety to predict macroscale behavior in brittle materials. We propose a method that integrates computational physics, machine learning and graph theory to make a paradigm shift from computationally intensive high-fidelity models to coarse-scale graphs without loss of critical structural information. We exploit the underlying discrete structure of fracture networks in systems considering flow through fractures and fracture propagation. We demonstrate that compact graph representations require significantly fewer degrees of freedom (dof) to capture micro-fracture information and further accelerate these models with Machine Learning. Our method has been shown to improve accuracy of predictions with up to four orders of magnitude speedup.