Orders Of Magnitude Speedup Research Articles

Performance bound analysis of analog circuits in frequency- and time-domain considering process variations

In this article, we propose a new performance bound analysis of analog circuits considering process variations. We model the variations of component values as intervals measured from tested chips and manufacture processes. The new method first applies a graph-based analysis approach to generate the symbolic transfer function of a linear(ized) analog circuit. Then the frequency response bounds (maximum and minimum) are obtained by performing nonlinear constrained optimization in which magnitude or phase of the transfer function is the objective function to be optimized subject to the ranges of process variational parameters. The response bounds given by the optimization-based method are very accurate and do not have the over-conservativeness issues of existing methods. Based on the frequency-domain bounds, we further develop a method to calculate the time-domain response bounds for any arbitrary input stimulus. Experimental results from several analog benchmark circuits show that the proposed method gives the correct bounds verified by Monte Carlo analysis while it delivers one order of magnitude speedup over Monte Carlo for both frequency-domain and time-domain bound analyses. We also show analog circuit yield analysis as an application of the frequency-domain variational bound analysis.

Subspace integration with local deformations

Subspace techniques greatly reduce the cost of nonlinear simulation by approximating deformations with a small custom basis. In order to represent the deformations well (in terms of a global metric), the basis functions usually have global support, and cannot capture localized deformations. While reduced-space basis functions can be localized to some extent, capturing truly local deformations would still require a very large number of precomputed basis functions, significantly degrading both precomputation and online performance. We present an efficient approach to handling local deformations that cannot be predicted, most commonly arising from contact and collisions, by augmenting the subspace basis with custom functions derived from analytic solutions to static loading problems. We also present a new cubature scheme designed to facilitate fast computation of the necessary runtime quantities while undergoing a changing basis. Our examples yield a two order of magnitude speedup over full-coordinate simulations, striking a desirable balance between runtime speeds and expressive ability.

An efficient method for analyzing on-chip thermal reliability considering process variations

This work provides an efficient statistical electrothermal simulator for analyzing on-chip thermal reliability under process variations. Using the collocation-based statistical modeling technique, first, the statistical interpolation polynomial for on-chip temperature distribution can be obtained by performing deterministic electrothermal simulation very few times and by utilizing polynomial interpolation. After that, the proposed simulator not only provides the mean and standard deviation profiles of on-chip temperature distribution, but also innovates the concept of thermal yield profile to statistically characterize the on-chip temperature distribution more precisely, and builds an efficient technique for estimating this figure of merit. Moreover, a mixed-mesh strategy is presented to further enhance the efficiency of the developed statistical electrothermal simulator. Experimental results demonstrate that (1) the developed statistical electrothermal simulator can obtain accurate approximations with orders of magnitude speedup over the Monte Carlo method; (2) comparing with a well-known cumulative distribution function estimation method, APEX [Li et al. 2004], the developed statistical electrothermal simulator can achieve 215× speedup with better accuracy; (3) the developed mixed-mesh strategy can achieve an order of magnitude faster over our baseline algorithm and still maintain an acceptable accuracy level.

Using Genome Query Language to uncover genetic variation.

With high-throughput DNA sequencing costs dropping <$1000 for human genomes, data storage, retrieval and analysis are the major bottlenecks in biological studies. To address the large-data challenges, we advocate a clean separation between the evidence collection and the inference in variant calling. We define and implement a Genome Query Language (GQL) that allows for the rapid collection of evidence needed for calling variants. We provide a number of cases to showcase the use of GQL for complex evidence collection, such as the evidence for large structural variations. Specifically, typical GQL queries can be written in 5-10 lines of high-level code and search large datasets (100 GB) in minutes. We also demonstrate its complementarity with other variant calling tools. Popular variant calling tools can achieve one order of magnitude speed-up by using GQL to retrieve evidence. Finally, we show how GQL can be used to query and compare multiple datasets. By separating the evidence and inference for variant calling, it frees all variant detection tools from the data intensive evidence collection and focuses on statistical inference. GQL can be downloaded from http://cseweb.ucsd.edu/~ckozanit/gql.

Surface and Curve Skeletonization of Large 3D Models on the GPU

We present a GPU-based framework for extracting surface and curve skeletons of 3D shapes represented as large polygonal meshes. We use an efficient parallel search strategy to compute point-cloud skeletons and their distance and feature transforms (FTs) with user-defined precision. We regularize skeletons by a new GPU-based geodesic tracing technique which is orders of magnitude faster and more accurate than comparable techniques. We reconstruct the input surface from skeleton clouds using a fast and accurate image-based method. We also show how to reconstruct the skeletal manifold structure as a polygon mesh and the curve skeleton as a polyline. Compared to recent skeletonization methods, our approach offers two orders of magnitude speed-up, high-precision, and low-memory footprints. We demonstrate our framework on several complex 3D models.

Efficient error-tolerant query autocompletion

Query autocompletion is an important feature saving users many keystrokes from typing the entire query. In this paper we study the problem of query autocompletion that tolerates errors in users' input using edit distance constraints. Previous approaches index data strings in a trie, and continuously maintain all the prefixes of data strings whose edit distance from the query are within the threshold. The major inherent problem is that the number of such prefixes is huge for the first few characters of the query and is exponential in the alphabet size. This results in slow query response even if the entire query approximately matches only few prefixes.In this paper, we propose a novel neighborhood generation-based algorithm, IncNGTrie, which can achieve up to two orders of magnitude speedup over existing methods for the error-tolerant query autocompletion problem. Our proposed algorithm only maintains a small set of active nodes, thus saving both space and time to process the query. We also study efficient duplicate removal which is a core problem in fetching query answers. In addition, we propose optimization techniques to reduce our index size, as well as discussions on several extensions to our method. The efficiency of our method is demonstrated against existing methods through extensive experiments on real datasets.

Scalable Incremental Network Programming for Multihop Wireless Sensors

We present a network programming mechanism that can flexibly and quickly re-task a large multi-hop network of wireless sensor nodes. Our mechanism allows each sensor node to be incrementally reprogrammed with heterogeneous images of native program code using Rsync block comparison algorithm, point-to-point routing with the BLIP IPv6 stack, and image volume management with Deluge2. With our re-tasking method, we demonstrate an order of magnitude speed-up on small code changes over non-incremental delivery. Our mechanism also scales sub-linearly in the diameter of the network. Collectively, these advancements qualitatively change the software life cycle of the embedded networked systems.

Know by a handful the whole sack: efficient sampling for top-k influential user identification in large graphs

Influence Maximization aims to find the top-K influential individuals to maximize the influence spread within a social network, which remains an important yet challenging problem. Most existing greedy algorithms mainly focus on computing the exact influence spread, leading to low computational efficiency and limiting their application to real-world social networks. While in this paper we show that through supervised sampling, we can efficiently estimate the influence spread at only negligible cost of precision, thus significantly reducing the execution time. Motivated by this, we propose ESMCE, a power-law exponent supervised Monte Carlo estimation method. In particular, ESMCE exploits the power-law exponent of the social network to guide the sampling, and employs multiple iterative steps to guarantee the estimation accuracy. Moreover, ESMCE shows excellent scalability and well suits large-scale social networks. Extensive experiments on six real-world social networks demonstrate that, compared with state-of-the-art greedy algorithms, ESMCE is able to achieve almost two orders of magnitude speedup in execution time with only negligible error (2.21 % on average) in influence spread.

Scalable Active Learning for Multiclass Image Classification

Machine learning techniques for computer vision applications like object recognition, scene classification, etc., require a large number of training samples for satisfactory performance. Especially when classification is to be performed over many categories, providing enough training samples for each category is infeasible. This paper describes new ideas in multiclass active learning to deal with the training bottleneck, making it easier to train large multiclass image classification systems. First, we propose a new interaction modality for training which requires only yes-no type binary feedback instead of a precise category label. The modality is especially powerful in the presence of hundreds of categories. For the proposed modality, we develop a Value-of-Information (VOI) algorithm that chooses informative queries while also considering user annotation cost. Second, we propose an active selection measure that works with many categories and is extremely fast to compute. This measure is employed to perform a fast seed search before computing VOI, resulting in an algorithm that scales linearly with dataset size. Third, we use locality sensitive hashing to provide a very fast approximation to active learning, which gives sublinear time scaling, allowing application to very large datasets. The approximation provides up to two orders of magnitude speedups with little loss in accuracy. Thorough empirical evaluation of classification accuracy, noise sensitivity, imbalanced data, and computational performance on a diverse set of image datasets demonstrates the strengths of the proposed algorithms.

TJJE: An efficient algorithm for top-k join on massive data

In many applications, top-k join is an important operation to return the k most important join tuples among the potentially huge answer space according to a given ranking function. PBRJ is an algorithm template that generalizes previous top-k join algorithms. In this paper, our analysis shows that PBRJ needs to maintain a large quantity of candidate tuples on massive data. Based on the analysis, this paper proposes a novel top-k join algorithm TJJE which is suitable for handling massive data. By some pre-computed information, TJJE first estimates an upper-bound on scan depth of each joined table. Then it determines the file that contains the join positional index pairs of the top-k join results. A novel algorithm is proposed to retrieve the required join tuples by a single sequential and selective scan on the joined tables. Finally, the top-k join results are obtained by a single scan on the retrieved join tuples. The correctness proof and cost analysis of TJJE are presented in this paper. Extensive experiments show that TJJE maintains up to three orders of magnitude fewer candidate tuples and obtains up to one order of magnitude speedup compared to PBRJ.

Spinning fast iterative data flows

Parallel dataflow systems are a central part of most analytic pipelines for big data. The iterative nature of many analysis and machine learning algorithms, however, is still a challenge for current systems. While certain types of bulk iterative algorithms are supported by novel dataflow frameworks, these systems cannot exploit computational dependencies present in many algorithms, such as graph algorithms. As a result, these algorithms are inefficiently executed and have led to specialized systems based on other paradigms, such as message passing or shared memory. We propose a method to integrate incremental iterations , a form of workset iterations, with parallel dataflows. After showing how to integrate bulk iterations into a dataflow system and its optimizer, we present an extension to the programming model for incremental iterations. The extension alleviates for the lack of mutable state in dataflows and allows for exploiting the sparse computational dependencies inherent in many iterative algorithms. The evaluation of a prototypical implementation shows that those aspects lead to up to two orders of magnitude speedup in algorithm runtime, when exploited. In our experiments, the improved dataflow system is highly competitive with specialized systems while maintaining a transparent and unified dataflow abstraction.

Efficient state merging in symbolic execution

Symbolic execution has proven to be a practical technique for building automated test case generation and bug finding tools. Nevertheless, due to state explosion, these tools still struggle to achieve scalability. Given a program, one way to reduce the number of states that the tools need to explore is to merge states obtained on different paths. Alas, doing so increases the size of symbolic path conditions (thereby stressing the underlying constraint solver) and interferes with optimizations of the exploration process (also referred to as search strategies). The net effect is that state merging may actually lower performance rather than increase it. We present a way to automatically choose when and how to merge states such that the performance of symbolic execution is significantly increased. First, we present query count estimation , a method for statically estimating the impact that each symbolic variable has on solver queries that follow a potential merge point; states are then merged only when doing so promises to be advantageous. Second, we present dynamic state merging , a technique for merging states that interacts favorably with search strategies in automated test case generation and bug finding tools. Experiments on the 96 GNU Coreutils show that our approach consistently achieves several orders of magnitude speedup over previously published results. Our code and experimental data are publicly available at http://cloud9.epfl.ch.

Medial Spheres for Shape Approximation

We study the problem of approximating a 3D solid with a union of overlapping spheres. In comparison with a state-of-the-art approach, our method offers more than an order of magnitude speedup and achieves a tighter approximation in terms of volume difference with the original solid while using fewer spheres. The spheres generated by our method are internal and tangent to the solid's boundary, which permits an exact error analysis, fast updates under local feature size preserving deformation, and conservative dilation. We show that our dilated spheres offer superior time and error performance in approximate separation distance tests than the state-of-the-art method for sphere set approximation for the class of (σ,θ)-fat solids. We envision that our sphere-based approximation will also prove useful for a range of other applications, including shape matching and shape segmentation.

OnlineCall: fast online parameter estimation and base calling for illumina's next-generation sequencing.

Next-generation DNA sequencing platforms are becoming increasingly cost-effective and capable of providing enormous number of reads in a relatively short time. However, their accuracy and read lengths are still lagging behind those of conventional Sanger sequencing method. Performance of next-generation sequencing platforms is fundamentally limited by various imperfections in the sequencing-by-synthesis and signal acquisition processes. This drives the search for accurate, scalable and computationally tractable base calling algorithms capable of accounting for such imperfections. Relying on a statistical model of the sequencing-by-synthesis process and signal acquisition procedure, we develop a computationally efficient base calling method for Illumina's sequencing technology (specifically, Genome Analyzer II platform). Parameters of the model are estimated via a fast unsupervised online learning scheme, which uses the generalized expectation-maximization algorithm and requires only 3 s of running time per tile (on an Intel i7 machine @3.07GHz, single core)-a three orders of magnitude speed-up over existing parametric model-based methods. To minimize the latency between the end of the sequencing run and the generation of the base calling reports, we develop a fast online scalable decoding algorithm, which requires only 9 s/tile and achieves significantly lower error rates than the Illumina's base calling software. Moreover, it is demonstrated that the proposed online parameter estimation scheme efficiently computes tile-dependent parameters, which can thereafter be provided to the base calling algorithm, resulting in significant improvements over previously developed base calling methods for the considered platform in terms of performance, time/complexity and latency. A C code implementation of our algorithm can be downloaded from http://www.cerc.utexas.edu/OnlineCall/.

Accelerated perturbation-resilient block-iterative projection methods with application to image reconstruction

We study the convergence of a class of accelerated perturbation-resilient block-iterative projection methods for solving systems of linear equations. We prove convergence to a fixed point of an operator even in the presence of summable perturbations of the iterates, irrespective of the consistency of the linear system. For a consistent system, the limit point is a solution of the system. In the inconsistent case, the symmetric version of our method converges to a weighted least-squares solution. Perturbation resilience is utilized to approximate the minimum of a convex functional subject to the equations. A main contribution, as compared to previously published approaches to achieving similar aims, is a more than an order of magnitude speed-up, as demonstrated by applying the methods to problems of image reconstruction from projections. In addition, the accelerated algorithms are illustrated to be better, in a strict sense provided by the method of statistical hypothesis testing, than their unaccelerated versions for the task of detecting small tumors in the brain from x-ray CT projection data.

General Parameterized Thermal Modeling for High-Performance Microprocessor Design

This paper proposes a new parameterized dynamic thermal modeling algorithm for emerging thermal-aware design and optimization for high-performance microprocessor design at architecture and package levels. Compared with existing behavioral thermal modeling algorithms, the proposed method can build the compact models from more general transient power and temperature waveforms used as training data. Such an approach can make the modeling process much easier and less restrictive than before and, thus, more amenable for practical measured data. The new method, called ParThermSID, consists of two steps. First, the response surface method based on second-order polynomials is applied to build the parameterized models at each time point for all of the given sampling nodes in the parameter space. Second, an improved subspace system identification method, called ThermSID, is employed to build the discrete state space models, by construction of the Hankel matrix and state space realization, for each time-varying coefficient of the polynomials generated in the first step. To overcome the overfitting problems of the subspace method, the new method employs an overfitting mitigation technique to improve model accuracy and predictive ability. Experimental results on a practical quad-core microprocessor show that the generated parameterized thermal model matches the given data very well. The compact models generated by ParThermSID also offer two orders of magnitude speedup over the commercial thermal analysis tool FloTHERM on the given example. The results also show that ThermSID is more accurate than the existing ThermPOF method.

Design and Verification of Five Port Router for Network on Chip Using VERILOG

The focus of this Paper is the actual implementation of Network Router and verifies the functionality of the five port router for network on chip using the latest verification methodologies, Hardware Verification Languages and EDA tools and qualifies the Design for Synthesis and implementation. This Router design contains Four output ports and one input port, it is packet based Protocol. This Design consists of Registers, FSM and FIFO's. For larger networks, where a direct-mapped approach is not feasible due to FPGA resource limitations, a virtualized time multiplexed approach was used. Compared to the provided software reference implementation, our direct-mapped approach achieves three orders of magnitude speedup, while our virtualized time multiplexed approach achieves one to two orders of magnitude speedup, depending on the network and router configuration.

Parallel collision detection of ellipsoids with applications in large scale multibody dynamics

This contribution describes a parallel approach for determining the collision state of a large collection of ellipsoids. Collision detection is required in granular dynamics simulation where it can combine with a differential variational inequality solver or discrete element method to approximate the time evolution of a collection of rigid bodies interacting through frictional contact. The approach proposed is structured on three levels. At the lowest level, the collision information associated with two colliding ellipsoids is obtained as the solution of a two-variable unconstrained optimization problem for which first and second order sensitivity information is derived analytically. Although this optimization approach suffices to resolve the collision problem between any two arbitrary ellipsoids, a less versatile but more efficient approach precedes it to gauge whether two ellipsoids are actually in contact and require the more costly optimization approach. This intermediate level draws on the analytical solution of a 3rd order polynomial obtained from the characteristic equation of two arbitrary ellipsoids. Finally, this intermediate level is invoked by the outer level only when a 3D spatial binning algorithm indicates that two ellipsoids share the same bin (box) and therefore could potentially collide. This multi-level approach is implemented in parallel and when executed on a ubiquitous Graphics Processing Unit (GPU) card scales linearly and yields a two orders of magnitude speedup over a similar algorithm executed on the Central Processing Unit (CPU). The GPU-based ellipsoid contact detection algorithm yields a 14-fold speedup over a CPU-based sphere contact detection algorithm implemented in the third party open source Bullet Physics Library (BPL). The proposed methodology provides the efficiency demanded by granular dynamics applications, which routinely handle scenarios with millions of collision events.

Top-k Attribute Difference q-clique Queries in Graph Data

The problem of dense subgraph discovery has important research meanings in many applications in real-world networks.This paper discusses a new problem about dense subgraph discovery,Top-k attribute difference q-clique queries,to find k q-cliques in which the dissimilarity between each vertices' attribute is as large as possible.The attribute difference q-clique is a dense subgraph combining both the characters of structure and attributes content,and the queries will find teams in which members' attribute(e.g.,research field or affiliation) are different from each other in a co-authorship network.This paper gives the attribute difference measure of q-clique and shows the problem of finding the maximum-difference q-clique is NP-Hard.This paper proposes a query algorithm AD-Qclique for the problem of attribute difference q-clique by using branch and bound strategy and optimizes the node visit order according to best-first order.This paper conducts extensive experiments through ACM author information dataset,which show that AD-Qclique obtains an order of magnitude speed-up comparing to BSL and all the authors in the result set have high H-index value and wide field of study.

Fast Multi-blind Modification Search through Tandem Mass Spectrometry

With great biological interest in post-translational modifications (PTMs), various approaches have been introduced to identify PTMs using MS/MS. Recent developments for PTM identification have focused on an unrestrictive approach that searches MS/MS spectra for all known and possibly even unknown types of PTMs at once. However, the resulting expanded search space requires much longer search time and also increases the number of false positives (incorrect identifications) and false negatives (missed true identifications), thus creating a bottleneck in high throughput analysis. Here we introduce MODa, a novel "multi-blind" spectral alignment algorithm that allows for fast unrestrictive PTM searches with no limitation on the number of modifications per peptide while featuring over an order of magnitude speedup in relation to existing approaches. We demonstrate the sensitivity of MODa on human shotgun proteomics data where it reveals multiple mutations, a wide range of modifications (including glycosylation), and evidence for several putative novel modifications. Based on the reported findings, we argue that the efficiency and sensitivity of MODa make it the first unrestrictive search tool with the potential to fully replace conventional restrictive identification of proteomics mass spectrometry data.