Multi-threaded Parallel Research Articles

A space-efficient alphabet-independent Four-Russians' lookup table and a multithreaded Four-Russians' edit distance algorithm

Given two strings X (|X|=m) and Y (|Y|=n) over an alphabet Σ, the edit distance between X and Y can be computed in O(mn/t) time with the help of the Four-Russians' lookup table whose block size is t. The Four-Russians' lookup table can be constructed in O((3|Σ|)2tt2) time using O((3|Σ|)2tt) space. However, the construction time and space requirement of the lookup table grow very fast as the alphabet size increases and thus it has been used only when |Σ| is very small. For example, when a string is a protein sequence, |Σ|=20 and thus it is almost impossible to use the Four-Russians' lookup table on typical workstations. In this paper, we present an efficient alphabet-independent Four-Russians' lookup table. It requires O(32t(2t)!t) space and can be constructed in O(32t(2t)!t2) time. Thus, the Four-Russians' lookup table can be constructed and used irrespective of the alphabet size. The time and space complexity were achieved by compacting the lookup table using a clever encoding of the preprocessed strings. Experimental results show that the space requirement of the lookup table is reduced to about 1/5,172,030 of its original size when |Σ|=26 and t=4. Furthermore, we present efficient multithreaded parallel algorithms for edit distance computation using the Four-Russians' lookup table. The parallel algorithm for lookup table construction runs in O(t) time and the parallel algorithm for edit distance computation between X and Y runs in O(m+n) time. Experiments performed on CUDA-supported GPU show that our algorithm runs about 942 times faster than the sequential version of the original Four-Russians' algorithm for 100 pairs of random strings of length approximately 1,000 when |Σ|=4 and t=4.

Thread-Aware Adaptive Prefetcher on Multicore Systems

Most processors employ hardware data prefetching techniques to hide memory access latencies. However, the prefetching requests from different threads on a multicore processor can cause severe interference with prefetching and/or demand requests of others. The data prefetching can lead to significant performance degradation due to shared resource contention on shared memory multicore systems. This article proposes a thread-aware data prefetching mechanism based on low-overhead runtime information to tune prefetching modes and aggressiveness, mitigating the resource contention in the memory system. Our solution has three new components: (1) a self-tuning prefetcher that uses runtime feedback to dynamically adjust data prefetching modes and arguments of each thread, (2) a filtering mechanism that informs the hardware about which prefetching request can cause shared data invalidation and should be discarded, and (3) a limiter thread acceleration mechanism to estimate and accelerate the critical thread which has the longest completion time in the parallel region of execution. On a set of multithreaded parallel benchmarks, our thread-aware data prefetching mechanism improves the overall performance of 64-core system by 13% over a multimode prefetch baseline system with two-level cache organization and conventional modified, exclusive, shared, and invalid-based directory coherence protocol. We compare our approach with the feedback directed prefetching technique and find that it provides 9% performance improvement on multicore systems, while saving the memory bandwidth consumption.

A Parallel Local Reconnection Approach for Tetrahedral Mesh Improvement

A multi-threaded parallel local reconnection algorithm is proposed for tetrahedral meshes. It defines a feature point within the region involved in each operation, and sorts the features points along a Hilbert curve. The decomposition of this Hilbert curve results in a load-balanced distribution of local operations. Meanwhile, the regions of concurrently executed local operations are separated far away, such that the possibility of interference is reduced to a very low level. Finally, a parallel mesh improver is developed by combining the proposed algorithm with a parallel mesh smoothing algorithm, and its effectiveness and efficiency is verified in various numerical experiments.

ITri: Index-based triangle listing in massive graphs

Triangle listing is a basic operator when dealing with many graph problems. However, in-memory algorithms do not work well with recently developed massive graphs such as social networks because these graphs cannot be accommodated in the memory. Thus, external memory-based algorithms have been proposed recently, but these approaches still require frequent multiple scans of the whole graph on the disk and large volumes of calculations are performed that involve the whole graph during every iteration. In this study, we propose a novel index-based method for listing triangles in massive graphs. First, we present new notions for the vertex range index and potential cone vertex index. Next, we propose an index join-based triangle listing algorithm. Our method accesses the indexed data asynchronously and joins them to list triangles using a multi-threaded parallel processing technique. Based on experiments, we demonstrate that our algorithm outperforms the state-of-the-art solution methods by three to eight times in terms of the wall clock time.

A parallel matrix-free conservative solution interpolation on unstructured tetrahedral meshes

This document presents an interpolation operator on unstructured tetrahedral meshes that satisfies the properties of mass conservation, P1P1-exactness (order 2) and maximum principle. Interpolation operators are important for many applications in scientific computing. For instance, in the context of anisotropic mesh adaptation for time-dependent problems, the interpolation stage becomes crucial as the error due to solution transfer accumulates throughout the simulation. This error can eventually spoil the overall solution accuracy. When dealing with conservation laws in CFD, solution accuracy requires enforcement of mass preservation throughout the computation, in particular in long time scale computations. In the proposed approach, the conservation property is achieved by local mesh intersection and quadrature formulae. Derivatives reconstruction is used to obtain a second order method. Algorithmically, our goal is to design a method which is robust and efficient. The robustness is mandatory to obtain a reliable method on real-life applications and to apply the operator to highly anisotropic meshes. The efficiency is achieved by designing a matrix-free operator which is highly parallel. A multi-thread parallelization is given in this work. Several numerical examples are presented to illustrate the efficiency of the proposed approach.

A Massively Parallel Computational Method of Reading Index Files for SOAPsnv.

SOAPsnv is the software used for identifying the single nucleotide variation in cancer genes. However, its performance is yet to match the massive amount of data to be processed. Experiments reveal that the main performance bottleneck of SOAPsnv software is the pileup algorithm. The original pileup algorithm's I/O process is time-consuming and inefficient to read input files. Moreover, the scalability of the pileup algorithm is also poor. Therefore, we designed a new algorithm, named BamPileup, aiming to improve the performance of sequential read, and the new pileup algorithm implemented a parallel read mode based on index. Using this method, each thread can directly read the data start from a specific position. The results of experiments on the Tianhe-2 supercomputer show that, when reading data in a multi-threaded parallel I/O way, the processing time of algorithm is reduced to 3.9s and the application program can achieve a speedup up to 100×. Moreover, the scalability of the new algorithm is also satisfying.

A Hybrid Technique to Secure E commerce Transaction with the Help of AES Encryption and Stenography in Image

Electronic commerce, commonly known as eCommerce is trade in products using the internet. Security at the e-commerce becomes more and more important. Electronic transaction security is a challenging task because of the insecure communication channel. In this paper, we proposed a new algorithm to make an e commerce transaction more secure. In this work we are using AES to encrypt data and this data is embedded to image edges to perform Stenography in this paper target on compress the time of encryption and decryption adopting parallel processing on which AES will be applied freely. In the results show the comparison between our work and previous work and we found that parallel processing and multi threading on the AES algorithm our results are quite improved, our proposed algorithm is taking much less time in encrypt and embedding or decrypt and extraction.

A SPATIOTEMPORAL AGGREGATION QUERY METHOD USING MULTI-THREAD PARALLEL TECHNIQUE BASED ON REGIONAL DIVISION

Abstract. Existing spatiotemporal database supports spatiotemporal aggregation query over massive moving objects datasets. Due to the large amounts of data and single-thread processing method, the query speed cannot meet the application requirements. On the other hand, the query efficiency is more sensitive to spatial variation then temporal variation. In this paper, we proposed a spatiotemporal aggregation query method using multi-thread parallel technique based on regional divison and implemented it on the server. Concretely, we divided the spatiotemporal domain into several spatiotemporal cubes, computed spatiotemporal aggregation on all cubes using the technique of multi-thread parallel processing, and then integrated the query results. By testing and analyzing on the real datasets, this method has improved the query speed significantly.

Parallel GPU-based Plane-Sweep Algorithm for Construction of iCPI-Trees

This article tackles the problem of efficient construction of iCPI trees, frequently used in co-location pattern discovery in spatial databases. It discusses the methods for parallelization of iCPI-tree construction and plane-sweep algorithms used in state-of-the-art algorithms for co-location pattern mining. The main contribution of this paper is threefold: (1) a general algorithm for parallel iCPI-tree construction is presented, (2) two variants of parallel plane-sweep algorithm (which can be used in conjunction with the aforementioned iCPI-tree construction algorithm) are introduced and (3) all three algorithms are implemented on CUDA GPU platform and their performance is tested against an efficient multithreaded parallel implementation of iCPI-tree construction on CPU. Experiments prove that our solutions allow for large speedups over CPU version of the algorithm. This paper is an extension of the conference paper (Andrzejewski & Boinski, 2014).

A multi-thread parallel computation method for dynamic simulation of molecular weight distribution of multisite polymerization

Molecular weight distribution (MWD) is an important quality index of polymer products. Many methods have been proposed to dynamically simulate the MWD of polymerization, but these methods are normally designed for serial computations. In this paper, a multi-thread parallel computation method was proposed for multisite free-radical polymerization. Analysis of the relationship among different subtasks revealed a combined parallel strategy by fully exploiting the parallel feature of the process. A good performance was obtained to accelerate the dynamic simulation of MWD based on Flory method. We theoretically analyzed the speedup ratio (SR) and parallel efficiency (PE). Results showed that software algorithm and hardware configuration exhibited a good match. The efficiency of the proposed parallel method was presented through industrial slurry processes that used high-density polyethylene (HDPE).

Full correlation matrix analysis (FCMA): An unbiased method for task-related functional connectivity

BackgroundThe analysis of brain imaging data often requires simplifying assumptions because exhaustive analyses are computationally intractable. Standard univariate and multivariate analyses of brain activity ignore interactions between regions and analyses of interactions (functional connectivity) reduce the computational challenge by using seed regions of interest or brain parcellations. New methodTo meet this challenge, we developed full correlation matrix analysis (FCMA), which leverages and optimizes algorithms from parallel computing and machine learning to efficiently analyze the pairwise correlations of all voxels in the brain during different cognitive tasks, with the goal of identifying task-related interactions in an unbiased manner. ResultsWhen applied to a localizer dataset on a small compute cluster, FCMA accelerated a naive, serial approach by four orders of magnitude, reducing running time from two years to one hour. In addition to this performance gain, FCMA emphasized different brain areas than existing methods. In particular, beyond replicating known category selectivity in visual cortex, FCMA also revealed a region of medial prefrontal cortex whose selectivity derived from differential patterns of functional connectivity across categories. Comparison with existing method(s)For benchmarking, we started with a naive approach and progressively built up to the complete FCMA procedure by adding optimized classifier algorithms, multi-threaded parallelism, and multi-node parallelism. To evaluate what can be learned with FCMA, we compared it against multivariate pattern analysis of activity and seed-based analysis of functional connectivity. ConclusionsFCMA demonstrates how advances in computer science can alleviate computational bottlenecks in neuroscience. We have released a software toolbox to help others evaluate FCMA.

A method of Level 1 frames-selection based on GPU for new vacuum solar telescope

The sun high resolution imaging observation is the most important way of New Vacuum Solar Telescope (NVST) located in Fuxian Lake of Yunnan. This paper mainly describes the method of the Level 1 data processing of NVST. Currently, the times used by NVST to calculate the 100 images is about 20 s, which is almost 1.7 times of 100 image acquisition time of 12 s. It cant meet the demand of real time processing. In this situation, we have decided to use GPU to process the data. Because the GPU has multi-threading parallelism and a superiority in image processing. We implemented the algorithm by GPU to short the processing time to meet the demand without changing the original algorithm. Thus, it ensures the accuracy of the original operation without affecting the subsequent operation. The experiment includes two main parts, frames-selection method and shift-and-add method. Frames-selection uses speckle interferometry (SI). It selects part of the images with better quality and then shifts-and-adds them. Not all of the image data are required. Usually we only concerned with the necessary part. So in the election process, we only calculated the frequency and the bandwidth. It will highly improve the accuracy of the selected frame. The Fourier transform spent the most time in this experiment. It runs three times in the frame selection and processes second times in the shift-and add. This part of the length of time consuming operation largely determines the efficiency of the program. The CUDA provides the standard library functions, which called CUBLAS and greatly improve the operating efficiency of the program. The library functions CUBLAS provides matrix operations. In this study, most of the matrix operations referenced standard library functions to achieve a large number of image matrix of the parallel arithmetic and other functions. The rest of them used a GPU multi-threading technology. In our experiments, 100 images calibration only takes 0.25 s, GPU runtime also only 0.6 s, the processing time of a single frame is only 4.7 ms. The time of processing image exposure spends only 12 ms. Therefore, it means that in the next frame of the image exposure time, the processing steps of the previous frame image exposure can be completed synchronously. In the current conditions, it meet the demand of real-time processing.

Fast Approach to Video Stabilization Based on Hierarchical Estimation of Global Motion with Parallel Computation

In this article, a fast approach of video stabilization is presented based on haerachical estimmation of global motion and multi-thread parallel computation. At first, the structure-texture decomposition is adopted to tackle the negative effect from illumination changes. Then according to pyramid structure of image texture a hierachical model is built to estimate global motion parameters with least square method. Afterwards Gaussian smoothing method is employed to eliminate error accumulation in motion compensation. Meanwhile the multi-thread parallel computation is used to speed up the processing efficiency so as to achieve a real time performance for videos with 25fps.

IMPROVING THE EFFICIENCY OF THE METHOD OF SPECULATIVE MULTI-THREADING FOR DYNAMIC PARALLELIZING

The possibilities of the method of speculative multithreading dynamic parallelization of sequential programs. It implemented its formalization and identified characteristics of the method, the account which will increase the efficiency of its use in automatic parallelizing. The results of experimental studies of the algorithm dynamically allocate speculative regions.

Numerical simulation of multi-combined effects of parameters on polymer charging characteristics due to electron irradiation

Charging characteristics of an insulator specimen due to non-penetrated electron irradiation have been attracting a great deal of attention in the fields such as scanning electron microscopy, electron probe analysis, and space irradiation. In this paper, we use a numerical simulation model based on Monte Carlo method for investigating the electron scattering. The elastic scattering is calculated with the Mott cross-section, and the inelastic scattering is simulated with Penn model and the fast secondary electron model according to the primary energy. The charge transport caused by the build-in electric field and charge density gradient is calculated with finite-difference time-domain method. Multi-combined effect of correlative parameters on charging characteristics is investigated by efficient multithreading parallel computing. During the irradiation, the landing energy of primary electrons decreases due to the negative surface potential, which makes the secondary electron yield increase. Variations of secondary electron current and sample current are presented to verify the validity of the simulation model by comparing with existing experimental results. Evolutions of leakage current, surface potential and internal space charge quantity are calculated under different conditions of incident electron current, primary energy and sample thickness. The results are presented in contour maps with different multi-parameter combinations, primary energy and sample mobility, primary energy and sample thickness, and primary energy and incident current. The balance state of charging will be determined by leakage current under conditions of a larger primary energy, sample mobility, incident current, or a less sample thickness, which is shown as the leakage current dominated mode. While in the cases of a lower primary energy, sample mobility, incident current, or a larger sample thickness, the balance state of charging is mainly dominated by secondary electron current, as the secondary electron current dominated mode. In other cases except the above two, the balance state will be determined by both leakage and secondary currents as the mixture mode. In the same mode, variations of charging characteristics with parameters are monotonic. When the change of a parameter makes the negative surface potential increase, the effect of this parameter on negative surface potential will be weakened, while the effects of other parameters on the negative potential will be enhanced. With the change of current dominated mode, the total charge quantity exhibits the local maximum with respect to the sample thickness, and the value of this maximum increases with primary energy. Moreover, the leakage current increases with incident current linearly. The presented results can be helpful for understanding regularities and mechanisms of charging due to electron irradiation, and estimating the charging intensity under different conditions of irradiation and sample material.

Shared memory parallelism for 3D Cartesian discrete ordinates solver

This paper describes the design and the performance of DOMINO, a 3D Cartesian SN solver that implements two nested levels of parallelism (multicore+SIMD) on shared memory computation nodes. DOMINO is written in C++, a multi-paradigm programming language that enables the use of powerful and generic parallel programming tools such as Intel TBB and Eigen. These two libraries allow us to combine multi-thread parallelism with vector operations in an efficient and yet portable way. As a result, DOMINO can exploit the full power of modern multi-core processors and is able to tackle very large simulations, that usually require large HPC clusters, using a single computing node. For example, DOMINO solves a 3D full core PWR eigenvalue problem involving 26 energy groups, 288 angular directions (S16), 46×106 spatial cells and 1×1012 DoFs within 11h on a single 32-core SMP node. This represents a sustained performance of 235 GFlops and 40.74% of the SMP node peak performance for the DOMINO sweep implementation. The very high Flops/Watt ratio of DOMINO makes it a very interesting building block for a future many-nodes nuclear simulation tool.

유한요소 구조해석 다중쓰레드 병렬 선형해법의 스케쥴링 및 부하 조절 기법 연구

본 논문은 최근에 널리 사용되는 다중코어 컴퓨팅 환경에서 병렬 다중프론트 해법의 스케쥴링 및 부하조절 기법에 대해 논의한다. 통상적으로 구조해석 문제들은 불규칙한 격자계와 혼재된 물성 때문에 병렬화 알고리즘 구현 시 병목현상을 일으키고 불필요한 유휴시간을 초래한다. 따라서 이를 극복하며 효율성을 향상시키기 위해 다중쓰레드 기반 환경에 걸맞는 작업 스케쥴링 및 부하 분산 기법의 적용이 필수적이다. 본 논문에서 제시된 정적, 동적 스케줄링 기법과 정적 전 임무 분산, 최소최대 임무 결합 등의 부하 분산 기법들에 대한 이론적, 실제 결과를 제시함으로서 그 유용성을 논의하고자 한다. In this paper, task scheduling and load balancing methods of multifrontal solution methods of finite element structural analysis in a modern multicore machine are introduced. Many structural analysis problems have generally irregular grid and many kinds of properties and materials. These irregularities and heterogeneities lead to bottleneck of parallelization and cause idle time to analysis. Therefore, task scheduling and load balancing are desired to reduce inefficiency. Several kinds of multithreaded parallelization methods are presented and comparison between static and dynamic task scheduling are shown. To reduce the idle time caused by irregular partitioned subdomains, computational load balancing methods, Balancing all tasks and minmax task pairing balancing, are invented. Theoretical and actual elapsed time are shown and the reason of their performance gap are discussed.

An automatic extraction method for individual tree crowns based on self-adaptive mutual information and tile computing

Forest data acquisition, which is of crucial importance for modeling global biogeochemical cycles and climate, makes a contribution to building the ecological Digital Earth (DE). Due to the complex calculations and large volumes of data associated with high-resolution images of large areas, accurate and effective extraction of individual tree crowns remains challenging. In this study, two GeoEye-1 panchromatic images of Beihai and Ningbo in China with areas of 5 and 25 km2, respectively, were used as experimental data to establish a novel method for the automatic extraction of individual tree crowns based on a self-adaptive mutual information (SMI) algorithm and tile computing technology (SMI-TCT). To evaluate the performance of the algorithm, four commonly used algorithms were also applied to extract the individual tree crowns. The overall accuracy of the proposed method for the two experimental areas was superior to that of the four other algorithms, with maximum extraction accuracies of 85.7% and 63.8%. Moreover, the results also indicated that the novel method was suitable for individual tree crowns extraction in sizeable areas because of the multithread parallel computing technology.

Playback system designed for X-Band SAR

SAR(Synthetic Aperture Radar) has extensive application because it is daylight and weather independent. In particular, X-Band SAR strip map, designed by Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, provides high ground resolution images, at the same time it has a large spatial coverage and a short acquisition time, so it is promising in multi-applications. When sudden disaster comes, the emergency situation acquires radar signal data and image as soon as possible, in order to take action to reduce loss and save lives in the first time. This paper summarizes a type of X-Band SAR playback processing system designed for disaster response and scientific needs. It describes SAR data workflow includes the payload data transmission and reception process. Playback processing system completes signal analysis on the original data, providing SAR level 0 products and quick image. Gigabit network promises radar signal transmission efficiency from recorder to calculation unit. Multi-thread parallel computing and ping pong operation can ensure computation speed. Through gigabit network, multi-thread parallel computing and ping pong operation, high speed data transmission and processing meet the SAR radar data playback real time requirement.

Scalable CAIM discretization on multiple GPUs using concurrent kernels

Class-attribute interdependence maximization (CAIM) is one of the state-of-the-art algorithms for discretizing data for which classes are known. However, it may take a long time when run on high-dimensional large-scale data, with large number of attributes and/or instances. This paper presents a solution to this problem by introducing a graphic processing unit (GPU)-based implementation of the CAIM algorithm that significantly speeds up the discretization process on big complex data sets. The GPU-based implementation is scalable to multiple GPU devices and enables the use of concurrent kernels execution capabilities of modern GPUs. The CAIM GPU-based model is evaluated and compared with the original CAIM using single and multi-threaded parallel configurations on 40 data sets with different characteristics. The results show great speedup, up to 139 times faster using four GPUs, which makes discretization of big data efficient and manageable. For example, discretization time of one big data set is reduced from 2 h to $$<$$ < 2 min.