114 resultados para Join Ordering
Resumo:
OpenMP是一种支持Fortran,C/C++的共享存储并行编程标准。它基于fork-join的并行执行模型,将程序划分为并行区和串行区。近几年来,OpenMP在SMP(Symmetric Multi-Processing)和多核体系结构的并行编程中得到了广泛的应用。随着多核处理器的发展,实际的应用程序如何充分利用多个处理器核来提高运算效率也成为研究的热点。 在科学计算中,循环结构是最核心的并行对象之一。考虑到负载平衡、调度开销、同步开销等多方面因素,OpenMP标准制定了Static调度、Dynamic调度、Guided调度和Runtime调度等不同策略。针对Guided调度策略不适合递减型循环结构的缺点,本文提出了一种改进的new_guided调度策略,并在OMPi编译器上加以实现。New_guided调度策略的主要思想是对前半部分的循环采用Static调度,后半部分的循环采用Guided调度。此外,本文针对不同的循环结构,在多核处理器上对不同的调度策略进行了评测。测试结果表明,在一般情况下,OpenMP默认的Static策略的调度性能最差;对于规则的循环结构和递增的循环结构,Dynamic调度策略、Guided调度策略和new_guided策略的性能差别不大;对于递减型的循环结构,Dynamic调度策略和new_guided策略的性能相当,要优于Guided调度策略;对于求解Mandelbrot集合这类计算量集中在中间的随机循环结构,Dynamic调度策略优于其它策略,new_guided策略的性能介于Dynamic调度和Guided调度之间。 随着多核处理器的问世和发展,多线程程序设计也已经成为一个不可回避的问题。稀疏矩阵向量乘(SpMV, Sparse Matrix-Vector Multiplication)是一个十分重要且经常被大量调用的科学计算内核。SpMV的存储访问一般都极不规则,导致现有的SpMV算法效率都比较低。目前,多核处理器芯片上的内核数量正在逐步增加。这使得在多核处理器上对SpMV进行并行化加速变得非常重要。本文介绍了稀疏矩阵的两种常用的存储格式CSR和BCSR,并采用OpenMP实现了SpMV的多核并行化。此外,本文还讨论了寄存器分块算法、压缩列索引等优化技术,以及不同调度策略对多线程并行后的SpMV的影响。在曙光天阔服务器S4800A1上的测试表明,大部分矩阵都取得了可扩展、甚至是超线性的加速比,但是对于部分规模较大的矩阵,加速效果并不明显。在我们的测试中,与基于CSR实现的多线程SpMV相比,采用寄存器分块算法优化后的SpMV运算速度平均提高了28.09%。在基于CSR实现的多线程SpMV中,采用列索引优化技术后的程序比优化前的速度平均提高了13.05%。此外,本文实现了一种基于非零元个数的调度策略。在该策略中,每个线程处理几乎相同数量的非零元。我们将它和OpenMP标准提供的三种调度策略进行了测试和分析。测试结果表明:与OpenMP提供的调度策略相比,基于非零元个数的调度策略能取得更好的负载平衡;Dynamic调度和Guided调度在多线程SpMV中的性能基本相当,均优于Static调度策略。
Resumo:
Reliable messaging is a key component necessary for mobile agent systems. Current researches focus on reliable one-to-one message delivery to mobile agents. But how to implement a group communication system for mobile agents remains an open issue, which is a powerful block that facilitates the development of fault-tolerant mobile agent systems. In this paper, we propose a group communication system for mobile agents (GCS-MA), which includes totally ordered multicast and membership management functions. We divide a group of mobile agents into several agent clusters,and each agent cluster consists of all mobile agents residing in the same sub-network and is managed by a special module, named coordinator. Then, all coordinators form a ring-based overlay for interchanging messages between clusters. We present a token-based algorithm, an intra-cluster messaging algorithm and an inter-cluster migration algorithm to achieve atomicity and total ordering properties of multicast messages, by building a membership protocol on top of the clustering and failure detection mechanisms. Performance issues of the proposed system have been analysed through simulations. We also describe the application of the proposed system in the context of the service cooperation middleware (SCM) project.
Resumo:
本文介绍了一种实现藏文字典序排序的方法,它针对藏文"大字丁字符集"编码方案.通过引入有(无)前加基字符的概念,它把待排序的藏字预处理为有(无)前加基字符、前加字符、基字(基字符或者字丁)、后加字符、再后加字符串后,再行比较,从而避免拆分字丁.本实现方法符合ISO/IEC14651标准语义.
Resumo:
The effects of five metal catalysts (K, Na, Ca, Mg, and Fe) on CO2 gasification reactivity of fir char were studied using thermal gravimetric analysis. The degree of carbonization, crystal structure and morphology of char samples was characterized by X-ray diffractometry (XRD) and scanning electron microscopy (SEM). The CO2 gasification reactivity of fir char was improved through the addition of metal catalysts, in the order K>Na>Ca>Fe>Mg. XRD analysis indicated that Na and Ca improved the formation of crystal structure, and that Mg enhanced the degree of carbon structure ordering. SEM analysis showed that spotted activation centers were distributed on the surface of char samples impregnated with catalysts. Moreover, a loose flake structure was observed on the surface of both K-char and Na-char. Finally, the kinetic parameters of CO2 gasification of char samples were calculated mathematically.
Resumo:
A simple, but important three-atom model was proposed at the solid/liquid interface, leading to a new criterion number, lambda, governing the boundary conditions (BCs) in nanoscale. The solid wall is considered as the face-centered-cubic (fcc) structure. The fluid is the liquid argon with the well-known LJ potential. Based on the concept, the two micro-systems have the same BCs if they have The same criterion number. The degree of the locking BCs is enhanced when lambda equals to 0.757. Such critical criterion number results in the substantial epitaxial ordering and one, two, or even three liquid layers are locked by the solid wall, depending on the coupling energy scale ratio of the solid and liquid atoms. With deviation from the critical criterion number, the flow approaches the slip BCs and there are little ordering structures within the liquid. Always at the same criterion number, the degree of the slip is decreased or the locking is enhanced with increasing the coupling energy scale ratio of the solid and liquid atoms. The above analysis is well confirmed by the molecular dynamics (MD) simulation. The slip length is well correlated in terms of the new criterion number. The future work is suggested to extend the present theory for other microstructures of the solid wall atoms and quasi-LJ potentials.
Resumo:
The relationship between liquid crystal orientational ordering and optical diffraction properties is investigated for a two-dimensional square photonic lattice fabricated in a polymer-dispersed liquid crystal (PDLC) composite. Modifications of the nematic director field in the liquid crystal domains were induced by an external applied voltage and by heating over the nematic-isotropic (N-I) phase transition. They were studied by optical polarization microscopy and by analysing far-field optical diffraction patterns. The intensities of various diffraction orders (from the zeroth up to the eighth diffraction order) were monitored with a CCD camera, and their variations were correlated with the modifications of the director field.
Resumo:
A theoretical model for the electronic structure of porous Si is presented. Three geometries of porous Si (wire with square cross section, pore with square cross section, and pore with circular cross section) along both the [001] and [110] directions are considered. It is found that the confinement geometry affects decisively the ordering of conduction-band states. Due to the quantum confinement effect, there is a mixing between the bulk X and GAMMA states, resulting in finite optical transition matrix elements, but smaller than the usual direct transition matrix elements by a factor of 10(-3). We found that the strengths of optical transitions are sensitive to the geometry of the structure. For (001) porous Si the structure with circular pores has much stronger optical transitions compared to the other two structures and it may play an important role in the observed luminescence. For this structure the energy difference between the direct and the indirect conduction-band minima is very small. Thus it is possible to observe photoluminescence from the indirect minimum at room temperature. For (110) porous Si of similar size of cross section the energy gap is smaller than that of (001) porous Si. The optical transitions for all three structures of (110) porous Si tend to be much stronger along the axis than perpendicular to the axis.
Resumo:
The intermittent illumination treatment by white light at elevated temperature is proved to be a convenient and efficient method for the improvement of the stability of hydrogenated amorphous silicon (a-Si:H) films. The effect of the treatment on electrical properties, light-induced degradation, and gap states of undoped a-Si:H films has been investigated in detail. With the increase of cycling number, the dark- as well as photo-conductivities in annealed state and light-soaked state approach each other, presenting an unique irreversible effect. The stabilization and ordering processes by the present treatment can not be achieved merely by annealing under the same conditions. It is shown that the treatment proposed here results in a shift to higher values of the energy barriers between defects and their precursors, and hence an improved stability of a-Si:H films. (C) 1996 American Institute of Physics.
Resumo:
Photoluminescence of GaInP under hydrostatic pressure is investigated. The Gamma valley of disordered GaInP shifts sublinearly upwards with respect to the top of the valence band with increasing pressure and this sublinearity is caused by the nonlinear relationship between lattice constant and hydrostatic pressure. The Gamma valleys of ordered GaInP rise more slowly than that of the disordered one and the relationship between the band gap and the pressure can not be explained in the same way. Taking into account the interactions between the Gamma valley and the folded L valleys, as well as, the X valleys, the experimental pressure dependences of the band gap of ordered GaInP epilayers are calculated and fitted quite well using first order perturbation theory. The results indicate that simultaneous ordering along [111] and [100] directions can occur in ordered GaInP. (C) 1996 American Institute of Physics.
Resumo:
Quantum dot (QD) lasers are expected to have superior properties over conventional quantum well lasers due to a delta-function like density of states resulting from three dimensional quantum confinements. QD lasers can only be realized till significant improvements in uniformity of QDs with free of defects and increasing QD density as well in recent years. In this paper, we first briefly give a review on the techniques for preparing QDs, and emphasis on strain induced self-organized quantum dot growth. Secondly, self-organized In(Ga)As/GaAs, InAlAs/GaAlAs and InAs/InAlAs Qds grown on both GaAs and InP substrates with different orientations by using MBE and the Stranski-Krastanow (SK) growth mode at our labs are presented. Under optimizing the growth conditions such as growth temperature, V/III ratio, the amount of InAs, InxGa1-xAs, InxAl1-xAs coverage, the composition x etc., controlling the thickness of the strained layers, for example, just slightly larger than the critical thickness and choosing the substrate orientation or patterned substrates as well, the sheet density of ODs can reach as high as 10(11) cm(-2), and the dot size distribution is controlled to be less than 10% (see Fig. 1). Those are very important to obtain the lower threshold current density (J(th)) of the QD Laser. How to improve the dot lateral ordering and the dot vertical alignment for realizing lasing from the ground states of the QDs and further reducing the Jth Of the QD lasers are also described in detail. Thirdly based on the optimization of the band engineering design for QD laser and the structure geometry and growth conditions of QDs, a 1W continuous-wave (cw) laser operation of a single composite sheet or vertically coupled In(Ga)As quantum dots in a GaAs matrix (see Fig. 2) and a larger than 10W semiconductor laser module consisted nineteen QD laser diodes are demonstrated. The lifetime of the QD laser with an emitting wavelength around 960nm and 0.613W cw operation at room temperature is over than 3000 hrs, at this point the output power was only reduced to 0.83db. This is the best result as we know at moment. Finally the future trends and perspectives of the QD laser are also discussed.
Resumo:
Self-assembled quantum dots and wires were obtained in the InxGa1-xAs/GaAs and InAs/In0.52Al0.48As/InP systems, respectively, using molecular beam epitaxy (MBE). Uniformity in the distribution, density, and spatial ordering of the nanostructures can be controlled to some extent by adjusting and optimizing the MBE growth parameters. In addition, some interesting observation on the InAs wire alignment on InP(001) is discussed. (C) 2003 Elsevier Science Ltd. All rights reserved.
Resumo:
Molecular beam epitaxy-grown self-assembled In(Ga)As/GaAs and InAs/InAlAs/InP quantum dots (QDs) and quantum wires (QWRs) have been studied. By adjusting growth conditions, surprising alignment. preferential elongation, and pronounced sequential coalescence of dots and wires under specific condition are realized. The lateral ordering of QDs and the vertical anti-correlation of QWRs are theoretically discussed. Room-temperature (RT) continuous-wave (CW) lasing at the wavelength of 960 nm with output power of 3.6 W from both uncoated facets is achieved fi-om vertical coupled InAs/GaAs QDs ensemble. The RT threshold current density is 218 A/cm(2). A RT CW output power of 0.6 W/facet ensures at least 3570 h lasing (only drops 0.83 dB). (C) 2001 Elsevier Science B.V, All rights reserved.
Resumo:
The growth of ordered self-assembled nanoislands on stepped substrates is studied systematically by kinetic Monte Carlo simulations. As the terrace width is small, the formation of nanoislands is confined in the steps and nanoislands ordered in lines or nanowires can be obtained. The Schwoebel barrier at the step edges has a great influence on the evolution of both the size and space distributions of the islands. When the terrace width is relatively large, self-ordering of nanoislands in the center regions of the terraces happens. An unexpected trend of the nanoisland self-ordering is found as the deposition thickness is larger than 0.2 ML, which can be related to the attractive migrations between nearby islands.
Resumo:
Molecular-dynamics simulations have been carried out to investigate the electric hysteresis of barium titanate nanofilm containing oxygen vacancy ordering array parallel to the {101} crystal plane. The results obtained show a significant weakening of polarization retention from non-zero value to zero as the size of the array was reduced to a critical level, which was attributed to the formation and motion of head-to-head domain wall structure under external field loading process. By comparing with materials containing isolated oxygen vacancies, it was found that the zero retention was due to the oxygen vacancy ordering array rather than to the concentration of oxygen vacancy. Copyright (C) EPLA, 2010
Resumo:
We report the first three-particle coincidence measurement in pseudorapidity (Delta eta) between a high transverse momentum (p(perpendicular to)) trigger particle and two lower p(perpendicular to) associated particles within azimuth |Delta phi| < 0.7 in root s(NN) = 200 GeV d + Au and Au + Au collisions. Charge ordering properties are exploited to separate the jetlike component and the ridge (long range Delta eta correlation). The results indicate that the correlation of ridge particles are uniform not only with respect to the trigger particle but also between themselves event by event in our measured Delta eta. In addition, the production of the ridge appears to be uncorrelated to the presence of the narrow jetlike component.