一种可扩展的三维半导体器件并行数值模拟算法

在基于漂移-扩散模型的三维半导体器件数值模拟中,通过有限体积法进行数值离散,采用完全耦合的牛顿迭代求解非线性代数方程组,并使用基于代数多重网格预条件子的GMRES方法求解牛顿迭代中的线性方程组,构造一种稳健且高度可扩展的非结构四面体网格上求解半导体方程的并行算法.基于PHG平台实现该算法的并行计算程序,并对PN结和MOS场效应晶体管等问题进行了最大网格规模达到5亿单元、最大并行规模达到1 024进程的大规模数值模拟实验,结果表明,该算法计算效率高,可扩展性好.     

封装结构强电磁脉冲多物理效应并行计算程序研制

介绍了自主研发的强电磁脉冲多物理效应并行计算程序JEMS-CDS-System的情况,该程序采用时域有限元方法,基于JAUMIN并行自适应结构网格支撑框架研制,并行效能高,可扩展性强,且支持动态负载平衡。通过算例测试表明,该程序对于键合线的电-热-应力失效过程的最高温度与范式等效应力计算结果与COMSOL软件计算结果吻合较好;SiP功率放大模块的热-应力耦合天河2高性能计算平台并行计算结果表明,该程序在CPU1024核时,具有38.1%并行效率。     

非定常Navier-Stokes方程基于两重网格离散的有限元并行算法

基于两重网格离散和区域分解技巧,提出三种求解非定常Navier-Stokes方程的有限元并行算法.算法的基本思想是在每一时间迭代步,在粗网格上采用Oseen迭代法求解非线性问题,在细网格上分别并行求解Oseen、Newton、Stokes线性问题以校正粗网格解.对于空间变量采用有限元离散,时间变量采用向后Euler格式离散.数值实验验证了算法的有效性.     

带阻尼项定常Navier-Stokes方程的并行两水平有限元算法

基于两重网格离散和区域分解技术，提出数值求解带阻尼项定常Navier-Stokes方程的三种并行两水平有限元算法。其基本思想是首先在粗网格上求解完全的非线性问题，以获得粗网格解，然后在重叠的局部细网格子区域上并行求解Stokes、 Oseen和Newton线性化的残差问题，最后在非重叠的局部细网格子区域上校正近似解。数值算例验证了算法的有效性。     

抛物方程的多重非均匀网格模型及其应用

提出了抛物方程的多重非均匀网格模型,以准确求解三维空间存在多辐射源的电波传播问题。通过对不同辐射源建立不同的坐标系,并对其仿真空间采用不同的非均匀网格划分,构建了抛物方程的多重非均匀网格模型。在此基础上,实现了三维多辐射源问题的并行计算。实例仿真了空间存在四个辐射源的电波传播特性。结果表明,抛物方程的多重非均匀网格模型能够准确求解多源的空间电磁场分布特性,且在该算例中,并行技术使得抛物方程的计算速度提升了2.41倍,极大地提高了抛物方程对三维多源问题的求解效率。     

抛物方程的多重非均匀网格模型及其应用

提出了抛物方程的多重非均匀网格模型,以准确求解三维空间存在多辐射源的电波传播问题。通过对不同辐射源建立不同的坐标系,并对其仿真空间采用不同的非均匀网格划分,构建了抛物方程的多重非均匀网格模型。在此基础上,实现了三维多辐射源问题的并行计算。实例仿真了空间存在四个辐射源的电波传播特性。结果表明,抛物方程的多重非均匀网格模型能够准确求解多源的空间电磁场分布特性,且在该算例中,并行技术使得抛物方程的计算速度提升了2.41倍,极大地提高了抛物方程对三维多源问题的求解效率。     

3维全电磁粒子软件NEPTUNE中的并行计算方法

介绍了NEPTUNE软件采用的一些并行计算方法：采用“块-网格片”二层并行区域分解方法,使计算规模能够扩展到上千个处理器核。基于复杂几何特征采用自适应技术并行生成结构网格,在原有规则区域的基础上剔除无效网格,大幅降低了存储量和并行执行时间。在经典的Boris和SOR迭代方法基础上,采用红黑排序和几何约束,提出了非规则区域上的Poisson方程并行求解方法。采用这些方法后,当使用NEPTUNE软件模拟MILO器件时,可在1 024个处理器核上获得51.8%的并行效率。     

基于JASMIN并行框架的2.5维粒子模拟程序NEPTUNE2D的研制

介绍了2.5维自主研制的并行电磁粒子模拟程序NEPTUNE2D初步研发情况。该程序基于JASMIN并行自适应结构网格支撑框架研制,并行效能高,可扩展性强,且支持动态负载平衡;采用新型PIC算法替代传统算法,避免求解泊松方程修正电场,更适用于大规模并行计算;程序支持r-z坐标系下的器件仿真,可应用于高功率微波器件、电真空器件的快速模拟设计。该程序现已完成电磁场更新、粒子推进、电磁场注入/引出、粒子发射/吸收等基本物理功能模块的研制,并通过同轴线、圆波导、同轴二极管及无箔二极管算例模拟验证了模块的正确性。最后,应用NEPTUNE2D程序设计了一个高效同轴相对论返波管,给出了粒子模拟结果和并行性能测试结果。     

基于JASMIN并行框架的2.5维粒子模拟程序NEPTUNE2D的研制北大核心CSCD

介绍了2.5维自主研制的并行电磁粒子模拟程序NEPTUNE2D初步研发情况。该程序基于JASMIN并行自适应结构网格支撑框架研制,并行效能高,可扩展性强,且支持动态负载平衡;采用新型PIC算法替代传统算法,避免求解泊松方程修正电场,更适用于大规模并行计算;程序支持r-z坐标系下的器件仿真,可应用于高功率微波器件、电真空器件的快速模拟设计。该程序现已完成电磁场更新、粒子推进、电磁场注入/引出、粒子发射/吸收等基本物理功能模块的研制,并通过同轴线、圆波导、同轴二极管及无箔二极管算例模拟验证了模块的正确性。最后,应用NEPTUNE2D程序设计了一个高效同轴相对论返波管,给出了粒子模拟结果和并行性能测试结果。     

隐-显积分因子间断Galerkin方法求解二维辐射扩散方程

提出一种求解二维非平衡辐射扩散方程的数值方法.空间离散上采用加权间断Galerkin有限元方法,其中数值流量的构造采用一种新的加权平均;时间离散上采用隐-显积分因子方法,将扩散系数线性化,然后用积分因子方法求解间断Galerkin方法离散后的非线性常微分方程组.数值试验中在非结构网格上求解了多介质的辐射扩散方程.结果表明:对于强非线性和强耦合的非线性扩散方程组,该方法是一种非常有效的数值算法.     

ARM架构下计算机图像并行化处理技术研究

为在图像处理与分析时具备良好的视觉效果,提高图像处理的速度,需要对ARM架构下计算机图像并行化处理技术进行研究。当前采用的方法是对各种变换频域图像特征提取与计算机图像集合特征的提取进行相结合,克服了当前对图像进行提取时存在图像形状描述的缺陷,提取图像特征向量维数相对较低。实验表明,通过对图像进行特征提取能很好的对图像效果进行展示,将图像的纹理特征进行详细的表述,将该方法应用到图像处理技术当中,具有良好的去噪效果及扩展性,该方法过程简单,但存在图像视觉效果较差的问题。为此,提出一种ARM架构下计算机图像并行化处理技术研究方法。该方法首先利用非局部均值去噪算法对图像进行去噪处理,然后结合图像去噪的结果利用小波变换对去噪图像进行边缘检测,最后采用非线性增强算法对图像进行增强完成对ARM架构下计算机图像并行化处理技术研究。实验结果表明,所提方法不仅提高图像处理速度,还提高图像视觉效果,具有广泛的应用价值。     

一种光电混合SIMD型计算机(OEHSC)系统结构及其算法的设计

本文基于光计算机结构和算法必须以“门”一级并行为基础的观点，提出了可实现性能较好的光电混合ＳＩＭＤ型计算机的结构和以位平面数据表示为基础的并行算法。通过性能分析，证明了光计算机计算能力的潜力及光电混合系统的有效性     

The Elastoplast Discontinuous Galerkin (EDG) method for the Navier–Stokes equations

The present work details the Elastoplast (this name is a translation from the French “sparadrap”, a concept first applied by Yves Morchoisne for Spectral methods [1]) Discontinuous Galerkin (EDG) method to solve the compressible Navier–Stokes equations. This method was first presented in 2009 at the ICOSAHOM congress with some Cartesian grid applications. We focus here on unstructured grid applications for which the EDG method seems very attractive. As in the Recovery method presented by van Leer and Nomura in 2005 for diffusion, jumps across element boundaries are locally eliminated by recovering the solution on an overlapping cell. In the case of Recovery, this cell is the union of the two neighboring cells and the Galerkin basis is twice as large as the basis used for one element. In our proposed method the solution is rebuilt through an L² projection of the discontinuous interface solution on a small rectangular overlapping interface element, named Elastoplast, with an orthogonal basis of the same order as the one in the neighboring cells. Comparisons on 1D and 2D scalar diffusion problems in terms of accuracy and stability with other viscous DG schemes are first given. Then, 2D results on acoustic problems, vortex problems and boundary layer problems both on Cartesian or unstructured triangular grids illustrate stability, precision and versatility of this method.     

On-line range images registration with GPGPU

This paper concerns implementation of algorithms in the two important aspects of modern 3D data processing: data registration and segmentation. Solution proposed for the first topic is based on the 3D space decomposition, while the latter on image processing and local neighbourhood search. Data processing is implemented by using NVIDIA compute unified device architecture (NIVIDIA CUDA) parallel computation. The result of the segmentation is a coloured map where different colours correspond to different objects, such as walls, floor and stairs. The research is related to the problem of collecting 3D data with a RGB-D camera mounted on a rotated head, to be used in mobile robot applications. Performance of the data registration algorithm is aimed for on-line processing. The iterative closest point (ICP) approach is chosen as a registration method. Computations are based on the parallel fast nearest neighbour search. This procedure decomposes 3D space into cubic buckets and, therefore, the time of the matching is deterministic. First technique of the data segmentation uses accele-rometers integrated with a RGB-D sensor to obtain rotation compensation and image processing method for defining pre-requisites of the known categories. The second technique uses the adapted nearest neighbour search procedure for obtaining normal vectors for each range point.     

基于JASMIN框架的辐射流体与粒子输运耦合计算

基于JASMIN框架的"联邦计算",将两个串行程序辐射流体RH2D与粒子输运Sn2D作为独立"邦元"耦合连接,形成的集成程序RHSn2D可以采用数千处理器并行模拟多物理耦合问题.集成程序RHSn2D中的邦元具有各自独立的网格划分与并行算法,同时借助框架技术,可以屏蔽邦元间的并行数据传递.算例表明,对于应用问题规模(90 720个网格单元,辐射流体100个Patch,粒子输运2 835个Patch,Sn方向48,16群),集成程序RHSn2D采用1 024个处理器可以达到36%的并行效率.     

AQFT from <Emphasis Type="Italic">n</Emphasis>-Functorial QFT

There are essentially two different approaches to the axiomatization of quantum field theory (QFT): algebraic QFT, going back to Haag and Kastler, and functorial QFT, going back to Atiyah and Segal. More recently, based on ideas by Baez and Dolan, the latter is being refined to “extended” functorial QFT by Freed, Hopkins, Lurie and others. The first approach uses local nets of operator algebras which assign to each patch an algebra “of observables”, the latter uses n-functors which assign to each patch a “propagator of states”.In this note we present an observation about how these two axiom systems are naturally related: we demonstrate under mild assumptions that every 2-dimensional extended Minkowskian QFT 2-functor (“parallel surface transport”) naturally yields a local net, whose locality derives from the 2-categorical exchange law, and which is covariant if the 2-functor is equivariant. This is obtained by postcomposing the propagation 2-functor with an operation that mimics the passage from the Schrödinger picture to the Heisenberg picture in quantum mechanics. The argument has a straightforward generalization to general Lorentzian structure, bare lightcone structure and higher dimensions. It does not, however, by itself imply anything about the existence of a vacuum state or about positive energy representations.     

基于几何光学的红外成像模型

提出一种基于光线追击法的CFD湍流场的红外成像模型,用“数字图像”模拟“真实目标”进行仿真计算。通过对比原始图与仿真结果,得出超高声速导弹附面层可引起降晰效应的结论,为弹载计算机的图像复原算法提供了理论依据。按照导引头的工作原理,把模型分为远距离、临界距离和近距离3种类型分别讨论,最后分析了点扩散函数与数字图像处理中降晰模板之间的关系。     

BGK方法在三维非结构网格上的初步应用

将BGK计算方法从二维拓展到三维并且应用于三维非结构网格,具有重要的理论价值和实用价值.采用旋转局部座标的方法,发展了一种针对三维非结构网格的BGK计算方法.在计算过程中,将最小二乘法应用于三维非结构网格的导数计算.对三维激波管和三维欠膨胀垂直冲击射流等两个算例进行了细致分析.这两个算例的计算结果表明,该方法在三维非结构网格上的初步应用是成功的 关键词： 气动BGK方法 三维 非结构网格     

Collisional gyrokinetic full-f particle-in-cell simulations on open field lines with PICLS

Applying gyrokinetic simulations in theoretical turbulence and transport studies for the plasma edge and scrape-off layer (SOL) presents significant challenges. To particularly account for steep density and temperature gradients in the SOL, the “full-f” code PICLS was developed. PICLS is a gyrokinetic particle-in-cell (PIC) code, is based on an electrostatic model with a linearized field equation, and uses kinetic electrons. In previously published results, we applied PICLS to the well-studied 1D parallel transport problem during an edge-localized mode (ELM) in the SOL without collisions. As an extension to this collision-less case and in preparation for 3D simulations, in this work, a collisional model will be introduced. The implemented Lenard–Bernstein collision operator and its Langevin discretization will be shown. Conservation properties of the collision operator, as well as a comparison of the collisional and non-collisional case, will be discussed.     

Observed two-dimensional tunnel bifurcations in an external electric field

The controllability problem for two-dimensional dissipative tunneling in the system of tunnel-coupled quantum dots (a quantum molecule), interacting quantum molecules, and the system “ACM/CTM cantilever tip-quantum dot” simulated by a 2D oscillator potential in a heat bath and an external electric field is investigated. The obtained results qualitatively correspond to the separate experimental volt-ampere characteristics (VACs) for the system “platinized ACM/CTM cantilever tip-gold quantum dot” obtained at Scientific-Research Physical-Technical Institute with Nizhnii Novgorod State University. The previously-predicted 2D tunnel bifurcations with dissipation for the case of interacting particles tunneling in parallel are found to be experimentally observed and stable.