2014年中国高性能计算机发展现状分析与展望

本文根据2014 年11 月发布的中国高性能计算机TOP100 排行榜的数据,对国内高性能计算机的发展现状从总体性能、制造商、行业领域等方面进行了深入分析。我们发现, 由于天河二号超级计算机的发布,中国TOP100 的平均Linpack 性能已经从两年前的落后一年变为比国际TOP500 平均Linpack 性能高的局面,且TOP100 的入门性能门槛也第一次超过TOP500。在此基础上,根据十三届排行榜积累的性能数据和能够得到的其他公开历史数据,对未来几年中国大陆高性能计算机的发展趋势进行了分析预测。根据新的数据和实际情况,经修正后我们认为,累计Linpack 性能将在2015 年到2016 年间达到100Petaflops, 比我们原来预期的晚;峰值100Petaflops 的机器将在2015 年到2016 年间出现;峰值Exaflops 的机器将在2019 年到2020 年间出现,比我们原来预期的晚2 年。单台机器峰值超过累计Linpack 性能的情况出现在2018 到2019 年间。     

Learning small gallery size for prediction of recognition performance on large populations

This paper addresses the estimation of a small gallery size that can generate the optimal error estimate and its confidence on a large population (relative to the size of the gallery) which is one of the fundamental problems encountered in performance prediction for object recognition. It uses a generalized two-dimensional prediction model that combines a hypergeometric probability distribution model with a binomial model and also considers the data distortion problem in large populations. Learning is incorporated in the prediction process in order to find the optimal small gallery size and to improve the prediction. The Chernoff and Chebychev inequalities are used as a guide to obtain the small gallery size. During the prediction, the expectation–maximization (EM) algorithm is used to learn the match score and the non-match score distributions that are represented as a mixture of Gaussians. The optimal size of the small gallery is learned by comparing it with the sizes obtained by the statistical approaches and at the same time the upper and lower bounds for the prediction on large populations are obtained. Results for the prediction are presented for the NIST-4 fingerprint database.     

Effective scheduling strategies for boosting performance on rule-based spam filtering frameworks

Despite the enormous importance of e-mail to current worldwide communication, the increase of spam deliveries has had a significant adverse effect for all its users. In order to adequately fight spam, both the filtering industry and scientific community have developed and deployed the fastest and most accurate filtering techniques. However, the increasing volume of new incoming messages needing classification together with the lack of adequate support for anti-spam services on the cloud, make filtering efficiency an absolute necessity. In this context, and given the extensive utilization and increasing significance of rule-based filtering frameworks for the anti-spam domain, this work studies and analyses the importance of both existing and novel scheduling strategies to make the most of currently available anti-spam filtering techniques. Results obtained from the experiments demonstrated that some scheduling alternatives resulted in time savings of up to 26% for filtering messages, while maintaining the same classification accuracy.     

石津灌区冬小麦种植肥料合理施用量试验研究

在河北省石津灌区节水改造示范区,分别对氮、磷、钾三种肥料不同用量进行分析对比试验,探讨了肥料施用量与冬小麦产量的关系,结果发现:在其他条件相对稳定情况下,要保证冬小麦产量在6 750kg/hm2以上,那么氮肥(折纯量)的合理施用量为210kg/hm2,磷肥(P2O5)为45⁷5kg/hm2,钾肥(K2O)为22.575kg/hm2,钾肥(K2O)为22.5⁴5kg/hm2。     

Adapting wave-front algorithms to efficiently utilize systems with deep communication hierarchies

Large-scale systems increasingly exhibit a differential between intra-chip and inter-chip communication performance especially in hybrid systems using accelerators. Processor-cores on the same socket are able to communicate at lower latencies, and with higher bandwidths, than cores on different sockets either within the same node or between nodes. A key challenge is to efficiently use this communication hierarchy and hence optimize performance. We consider here the class of applications that contains wave-front processing. In these applications data can only be processed after their upstream neighbors have been processed. Similar dependencies result between processors in which communication is required to pass boundary data downstream and whose cost is typically impacted by the slowest communication channel in use. In this work we develop a novel hierarchical wave-front approach that reduces the use of slower communications in the hierarchy but at the cost of additional steps in the parallel computation and higher use of on-chip communications. This tradeoff is explored using a performance model. An implementation using the reverse-acceleration programming model on the petascale Roadrunner system demonstrates a 27% performance improvement at full system-scale on a kernel application. The approach is generally applicable to large-scale multi-core and accelerated systems where a differential in communication performance exists.     

MLMIN: A multicore processor and parallel computer network topology for multicast

In future, multicore processors with hundreds of cores will collaborate on a single chip. Then, more advanced network-on-chip (NoC) topologies will be needed than today's shared busses for dual core processors. Multistage interconnection networks, which are already used in parallel computers, seem to be a promising alternative. In this paper, a new network topology is introduced that particularly applies to multicast traffic in multicore systems and parallel computers. Those multilayer multistage interconnection networks are described by defining the main parameters of such a topology. Performance and costs of the new architecture are determined and compared to other network topologies. Network traffic consisting of constant size packets and of varying size packets is investigated. It is shown that all kinds of multicast traffic particularly benefit from the new topology.     

Is conversion efficiency still relevant to qualify advanced multi‐junction solar cells?

For better conversion of sunlight into electricity, advanced architectures of multi‐junction (MJ) solar cells include increasing numbers of subcells. The Achilles' heel of these cells lies in their increased sensitivity to the spectral distribution of sunlight, which is likely to significantly alter their performance during real working operation. This study investigates the capacity of MJ solar cells comprising up to 10 subcells to accommodate a wide range of spectral characteristics of the incident radiation. A systematic study is performed, aimed at a realistic estimation of the energy output of MJ‐based concentrating photovoltaic systems at characteristic locations selected to represent a large range of climatic conditions. We show that optimal MJ architectures could have between 4 and 7 subcells. Beyond seven subcells, the slight gains in peak efficiency are likely outweighed by detrimental increases in dependence on local conditions and in annual yield variability. The relevance of considering either conversion efficiency or modeled energy output as the most appropriate indicator of the cell performance, when considering advanced architectures of MJ solar cells, is also discussed. Copyright © 2016 John Wiley & Sons, Ltd.     

大豆分离蛋白质制备工艺技术

对所设计的大豆分离蛋白质的制备工艺、所用设备、投资和经济效益进行了介绍，并对我国目前大豆分离蛋白的开发与利用作了客观的分析     

Study of inductively coupled Cl2/BCl3 plasma process for high etch rate selective etching of via-holes in GaAs

We have investigated the selective etching of 50 μm diameter via-holes for etch depth >200 μm using 30 μm thick photo resist mask in Inductively Coupled Plasma system with Cl₂/BCl₃ chemistry. Resultant etch rate/etch profiles are studied as a function of ICP process parameters and photo resist mask sidewall profile. Etch yield and aspect ratio variation with process pressure and substrate bias is also investigated at constant ICP power. The etch yield of ICP process increased with pressure due to reactant limited etch mechanism and reached a maximum of ∼19 for 200 μm depth at 50 mTorr pressure, 950 W coil power, 80 W substrate bias with an etch rate ∼4.9 μm/min. Final aspect ratio of etched holes is increased with pressure from 1.02 at 20 mTorr to 1.38 at 40 mTorr respectively for fixed etch time and then decreased to 1.24 at 50 mTorr pressure. The resultant final etch profile and undercut is found to have a strong dependence on the initial slope of photo resist mask sidewall angle and its selectivity in the pressure range of 20-50mTorr.