Fast finite difference Poisson solvers on heterogeneous architectures

In this paper we propose and evaluate a set of new strategies for the solution of three dimensional separable elliptic problems on CPU–GPU platforms. The numerical solution of the system of linear equations arising when discretizing those operators often represents the most time consuming part of larger simulation codes tackling a variety of physical situations. Incompressible fluid flows, electromagnetic problems, heat transfer and solid mechanic simulations are just a few examples of application areas that require efficient solution strategies for this class of problems. GPU computing has emerged as an attractive alternative to conventional CPUs for many scientific applications. High speedups over CPU implementations have been reported and this trend is expected to continue in the future with improved programming support and tighter CPU–GPU integration. These speedups by no means imply that CPU performance is no longer critical. The conventional CPU-control–GPU-compute pattern used in many applications wastes much of CPU’s computational power. Our proposed parallel implementation of a classical cyclic reduction algorithm to tackle the large linear systems arising from the discretized form of the elliptic problem at hand, schedules computing on both the GPU and the CPUs in a cooperative way. The experimental result demonstrates the effectiveness of this approach.     

Fast computation of computer-generated hologram using Xeon Phi coprocessor

We report fast computation of computer-generated holograms (CGHs) using Xeon Phi coprocessors, which have massively x86-based processors on one chip, recently released by Intel. CGHs can generate arbitrary light wavefronts, and therefore, are promising technology for many applications: for example, three-dimensional displays, diffractive optical elements, and the generation of arbitrary beams. CGHs incur enormous computational cost. In this paper, we describe the implementations of several CGH generating algorithms on the Xeon Phi, and the comparisons in terms of the performance and the ease of programming between the Xeon Phi, a CPU and graphics processing unit (GPU).     

Mobile augmented reality based context-aware library management system

Mobile augmented reality has gained popularity in recent years due to the technological advances of smartphones and other mobile devices. One particular area in which mobile augmented reality is being used is library management. However, current mobile augmented reality solutions in this domain are lacking in context-awareness. It has been suggested in the literature that agent programming may be suitable at overcoming this problem, but little research has been conducted using modern mobile augmented reality applications with agents. This paper aims to bridge this gap through the development of an agent-based, mobile augmented reality prototype, titled Libagent. Libagent was subjected to five experiments to determine its suitability, efficiency, and accuracy for library management. The results of these experiments indicate that agent-based mobile augmented reality is a promising tool for context-aware library management.     

Detecting stable locality-aware predicates

In a large-scale locality-driven network such as in modular robotics and wireless sensor networks, knowing the state of a local area is sometimes necessary due to either interactions being local and driven by neighborhood proximity or the users being interested in the state of a certain region. We define locality-aware predicates (LAP) that aim at detecting a predicate within a specified area. We model the area of interest as the set of processes that are within a breadth-first search tree (BFST) of height k$k$ rooted at the initiator process. Although a locality-aware predicate specifies a predicate only within a local area, observing the area consistently requires considering the entire system in a consistent manner. This raises the challenge of making the complexities of the corresponding predicate detection algorithms scale-free, i.e., independent of the size of the system. Since all existing algorithms for getting a consistent view of the system require either a global snapshot of the entire system or vector clocks of the size of the system, a new solution is needed. We focus on stable LAP, which are those LAP that remain true once they become true. We propose a scale-free algorithm to detect stable LAP within a k$k$-height BFST. Our algorithm can detect both stable conjunctive LAP and stable relational LAP. In the process of designing our algorithm, we also propose the first distributed algorithm for building a BFST within an area of interest in a graph, and the first distributed algorithm for recording a consistent sub-cut within the area of interest. This paper demonstrates that LAPs are a natural fit for detecting distributed properties in large-scale distributed systems, and stable LAPs can be practically detected at low cost.     

Kokkos: Enabling manycore performance portability through polymorphic memory access patterns

The manycore revolution can be characterized by increasing thread counts, decreasing memory per thread, and diversity of continually evolving manycore architectures. High performance computing (HPC) applications and libraries must exploit increasingly finer levels of parallelism within their codes to sustain scalability on these devices. A major obstacle to performance portability is the diverse and conflicting set of constraints on memory access patterns across devices. Contemporary portable programming models address manycore parallelism (e.g., OpenMP, OpenACC, OpenCL) but fail to address memory access patterns. The Kokkos C++ library enables applications and domain libraries to achieve performance portability on diverse manycore architectures by unifying abstractions for both fine-grain data parallelism and memory access patterns. In this paper we describe Kokkos’ abstractions, summarize its application programmer interface (API), present performance results for unit-test kernels and mini-applications, and outline an incremental strategy for migrating legacy C++ codes to Kokkos. The Kokkos library is under active research and development to incorporate capabilities from new generations of manycore architectures, and to address a growing list of applications and domain libraries.     

Proactive scheduling in distributed computing—A reinforcement learning approach

In distributed computing such as grid computing, online users submit their tasks anytime and anywhere to dynamic resources. Task arrival and execution processes are stochastic. How to adapt to the consequent uncertainties, as well as scheduling overhead and response time, are the main concern in dynamic scheduling. Based on the decision theory, scheduling is formulated as a Markov decision process (MDP). To address this problem, an approach from machine learning is used to learn task arrival and execution patterns online. The proposed algorithm can automatically acquire such knowledge without any aforehand modeling, and proactively allocate tasks on account of the forthcoming tasks and their execution dynamics. Under comparison with four classic algorithms such as Min–Min, Min–Max, Suffrage, and ECT, the proposed algorithm has much less scheduling overhead. The experiments over both synthetic and practical environments reveal that the proposed algorithm outperforms other algorithms in terms of the average response time. The smaller variance of average response time further validates the robustness of our algorithm.     

Detecting similarities in virtual machine behavior for cloud monitoring using smoothed histograms

The growing size and complexity of cloud systems determine scalability issues for resource monitoring and management. While most existing solutions consider each Virtual Machine (VM) as a black box with independent characteristics, we embrace a new perspective where VMs with similar behaviors in terms of resource usage are clustered together. We argue that this new approach has the potential to address scalability issues in cloud monitoring and management. In this paper, we propose a technique to cluster VMs starting from the usage of multiple resources, assuming no knowledge of the services executed on them. This innovative technique models VMs behavior exploiting the probability histogram of their resources usage, and performs smoothing-based noise reduction and selection of the most relevant information to consider for the clustering process. Through extensive evaluation, we show that our proposal achieves high and stable performance in terms of automatic VM clustering, and can reduce the monitoring requirements of cloud systems.     

A uniform approach for programming distributed heterogeneous computing systems

Large-scale compute clusters of heterogeneous nodes equipped with multi-core CPUs and GPUs are getting increasingly popular in the scientific community. However, such systems require a combination of different programming paradigms making application development very challenging.In this article we introduce libWater, a library-based extension of the OpenCL programming model that simplifies the development of heterogeneous distributed applications. libWater consists of a simple interface, which is a transparent abstraction of the underlying distributed architecture, offering advanced features such as inter-context and inter-node device synchronization. It provides a runtime system which tracks dependency information enforced by event synchronization to dynamically build a DAG of commands, on which we automatically apply two optimizations: collective communication pattern detection and device-host-device copy removal.We assess libWater’s performance in three compute clusters available from the Vienna Scientific Cluster, the Barcelona Supercomputing Center and the University of Innsbruck, demonstrating improved performance and scaling with different test applications and configurations.     

Self-scaling cooperative discovery of service compositions in unstructured P2P networks

We propose an efficient technique for improving the performance of automatic and cooperative compositions in unstructured Peer-to-Peer networks during service discovery. The technique exploits a probabilistic forwarding algorithm that uses different sources of knowledge, such as network density and service grouping, to reduce the amount of messages exchanged in the network. The technique, analysed in several network configurations by using a simulator to observe resolution time, recall and message overhead, presents good performance especially in dense and large-scale service networks.     

Effects of portable computing devices on posture,muscle activation levels and efficiency

Very little research exists on ergonomic exposures when using portable computing devices. This study quantified muscle activity (forearm and neck), posture (wrist, forearm and neck), and performance (gross typing speed and error rates) differences across three portable computing devices (laptop, netbook, and slate computer) and two work settings (desk and computer) during data entry tasks. Twelve participants completed test sessions on a single computer using a test–rest–test protocol (30 min of work at one work setting, 15 min of rest, 30 min of work at the other work setting). The slate computer resulted in significantly more non-neutral wrist, elbow and neck postures, particularly when working on the sofa. Performance on the slate computer was four times less than that of the other computers, though lower muscle activity levels were also found. Potential or injury or illness may be elevated when working on smaller, portable computers in non-traditional work settings.