GPRM: a high performance programming framework for manycore processors

Processors with large numbers of cores are becoming commonplace. In order to utilise the available resources in such systems, the programming paradigm has to move towards increased parallelism. However, increased parallelism does not necessarily lead to better performance. Parallel programming models have to provide not only flexible ways of defining parallel tasks, but also efficient methods to manage the created tasks. Moreover, in a general-purpose system, applications residing in the system compete for the shared resources. Thread and task scheduling in such a multiprogrammed multithreaded environment is a significant challenge. In this thesis, we introduce a new task-based parallel reduction model, called the Glasgow Parallel Reduction Machine (GPRM). Our main objective is to provide high performance while maintaining ease of programming. GPRM supports native parallelism; it provides a modular way of expressing parallel tasks and the communication patterns between them. Compiling a GPRM program results in an Intermediate Representation (IR) containing useful information about tasks, their dependencies, as well as the initial mapping information. This compile-time information helps reduce the overhead of runtime task scheduling and is key to high performance. Generally speaking, the granularity and the number of tasks are major factors in achieving high performance. These factors are even more important in the case of GPRM, as it is highly dependent on tasks, rather than threads. We use three basic benchmarks to provide a detailed comparison of GPRM with Intel OpenMP, Cilk Plus, and Threading Building Blocks (TBB) on the Intel Xeon Phi, and with GNU OpenMP on the Tilera TILEPro64. GPRM shows superior performance in almost all cases, only by controlling the number of tasks. GPRM also provides a low-overhead mechanism, called “Global Sharing”, which improves performance in multiprogramming situations. We use OpenMP, as the most popular model for shared-memory parallel programming as the main GPRM competitor for solving three well-known problems on both platforms: LU factorisation of Sparse Matrices, Image Convolution, and Linked List Processing. We focus on proposing solutions that best fit into the GPRM’s model of execution. GPRM outperforms OpenMP in all cases on the TILEPro64. On the Xeon Phi, our solution for the LU Factorisation results in notable performance improvement for sparse matrices with large numbers of small blocks. We investigate the overhead of GPRM’s task creation and distribution for very short computations using the Image Convolution benchmark. We show that this overhead can be mitigated by combining smaller tasks into larger ones. As a result, GPRM can outperform OpenMP for convolving large 2D matrices on the Xeon Phi. Finally, we demonstrate that our parallel worksharing construct provides an efficient solution for Linked List processing and performs better than OpenMP implementations on the Xeon Phi. The results are very promising, as they verify that our parallel programming framework for manycore processors is flexible and scalable, and can provide high performance without sacrificing productivity.

A future dynamically reconfigurable automotive software system

Embedded software systems in vehicles are of rapidly increasing commercial importance for the automotive industry. Current systems employ a static run-time environment; due to the difficulty and cost involved in the development of dynamic systems in a high-integrity embedded control context. A dynamic system, referring to the system configuration, would greatly increase the flexibility of the offered functionality and enable customised software configuration for individual vehicles, adding customer value through plug-and-play capability, and increased quality due to its inherent ability to adjust to changes in hardware and software. We envisage an automotive system containing a variety of components, from a multitude of organizations, not necessarily known at development time. The system dynamically adapts its configuration to suit the run-time system constraints. This paper presents our vision for future automotive control systems that will be regarded in an EU research project, referred to as DySCAS (Dynamically Self-Configuring Automotive Systems). We propose a self-configuring vehicular control system architecture, with capabilities that include automatic discovery and inclusion of new devices, self-optimisation to best-use the processing, storage and communication resources available, self-diagnostics and ultimately self-healing. Such an architecture has benefits extending to reduced development and maintenance costs, improved passenger safety and comfort, and flexible owner customisation. Specifically, this paper addresses the following issues: The state of the art of embedded software systems in vehicles, emphasising the current limitations arising from fixed run-time configurations; and the benefits and challenges of dynamic configuration, giving rise to opportunities for self-healing, self-optimisation, and the automatic inclusion of users’ Consumer Electronic (CE) devices. Our proposal for a dynamically reconfigurable automotive software system platform is outlined and a typical use-case is presented as an example to exemplify the benefits of the envisioned dynamic capabilities.

Integrated visual perception architecture for robotic clothes perception and manipulation

This thesis proposes a generic visual perception architecture for robotic clothes perception and manipulation. This proposed architecture is fully integrated with a stereo vision system and a dual-arm robot and is able to perform a number of autonomous laundering tasks. Clothes perception and manipulation is a novel research topic in robotics and has experienced rapid development in recent years. Compared to the task of perceiving and manipulating rigid objects, clothes perception and manipulation poses a greater challenge. This can be attributed to two reasons: firstly, deformable clothing requires precise (high-acuity) visual perception and dexterous manipulation; secondly, as clothing approximates a non-rigid 2-manifold in 3-space, that can adopt a quasi-infinite configuration space, the potential variability in the appearance of clothing items makes them difficult to understand, identify uniquely, and interact with by machine. From an applications perspective, and as part of EU CloPeMa project, the integrated visual perception architecture refines a pre-existing clothing manipulation pipeline by completing pre-wash clothes (category) sorting (using single-shot or interactive perception for garment categorisation and manipulation) and post-wash dual-arm flattening. To the best of the author’s knowledge, as investigated in this thesis, the autonomous clothing perception and manipulation solutions presented here were first proposed and reported by the author. All of the reported robot demonstrations in this work follow a perception-manipulation method- ology where visual and tactile feedback (in the form of surface wrinkledness captured by the high accuracy depth sensor i.e. CloPeMa stereo head or the predictive confidence modelled by Gaussian Processing) serve as the halting criteria in the flattening and sorting tasks, respectively. From scientific perspective, the proposed visual perception architecture addresses the above challenges by parsing and grouping 3D clothing configurations hierarchically from low-level curvatures, through mid-level surface shape representations (providing topological descriptions and 3D texture representations), to high-level semantic structures and statistical descriptions. A range of visual features such as Shape Index, Surface Topologies Analysis and Local Binary Patterns have been adapted within this work to parse clothing surfaces and textures and several novel features have been devised, including B-Spline Patches with Locality-Constrained Linear coding, and Topology Spatial Distance to describe and quantify generic landmarks (wrinkles and folds). The essence of this proposed architecture comprises 3D generic surface parsing and interpretation, which is critical to underpinning a number of laundering tasks and has the potential to be extended to other rigid and non-rigid object perception and manipulation tasks. The experimental results presented in this thesis demonstrate that: firstly, the proposed grasp- ing approach achieves on-average 84.7% accuracy; secondly, the proposed flattening approach is able to flatten towels, t-shirts and pants (shorts) within 9 iterations on-average; thirdly, the proposed clothes recognition pipeline can recognise clothes categories from highly wrinkled configurations and advances the state-of-the-art by 36% in terms of classification accuracy, achieving an 83.2% true-positive classification rate when discriminating between five categories of clothes; finally the Gaussian Process based interactive perception approach exhibits a substantial improvement over single-shot perception. Accordingly, this thesis has advanced the state-of-the-art of robot clothes perception and manipulation.

Airborne vision-based collision-detection system

Machine vision represents a particularly attractive solution for sensing and detecting potential collision-course targets due to the relatively low cost, size, weight, and power requirements of vision sensors (as opposed to radar and TCAS). This paper describes the development and evaluation of a real-time vision-based collision detection system suitable for fixed-wing aerial robotics. Using two fixed-wing UAVs to recreate various collision-course scenarios, we were able to capture highly realistic vision (from an onboard camera perspective) of the moments leading up to a collision. This type of image data is extremely scarce and was invaluable in evaluating the detection performance of two candidate target detection approaches. Based on the collected data, our detection approaches were able to detect targets at distances ranging from 400m to about 900m. These distances (with some assumptions about closing speeds and aircraft trajectories) translate to an advanced warning of between 8-10 seconds ahead of impact, which approaches the 12.5 second response time recommended for human pilots. We overcame the challenge of achieving real-time computational speeds by exploiting the parallel processing architectures of graphics processing units found on commercially-off-the-shelf graphics devices. Our chosen GPU device suitable for integration onto UAV platforms can be expected to handle real-time processing of 1024 by 768 pixel image frames at a rate of approximately 30Hz. Flight trials using manned Cessna aircraft where all processing is performed onboard will be conducted in the near future, followed by further experiments with fully autonomous UAV platforms.

Parallel linear system solution and its application to railway power network simulation

The Streaming SIMD extension (SSE) is a special feature embedded in the Intel Pentium III and IV classes of microprocessors. It enables the execution of SIMD type operations to exploit data parallelism. This article presents improving computation performance of a railway network simulator by means of SSE. Voltage and current at various points of the supply system to an electrified railway line are crucial for design, daily operation and planning. With computer simulation, their time-variations can be attained by solving a matrix equation, whose size mainly depends upon the number of trains present in the system. A large coefficient matrix, as a result of congested railway line, inevitably leads to heavier computational demand and hence jeopardizes the simulation speed. With the special architectural features of the latest processors on PC platforms, significant speed-up in computations can be achieved.

A parallel computation of power system equations

Streaming SIMD Extensions (SSE) is a unique feature embedded in the Pentium III and P4 classes of microprocessors. By fully exploiting SSE, parallel algorithms can be implemented on a standard personal computer and a theoretical speedup of four can be achieved. In this paper, we demonstrate the implementation of a parallel LU matrix decomposition algorithm for solving power systems network equations with SSE and discuss advantages and disadvantages of this approach.

Real-time simulation for power systems based on parallel computing - an empirical study

Abstract Computer simulation is a versatile and commonly used tool for the design and evaluation of systems with different degrees of complexity. Power distribution systems and electric railway network are areas for which computer simulations are being heavily applied. A dominant factor in evaluating the performance of a software simulator is its processing time, especially in the cases of real-time simulation. Parallel processing provides a viable mean to reduce the computing time and is therefore suitable for building real-time simulators. In this paper, we present different issues related to solving the power distribution system with parallel computing based on a multiple-CPU server and we will concentrate, in particular, on the speedup performance of such an approach.

MAVwork : a framework for unified interfacing between micro aerial vehicles and visual controllers

Debugging control software for Micro Aerial Vehicles (MAV) can be risky out of the simulator, especially with professional drones that might harm people around or result in a high bill after a crash. We have designed a framework that enables a software application to communicate with multiple MAVs from a single unified interface. In this way, visual controllers can be first tested on a low-cost harmless MAV and, after safety is guaranteed, they can be moved to the production MAV at no additional cost. The framework is based on a distributed architecture over a network. This allows multiple configurations, like drone swarms or parallel processing of drones' video streams. Live tests have been performed and the results show comparatively low additional communication delays, while adding new functionalities and flexibility. This implementation is open-source and can be downloaded from github.com/uavster/mavwork

Parallel ABM for electricity distribution grids : a case study

This paper introduces a parallel implementation of an agent-based model applied to electricity distribution grids. A fine-grained shared memory parallel implementation is presented, detailing the way the agents are grouped and executed on a multi-threaded machine, as well as the way the model is built (in a composable manner) which is an aid to the parallelisation. Current results show a medium level speedup of 2.6, but improvements are expected by incor-porating newer distributed or parallel ABM schedulers into this implementa-tion. While domain-specific, this parallel algorithm can be applied to similarly structured ABMs (directed acyclic graphs).

Visual search and eye movements: Studies of perceptual span

In visual search one tries to find the currently relevant item among other, irrelevant items. In the present study, visual search performance for complex objects (characters, faces, computer icons and words) was investigated, and the contribution of different stimulus properties, such as luminance contrast between characters and background, set size, stimulus size, colour contrast, spatial frequency, and stimulus layout were investigated. Subjects were required to search for a target object among distracter objects in two-dimensional stimulus arrays. The outcome measure was threshold search time, that is, the presentation duration of the stimulus array required by the subject to find the target with a certain probability. It reflects the time used for visual processing separated from the time used for decision making and manual reactions. The duration of stimulus presentation was controlled by an adaptive staircase method. The number and duration of eye fixations, saccade amplitude, and perceptual span, i.e., the number of items that can be processed during a single fixation, were measured. It was found that search performance was correlated with the number of fixations needed to find the target. Search time and the number of fixations increased with increasing stimulus set size. On the other hand, several complex objects could be processed during a single fixation, i.e., within the perceptual span. Search time and the number of fixations depended on object type as well as luminance contrast. The size of the perceptual span was smaller for more complex objects, and decreased with decreasing luminance contrast within object type, especially for very low contrasts. In addition, the size and shape of perceptual span explained the changes in search performance for different stimulus layouts in word search. Perceptual span was scale invariant for a 16-fold range of stimulus sizes, i.e., the number of items processed during a single fixation was independent of retinal stimulus size or viewing distance. It is suggested that saccadic visual search consists of both serial (eye movements) and parallel (processing within perceptual span) components, and that the size of the perceptual span may explain the effectiveness of saccadic search in different stimulus conditions. Further, low-level visual factors, such as the anatomical structure of the retina, peripheral stimulus visibility and resolution requirements for the identification of different object types are proposed to constrain the size of the perceptual span, and thus, limit visual search performance. Similar methods were used in a clinical study to characterise the visual search performance and eye movements of neurological patients with chronic solvent-induced encephalopathy (CSE). In addition, the data about the effects of different stimulus properties on visual search in normal subjects were presented as simple practical guidelines, so that the limits of human visual perception could be taken into account in the design of user interfaces.

Small Log Processing Resource Specification

Sawing studies of mixed species and age plantation hardwoods representing potential early resource flows available to plantation sawlog processing industry. Provide information needed to underpin a business case to invest in small log sawing infrastucture.

Modelling and analysis of the variance in parallelism in parallel computations

In this paper, we introduce an analytical technique based on queueing networks and Petri nets for making a performance analysis of dataflow computations when executed on the Manchester machine. This technique is also applicable for the analysis of parallel computations on multiprocessors. We characterize the parallelism in dataflow computations through a four-parameter characterization, namely, the minimum parallelism, the maximum parallelism, the average parallelism and the variance in parallelism. We observe through detailed investigation of our analytical models that the average parallelism is a good characterization of the dataflow computations only as long as the variance in parallelism is small. However, significant difference in performance measures will result when the variance in parallelism is comparable to or higher than the average parallelism.

Performability Analysis of Fork-Join Queueing Systems

Fork-join queueing systems offer a natural modelling paradigm for parallel processing systems and for assembly operations in automated manufacturing. The analysis of fork-join queueing systems has been an important subject of research in recent years. Existing analysis methodologies-both exact and approximate-assume that the servers are failure-free. In this study, we consider fork-join queueing systems in the presence of server failures and compute the cumulative distribution of performability with respect to the response time of such systems. For this, we employ a computational methodology that uses a recent technique based on randomization. We compare the performability of three different fork-join queueing models proposed in the literature: the distributed model, the centralized splitting model, and the split-merge model. The numerical results show that the centralized splitting model offers the highest levels of performability, followed by the distributed splitting and split-merge models.