Conference Paper

A hybrid memory organization to enhance task migration and dynamic task allocation in NoC-based MPSoCs.

DOI: 10.1145/1284480.1284557 Conference: Proceedings of the 20th Annual Symposium on Integrated Circuits and Systems Design, SBCCI 2007, Copacabana, Rio de Janeiro, Brazil, September 3-6, 2007
Source: DBLP

ABSTRACT Regarding MPSoCs, dynamic task allocation and task migration are still open research areas and, for both of them, there is no clear memory organization winner. While bus-connected systems commonly use a shared memory paradigm, NoC-based ones easily allow the exploration of distributed memory alternatives. This paper proposes a hybrid memory organization for NoC-based systems as the way to minimize the energy spent during the code transfer when task migration or dynamic task allocation needs to be performed. In our hybrid approach, the code can be transferred from the node where the task was originally running or from a memory positioned at the center of the system. The choice between the two options is done at runtime in a very intuitive way, based on the distance between the nodes involved on the transfer. Results are very encouraging and indicate that the proposed hybrid organization reduces the code transfer energy by 24% and 10% on average, as compared to global- and distributed-only memory organizations, respectively.

Download full-text

Full-text

Available from: Eduardo Wenzel Brião, Mar 25, 2014
0 Followers
 · 
111 Views
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In deep submicron circuits, high temperatures have created critical issues in reliability, timing, performance, coolings costs and leakage power. Task migration techniques have been proposed to manage efficiently the thermal distribution in multi-processor systems but at the cost of important performance penalties. While traditional techniques have focused on reducing the average temperature of the chip, they have not considered the effect that temperature gradients have in system reliability. In this work, we explore the benefits of thermal-aware task migration techniques for embedded multi-processor systems. We show the implementation issues of task migration policies on next generation architectural template of distributed memory multicore systems and we discuss the programmer’s implications. Built on top of this programming model, we propose several policies that are able to reduce the average temperature of the chip and the thermal gradients with a negligible performance overhead. With our techniques, hot spots and temperature gradients are decreased up to 30% with respect to state-of-the-art thermal management approaches.
    12/2010: pages 83-115;
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The continuing employment of distributed embedded systems in various safety critical applications requires the demonstration of a high degrees of dependability. Fault-tolerance is the primary means to achieve dependability at runtime and has been an active research area for decades. In addition to traditional hardware and software redundancy, hybrid and reconfiguration-based approaches to fault-tolerance are evolving. Software simulation techniques are a common starting point for evaluating reconfigurable computing approaches. Simulations, however, typically make an amount of unrealistic assumptions. In this paper, we present a test-bed for the testing and evaluation of reconfiguration-based fault-tolerance approaches. The test-bed uses 16-bit microcontrollers as processing elements. A real-time operating system is used for task scheduling and inter-task communication. The test-beds’ processing elements communicate through a CAN network. Monitoring, user interface, environmental interaction, and fault-injection abilities are provided through a graphical user interface.
    The International Conference on Information and Communications Systems (ICICS2009), Amman, Jordan; 12/2009
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: En los circuitos tecnológicos diseñados con tecnologías de fabricación por debajo de la micra, las altas temperaturas provocan fallos críticos en la fiabilidad, el temporizado, los costes de refrigeración y la potencia de pérdidas. Hasta el momento se han propuesto varias técnicas de migración de tareas para manejar eficientemente la distribución térmica en los sistemas multiprocesador, pero con un alto coste en cuanto a la eficiencia del sistema. Aunque las técnicas tradicionales se han centrado en reducir la temperatura media del chip, no han considerado los efectos que los gradientes térmicos tienen en la fiabilidad del sistema. En este trabajo, se exploran los beneficios de las técnicas de migración de tareas basadas en la temperatura en sistemas empotrados multiprocesador. En particular se proponen algunas políticas que son capaces de reducir la temperatura media del chip y los gradientes térmicos con un impacto en el rendimiento prácticamente despreciable. Con nuestras técnicas, la aparición de puntos calientes (hot spots) y variaciones térmicas se ven drásticamente reducidos con respecto a otras propuestas, con lo que la fiabilidad del sistema se mejora significativamente cuando la comparamos con las técnicas de migración de tareas tradicionales [ABSTRACT] In deep submicron circuits, high temperatures have created critical issues in reliability, timing, performance, cooling costs and leakage power. Task migration techniques have been proposed to manage efficiently the thermal distribution in multi-processor systems but at the cost of important performance penalties. While traditional techniques have focused on reducing the average temperature of the chip, they have not considered the effect that temperature gradients have in system reliability. In this work, we explore the benefits of thermal-aware task migration techniques for embedded multi-processor systems. We propose several policies that are able to reduce the average temperature of the chip and the thermal gradients with a negligible performance overhead. With our techniques, hot spots and temperature variations are decreased, and the reliability of the system is significantly improved when compared to traditional task migration techniques.