Commissariat &#xc0; L'&#xc9;nergie Atomique Et Aux &#xc9;nergies Alternatives

Karim Ben Chehida

Frédéric Blanc

Farhat Thabet

Paul-Antoine Arras

Frédérique Blanc

A method for managing a computation task on a functionally asymmetric multi-core processor includes a plurality of cores at least one of which comprises at least one hardware extension for executing specialized instructions, comprising the following steps: a) starting the execution of the computation task on a core of the processor; b) monitoring a parameter indicative of a quality of service of the computation task, and at least a number of specialized instructions loaded by the core; c) identifying instants splitting an application period of the computation task into a predetermined number of portions; d) computing costs or gains in quality of service and in energy consumption corresponding to different management options of the computation task; and e) making a management choice according to the costs or gains thus computed. Computer program product, processor and computer system for implementing such a method are also provided.

Method for managing computation tasks on a functionally asymmetric multi-core processor

The invention relates to a method for the on-line management of energy consumption for multiprocessor systems, the method executing at least one application according to a chart of tasks, wherein the method includes, for each application: a first phase for the off-line characterization of the variation of the potential rate of parallelism of execution of the tasks as a function of time, this characterization being based on the worst-case task execution times; and a second phase for the on-line detection and exploitation of the inactivity intervals and of the potential time excesses. A DPM technique makes it possible to determine the duration of the inactivity interval during which a processor may remain inactive according to the potential rate of parallelism characterized in the worst case.

Method for managing energy consumption for multiprocessor systems using an offline phase for characterization of a variation of a potential rate of parallelism for task scheduling and an online phase for detection and exploitation of potential inactivity intervals

Growing community of inventors

Suresnes, France

Karim Ben Chehida