Dynamic Resource Allocation in Multimedia Applications
by: Martijn van den Heuvel

SAN I am a researcher at the System Architecture and Networks (SAN) group at the Eindhoven University of Technology (TU/e) This page describes my master project, finished as of July 2009.

Contact:

email: m.m.h.p.v.d.heuvel (at) tue.nl
web: [Main page]

Introduction

Scalable video algorithms using novel priority processing can guarantee real-time performance on programmable platforms even with limited resources. According to the priority processing principle, as developed at BTU, scalable video algorithms follows a priority order. Hence, important image parts are processed first and less important parts are subsequently processed in a decreasing order of importance. After creation of an initial output by a basic function, processing can be terminated at an arbitrary moment in time, yielding the best output for given resources. This characteristic of priority processing allowing preliminary termination, requires appropriate mechanisms.
HTAS

Dynamic Resource Allocation

Dynamic resource allocation is required to maximize the overall output quality of competing, independent priority-processing algorithms that are executed on a shared platform. To distribute the available processor time, among these competing algorithms, a scheduling policy has been developed. This policy aims at maximizing the total progress of the algorithms on a frame-basis. Allocation of processor resources to priority processing algorithms can be established by means of different mechanisms.
The allocation of the processor requires monitoring mechanisms to provide the control component with information about the progress of the individual algorithms relative to the amount of consumed time. Based on this information, the control component decides which algorithm to allocate the processor for a fixed time slice.
The priority processing applications are prototyped in a Matlab/Simulink environment and executed on a multi-core platform under Microsoft Windows XP. This platform is used to compare the performance of different implementations of the above described mechanisms.

Project Goals

Persons Involved

Deliverables

External Links