Energy Efficiency and Schedulability with Resource Sharing over Multiple Processing Elements
Date Issued
2007
Date
2007
Author(s)
Chen, Ya-Shu
DOI
en-US
Abstract
The designs of embedded systems are complicated by various
limitations on resource supports, such as computing power, even
though their potential application domains grow exponentially. In
order to support applications with strong resource demands, many
embedded systems are now equipped with multiple processing elements
with particular functionalities, such as H.264 decoding. Since many
embedded systems are running on mobile devices, their designs must
explore tradeoffs among cost, performance, and energy efficiency.
This thesis investigates resource scheduling issues of embedded
systems with multiple processing elements with respect to the above
tradeoffs. We first explore scheduling anomaly problems in resource
sharing when there is more than one processing elements, referred to
as active resources. Rules are then proposed for anomaly prevention.
The work is later extended by considering the blocking behavior of
tasks that share multiple active resources. The cost optimization
problem is resolved under the schedulability requirements of task
executions. When energy efficiency is under considerations, we
propose real-time resource synchronization protocols with an
objective to minimize the number of processor frequency switchings
and the total energy consumption. Protocols are first proposed for
the access synchronization of passive resources, e.g., semaphores
and memory, that do not own any computing capability and are used by
tasks in a passive way. We then address synchronization issues over
active and passive resources.
Subjects
省電
資源管理
系統晶片
排程
嵌入式系統
同步協定
Energy Efficiency
Resource Management
SoC
Scheduling
Embedded System
Synchronization Protocol
SDGs
Type
thesis