实时调度算法在多核平台上的实现1
需积分: 0 95 浏览量
2022-08-03
11:52:46
上传
评论
收藏 307KB PDF 举报
Implementation of Real-Time Scheduling Algorithm on Multi-Core Platform
Xuemei Zhang
School of Computer Science and Engineering
Xi'an Technological University
Xi'an, 710021, China
E-mail: 875781076@qq.com
Jinghui Li
School of Computer Science and Engineering
Xi'an Technological University
Xi'an, 710021, China
E-mail: 715319152@qq.com
Shujuan Huang
School of Computer Science and Engineering
Xi'an Technological University
Xi'an, 710021, China
E-mail: 349242386@qq.com
Abstract—With the increasing demand for computing power
in embedded systems, multi-core processor architectures have
become increasingly common
[1]
. This paper studies and
analyzes various real-time scheduling algorithms and test
platform LITMUSRT under the embedded multi-core
platform, and proposes a multi-core scheduling method for
real-time tasks with dependencies. The static real-time
scheduling algorithm RM (Rate-Monotonic), the algorithm
and the proposed multi-core scheduling method for real-time
tasks with dependencies are implemented in LITMUSRT. Test
cases show that the proposed multi-core scheduling method
with dependencies can be implemented in the actual Linux
environment and can meet the real-time requirements of tasks
in real-time scheduling.
Keywords-Component; LITMUSRT; Multi-Core Systems;
Rate Monotic Algorithm; Real-Time Scheduling Algorithm
I. INTRODUCTION
In 1973, Liu and Layland proposed a static priority
scheduling algorithm-Rate Monotonie (RM) scheduling
algorithm
[2]
, but the algorithm was based on an ideal
scheduling model based on a series of ideal assumptions,
and In application, consider the influence of various factors.
In addition, the real-time scheduling algorithm that has been
implemented on the LITMUS
RT
platform does not consider
the dependencies between tasks, nor does it consider the
operation of safety-critical tasks
[3]
.
This paper implements the RM scheduling algorithm and
the scheduling algorithm of real-time periodic tasks with
dependencies by modifying the code of the LITMUS
RT
platform, and implements the operation of real-time periodic
tasks with safety-critical factors on the platform, which
solves the research field of multi-core real-time scheduling
Some practical problems.
II. LITMUSRT
PLATFORM AND SCHEDULING
ALGORITHM
A. LITMUSRT platform
LITMUS
RT
platform was researched and developed by
Professor James H. Derson and his multi-verification time
scheduling team in the United States. The original design of
LITMUS
RT
platform is to compare the performance,
constraint and application scope of different types of
multi-core scheduling algorithms in a unified real-time
system. This platform can perform scheduling test for tasks
with the same load for different scheduling algorithms, and
then compare and analyze some index data. It is based on
Linux kernel. LITMUS
RT
has modified Linux kernel,
supported random task model, modular scheduling algorithm
plug-in, and provided a new synchronization algorithm. The
current version is version 2011.1, supporting X86, ARM and
other architectures.
The platform mainly consists of five parts: the
underlying data structure, the scheduling plug-in, the
synchronous processing mechanism, the scheduling trace,
and the system call.
The underlying data structure and scheduling tracking
are part of LITMUS
RT
support. The underlying data
structure is responsible for the implementation of the data
structure of the scheduling plug-in and synchronous
processing mechanism, and the scheduling tracking part uses
Linux counters to record the scheduling information.
The scheduling plug-in and synchronous processing
mechanism are part of LITMUS
RT
processing. The
scheduling plug-in includes the implementation of various
scheduling algorithms and is integrated into the plug-in form
66
2020 International Conference on Computer Network, Electronic and Automation (ICCNEA)
978-1-7281-7083-1/20/$31.00 ©2020 IEEE
DOI 10.1109/ICCNEA50255.2020.00023
Authorized licensed use limited to: University of Shanghai For Science and Technology. Downloaded on December 28,2020 at 04:17:59 UTC from IEEE Xplore. Restrictions apply.
评论0