Skip to content

shambakey1/sched_test_app_sh

Repository files navigation

Readme for Sched_Test_App
Written by Matthew Dellinger <[email protected]>, January 21, 2010

Copyright 2008-2010 by Matthew Dellinger.
Redistribution of this file is permitted under the GNU Public License.

Table of Contents
~~~~~~~~~~~~~~~~~
1.    General Information
2.    Known Bugs
3.    Interface Options
3.1.  Scheduling Algorithm
3.2.  Locking
3.3.  Run Properties
3.4.  Taskset
4.    Setup
5.    Coding Style
6.    Update History


1. General Information
~~~~~~~~~~~~~~~~~~~~~
Sched_Test_App is a scheduling algorithm test application for use with the 
vt_rt_sched scheduling framework. It generates a series of periodic threads
which are given real-time linux priorities and records statistics about each
thread.

2. Known Bugs
~~~~~~~~~~~~~~~~~~~~~
None

3. Interface Options
~~~~~~~~~~~~~~~~~~~~~
The interface options are divided up into several groups.

The main panel displays all output from the program, including error messages. It
can be cleared with the clear button.

The panel on the left contains all the output formatting controls.

The panels on the right contain all the necessary setting for a run, which can be
started using the run button in the lower right hand corner.

The interface via the console offers all the same options, so a separate discussion
of that is not included. Running the compiled binary with any arguments will
print a basic help file to the console.

3.1 Scheduling Algorithm
~~~~~~~~~~~~~~~~~~~~~
This panel contains the interface for selecting a scheduling algorithm. The
algorithm is selectable via the top combobox.

The options below the combobox are flags sent to the kernel scheduler which 
affect the behaviour of a given algorithm. If you are using this application in 
ECE 4550 or ECE 5550, you do not need to worry about these.

Abort tasks on idle will run aborted tasks only when there are no other real-time
tasks to be run. This incurs a higher memory and scheduling overhead, but decreases
the amount of time spent executing non-utility accruing tasks and minimizes
context switches.

Use HUA handlers will create HUA abort handlers for any task that has an abort
handler utility defined in the task set description file. The handler's deadline
is always infinite. If no handler utility is present for a task, it will be aborted
as normal by the kernel.

Use PI enables priority inheritance if the algorithm supports it. This option does
nothing if the algorithm does not support PI or if the algorithm has another
resource management scheme (such as DASA).

Deadlock prevention prevents deadlocks from occuring in resource algorithms like 
PI. This generally incurs a higher scheduling overhead. This does nothing if the
algorithm already has a deadlock prevention mechanism (like DASA).

3.2 Locking
~~~~~~~~~~~~~~~~~~~~~
Resources will only be used if locking is enabled. When it is enabled, all 
locks listed in the task set descriptor file will be locked, and the task will
then burn the processor for the critical section length. All locks will then
be released, and the processor will be burned for the remainder of its execution
time.

3.2 Run Properties
~~~~~~~~~~~~~~~~~~~~~
Execution time specifies the length of the run in seconds. This should always be
a multiple of the task set period.

CPU usage specifies the amount of processor usage to use for the run. Modifying
this changes the execution time of the task.

End usage and Iteration: When doing a batch run, these specify the end processor
usage and the step size.

Excel or Gnuplot output format the output appropriately for input to those
programs.

Verbose output lists every action of every task, rather than just the final
results. This is currently unstable due to the lack of thread-safety in streams.

Log to file will log the output to an automatically named file at the end of
every run. When used in conjunction with a batch run, this will output between
every iteration.

3.3 Taskset
~~~~~~~~~~~~~~~~~~~~~
The taskset must be loaded from a file before any runs can be made. The number
of tasks, the processor usage of the taskset (i.e. the actual processor usage
if processor usage is specified as 100), the task set's period, and the number
of locks are all displayed. A sample taskset with five tasks and five locks is
shown below.

#the number of locks
L	5
#The tasks
#	Period (us)	Usage (us)	Utility		Locks	HUA Utility
T	500000		150000		17		2,3	2
T	1000000		230000		4		1,4	2
T	1500000		410000		24		0,2	2
T	3000000		300000		39		4,0	2
T	5000000		500000		18		1,3	2

4. Setup
~~~~~~~~~~~~~~~~~~~~~
To compile Sched_Test_App, you will need the Qt libraries installed. 
On Ubuntu this can be done with

sudo apt-get install qt4-dev-tools
sudo apt-get install qt4-qmake

You can then open the included kdevelop3 project and build the project, or build
with qmake followed by make.

If you compile it another way, note that the -O2 flag will result in the burn
loop being optimized out.

To install, run the included install script.

You will have to sudo to run, as the call to set the scheduler requires it. Also,
if you debug the application in kdevelop, you will notice that the results are
incorrect, even if you sudo kdevelop.

6. Code Formatting
~~~~~~~~~~~~~~~~~~~~~
All code in this application follows Google's C++ Style Guide
http://google-styleguide.googlecode.com/svn/trunk/cppstyleguide.xml

While it is by no means perfect, it provides a reasonable standard. A good
discussion of its merits and flaws can be found here.

http://groups.google.com/group/comp.lang.c++.moderated/browse_thread/thread/d7b0a5c663467471/ec4c733dd30b34e3

6. Update History
~~~~~~~~~~~~~~~~~~~~~
Written by Matthew Dellinger <[email protected]>, January 21, 2010
Updated by Matthew Dellinger <[email protected]>, May 18, 2010
Updated by Matthew Dellinger <[email protected]>, July 22, 2010

About

My modification to the general version of sched_test_app

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published