https://github.com/aliaksandrsiarohin/rt_simulator

Simulator of real time taskset
https://github.com/aliaksandrsiarohin/rt_simulator

Last synced: about 1 month ago
JSON representation

Simulator of real time taskset

• Host: GitHub
• URL: https://github.com/aliaksandrsiarohin/rt_simulator
• Owner: AliaksandrSiarohin
• Created: 2016-02-03T21:45:42.000Z (almost 9 years ago)
• Default Branch: master
• Last Pushed: 2016-02-03T21:49:00.000Z (almost 9 years ago)
• Last Synced: 2023-08-04T13:56:17.135Z (over 1 year ago)
• Language: C++
• Size: 47.9 KB
• Stars: 1
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Real Time Simulator 

This is a project for Real Time Systems course. The main goal of this project is to create set real time tasks (with configurable execution times and relative periods) measure worst case response time and compare this response time with theoretical response time computed using RTA.

## Requirments

* libgonfig (I installed it like this sudo apt-get install libconfig++8-dev)

* gnuplot (Only if run with --plot (-p) option)

* astyle (Only for code formatting e.g. make indent)

## Usage

Usage: simulator [\-f file] [\-o directory] [\-s number] [\-plc]

* \-f, \-\-file      (file with tasks parameters (default task_spec.cfg))

* \-o, \-\-output    (directory to which put simulated response times of the tasks (default no output))

* \-s, \-\-simtime   (time of simulation in seconds (default 10s))

* \-p, \-\-plot      (plot task CDF (using gnuplot))

* \-c, \-\-counting  (use counting to model execution time (instead of busy wait on timer))

* \-h, \-\-help     (run help command)

### File option

File is file in libconfig format (see example task_spec.cfg). Here you can specify tasks parameters, e.g.

to which CPU task will be assigned (CPU option), task priority (priority option), inter arrival times (period option) for this option you should specify a pmf, and execution time (execution option) also pmf should be specified.

## Execution time modeling

To model execution times there are 2 approaches:

Counting:

```cpp

long long countTo = executionTime / HowLongOneAdditionTake;

for (long long i = 0; i < countTo; ++i) {}

```

Timer busy waiting:

```cpp

clock_gettime(CLOCK_THREAD_CPUTIME_ID, &start_time);

while (start_time + executionTime > current_time) {

  clock_gettime(CLOCK_THREAD_CPUTIME_ID, &current_time);

}

```

Roughly counting gives an error +-0.5ms, and timer busy waiting gives an error +0.3ms.

So the timer busy waiting is default option (you can specify it with -c option).

## Latency generation

Latency can be generated by running latency.sh script, which runs number of following commands:

```bash

find / -exec sync \; -exec ls {} \; &> $tmp_file &

```

Then run simulator, and when simulation is complete kill all the processes.

## Sample result 

Result for Task Set Task1((100ms, 200ms), (10ms, 15ms, 20ms)), Task2((60ms, 50ms), (10ms, 19ms)), Task3((50ms, 10ms)).

First is inter arrival time, second is execution time. Everything with equal probabilities. 


Priority of Task1 - 81, Task2 - 82, Task3 - 83.

#### Responce time analysis

* Task1: worst case response time 49ms

* Task2: worst case response time 29ms

* Task3: worst case response time 10ms

#### Run simulator

* Task1: worst case response time 49.248217ms

* Task2: worst case response time 29.205995ms

* Task3: worst case response time 10.228918ms

#### Run latency.sh

* Task1: worst case response time 62.108559ms

* Task2: worst case response time 33.716152ms

* Task3: worst case response time 14.685030ms

* Worst case kernel latency: 4.716000ms