Probabilisitc Schedulability Test in SIES'17

• Most of codes are refined in RTCSA'18 repo: https://github.com/kuanhsunchen/MissRateSimulator
• The paper can be found in https://ieeexplore.ieee.org/abstract/document/7993392/

Proposed and compared approaches:

• EPST.py contains the proposed approaches.
• bounds.py contains different bounds related to the Chernoff bound.
• task-generator.py contains the task generating routines, especially for generating MATLAB input [1].
• sort_task_set.py contains the task generating routines.
• cprta.py contains the approaches implemented in Python [1].

Experimental setup in the paper:

• timing.py contains the time measument routines.
• taskset-matlabs100u60.m / taskset-p10s100u60.txt are the input of shown results for utilization 60 generated by Uunifast.
• taskset-matlabs100u70.m / taskset-p10s100u70.txt are the input of shown results for utilization 70 generated by Uunifast.

Heuristic example:

• simple_test.py heuristically demonstrates the usage of the proposed approaches.
• simple_approxError.py heuristically demonstrates the approx. error we made with given inputs.

Error in my paper [2]

• in simple_test.py, P^A_1 and P^A_2 in the paper was reported as 10^-5. However in fact the results was all based on 10^-6.
• Which real value s should be selected is out of scope of this paper. Currently I use the range from (0, 10] with step 0.1 to get the minimum. Since any s larger than 0 is sutiable for the Chernoff bound, it is another story to find out a specific s giving the optimal result.

Matlab files (These can be found in matlab folder):

• scriptSimulations.m is the modified simulation executable from [1].
• dmp.m is almost the original one but the redundant output behavior is disable for readability.
• givenTaskSetFunction.m contains the function for using our generated input. The given input from taskset-matlabs100u60.m / taskset-matlabs100u70.m should be renamed as "INPUT.m" in tehe same folder.

Legacy files:

• mp-frVsProb.py is used to generate the results on multiprocessors.
• mp-frVsProb_dml.py is used to generate the results for l-consecutive misses on multiprocessors.

Reference

• [1] D. Maxim and L. Cucu-Grosjean. Response time analysis for fixed-priority tasks with multiple probabilistic parameters. RTSS'13. The source code of CPRTA and the resampling feature is from here: https://who.rocq.inria.fr/Dorin.Maxim/tool.html
• [2] K. H. Chen and J. J. Chen, "Probabilistic schedulability tests for uniprocessor fixed-priority scheduling under soft errors," 2017 12th IEEE International Symposium on Industrial Embedded Systems (SIES), Toulouse, France, 2017, pp. 1-8.