RTSS’17 Supplemental Materials

This page provides the implementation of the exact and sustainable schedulability test for non-preemptive job sets and periodic tasks discussed in the paper:

Git Repository

For convenience, the full implementation is provided as a Git repository on Github.

If you want to build on this work, please use the latest version from the repository.

Implementation used in Paper

The source code of the implementation used for the experiments is provided as a ZIP archive.

Dependencies

Build Instructions

These instructions assume a Linux or macOS host.

To compile the tool, first generate an appropriate Makefile with cmake and then use it to actually build the source tree.

# (1) create a build directory
make build
# (2) enter the build directory
cd build
# (3) generate the Makefile
cmake ..
# (4) build everything
make -j

The last step yields two binaries:

Unit Tests

The tool comes with a test driver (based on C++ doctest) named runtests. After compiling everything, just run the tool to make sure everything works.

$ ./runtests 
[doctest] doctest version is "1.2.1"
[doctest] run with "--help" for options
===============================================================================
[doctest] test cases:     27 |     27 passed |      0 failed |      0 skipped
[doctest] assertions:    260 |    260 passed |      0 failed |
[doctest] Status: SUCCESS!

Input Format

The tool operates on CSV files with a fixed column order. Each input CSV file is expected to contain a set of jobs, where each row specifies one job. The following columns are expected.

  1. Task ID — an arbitrary numeric ID to identify the task to which a job belongs

All numeric parameters can be 64-bit integers (preferred) or floating point values (slower, not recommended).

For example input files, see the files in the examples/ folder (e.g., examples/fig1a.csv).

Analyzing a Job Set

To run the tool on a given job set, just pass the filename as an argument. For example:

$ build/nptest examples/fig1a.csv 
examples/fig1a.csv,  0,  9,  6,  5,  1,  0.000329,  820.000000,  0

The output format is explained below.

If no input files are provided, or if - is provided as an input file name, nptest will read the job set description from standard input, which allows applying filters etc. For example:

$ cat examples/fig1a.csv | grep -v '3, 9' | build/nptest 
-,  1,  8,  9,  8,  0,  0.000069,  816.000000,  0

To activate an idle-time insertion policy (IIP, see paper for details), use the -i option. For example, to use the CW-EDF IIP, pass the -i CW option:

$ build/nptest -i CW examples/fig1a.csv 
examples/fig1a.csv,  1,  9,  10,  9,  0,  0.000121,  848.000000,  0

When analyzing a job set with dense-time parameters (i.e., time values specified as floating-point numbers), the option -t dense must be passed.

See the builtin help (nptest -h) for further options.

Output Format

The output is also provided in CSV format and consists of the following columns:

  1. The input file name.
  2. The schedulability result: 1 if the job is is schedulable, 0 if it is not.
  3. The number of jobs in the job set.
  4. The number of states that were explored.
  5. The number of edges that were discovered.
  6. The maximum “exploration front width” of the schedule graph, which is the maximum number of unprocessed states that are queued for exploration (at any point in time).
  7. The CPU time used in the analysis (in seconds).
  8. The peak amount of memory used (as reported by getrusage()), divided by 1024. Due to non-portable differences in getrusage(), on Linux this reports the memory usage in megabytes, whereas on macOS it reports the memory usage in kilobytes.