MSL accuracy scoreboard
Every example of the Modelica Standard Library that has a Modelica Association reference result is simulated with DAEdalus and compared against that reference, signal by signal.
By package
Select a package to filter the table.All examples
| Model | Status | Time (s) | Detail |
|---|
How the comparison works
For each example the reference results list a set of comparison signals and give their values computed with Dymola. DAEdalus simulates the example with its default settings over the same interval, writes those signals on the reference time grid, and checks each one with a tube test in the style of csv-compare, the tool used for MSL releases. Each reference point gets a box with a half-width of 2e-3 × the time span in time and 2e-3 × the signal range in value. The DAEdalus value must stay inside the union of those boxes.
- Pass: every comparison signal stays inside its tube.
- Differ: the model simulates, but at least one signal leaves its tube or is missing from the output.
- Error: translation or simulation failed. The message shown is the last line DAEdalus printed.
- Timeout: no result within the time limit.
The comparison script is tools/refcompare.py. It runs nightly in CI, and the raw data is
msl-4.1.0.json.