* bench/stutter/stutter_bench.sh: Add headers in the CSV files, and also run stutter_invariance_randomgraph. * bench/stutter/stutter_invariance_formulas.cc: Remove space from CSV output. * bench/stutter/stutter_invariance_randomgraph.cc: Likewise, plus fix the call to is_stutter_invariant(), and return an average time. * bench/stutter/stutter.ipynb: Adjust. * bench/stutter/README: Simplify. * bench/stutter/Makefile.am: Distribute the script and python notebook.
18 lines
599 B
Text
18 lines
599 B
Text
This benchmark measures the performance of different algorithms to
|
|
check if property (expressed as a formula or as a deterministic TGBA)
|
|
is stutter-invariant. When the benchmark is run on formulas, the
|
|
translation time is not included in the measured time.
|
|
|
|
|
|
To reproduce the benchmark is to run
|
|
|
|
% ./stutter_bench.sh
|
|
|
|
to create bench_formulas.csv and bench_randgraph.csv, and then
|
|
explore these data the provided ipython notbook
|
|
|
|
% ipython notebook --pylab=inline stutter.ipynb
|
|
|
|
|
|
The time in bench_formulas.csv is reported in microseconds, while the
|
|
time in bench_randgraph.csv is in seconds.
|