+covexp

covexp: Rapid Experiment on Data/Simulink Models

covexp is a MATLAB package which runs your experiments on Simulink programs (i.e. subjects), caches them and aggregates results.

Why call covexp?

We initially created it to experiment with Simulink model coverage. Now we use it for general-purpose experiments on some subjects.

Running

Clone this git repo, cd to the cloned directory, then in a MATLAB prompt:

covexp.covcollect();

Configuring

Edit ../covcfg.m which is self-documented.

Parallel run utilizing multiple cores

set PARFOR = true

Results

Results on individual subjects are cached in the disc with the covdata.mat suffix

Experiments

To create an experiment, write a function and put that function name in covcfg.m's EXPERIMENTS field.

Existing experiments are in +covexp/+experiments

If your experiment would return result, you need to initialize data-structures. Initialize results that the experiment would return using a function in , and point to that function in covcfg.m using EXP_INITS

Existing data-structure initialization functions are inside +covexp/+experiments/+ds_init

• Experiment should not throw errors. If an experiment throws, subsequent experiments for the same subject will not run.
• In Parallel mode other models will be run, but the touched file will not be cleared so that you know which model threw.
• In serial mode the script will stop sot that you can fix the bug.
• Do not close the Simulink model inside experiments

Running selected experiments

Choose which experiments you want to run in DO_THESE_EXPERIMENTS configuration. You can pass in an array of experiment ids. Experiments would be perfomed sequentially on a subject.

Simulink specific/advanced information

Copying cached results created elsewhere

In some other machine, issue following

tar -cjvf backup.tar.bz2 *covdata.mat *_pp.slx

Here, we are also copying the pre-processed _pp.slx files.

Next, copy the tar.bz2 file in your machine and extract:

tar -xjvf backup.tar.bz2 --overwrite

Fixing subject locations

Since the cached covdata.mat files were created in a different machine, they contain absolute directory locations for that machine. To fix these, run the fix_input_loc (5th) experiment

Results

Explain how to interpret the cached results

• simdur : duration to simulate the original model
• duration : duration to collect coverage