Running Experiments

If you start narjillos with the -fast option, then it runs in Fast Mode: no graphics, parallel calculations. It's quite fast, especially if you run it on multiple cores.

Run narjillos --help to see other command-line arguments. In particular, you can use -p (or --persistent) to save the experiment to an .exp file every 10 minutes. If you interrupt the experiment for any reason, you can pass the name of this file as an argument to narjillos and pick up the experiment from the last saved state. This allows you to run very long experiments. My tipical workflow is:

1. I start an experiment with narjillos --fast --persistent and let it run for a few hours.

2. After a while, I kill the process, that leaves behind an .exp file (say, 774012997-0.3.0.exp).

3. I load the file in narjillos to see what is happening:

    narjillos 774012997-0.3.0
   

4. If I see interesting stuff, I exit the program and continue the experiment in Fast Mode (when you load from a file, it's automatically persisted, so you don't need the --persistent option):

    narjillos 774012997-0.3.0 --fast
   

5. I go back to 2.

Experiments are deterministic. Each experiment has a name composed of a random seed, plus the current version of the application. Experiments with the same identifier will give exactly the same results over time.

If you have the soul of a scientist in you, see the page on Analyzing Data for more cool stuff.