Rerunning an Experiment: The labrerun Tool

Once an experiment has been performed using labrun it is easy to rerun the same experiment later.

labrerun [<options>] <log_file>

In its simplest form, labrerun takes a log file <log_file> and then repeats the experiment specified in this log file. labrerun will first check out the sources that were used when the experiment was performed. Then, it will reconstruct all parameters that are needed by a call of labrun in order to perform the same experiment. The environment variables will be set as recorded in <log_file>. New .[out|log|err] files for the rerun will be generated by default in the directory containing the original log file. Several options are provided by labrerun; many of them are the same as for labrun.

$\Rightarrow$ Tutorial 11:

Rerun the experiment (2b) from Tutuorial . (If you like, you can even modify the source code of sort-demo, perform a new set of expriments and after that repeat the experiment (2b).)

Solution:

labrerun ./lab_log/sort-demo-10000-10-O3.log

Sometimes it might be desirable to keep the CVS sources that are checked out by labrerun. (By default, the latest versions of the sources are retrieved from CVS after labrerun.) This can be achieved by using the option --keep. The --nocvs option specifies that the current source files should be used instead of those specified in the log file. In cases where one wants to disregard certain enviornment variable definitions given in <log_file>, the --ignore option comes in handy.

$\Rightarrow$ Tutorial 12:

(1) Rerun experiment (2b) from Tutuorial  but keep the sources that are checked out. (2) Rerun experiment (1a) from Tutorial  but use the current sources. (3) Restore the latest version of the source code in the directory.

Solution:

labrerun --keep ./lab_log/sort-demo-10000-10-O3.log
labrerun --nocvs ./lab_log/sort-demo-5000-10-no-O.log
cvs up -A