Multiple Data Sets, Compilations, or Machines

Though conducting an experiment using multiple data sets or multiple sets of compilation flags can be easily accomplished by issuing multiple separate experiment commands, it is a bit tedious to do things this way. One can easily forget which data sets have been tested and which have not. We have therefore provided ways around (some of) the tedium and (some of) the forgetfulness. For the tools labrun (Section ) and labrerun (Section ), the keyword NEX (for ``new experiment'', or simply ``next'' if you prefer) can be used in the command line to indicate that what follows is the input for a new experiment. Thus multiple experiments can be specified at once. See Section  and the examples accompanying the description of labrun in Section  for examples.

Any number of NEX's are possible. This feature is most useful when command-line arguments are provided via a file instead of on the command line itself (Section ).

For running experiments on different machines, our tools can be used quite easily in conjunction with the ssh command. For example, one could compile and execute a program on a specific machine turing using the command

   ssh turing labmex fun 123

See the documentation for ssh for more details on remote execution of commands. Using ssh in conjunction with the NEX keyword allows one to run an experiment on several different machines quite easily. See the examples accompanying the description of labrun in Section  for an example.