In the serial version of LUCAS, the main program directed the classes in Figure 7 to perform the desired number of replicates and time steps of simulation. In the parallel version, this responsibility is passed to the PVM class (Section 4.2.6) and the main program determines which incarnation of pLUCAS is currently running. Under PVM it is typical to have a single executable perform many different functions. pLUCAS has three major modes: master, servant and send data, each of which are examined in Section 4.2.6 in more detail.
The Stats class also had to be altered in the distributed version of LUCAS because all of the statistics collected each time step were originally stored in a single file. The implicit parallelism of pLUCAS required that each servant create its own, independent statistical files which are later reassembled by the master.
The final major alteration necessary to make LUCAS a parallel application was the creation and distribution of unique seeds for the random number generator. As discussed in Section 3.3.3, the R250 generator uses the standard Lehmer linear congruential generator to create seeds. This seed must be unique to each servant process to insure that a replicate is not repeated. Two hundred fifty integers are used to initialize the R250 generator, so a new seed for the LCG is generated simply by taking the value of the sequence after every 250 iterations.