Several design algorithms use random numbers to make arbitrary decisions. The graph algorithms use random numbers to break ties in locating department nodes in the adjacency graph. The layout improvement algorithms based on simulated annealing use random numbers to select the next two departments to be tested for a possible exchange. The sequence of random numbers is determined by its starting point, called the seed. An algorithm will always produce the same results if you select the same algorithm options and set the same random number seed. You can set the random number generator seed with the Seed command of the Edit menu.
If an algorithm has executed multiple replications then the seed reported in the Statistics and Notes windows is the seed that yielded the best graph or layout among all the replications. An algorithm that does not use random numbers or that only executes one replication will always return the same seed as it was started with.
In order to repeat the results for an algorithm, you must set the seed to the value reported at the completion of the algorithm and set the number of replications to one. Then run the algorithm again with all other parameters unchanged.
The actual random numbers are generated from the seed with a standard pseudo random number generation process. This process is dependent on the combination of processor and compiler, so computers with different processors and compiler version may give different random number sequences. SPIRAL has no control over this process.