The programs were all written in ANSI C, and the simulations were run on a number of IBM RS/6000 workstations, and on a 12-processor IBM SP2, which uses RS/6000 processors. For the SP2, individual jobs were run separately on different processors using different sets of parameters, or different cooling schedules, or just different random numbers. Note that with multiple processors, it is possible to do many different annealing runs (using different random numbers) and then chose the result with the lowest cost. In this way, parallel processing is used not to speed up the time to obtain a result, but to obtain a better result in the same amount of time.
For the first and second semester data, the average time for running the preprocessor (the expert system) was about three hours, and the average time for each simulated annealing run was approximately 22 hours for geometric cooling and about 15 hours using adaptive cooling with reheating, which means the program could be run overnight. This is much faster than the current semi-automated procedure used at Syracuse University to produce a produce a final schedule for a semester, which involves substantial manual effort to iterate to a solution, and can take a few weeks.