How many times faster a program is when run in parallel than when run on one processor. Speedup can be computed by dividing the running time TP of the program on P processors by its running time T1 on one processor.