In contrast, software prefetching can determine nonunit strides as easily as it can determine unit strides. Assume prefetches write directly into the cache and that there is no ``pollution'' (overwriting data that must be used before the data that are prefetched). For best performance given a nonunit stride prefetcher, in the steady state of the inner loop, how many prefetches must be outstanding at a given time?

Use a physical stopwatch to record the length of time it takes to run the program. calculate the difference obtained by calls to the method system.currenttimemillis() just before the start of the algorithm and just after the end of the algorithm. calculate the difference obtained by calls to the method system.currenttimemillis() at the start of the program and at the end of the program so that the elapsed time includes the display of the result. use the value returned by the method system.currenttimemillis() just after the end of the algorithm as the elapsed time.