After you have instrumented your binary with discover, you run the binary the same way you normally would. Typically, if a particular combination of input causes your program to behave unexpectedly, you would instrument it with discover and run it with the same input to investigate potential memory problems. While the instrumented program is running, discover writes information about any memory problems it finds to the specified output files in the selected formats (text, HTML, or both). For information about interpreting the reports, see Analyzing discover Reports.
Because of the overhead of instrumentation, your program is likely to run significantly slower after you instrument it. Depending on the frequency of memory access, it might run as much as 50 times slower.