Sampling profilers create a profile by looking at the running program's state at even intervals—typically, every 10 ms. Sampling profilers usually have a minimum impact on the program's actual performance, and it's also possible to build the program in release mode with all optimizations turned on. A drawback of sampling profilers is their inaccuracy and statistical approach, which is usually not a problem as long as you are aware of it. The following figure shows a sampling session of a running program with five functions: main(), f1(), f2(), f3(), and f4(). The t1-t10 labels indicate when each sample was taken. The boxes indicate the entry and exit point of each executing function:
Example of a sampling profiler session ...
