Profiling tools are critical for trying to understand performance bottlenecks in code.
Without real data about what a program is doing while running, detecting bottlenecks is at best a process of trial and error for both users and developers. While thoughtful testing of various program configurations along with measuring time elapsed for decrete runs can often be enough - why not learn faster and funner ways to rapidly collect real data about what a program is doing?
Actual data about program execution like which functions are being called while a program is active helps point to hot parts of the code where most time may be being spent. While users of applications may not easily understand the output of profiling tools, being equipped to generate profiling output can be extremely useful for sharing with developers, since the time to set up robust test cases for developers is can be greater than the time it takes to understand and optimize slow code paths. Therefore it can be invaluable to get
