As noted in a previous section on algorithms and performance, it is important to take an empirical perspective on run time optimization through the use of various performance assessment tools. Support for code profiling and timing is available within the Python and IPython interpreters; we've already seen several uses of the IPython %timeit function. Profiling provides information on which functions are called, the number of times they are called, and the cumulative and per-call run time associated with each function. Profiling is thus especially useful in identifying code "hotspots" that account for a large fraction of the overall program run time. Timing (or benchmarking) involves characterizing how long an entire program — or specific functions within that program — take to execute, perhaps as a function of the size of the relevant data being operated upon. Profiling can be used to get an overall picture of where one should best focus one's efforts on performance optimization, and then more targeted timing tests can be used to evaluate the efficacy of different optimization strategies.