The Julia runtime includes a built-in profiler that can be used to measure which lines of code contribute the most to the total execution time of a codebase. It can therefore be used to identify bottlenecks in code, which can in turn be used to prioritize optimization efforts.
This built-in system is what is known as a sampling profiler. Its work is to inspect the call stack of the running system every few milliseconds (by default, 1 millisecond on UNIX and 10 milliseconds on Windows), and identify each line of code that contributes to this call stack. The idea is that the lines of code that are executed most often are found more often on the call stack. Hence, over many such samples, the count of how often each line of code is encountered will be a measure of how often this code runs.
The primary advantage of a sampling profiler is that it can run without modifying the source program, and thus has a very minimal overhead. The program runs at almost full speed when being...