Preparing the profiler and processing the results

When analyzing profiling results, you may often want to perform some preparation, cleanup, and processing. For instance, if your code mostly spends time spinning around, you might want to filter that out. Before even starting the profiler, be sure to compile or download as many debug symbols as you can, both for your code, your dependencies, even the OS libraries, and kernel. Also, it's essential you disable frame pointer optimizations. On GCC and Clang, you can do so by passing the -fno-omit-frame-pointer flag. It won't affect performance much but will give you much more data about the execution of your code. When it comes to post-processing of the results, when using perf, it's usually a good idea to create flame graphs from the results. Brendan Gregg's tool from the Further reading section is great for that. Flame graphs are a simple and effective tool to see where the execution takes too much time, as the width...