wallClockProfiler | A wall-clock stack-sampling profiler

wallClockProfiler is a C++ library. wallClockProfiler has no bugs, it has no vulnerabilities and it has low support. However wallClockProfiler has a Non-SPDX License. You can download it from GitHub.

Sadly, this is apparently the ONLY profiler on Linux that actually works. By "actually works," I mean that a profiler is supposed to give you a rough estimate where your program is spending the most time. Slow programs are slow for a variety of reasons, not just because of inefficient algorithms that waste CPU cycles. Lots of programs are I/O bound. If you profile such a program, you want to know where the I/O hotspots are, and you don't care where the CPU hotspots are, because they are tiny, by comparison. As far as I can tell, after way too many hours of researching and testing every available tool, there is nothing on Linux that actually does this well. I had a simple test program, written in C++, that looped over a big file on disk, fseeking randomly and reading from the file a bunch of times. Obviously, such a program is spending almost all of its time in fseek. But not a single existing profiler would tell me this. Not perf. Not gperftools. Not OProfile. Obviously not gprof. None of the usual suspects. Some of them were telling me that a good portion of the running time was spent in rand (which was used to pick the random offset to fseek to). The truth was, this was only about 1% of the simple program's wall-clock runtime. If I want to speed this program up, optimizing rand is going to be a waste of time. Valgrind's callgrind can provide better results than the others, but it makes the program 10x slower (or something like that), so it's useless in production environments (for example, profiling a live server that has real users connecting to it).