https://github.com/andreas-abel/nanoBench



I don't really have any experience with this tool, but it seems really cool. When I get more time I'll have to try it out. Anyone have some experience to share?

I’ve read through the description and… What is the purpose of measuring pieces of code unrelated to the rest of the project? To get values that are (sort of) documented for a particular processor? The piece of code may work in a completely different manner in the real code due to caching, multithreading, etc. It doesn’t seem to be suggested as a nano-replacement for a profiler, just some tool that generates a customizable loop and measures it.

Multithreading is certainly at a higher granularity and has no bearing on testing fragments of code. Caching effects can be reduce for inner loops. If that relates to one's application can only be determined by the goals of the project. Fitting this little tool into an existing workflow is also highly subjective.

https://ftp.gnu.org/pub/gnu/superopt/

Then there are tools which might benefit from integration with more detailed instruction level information.

I'm a big fan of premature optimization so it's right up my alley.