Benchmarking different compilers?

[Qqwy]

Hi there!

I want to benchmark how a piece of code (in this example: some coroutine stuff) is optimized differently by clang++ and g++.

What would be the best way to benchmark this?

Should I write one piece of code that is wrapped by nanobench, compile it to different binaries using different compilers/compiler settings, and then run them separately and manually combine the results?

Or should I wrap the same code by different namespaces, compile them using different compilers/compiler settings to a bunch of .so files and link them together, so they can be run from a single nanobench setup?

Both of these techniques seem a little involved. Is there anyone who has used nanobench for this kind of task before that might have some advice to offer?

Thank you for your help,

~Qqwy