I tried to do something similar for building highly optimized and slimmed down (only use flags that are needed) for container images (installing with different

I tried to do something similar for building highly optimized and slimmed down (only use flags that are needed) for container images (installing with different root path) but got some library issues in resulting system and gave up. maybe I'll try the same with prefix. I'm curious if compiling everything with exact expected CPU and LTO would result in measurable performance gains.