Actual benefit of compiling for CPUTYPE?

[zirias@]

Seeing this thread: https://forums.freebsd.org/threads/different-cputype-for-clang-and-gcc.75144/ I was wondering whether there are some actual measurement results (on x86_64) showing a performance improvement when compiling for some specific CPU types? So far, I never used this option as I don't think it's worth the time to build all the packages for every single machine (as they all have different CPUs here) -- is this assumption still correct? Or are there some ports that significantly benefit so it makes sense to enable CPUTYPE just for them?

 

[tingo]

Yes, that assumption is still correct, IMHO. You might get a measurable improvement, but the benefit is far too small to justify the time spent.

 

[diizzy]

You can get quite a bit of improvment but it's very library/application dependant so you have to evalute (do benchmarks).

 

[Alain De Vos]

Compiling with CPUTYPE can be usefull in two situations:
1. You have computer.
2. You identical computers.