[duped]

The problem is that most default linkers don't scale well with core count, and throwing hardware at the problem is kind of a waste of money since single core execution speeds only present marginal gains. If you use a linker like mold instead of ld you'll see orders of magnitude higher performance.

Also throw the guy a few bucks if you're on MacOS

[sagarm]

Clang's lld also uses multiple threads. The author of mold also contributed (contributed?) to lld.