[lld] [lld] Add target support for SystemZ (s390x) (PR #75643)

[Ulrich Weigand via llvm-commits]

uweigand wrote:

> FYI: Slightly off-topic, but we use an alternative PLT code sequence in mold so that we use only 16 bytes instead of 32 for each PLT entry. That's something you might want to consider. You can see the code here: https://github.com/rui314/mold/blob/4cdfc7e72c5c88010de0e43952b19fd48647848e/elf/arch-s390x.cc#L72-L76

@rui314 Maybe I'm missing something here, but doesn't this code still generate 32 bytes per function, 16 in the .plt section and 16 in the .plt.got section?  In bfd (and lld) we currently do not use any separate .plt.got ...   I guess there might still be some icache benefits though.


https://github.com/llvm/llvm-project/pull/75643