[llvm] [X86] fold mov dec/inc to lea +- 1 (PR #185194)

[Takashi Idobe via llvm-commits]

Takashiidobe wrote:

The other todo (outside of other review comments) is whether or not this code is faster on slow lea machines. On a zen 4 and an alder lake this optimization was faster according to exegesis but llvm-mca thinks both code is pretty much the same cycle wise for those two cpus so without a real machine it's hard to say. 

I might be able to dust off an old imac I have laying around that but how does one generally test performance in similar cases?

https://github.com/llvm/llvm-project/pull/185194