[llvm] [AArch64] optimize lowering for icmp on i128 when RHS is an immediate (PR #181822)

[Cheng Lingfei via llvm-commits]

================

----------------
clingfei wrote:

Yes, that seems like a real concern. The cutoff is probably microarchitecture-dependent, as the transform reduces instruction count but increases the critical-path dependency through the CCMP chain. Do you think I need to perform tests on certain CPUs to find a reasonable conservative threshold?

https://github.com/llvm/llvm-project/pull/181822