[llvm] [LoopBoundSplit] Check if latch is operand of PHI induction variable (PR #167227)

[Nikita Popov via llvm-commits]

https://github.com/nikic commented:

I think the deeper problem here is that we're missing a check that AddRecSCEV is an addrec for the correct loop? I think this is a symptom of that.

(By the way, the way things are split across analyzeICmp and hasProcessableCondition is unintuitive and it would make more sense to merge these functions instead of first setting things to incorrect values and then bailing out later...)

https://github.com/llvm/llvm-project/pull/167227