[llvm] [DA] Add tests for nsw doesn't hold on entier iteration (PR #162281)

[Ryotaro Kasuga via llvm-commits]

kasuga-fj wrote:

I'm just suggesting that different topics should be discussed in separate threads. I raised the original issue to show the potential inconsistency between what current nsw-based monotonicity checker guarantees and what the dependence tests implicitly assume. However, you've been focusing mostly on your new approach to Strong SIV. I feel that the original topic I'd like to discuss is getting lost in the flow of your posts. If you want to propose a new approach, you should open another PR.

---

As for your new approach, it looks like a special case of Exact SIV.

https://github.com/llvm/llvm-project/pull/162281