[llvm] [LSV] Enhance LoadStoreVectorizer to Handle Disjoint Flag in OR Instructions and Restore Vectorization Opportunities (PR #96495)

[Nikita Popov via llvm-commits]

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

----------------
nikic wrote:

Is it possible to further reduce this? It's hard to understand test cases with so many instructions.

I can see that `or disjoint` is indeed missing -- but only if `-mtriple=nvptx64-nvidia-cuda` is used. Why does adding the nvptx triple break the disjoint inference

https://github.com/llvm/llvm-project/pull/96495