[llvm] [AMDGPU] Add target feature require-naturally-aligned-buffer-access (PR #115479)

[Piotr Sobczak via llvm-commits]

piotrAMD wrote:

Added a codegen test that demonstrates a problem with under-aligned memory accesses (and the new target feature). This will need a different patch to fix, adding a FIXME in the test.
I would expect in that case the load/stores get split. We already have a path in the lowering for regular loads (SplitVectorLoad), but this is too late for buffer loads because they get converted to intrinsics in amdgpu-lower-buffer-fat-pointers. Looks that pass would need to be extended to address this case.

https://github.com/llvm/llvm-project/pull/115479