[llvm] AMDGPU: Shrink extracted dword vector loads (PR #203484)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> For the sequencing point, I will pause this SDAG PR for now. I have updated the VectorCombine PR based on the reviews, but have not had further feedback there yet.

Which VectorCombine are you referring to?

https://github.com/llvm/llvm-project/pull/203484