[clang] [llvm] [SROA] Vector promote some memsets (PR #133301)

[via cfe-commits]

macurtis-amd wrote:

> To make the question more precise, would it make sense to change AMDGPUs unaligned access hook such that https://godbolt.org/z/Gv1j4vjqE will look the same as on X86?
> 
> That should also fix the motivating problem here.

I'll investigate this as an alternative. Thanks.

https://github.com/llvm/llvm-project/pull/133301