[llvm] [AMDGPU] Overload resource descriptor in image intrinsics. (PR #107255)

[Jay Foad via llvm-commits]

jayfoad wrote:

I think the only real use case for this is allowing either v4i32 or v8i32 for an image descriptor, which would allow implementing support for the r128 bit in image instructions. But I guess it doesn't hurt to add the extra flexibility of overloading the sampler descriptor type too.

https://github.com/llvm/llvm-project/pull/107255