[llvm] [AMDGPU] Add alignment attr & propagate alignment through make.buffer.rsrc inst (PR #145278)

[Matt Arsenault via llvm-commits]

================
@@ -138,6 +138,18 @@ static bool funcRequiresHostcallPtr(const Function &F) {
          F.hasFnAttribute(Attribute::SanitizeMemTag);
 }
 
+static bool isAlignAndMakeBuffer(const AbstractAttribute *AA,
+                                 const Instruction *I) {
+  if (isa<AAAlign>(AA)) {
----------------
arsenm wrote:

Probably shouldn't need to ever identify AAs like this? 

https://github.com/llvm/llvm-project/pull/145278