[clang] [llvm] [AMDGPU] Add initial support for VGPR as memory (PR #205435)

[Matt Arsenault via llvm-commits]

================
@@ -1779,6 +1780,17 @@ bool hasValueInRangeLikeMetadata(const MDNode &MD, int64_t Val) {
   return false;
 }
 
+AllocatedVGPRsMetadata AllocatedVGPRsMetadata::get(const AllocaInst &Alloca) {
+  const MDNode *MD = Alloca.getMetadata("amdgpu.allocated.vgprs");
+  assert(MD && MD->getNumOperands() == 2 &&
+         "expected !amdgpu.allocated.vgprs metadata with 2 operands");
----------------
arsenm wrote:

IR structural rules should be enforced by the verifier 

https://github.com/llvm/llvm-project/pull/205435