[clang] [CIR] Wire AMDGPU into the call-convention lowering pass (PR #220197)

[Steffen Larsen via cfe-commits]

================
@@ -615,19 +618,20 @@ static const llvm::abi::Type *mapCIRType(mlir::Type type,
       })
       .Default([](mlir::Type) -> const llvm::abi::Type * {
         llvm_unreachable(
-            "mapCIRType: type not pre-filtered by classifyX86_64Function");
+            "mapCIRType: type not pre-filtered by classifyAbiSignature");
       });
 }
 
 /// Convert an llvm::abi::ArgInfo into the ArgClassification consumed by
 /// CIRABIRewriteContext.
 ///
 /// Direct: the value passes in register(s).  A coercion is forwarded in the
-/// four cases where the value has to be rebuilt on the wire: an aggregate
+/// five cases where the value has to be rebuilt on the wire: an aggregate
----------------
steffenlarsen wrote:

Personally I would just remove the number here. "cases" expresses the plural nature, so very little value is lost and we won't have to maintain a number in comments.

https://github.com/llvm/llvm-project/pull/220197