[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

[Alex Voicu via llvm-commits]

================
@@ -15576,6 +15609,38 @@ static bool isOverflowingIntegerType(ASTContext &Ctx, QualType T) {
   return Ctx.getIntWidth(T) >= Ctx.getIntWidth(Ctx.IntTy);
 }
 
+static Expr *ExpandAMDGPUPredicateBI(ASTContext &Ctx, CallExpr *CE) {
+  if (!CE->getBuiltinCallee())
+    return CXXBoolLiteralExpr::Create(Ctx, false, Ctx.BoolTy, CE->getExprLoc());
+
+  if (Ctx.getTargetInfo().getTriple().isSPIRV()) {
+    CE->setType(Ctx.getLogicalOperationType());
+    return CE;
+  }
+
+  bool P = false;
+  auto &TI = Ctx.getTargetInfo();
+
+  if (CE->getDirectCallee()->getName() == "__builtin_amdgcn_processor_is") {
----------------
AlexVlx wrote:

These are both true, and I thank you for the feedback. As a relatively weak retort, I will note that:
- I went for the names because it felt a bit icky to add the AMDGPU specific builtin header, considering we're trying to limit the scope of these; also I did not feel confident enough to make these generic Clang BIs (for good reason, as the review shows:));
- The call to this function comes after having already checked that the Callee is one of the predicates, `IsAMDGPUPredicateBI` and `ValidateAMDGPUPredicateBI` get called before, so the precondition that we are indeed dealing with the magical BIs is established; furthermore, we're already checking upon entry that the Callee is indeed a builtin, and I *believe* that builtins always have non-elaborated names which can always be obtained via getName - I could be wrong here.
Having said that, using the Builtin IDs would indeed be nicer, so I can switch to that, thank you for the suggestion.

https://github.com/llvm/llvm-project/pull/134016