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

[Aaron Ballman via llvm-commits]

================
@@ -9102,6 +9102,15 @@ bool InitializationSequence::Diagnose(Sema &S,
 
   case FK_ConversionFailed: {
     QualType FromType = OnlyArg->getType();
+    // __amdgpu_feature_predicate_t can be explicitly cast to the logical op
+    // type, although this is almost always an error and we advise against it
----------------
AaronBallman wrote:

Let me see if I'm on the same page now. The backend will generate code for gfx900 and the programmer will guard that block of code with `if (__builtin_amdgcn_processor_is("gfx900"))`. So if the predicate and the block it controls become disjointed somehow, the user can get incorrect behavior at runtime. Am I on the right track?

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