Truncation warning message incorrectly references an invalid hlsl type

[dmpots]

Description Dxc issues a confusing warning when an hlsl intrinsic that returns a 64-bit type is cast to a 32-bit type.

Steps to Reproduce https://godbolt.org/z/vo84G6nac

// dxc /Tvs_6_0 -spirv longlong.hlsl
uint main() : A {
   return vk::ReadClock(vk::SubgroupScope);
}

Actual Behavior Produces the warning

warning: conversion from larger type 'unsigned long long' to smaller type 'uint', possible loss of data [-Wconversion]
   return vk::ReadClock(vk::SubgroupScope);

The unsinged long long type is not valid in hlsl. I would expect the warning to refer to the type by the hlsl name of uint64_t.

It seems to only refer to the long long type for intrinsics. If I have variables that are uint64_t that are cast to uint the warning message refers to the correct type.

Environment

• dxcompiler.dll: 1.8 - 1.7.0.4383 (ad3958c2d); dxil.dll: 1.6(101.6.2104.33)
• Windows 11

[llvm-beanz]

The fix for this is extremely straightforward:

diff --git a/tools/clang/lib/AST/Type.cpp b/tools/clang/lib/AST/Type.cpp
index 06db4747ffe..e114f856859 100644
--- a/tools/clang/lib/AST/Type.cpp
+++ b/tools/clang/lib/AST/Type.cpp
@@ -2500,13 +2500,13 @@ StringRef BuiltinType::getName(const PrintingPolicy &Policy) const {
   case Short:             return /* "short" */ "int16_t" /* HLSL Change */;
   case Int:               return "int";
   case Long:              return "long";
-  case LongLong:          return "long long";
+  case LongLong:          return "int64_t";
   case Int128:            return "__int128";
   case UChar:             return "unsigned char";
   case UShort:            return /* "unsigned short" */ "uint16_t" /* HLSL Change */;
   case UInt:              return "unsigned int";
   case ULong:             return "unsigned long";
-  case ULongLong:         return "unsigned long long";
+  case ULongLong:         return "uint64_t";
   case UInt128:           return "unsigned __int128";
   // HLSL Changes begin
   case HalfFloat:

The unfortunate problem is that we have a significant number of tests that break if we make this change. We'll need to find time to update the tests appropriately.