[clang] [X86][Clang] Add AVX512 Integer Comparison Intrinsics for constexpr Evaluation (PR #164026)

[Simon Pilgrim via cfe-commits]

================
@@ -15766,6 +15766,89 @@ bool IntExprEvaluator::VisitBuiltinCallExpr(const CallExpr *E,
     unsigned Idx = static_cast<unsigned>(IdxAPS.getZExtValue() & (N - 1));
     return Success(Vec.getVectorElt(Idx).getInt(), E);
   }
+
+  case clang::X86::BI__builtin_ia32_cmpb128_mask:
+  case clang::X86::BI__builtin_ia32_cmpw128_mask:
+  case clang::X86::BI__builtin_ia32_cmpd128_mask:
+  case clang::X86::BI__builtin_ia32_cmpq128_mask:
+  case clang::X86::BI__builtin_ia32_cmpb256_mask:
+  case clang::X86::BI__builtin_ia32_cmpw256_mask:
+  case clang::X86::BI__builtin_ia32_cmpd256_mask:
+  case clang::X86::BI__builtin_ia32_cmpq256_mask:
+  case clang::X86::BI__builtin_ia32_cmpb512_mask:
+  case clang::X86::BI__builtin_ia32_cmpw512_mask:
+  case clang::X86::BI__builtin_ia32_cmpd512_mask:
+  case clang::X86::BI__builtin_ia32_cmpq512_mask:
+  case clang::X86::BI__builtin_ia32_ucmpb128_mask:
+  case clang::X86::BI__builtin_ia32_ucmpw128_mask:
+  case clang::X86::BI__builtin_ia32_ucmpd128_mask:
+  case clang::X86::BI__builtin_ia32_ucmpq128_mask:
+  case clang::X86::BI__builtin_ia32_ucmpb256_mask:
+  case clang::X86::BI__builtin_ia32_ucmpw256_mask:
+  case clang::X86::BI__builtin_ia32_ucmpd256_mask:
+  case clang::X86::BI__builtin_ia32_ucmpq256_mask:
+  case clang::X86::BI__builtin_ia32_ucmpb512_mask:
+  case clang::X86::BI__builtin_ia32_ucmpw512_mask:
+  case clang::X86::BI__builtin_ia32_ucmpd512_mask:
+  case clang::X86::BI__builtin_ia32_ucmpq512_mask: {
+    assert(E->getNumArgs() == 4);
+
+    bool IsUnsigned =
+        (BuiltinOp >= clang::X86::BI__builtin_ia32_ucmpb128_mask &&
+         BuiltinOp <= clang::X86::BI__builtin_ia32_ucmpq512_mask);
----------------
RKSimon wrote:

@sskzakaria This should have been:
```suggestion
         BuiltinOp <= clang::X86::BI__builtin_ia32_ucmpw512_mask);
```
As the builtins are alpha sorted not bitwidth - I think its the source of the problem that @shri-acha is seeing - I'll create a patch shortly

https://github.com/llvm/llvm-project/pull/164026