Formatting changes, don't use saturated add

Author: Temperz87

clang/include/clang/Basic/BuiltinsX86.td

@@ -1354,13 +1354,10 @@ let Features = "avx512bitalg", Attributes = [NoThrow, Const, RequiredVectorWidth
   def vpshufbitqmb512_mask : X86Builtin<"unsigned long long int(_Vector<64, char>, _Vector<64, char>, unsigned long long int)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr, RequiredVectorWidth<512>] in {
-  def pmulhrsw512 : X86Builtin<"_Vector<32, short>(_Vector<32, short>, _Vector<32, short>)">;
-}
-
 let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr, RequiredVectorWidth<512>] in {
   def pavgb512 : X86Builtin<"_Vector<64, unsigned char>(_Vector<64, unsigned char>, _Vector<64, unsigned char>)">;
   def pavgw512 : X86Builtin<"_Vector<32, unsigned short>(_Vector<32, unsigned short>, _Vector<32, unsigned short>)">;
+  def pmulhrsw512 : X86Builtin<"_Vector<32, short>(_Vector<32, short>, _Vector<32, short>)">;
   def pmulhuw512 : X86Builtin<"_Vector<32, unsigned short>(_Vector<32, unsigned short>, _Vector<32, unsigned short>)">;
   def pmulhw512 : X86Builtin<"_Vector<32, short>(_Vector<32, short>, _Vector<32, short>)">;
 }

clang/lib/AST/ByteCode/InterpBuiltin.cpp

@@ -3430,11 +3430,10 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const CallExpr *Call,
     return interp__builtin_elementwise_int_binop(
       S, OpPC, Call, [](const APSInt &LHS, const APSInt &RHS) {        
         unsigned Width = LHS.getBitWidth();
-
         APInt Mul = llvm::APIntOps::mulhs(LHS, RHS);
         Mul = Mul.relativeLShr(14);
-        Mul = Mul.sadd_sat(APInt(Width, 1, true));
-        return APInt(Mul.relativeLShr(1));
+        Mul = Mul + APInt(Width, 1, true);
+        return Mul.relativeLShr(1);
       });
 
   case clang::X86::BI__builtin_ia32_pavgb128:

clang/lib/AST/ExprConstant.cpp

@@ -67,7 +67,6 @@
 #include <cstring>
 #include <functional>
 #include <limits>
-#include <llvm/ADT/APSInt.h>
 #include <optional>
 
 #define DEBUG_TYPE "exprconstant"
@@ -11733,8 +11732,8 @@ bool VectorExprEvaluator::VisitCallExpr(const CallExpr *E) {
       unsigned Width = LHS.getBitWidth();
       APInt Mul = llvm::APIntOps::mulhs(LHS, RHS);
       Mul = Mul.relativeLShr(14);
-      Mul = Mul.sadd_sat(APInt(Width, 1, true));
-      return APInt(Mul.relativeLShr(1));
+      Mul = Mul + APInt(Width, 1, true);
+      return Mul.relativeLShr(1);
     });
 
   case clang::X86::BI__builtin_ia32_pmulhuw128:

clang/lib/Headers/tmmintrin.h

@@ -799,6 +799,7 @@ _mm_sign_pi32(__m64 __a, __m64 __b)
 }
 
 #undef __anyext128
+#undef __zext128
 #undef __trunc64
 #undef __DEFAULT_FN_ATTRS
 #undef __DEFAULT_FN_ATTRS_CONSTEXPR