[Headers][X86] Allow basic AVX512 predicate ops to be used in constexpr

clang/include/clang/Basic/BuiltinsX86.td

@@ -1219,15 +1219,15 @@ let Features = "avx512f", Attributes = [NoThrow, RequiredVectorWidth<512>] in {
   def scatterdiv16si : X86Builtin<"void(void *, unsigned char, _Vector<8, long long int>, _Vector<8, int>, _Constant int)">;
 }
 
-let Features = "avx512dq", Attributes = [NoThrow, Const] in {
+let Features = "avx512dq", Attributes = [NoThrow, Const, Constexpr] in {
   def knotqi : X86Builtin<"unsigned char(unsigned char)">;
 }
 
-let Features = "avx512f", Attributes = [NoThrow, Const] in {
+let Features = "avx512f", Attributes = [NoThrow, Const, Constexpr] in {
   def knothi : X86Builtin<"unsigned short(unsigned short)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const] in {
+let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr] in {
   def knotsi : X86Builtin<"unsigned int(unsigned int)">;
   def knotdi : X86Builtin<"unsigned long long int(unsigned long long int)">;
 }
@@ -3076,51 +3076,51 @@ let Features = "avx512dq", Attributes = [NoThrow, Const, RequiredVectorWidth<128
   def fpclassss_mask : X86Builtin<"unsigned char(_Vector<4, float>, _Constant int, unsigned char)">;
 }
 
-let Features = "avx512dq", Attributes = [NoThrow, Const] in {
+let Features = "avx512dq", Attributes = [NoThrow, Const, Constexpr] in {
   def kaddqi : X86Builtin<"unsigned char(unsigned char, unsigned char)">;
   def kaddhi : X86Builtin<"unsigned short(unsigned short, unsigned short)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const] in {
+let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr] in {
   def kaddsi : X86Builtin<"unsigned int(unsigned int, unsigned int)">;
   def kadddi : X86Builtin<"unsigned long long int(unsigned long long int, unsigned long long int)">;
 }
 
-let Features = "avx512dq", Attributes = [NoThrow, Const] in {
+let Features = "avx512dq", Attributes = [NoThrow, Const, Constexpr] in {
   def kandqi : X86Builtin<"unsigned char(unsigned char, unsigned char)">;
 }
 
-let Features = "avx512f", Attributes = [NoThrow, Const] in {
+let Features = "avx512f", Attributes = [NoThrow, Const, Constexpr] in {
   def kandhi : X86Builtin<"unsigned short(unsigned short, unsigned short)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const] in {
+let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr] in {
   def kandsi : X86Builtin<"unsigned int(unsigned int, unsigned int)">;
   def kanddi : X86Builtin<"unsigned long long int(unsigned long long int, unsigned long long int)">;
 }
 
-let Features = "avx512dq", Attributes = [NoThrow, Const] in {
+let Features = "avx512dq", Attributes = [NoThrow, Const, Constexpr] in {
   def kandnqi : X86Builtin<"unsigned char(unsigned char, unsigned char)">;
 }
 
-let Features = "avx512f", Attributes = [NoThrow, Const] in {
+let Features = "avx512f", Attributes = [NoThrow, Const, Constexpr] in {
   def kandnhi : X86Builtin<"unsigned short(unsigned short, unsigned short)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const] in {
+let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr] in {
   def kandnsi : X86Builtin<"unsigned int(unsigned int, unsigned int)">;
   def kandndi : X86Builtin<"unsigned long long int(unsigned long long int, unsigned long long int)">;
 }
 
-let Features = "avx512dq", Attributes = [NoThrow, Const] in {
+let Features = "avx512dq", Attributes = [NoThrow, Const, Constexpr] in {
   def korqi : X86Builtin<"unsigned char(unsigned char, unsigned char)">;
 }
 
-let Features = "avx512f", Attributes = [NoThrow, Const] in {
+let Features = "avx512f", Attributes = [NoThrow, Const, Constexpr] in {
   def korhi : X86Builtin<"unsigned short(unsigned short, unsigned short)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const] in {
+let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr] in {
   def korsi : X86Builtin<"unsigned int(unsigned int, unsigned int)">;
   def kordi : X86Builtin<"unsigned long long int(unsigned long long int, unsigned long long int)">;
 }
@@ -3160,28 +3160,28 @@ let Features = "avx512f", Attributes = [NoThrow, Const] in {
   def kunpckhi : X86Builtin<"unsigned short(unsigned short, unsigned short)">;
 }
 
-let Features = "avx512dq", Attributes = [NoThrow, Const] in {
+let Features = "avx512dq", Attributes = [NoThrow, Const, Constexpr] in {
   def kxnorqi : X86Builtin<"unsigned char(unsigned char, unsigned char)">;
 }
 
-let Features = "avx512f", Attributes = [NoThrow, Const] in {
+let Features = "avx512f", Attributes = [NoThrow, Const, Constexpr] in {
   def kxnorhi : X86Builtin<"unsigned short(unsigned short, unsigned short)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const] in {
+let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr] in {
   def kxnorsi : X86Builtin<"unsigned int(unsigned int, unsigned int)">;
   def kxnordi : X86Builtin<"unsigned long long int(unsigned long long int, unsigned long long int)">;
 }
 
-let Features = "avx512dq", Attributes = [NoThrow, Const] in {
+let Features = "avx512dq", Attributes = [NoThrow, Const, Constexpr] in {
   def kxorqi : X86Builtin<"unsigned char(unsigned char, unsigned char)">;
 }
 
-let Features = "avx512f", Attributes = [NoThrow, Const] in {
+let Features = "avx512f", Attributes = [NoThrow, Const, Constexpr] in {
   def kxorhi : X86Builtin<"unsigned short(unsigned short, unsigned short)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const] in {
+let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr] in {
   def kxorsi : X86Builtin<"unsigned int(unsigned int, unsigned int)">;
   def kxordi : X86Builtin<"unsigned long long int(unsigned long long int, unsigned long long int)">;
 }

clang/lib/AST/ByteCode/InterpBuiltin.cpp

@@ -666,6 +666,16 @@ static bool interp__builtin_abs(InterpState &S, CodePtr OpPC,
   return true;
 }
 
+static bool interp__builtin_knot(InterpState &S, CodePtr OpPC,
+                                 const InterpFrame *Frame,
+                                 const CallExpr *Call) {
+  APSInt Val =
+      popToAPSInt(S.Stk, *S.getContext().classify(Call->getArg(0)->getType()));
+  APInt Result = ~Val;
+  pushInteger(S, APSInt(std::move(Result), true), Call->getType());
+  return true;
+}
+
 static bool interp__builtin_popcount(InterpState &S, CodePtr OpPC,
                                      const InterpFrame *Frame,
                                      const CallExpr *Call) {
@@ -3607,6 +3617,60 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const CallExpr *Call,
   case X86::BI__builtin_ia32_selectpd_512:
     return interp__builtin_select(S, OpPC, Call);
 
+  case X86::BI__builtin_ia32_kandqi:
+  case X86::BI__builtin_ia32_kandhi:
+  case X86::BI__builtin_ia32_kandsi:
+  case X86::BI__builtin_ia32_kanddi:
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call,
+        [](const APSInt &LHS, const APSInt &RHS) { return LHS & RHS; });
+
+  case X86::BI__builtin_ia32_kandnqi:
+  case X86::BI__builtin_ia32_kandnhi:
+  case X86::BI__builtin_ia32_kandnsi:
+  case X86::BI__builtin_ia32_kandndi:
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call,
+        [](const APSInt &LHS, const APSInt &RHS) { return ~LHS & RHS; });
+
+  case X86::BI__builtin_ia32_korqi:
+  case X86::BI__builtin_ia32_korhi:
+  case X86::BI__builtin_ia32_korsi:
+  case X86::BI__builtin_ia32_kordi:
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call,
+        [](const APSInt &LHS, const APSInt &RHS) { return LHS | RHS; });
+
+  case X86::BI__builtin_ia32_kxnorqi:
+  case X86::BI__builtin_ia32_kxnorhi:
+  case X86::BI__builtin_ia32_kxnorsi:
+  case X86::BI__builtin_ia32_kxnordi:
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call,
+        [](const APSInt &LHS, const APSInt &RHS) { return ~(LHS ^ RHS); });
+
+  case X86::BI__builtin_ia32_kxorqi:
+  case X86::BI__builtin_ia32_kxorhi:
+  case X86::BI__builtin_ia32_kxorsi:
+  case X86::BI__builtin_ia32_kxordi:
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call,
+        [](const APSInt &LHS, const APSInt &RHS) { return LHS ^ RHS; });
+
+  case X86::BI__builtin_ia32_knotqi:
+  case X86::BI__builtin_ia32_knothi:
+  case X86::BI__builtin_ia32_knotsi:
+  case X86::BI__builtin_ia32_knotdi:
+    return interp__builtin_knot(S, OpPC, Frame, Call);
+
+  case X86::BI__builtin_ia32_kaddqi:
+  case X86::BI__builtin_ia32_kaddhi:
+  case X86::BI__builtin_ia32_kaddsi:
+  case X86::BI__builtin_ia32_kadddi:
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call,
+        [](const APSInt &LHS, const APSInt &RHS) { return LHS + RHS; });
+
   case Builtin::BI__builtin_elementwise_fshl:
     return interp__builtin_elementwise_triop(S, OpPC, Call,
                                              llvm::APIntOps::fshl);

clang/lib/AST/ExprConstant.cpp

@@ -13588,6 +13588,20 @@ static bool getBuiltinAlignArguments(const CallExpr *E, EvalInfo &Info,
 
 bool IntExprEvaluator::VisitBuiltinCallExpr(const CallExpr *E,
                                             unsigned BuiltinOp) {
+
+  auto HandleMaskBinOp =
+      [&](llvm::function_ref<APSInt(const APSInt &, const APSInt &)> Fn)
+      -> bool {
+    APValue LHS, RHS;
+    if (!Evaluate(LHS, Info, E->getArg(0)) ||
+        !Evaluate(RHS, Info, E->getArg(1)))
+      return false;
+
+    APSInt ResultInt = Fn(LHS.getInt(), RHS.getInt());
+
+    return Success(APValue(ResultInt), E);
+  };
+
   switch (BuiltinOp) {
   default:
     return false;
@@ -14687,6 +14701,65 @@ bool IntExprEvaluator::VisitBuiltinCallExpr(const CallExpr *E,
         Result.setBitVal(P++, Val[I]);
     return Success(Result, E);
   }
+
+  case X86::BI__builtin_ia32_kandqi:
+  case X86::BI__builtin_ia32_kandhi:
+  case X86::BI__builtin_ia32_kandsi:
+  case X86::BI__builtin_ia32_kanddi: {
+    return HandleMaskBinOp(
+        [](const APSInt &LHS, const APSInt &RHS) { return LHS & RHS; });
+  }
+
+  case X86::BI__builtin_ia32_kandnqi:
+  case X86::BI__builtin_ia32_kandnhi:
+  case X86::BI__builtin_ia32_kandnsi:
+  case X86::BI__builtin_ia32_kandndi: {
+    return HandleMaskBinOp(
+        [](const APSInt &LHS, const APSInt &RHS) { return ~LHS & RHS; });
+  }
+
+  case X86::BI__builtin_ia32_korqi:
+  case X86::BI__builtin_ia32_korhi:
+  case X86::BI__builtin_ia32_korsi:
+  case X86::BI__builtin_ia32_kordi: {
+    return HandleMaskBinOp(
+        [](const APSInt &LHS, const APSInt &RHS) { return LHS | RHS; });
+  }
+
+  case X86::BI__builtin_ia32_kxnorqi:
+  case X86::BI__builtin_ia32_kxnorhi:
+  case X86::BI__builtin_ia32_kxnorsi:
+  case X86::BI__builtin_ia32_kxnordi: {
+    return HandleMaskBinOp(
+        [](const APSInt &LHS, const APSInt &RHS) { return ~(LHS ^ RHS); });
+  }
+
+  case X86::BI__builtin_ia32_kxorqi:
+  case X86::BI__builtin_ia32_kxorhi:
+  case X86::BI__builtin_ia32_kxorsi:
+  case X86::BI__builtin_ia32_kxordi: {
+    return HandleMaskBinOp(
+        [](const APSInt &LHS, const APSInt &RHS) { return LHS ^ RHS; });
+  }
+
+  case X86::BI__builtin_ia32_knotqi:
+  case X86::BI__builtin_ia32_knothi:
+  case X86::BI__builtin_ia32_knotsi:
+  case X86::BI__builtin_ia32_knotdi: {
+    APSInt Val;
+    if (!EvaluateInteger(E->getArg(0), Val, Info))
+      return false;
+    APSInt Result = ~Val;
+    return Success(APValue(Result), E);
+  }
+
+  case X86::BI__builtin_ia32_kaddqi:
+  case X86::BI__builtin_ia32_kaddhi:
+  case X86::BI__builtin_ia32_kaddsi:
+  case X86::BI__builtin_ia32_kadddi: {
+    return HandleMaskBinOp(
+        [](const APSInt &LHS, const APSInt &RHS) { return LHS + RHS; });
+  }
   }
 }
 

clang/lib/Headers/avx512bwintrin.h

@@ -32,68 +32,63 @@ typedef unsigned long long __mmask64;
 #define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS
 #endif
 
-static __inline __mmask32 __DEFAULT_FN_ATTRS
-_knot_mask32(__mmask32 __M)
-{
+static __inline __mmask32
+    __DEFAULT_FN_ATTRS_CONSTEXPR _knot_mask32(__mmask32 __M) {
   return __builtin_ia32_knotsi(__M);
 }
 
-static __inline __mmask64 __DEFAULT_FN_ATTRS _knot_mask64(__mmask64 __M) {
+static __inline __mmask64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_knot_mask64(__mmask64 __M) {
   return __builtin_ia32_knotdi(__M);
 }
 
-static __inline__ __mmask32 __DEFAULT_FN_ATTRS
-_kand_mask32(__mmask32 __A, __mmask32 __B)
-{
+static __inline__ __mmask32 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kand_mask32(__mmask32 __A, __mmask32 __B) {
   return (__mmask32)__builtin_ia32_kandsi((__mmask32)__A, (__mmask32)__B);
 }
 
-static __inline__ __mmask64 __DEFAULT_FN_ATTRS _kand_mask64(__mmask64 __A,
-                                                            __mmask64 __B) {
+static __inline__ __mmask64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kand_mask64(__mmask64 __A, __mmask64 __B) {
   return (__mmask64)__builtin_ia32_kanddi((__mmask64)__A, (__mmask64)__B);
 }
 
-static __inline__ __mmask32 __DEFAULT_FN_ATTRS
-_kandn_mask32(__mmask32 __A, __mmask32 __B)
-{
+static __inline__ __mmask32 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kandn_mask32(__mmask32 __A, __mmask32 __B) {
   return (__mmask32)__builtin_ia32_kandnsi((__mmask32)__A, (__mmask32)__B);
 }
 
-static __inline__ __mmask64 __DEFAULT_FN_ATTRS _kandn_mask64(__mmask64 __A,
-                                                             __mmask64 __B) {
+static __inline__ __mmask64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kandn_mask64(__mmask64 __A, __mmask64 __B) {
   return (__mmask64)__builtin_ia32_kandndi((__mmask64)__A, (__mmask64)__B);
 }
 
-static __inline__ __mmask32 __DEFAULT_FN_ATTRS
-_kor_mask32(__mmask32 __A, __mmask32 __B)
-{
+static __inline__ __mmask32 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kor_mask32(__mmask32 __A, __mmask32 __B) {
   return (__mmask32)__builtin_ia32_korsi((__mmask32)__A, (__mmask32)__B);
 }
 
-static __inline__ __mmask64 __DEFAULT_FN_ATTRS _kor_mask64(__mmask64 __A,
-                                                           __mmask64 __B) {
+static __inline__ __mmask64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kor_mask64(__mmask64 __A, __mmask64 __B) {
   return (__mmask64)__builtin_ia32_kordi((__mmask64)__A, (__mmask64)__B);
 }
 
-static __inline__ __mmask32 __DEFAULT_FN_ATTRS
-_kxnor_mask32(__mmask32 __A, __mmask32 __B)
-{
+static __inline__ __mmask32 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kxnor_mask32(__mmask32 __A, __mmask32 __B) {
   return (__mmask32)__builtin_ia32_kxnorsi((__mmask32)__A, (__mmask32)__B);
 }
 
-static __inline__ __mmask64 __DEFAULT_FN_ATTRS _kxnor_mask64(__mmask64 __A,
-                                                             __mmask64 __B) {
+static __inline__ __mmask64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kxnor_mask64(__mmask64 __A, __mmask64 __B) {
   return (__mmask64)__builtin_ia32_kxnordi((__mmask64)__A, (__mmask64)__B);
 }
 
-static __inline__ __mmask32 __DEFAULT_FN_ATTRS
-_kxor_mask32(__mmask32 __A, __mmask32 __B)
-{
+static __inline__ __mmask32 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kxor_mask32(__mmask32 __A, __mmask32 __B) {
   return (__mmask32)__builtin_ia32_kxorsi((__mmask32)__A, (__mmask32)__B);
 }
 
-static __inline__ __mmask64 __DEFAULT_FN_ATTRS _kxor_mask64(__mmask64 __A,
-                                                            __mmask64 __B) {
+static __inline__ __mmask64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kxor_mask64(__mmask64 __A, __mmask64 __B) {
   return (__mmask64)__builtin_ia32_kxordi((__mmask64)__A, (__mmask64)__B);
 }
 
@@ -165,14 +160,13 @@ _ktest_mask64_u8(__mmask64 __A, __mmask64 __B, unsigned char *__C) {
   return (unsigned char)__builtin_ia32_ktestzdi(__A, __B);
 }
 
-static __inline__ __mmask32 __DEFAULT_FN_ATTRS
-_kadd_mask32(__mmask32 __A, __mmask32 __B)
-{
+static __inline__ __mmask32 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kadd_mask32(__mmask32 __A, __mmask32 __B) {
   return (__mmask32)__builtin_ia32_kaddsi((__mmask32)__A, (__mmask32)__B);
 }
 
-static __inline__ __mmask64 __DEFAULT_FN_ATTRS _kadd_mask64(__mmask64 __A,
-                                                            __mmask64 __B) {
+static __inline__ __mmask64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_kadd_mask64(__mmask64 __A, __mmask64 __B) {
   return (__mmask64)__builtin_ia32_kadddi((__mmask64)__A, (__mmask64)__B);
 }