[HLSL][SPIRV][DXIL] Implement WaveActiveSum intrinsic (#118580)

```    - add clang builtin to Builtins.td
      - link builtin in hlsl_intrinsics
      - add codegen for spirv intrinsic and two directx intrinsics to retain
        signedness information of the operands in CGBuiltin.cpp
      - add semantic analysis in SemaHLSL.cpp
      - add lowering of spirv intrinsic to spirv backend in
        SPIRVInstructionSelector.cpp
      - add lowering of directx intrinsics to WaveActiveOp dxil op in
    DXIL.td

      - add test cases to illustrate passespendent pr merges.
```
Resolves #70106

---------

Co-authored-by: Finn Plummer <[email protected]>

Author: adam-yang

clang/include/clang/Basic/Builtins.td

@@ -4795,6 +4795,12 @@ def HLSLWaveActiveCountBits : LangBuiltin<"HLSL_LANG"> {
   let Prototype = "unsigned int(bool)";
 }
 
+def HLSLWaveActiveSum : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_wave_active_sum"];
+  let Attributes = [NoThrow, Const];
+  let Prototype = "void (...)";
+}
+
 def HLSLWaveGetLaneIndex : LangBuiltin<"HLSL_LANG"> {
   let Spellings = ["__builtin_hlsl_wave_get_lane_index"];
   let Attributes = [NoThrow, Const];

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -9303,7 +9303,7 @@ def err_typecheck_expect_scalar_or_vector : Error<
   "invalid operand of type %0 where %1 or "
   "a vector of such type is required">;
 def err_typecheck_expect_any_scalar_or_vector : Error<
-  "invalid operand of type %0 where a scalar or vector is required">;
+  "invalid operand of type %0%select{| where a scalar or vector is required}1">;
 def err_typecheck_expect_flt_or_vector : Error<
   "invalid operand of type %0 where floating, complex or "
   "a vector of such types is required">;

clang/lib/CodeGen/CGBuiltin.cpp

@@ -19186,6 +19186,23 @@ static Intrinsic::ID getFirstBitHighIntrinsic(CGHLSLRuntime &RT, QualType QT) {
   return RT.getFirstBitUHighIntrinsic();
 }
 
+// Return wave active sum that corresponds to the QT scalar type
+static Intrinsic::ID getWaveActiveSumIntrinsic(llvm::Triple::ArchType Arch,
+                                               CGHLSLRuntime &RT, QualType QT) {
+  switch (Arch) {
+  case llvm::Triple::spirv:
+    return llvm::Intrinsic::spv_wave_reduce_sum;
+  case llvm::Triple::dxil: {
+    if (QT->isUnsignedIntegerType())
+      return llvm::Intrinsic::dx_wave_reduce_usum;
+    return llvm::Intrinsic::dx_wave_reduce_sum;
+  }
+  default:
+    llvm_unreachable("Intrinsic WaveActiveSum"
+                     " not supported by target architecture");
+  }
+}
+
 Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
                                             const CallExpr *E,
                                             ReturnValueSlot ReturnValue) {
@@ -19498,6 +19515,23 @@ case Builtin::BI__builtin_hlsl_elementwise_isinf: {
         Intrinsic::getOrInsertDeclaration(&CGM.getModule(), ID),
         ArrayRef{OpExpr});
   }
+  case Builtin::BI__builtin_hlsl_wave_active_sum: {
+    // Due to the use of variadic arguments, explicitly retreive argument
+    Value *OpExpr = EmitScalarExpr(E->getArg(0));
+    llvm::FunctionType *FT = llvm::FunctionType::get(
+        OpExpr->getType(), ArrayRef{OpExpr->getType()}, false);
+    Intrinsic::ID IID = getWaveActiveSumIntrinsic(
+        getTarget().getTriple().getArch(), CGM.getHLSLRuntime(),
+        E->getArg(0)->getType());
+
+    // Get overloaded name
+    std::string Name =
+        Intrinsic::getName(IID, ArrayRef{OpExpr->getType()}, &CGM.getModule());
+    return EmitRuntimeCall(CGM.CreateRuntimeFunction(FT, Name, {},
+                                                     /*Local=*/false,
+                                                     /*AssumeConvergent=*/true),
+                           ArrayRef{OpExpr}, "hlsl.wave.active.sum");
+  }
   case Builtin::BI__builtin_hlsl_wave_get_lane_index: {
     // We don't define a SPIR-V intrinsic, instead it is a SPIR-V built-in
     // defined in SPIRVBuiltins.td. So instead we manually get the matching name

clang/lib/Headers/hlsl/hlsl_intrinsics.h

@@ -2468,6 +2468,105 @@ __attribute__((convergent)) double3 WaveReadLaneAt(double3, int32_t);
 _HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_read_lane_at)
 __attribute__((convergent)) double4 WaveReadLaneAt(double4, int32_t);
 
+//===----------------------------------------------------------------------===//
+// WaveActiveSum builtins
+//===----------------------------------------------------------------------===//
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) half WaveActiveSum(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) half2 WaveActiveSum(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) half3 WaveActiveSum(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) half4 WaveActiveSum(half4);
+
+#ifdef __HLSL_ENABLE_16_BIT
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int16_t WaveActiveSum(int16_t);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int16_t2 WaveActiveSum(int16_t2);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int16_t3 WaveActiveSum(int16_t3);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int16_t4 WaveActiveSum(int16_t4);
+
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint16_t WaveActiveSum(uint16_t);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint16_t2 WaveActiveSum(uint16_t2);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint16_t3 WaveActiveSum(uint16_t3);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint16_t4 WaveActiveSum(uint16_t4);
+#endif
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int WaveActiveSum(int);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int2 WaveActiveSum(int2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int3 WaveActiveSum(int3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int4 WaveActiveSum(int4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint WaveActiveSum(uint);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint2 WaveActiveSum(uint2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint3 WaveActiveSum(uint3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint4 WaveActiveSum(uint4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int64_t WaveActiveSum(int64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int64_t2 WaveActiveSum(int64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int64_t3 WaveActiveSum(int64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) int64_t4 WaveActiveSum(int64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint64_t WaveActiveSum(uint64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint64_t2 WaveActiveSum(uint64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint64_t3 WaveActiveSum(uint64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) uint64_t4 WaveActiveSum(uint64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) float WaveActiveSum(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) float2 WaveActiveSum(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) float3 WaveActiveSum(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) float4 WaveActiveSum(float4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) double WaveActiveSum(double);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) double2 WaveActiveSum(double2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) double3 WaveActiveSum(double3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_sum)
+__attribute__((convergent)) double4 WaveActiveSum(double4);
+
 //===----------------------------------------------------------------------===//
 // sign builtins
 //===----------------------------------------------------------------------===//

clang/lib/Sema/SemaHLSL.cpp

@@ -1859,7 +1859,24 @@ static bool CheckAnyScalarOrVector(Sema *S, CallExpr *TheCall,
         (VTy && VTy->getElementType()->isScalarType()))) {
     S->Diag(TheCall->getArg(0)->getBeginLoc(),
             diag::err_typecheck_expect_any_scalar_or_vector)
-        << ArgType;
+        << ArgType << 1;
+    return true;
+  }
+  return false;
+}
+
+static bool CheckWaveActive(Sema *S, CallExpr *TheCall) {
+  QualType BoolType = S->getASTContext().BoolTy;
+  assert(TheCall->getNumArgs() >= 1);
+  QualType ArgType = TheCall->getArg(0)->getType();
+  auto *VTy = ArgType->getAs<VectorType>();
+  // is the bool or vector<bool>
+  if (S->Context.hasSameUnqualifiedType(ArgType, BoolType) ||
+      (VTy &&
+       S->Context.hasSameUnqualifiedType(VTy->getElementType(), BoolType))) {
+    S->Diag(TheCall->getArg(0)->getBeginLoc(),
+            diag::err_typecheck_expect_any_scalar_or_vector)
+        << ArgType << 0;
     return true;
   }
   return false;
@@ -2156,6 +2173,20 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
     TheCall->setType(ArgTyA);
     break;
   }
+  case Builtin::BI__builtin_hlsl_wave_active_sum: {
+    if (SemaRef.checkArgCount(TheCall, 1))
+      return true;
+
+    // Ensure input expr type is a scalar/vector and the same as the return type
+    if (CheckAnyScalarOrVector(&SemaRef, TheCall, 0))
+      return true;
+    if (CheckWaveActive(&SemaRef, TheCall))
+      return true;
+    ExprResult Expr = TheCall->getArg(0);
+    QualType ArgTyExpr = Expr.get()->getType();
+    TheCall->setType(ArgTyExpr);
+    break;
+  }
   // Note these are llvm builtins that we want to catch invalid intrinsic
   // generation. Normal handling of these builitns will occur elsewhere.
   case Builtin::BI__builtin_elementwise_bitreverse: {

clang/test/CodeGenHLSL/builtins/WaveActiveSum.hlsl

@@ -0,0 +1,45 @@
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN:   dxil-pc-shadermodel6.3-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN:   FileCheck %s --check-prefixes=CHECK,CHECK-DXIL
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN:   spirv-pc-vulkan-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN:   FileCheck %s --check-prefixes=CHECK,CHECK-SPIRV
+
+// Test basic lowering to runtime function call.
+
+// CHECK-LABEL: test_int
+int test_int(int expr) {
+  // CHECK-SPIRV:  %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.wave.reduce.sum.i32([[TY]] %[[#]])
+  // CHECK-DXIL:  %[[RET:.*]] = call [[TY:.*]] @llvm.dx.wave.reduce.sum.i32([[TY]] %[[#]])
+  // CHECK:  ret [[TY]] %[[RET]]
+  return WaveActiveSum(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.wave.reduce.sum.i32([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare spir_func [[TY]] @llvm.spv.wave.reduce.sum.i32([[TY]]) #[[#attr:]]
+
+// CHECK-LABEL: test_uint64_t
+uint64_t test_uint64_t(uint64_t expr) {
+  // CHECK-SPIRV:  %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.wave.reduce.sum.i64([[TY]] %[[#]])
+  // CHECK-DXIL:  %[[RET:.*]] = call [[TY:.*]] @llvm.dx.wave.reduce.usum.i64([[TY]] %[[#]])
+  // CHECK:  ret [[TY]] %[[RET]]
+  return WaveActiveSum(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.wave.reduce.usum.i64([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare spir_func [[TY]] @llvm.spv.wave.reduce.sum.i64([[TY]]) #[[#attr:]]
+
+// Test basic lowering to runtime function call with array and float value.
+
+// CHECK-LABEL: test_floatv4
+float4 test_floatv4(float4 expr) {
+  // CHECK-SPIRV:  %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn spir_func [[TY1:.*]] @llvm.spv.wave.reduce.sum.v4f32([[TY1]] %[[#]]
+  // CHECK-DXIL:  %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn [[TY1:.*]] @llvm.dx.wave.reduce.sum.v4f32([[TY1]] %[[#]])
+  // CHECK:  ret [[TY1]] %[[RET1]]
+  return WaveActiveSum(expr);
+}
+
+// CHECK-DXIL: declare [[TY1]] @llvm.dx.wave.reduce.sum.v4f32([[TY1]]) #[[#attr]]
+// CHECK-SPIRV: declare spir_func [[TY1]] @llvm.spv.wave.reduce.sum.v4f32([[TY1]]) #[[#attr]]
+
+// CHECK: attributes #[[#attr]] = {{{.*}} convergent {{.*}}}

clang/test/SemaHLSL/BuiltIns/WaveActiveSum-errors.hlsl

@@ -0,0 +1,28 @@
+// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.6-library %s -emit-llvm-only -disable-llvm-passes -verify
+
+int test_too_few_arg() {
+  return __builtin_hlsl_wave_active_sum();
+  // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float2 test_too_many_arg(float2 p0) {
+  return __builtin_hlsl_wave_active_sum(p0, p0);
+  // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+bool test_expr_bool_type_check(bool p0) {
+  return __builtin_hlsl_wave_active_sum(p0);
+  // expected-error@-1 {{invalid operand of type 'bool'}}
+}
+
+bool2 test_expr_bool_vec_type_check(bool2 p0) {
+  return __builtin_hlsl_wave_active_sum(p0);
+  // expected-error@-1 {{invalid operand of type 'bool2' (aka 'vector<bool, 2>')}}
+}
+
+struct S { float f; };
+
+S test_expr_struct_type_check(S p0) {
+  return __builtin_hlsl_wave_active_sum(p0);
+  // expected-error@-1 {{invalid operand of type 'S' where a scalar or vector is required}}
+}

llvm/include/llvm/IR/IntrinsicsDirectX.td

@@ -105,6 +105,8 @@ def int_dx_wave_active_countbits : DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i1
 def int_dx_wave_all : DefaultAttrsIntrinsic<[llvm_i1_ty], [llvm_i1_ty], [IntrConvergent, IntrNoMem]>;
 def int_dx_wave_any : DefaultAttrsIntrinsic<[llvm_i1_ty], [llvm_i1_ty], [IntrConvergent, IntrNoMem]>;
 def int_dx_wave_getlaneindex : DefaultAttrsIntrinsic<[llvm_i32_ty], [], [IntrConvergent, IntrNoMem]>;
+def int_dx_wave_reduce_sum : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+def int_dx_wave_reduce_usum : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
 def int_dx_wave_is_first_lane : DefaultAttrsIntrinsic<[llvm_i1_ty], [], [IntrConvergent]>;
 def int_dx_wave_readlane : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>, llvm_i32_ty], [IntrConvergent, IntrNoMem]>;
 def int_dx_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;

llvm/include/llvm/IR/IntrinsicsSPIRV.td

@@ -91,6 +91,7 @@ let TargetPrefix = "spv" in {
   def int_spv_wave_active_countbits : DefaultAttrsIntrinsic<[llvm_i32_ty], [llvm_i1_ty], [IntrConvergent, IntrNoMem]>;
   def int_spv_wave_all : DefaultAttrsIntrinsic<[llvm_i1_ty], [llvm_i1_ty], [IntrConvergent, IntrNoMem]>;
   def int_spv_wave_any : DefaultAttrsIntrinsic<[llvm_i1_ty], [llvm_i1_ty], [IntrConvergent, IntrNoMem]>;
+  def int_spv_wave_reduce_sum : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
   def int_spv_wave_is_first_lane : DefaultAttrsIntrinsic<[llvm_i1_ty], [], [IntrConvergent]>;
   def int_spv_wave_readlane : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>, llvm_i32_ty], [IntrConvergent, IntrNoMem]>;
   def int_spv_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;

llvm/lib/Target/DirectX/DXIL.td

@@ -303,6 +303,14 @@ defvar BarrierMode_GroupMemoryBarrierWithGroupSync  = 9;
 defvar BarrierMode_AllMemoryBarrier                 = 10;
 defvar BarrierMode_AllMemoryBarrierWithGroupSync    = 11;
 
+defvar WaveOpKind_Sum     = 0;
+defvar WaveOpKind_Product = 1;
+defvar WaveOpKind_Min     = 2;
+defvar WaveOpKind_Max     = 3;
+
+defvar SignedOpKind_Signed   = 0;
+defvar SignedOpKind_Unsigned = 1;
+
 // Intrinsic arg selection
 class IntrinArgSelectType;
 def IntrinArgSelect_Index : IntrinArgSelectType;
@@ -340,7 +348,7 @@ class IntrinArgI32  <int value> : IntrinArgSelect<IntrinArgSelect_I32,   value>;
 //     IntrinSelect<int_dx_wave_active_usum,
 //       [ IntrinArgIndex<0>, IntrinArgI8<0>, IntrinArgI8<1> ]
 //     >,
-//     IntrinSelect<int_dx_wave_active_sum,
+//     IntrinSelect<int_dx_wave_reduce_sum,
 //       [ IntrinArgIndex<0>, IntrinArgI8<0>, IntrinArgI8<0> ]
 //     >,
 //   ]
@@ -991,6 +999,24 @@ def WaveActiveAnyTrue : DXILOp<113, waveAnyTrue> {
   let stages = [Stages<DXIL1_0, [all_stages]>];
 }
 
+def WaveActiveOp : DXILOp<119, waveActiveOp> {
+  let Doc = "returns the result of the operation across waves";
+  let intrinsics = [
+    IntrinSelect<
+        int_dx_wave_reduce_sum,
+        [ IntrinArgIndex<0>, IntrinArgI8<WaveOpKind_Sum>, IntrinArgI8<SignedOpKind_Signed> ]>,
+    IntrinSelect<
+        int_dx_wave_reduce_usum,
+        [ IntrinArgIndex<0>, IntrinArgI8<WaveOpKind_Sum>, IntrinArgI8<SignedOpKind_Unsigned> ]>,
+  ];
+
+  let arguments = [OverloadTy, Int8Ty, Int8Ty];
+  let result = OverloadTy;
+  let overloads = [Overloads<DXIL1_0, [HalfTy, FloatTy, DoubleTy, Int16Ty, Int32Ty, Int64Ty]>];
+  let stages = [Stages<DXIL1_0, [all_stages]>];
+  let attributes = [Attributes<DXIL1_0, [ReadNone]>];
+}
+
 def WaveIsFirstLane :  DXILOp<110, waveIsFirstLane> {
   let Doc = "returns 1 for the first lane in the wave";
   let intrinsics = [ IntrinSelect<int_dx_wave_is_first_lane> ];