[HLSL][SPIRV][DXIL] Implement WaveActiveMax intrinsic

clang/include/clang/Basic/Builtins.td

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

clang/lib/CodeGen/CGBuiltin.cpp

@@ -19295,6 +19295,25 @@ static Intrinsic::ID getWaveActiveSumIntrinsic(llvm::Triple::ArchType Arch,
   }
 }
 
+// Return wave active sum that corresponds to the QT scalar type
+static Intrinsic::ID getWaveActiveMaxIntrinsic(llvm::Triple::ArchType Arch,
+                                               CGHLSLRuntime &RT, QualType QT) {
+  switch (Arch) {
+  case llvm::Triple::spirv:
+    if (QT->isUnsignedIntegerType())
+      return llvm::Intrinsic::spv_wave_reduce_umax;
+    return llvm::Intrinsic::spv_wave_reduce_max;
+  case llvm::Triple::dxil: {
+    if (QT->isUnsignedIntegerType())
+      return llvm::Intrinsic::dx_wave_reduce_umax;
+    return llvm::Intrinsic::dx_wave_reduce_max;
+  }
+  default:
+    llvm_unreachable("Intrinsic WaveActiveMax"
+                     " not supported by target architecture");
+  }
+}
+
 Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
                                             const CallExpr *E,
                                             ReturnValueSlot ReturnValue) {
@@ -19624,6 +19643,23 @@ case Builtin::BI__builtin_hlsl_elementwise_isinf: {
                                                      /*AssumeConvergent=*/true),
                            ArrayRef{OpExpr}, "hlsl.wave.active.sum");
   }
+  case Builtin::BI__builtin_hlsl_wave_active_max: {
+    // 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 = getWaveActiveMaxIntrinsic(
+        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.max");
+  }
   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);
 
+//===----------------------------------------------------------------------===//
+// WaveActiveMax builtins
+//===----------------------------------------------------------------------===//
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) half WaveActiveMax(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) half2 WaveActiveMax(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) half3 WaveActiveMax(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) half4 WaveActiveMax(half4);
+
+#ifdef __HLSL_ENABLE_16_BIT
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int16_t WaveActiveMax(int16_t);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int16_t2 WaveActiveMax(int16_t2);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int16_t3 WaveActiveMax(int16_t3);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int16_t4 WaveActiveMax(int16_t4);
+
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint16_t WaveActiveMax(uint16_t);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint16_t2 WaveActiveMax(uint16_t2);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint16_t3 WaveActiveMax(uint16_t3);
+_HLSL_AVAILABILITY(shadermodel, 6.0)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint16_t4 WaveActiveMax(uint16_t4);
+#endif
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int WaveActiveMax(int);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int2 WaveActiveMax(int2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int3 WaveActiveMax(int3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int4 WaveActiveMax(int4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint WaveActiveMax(uint);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint2 WaveActiveMax(uint2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint3 WaveActiveMax(uint3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint4 WaveActiveMax(uint4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int64_t WaveActiveMax(int64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int64_t2 WaveActiveMax(int64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int64_t3 WaveActiveMax(int64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) int64_t4 WaveActiveMax(int64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint64_t WaveActiveMax(uint64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint64_t2 WaveActiveMax(uint64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint64_t3 WaveActiveMax(uint64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) uint64_t4 WaveActiveMax(uint64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) float WaveActiveMax(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) float2 WaveActiveMax(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) float3 WaveActiveMax(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) float4 WaveActiveMax(float4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) double WaveActiveMax(double);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) double2 WaveActiveMax(double2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) double3 WaveActiveMax(double3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_wave_active_max)
+__attribute__((convergent)) double4 WaveActiveMax(double4);
+
 //===----------------------------------------------------------------------===//
 // WaveActiveSum builtins
 //===----------------------------------------------------------------------===//

clang/lib/Sema/SemaHLSL.cpp

@@ -2430,6 +2430,7 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
     TheCall->setType(ArgTyA);
     break;
   }
+  case Builtin::BI__builtin_hlsl_wave_active_max:
   case Builtin::BI__builtin_hlsl_wave_active_sum: {
     if (SemaRef.checkArgCount(TheCall, 1))
       return true;

clang/test/CodeGenHLSL/builtins/WaveActiveMax.hlsl

@@ -0,0 +1,46 @@
+// 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.max.i32([[TY]] %[[#]])
+  // CHECK-DXIL:  %[[RET:.*]] = call [[TY:.*]] @llvm.dx.wave.reduce.max.i32([[TY]] %[[#]])
+  // CHECK:  ret [[TY]] %[[RET]]
+  return WaveActiveMax(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.wave.reduce.max.i32([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare spir_func [[TY]] @llvm.spv.wave.reduce.max.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.umax.i64([[TY]] %[[#]])
+  // CHECK-DXIL:  %[[RET:.*]] = call [[TY:.*]] @llvm.dx.wave.reduce.umax.i64([[TY]] %[[#]])
+  // CHECK:  ret [[TY]] %[[RET]]
+  return WaveActiveMax(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.wave.reduce.umax.i64([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare spir_func [[TY]] @llvm.spv.wave.reduce.umax.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.max.v4f32([[TY1]] %[[#]]
+  // CHECK-DXIL:  %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn [[TY1:.*]] @llvm.dx.wave.reduce.max.v4f32([[TY1]] %[[#]])
+  // CHECK:  ret [[TY1]] %[[RET1]]
+  return WaveActiveMax(expr);
+}
+
+// CHECK-DXIL: declare [[TY1]] @llvm.dx.wave.reduce.max.v4f32([[TY1]]) #[[#attr]]
+// CHECK-SPIRV: declare spir_func [[TY1]] @llvm.spv.wave.reduce.max.v4f32([[TY1]]) #[[#attr]]
+
+// CHECK: attributes #[[#attr]] = {{{.*}} convergent {{.*}}}
+

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

@@ -0,0 +1,29 @@
+// 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_max();
+  // 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_max(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_max(p0);
+  // expected-error@-1 {{invalid operand of type 'bool'}}
+}
+
+bool2 test_expr_bool_vec_type_check(bool2 p0) {
+  return __builtin_hlsl_wave_active_max(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_max(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_max : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+def int_dx_wave_reduce_umax : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [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]>;

llvm/include/llvm/IR/IntrinsicsSPIRV.td

@@ -91,6 +91,8 @@ 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_umax : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+  def int_spv_wave_reduce_max : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [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]>;

llvm/lib/Target/DirectX/DXIL.td

@@ -1000,6 +1000,16 @@ def WaveActiveOp : DXILOp<119, waveActiveOp> {
                    IntrinArgIndex<0>, IntrinArgI8<WaveOpKind_Sum>,
                    IntrinArgI8<SignedOpKind_Unsigned>
                  ]>,
+    IntrinSelect<int_dx_wave_reduce_max,
+                 [
+                   IntrinArgIndex<0>, IntrinArgI8<WaveOpKind_Max>,
+                   IntrinArgI8<SignedOpKind_Signed>
+                 ]>,
+    IntrinSelect<int_dx_wave_reduce_umax,
+                 [
+                   IntrinArgIndex<0>, IntrinArgI8<WaveOpKind_Max>,
+                   IntrinArgI8<SignedOpKind_Unsigned>
+                 ]>,
   ];
 
   let arguments = [OverloadTy, Int8Ty, Int8Ty];
@@ -1008,7 +1018,7 @@ def WaveActiveOp : DXILOp<119, waveActiveOp> {
     Overloads<DXIL1_0, [HalfTy, FloatTy, DoubleTy, Int16Ty, Int32Ty, Int64Ty]>
   ];
   let stages = [Stages<DXIL1_0, [all_stages]>];
-  let attributes = [Attributes<DXIL1_0, [ReadNone]>];
+  let attributes = [Attributes<DXIL1_0, []>];
 }
 
 def WaveAllBitCount : DXILOp<135, waveAllOp> {

llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp

@@ -40,6 +40,8 @@ bool DirectXTTIImpl::isTargetIntrinsicTriviallyScalarizable(
   switch (ID) {
   case Intrinsic::dx_frac:
   case Intrinsic::dx_rsqrt:
+  case Intrinsic::dx_wave_reduce_max:
+  case Intrinsic::dx_wave_reduce_umax:
   case Intrinsic::dx_wave_reduce_sum:
   case Intrinsic::dx_wave_reduce_usum:
   case Intrinsic::dx_wave_readlane:

llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp

@@ -215,6 +215,9 @@ class SPIRVInstructionSelector : public InstructionSelector {
   bool selectDot4AddPackedExpansion(Register ResVReg, const SPIRVType *ResType,
                                     MachineInstr &I) const;
 
+  bool selectWaveReduceMax(Register ResVReg, const SPIRVType *ResType,
+                           MachineInstr &I, bool IsUnsigned) const;
+
   bool selectWaveReduceSum(Register ResVReg, const SPIRVType *ResType,
                            MachineInstr &I) const;
 
@@ -2132,6 +2135,34 @@ bool SPIRVInstructionSelector::selectWaveActiveCountBits(
   return Result;
 }
 
+bool SPIRVInstructionSelector::selectWaveReduceMax(Register ResVReg,
+                                                   const SPIRVType *ResType,
+                                                   MachineInstr &I,
+                                                   bool IsUnsigned) const {
+  assert(I.getNumOperands() == 3);
+  assert(I.getOperand(2).isReg());
+  MachineBasicBlock &BB = *I.getParent();
+  Register InputRegister = I.getOperand(2).getReg();
+  SPIRVType *InputType = GR.getSPIRVTypeForVReg(InputRegister);
+
+  if (!InputType)
+    report_fatal_error("Input Type could not be determined.");
+
+  SPIRVType *IntTy = GR.getOrCreateSPIRVIntegerType(32, I, TII);
+  // Retreive the operation to use based on input type
+  bool IsFloatTy = GR.isScalarOrVectorOfType(InputRegister, SPIRV::OpTypeFloat);
+  auto IntegerOpcodeType =
+      IsUnsigned ? SPIRV::OpGroupNonUniformUMax : SPIRV::OpGroupNonUniformSMax;
+  auto Opcode = IsFloatTy ? SPIRV::OpGroupNonUniformFMax : IntegerOpcodeType;
+  return BuildMI(BB, I, I.getDebugLoc(), TII.get(Opcode))
+      .addDef(ResVReg)
+      .addUse(GR.getSPIRVTypeID(ResType))
+      .addUse(GR.getOrCreateConstInt(SPIRV::Scope::Subgroup, I, IntTy, TII))
+      .addImm(SPIRV::GroupOperation::Reduce)
+      .addUse(I.getOperand(2).getReg())
+      .constrainAllUses(TII, TRI, RBI);
+}
+
 bool SPIRVInstructionSelector::selectWaveReduceSum(Register ResVReg,
                                                    const SPIRVType *ResType,
                                                    MachineInstr &I) const {
@@ -3086,6 +3117,10 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
     return selectWaveOpInst(ResVReg, ResType, I, SPIRV::OpGroupNonUniformAny);
   case Intrinsic::spv_wave_is_first_lane:
     return selectWaveOpInst(ResVReg, ResType, I, SPIRV::OpGroupNonUniformElect);
+  case Intrinsic::spv_wave_reduce_umax:
+    return selectWaveReduceMax(ResVReg, ResType, I, /*IsUnsigned*/ true);
+  case Intrinsic::spv_wave_reduce_max:
+    return selectWaveReduceMax(ResVReg, ResType, I, /*IsUnsigned*/ false);
   case Intrinsic::spv_wave_reduce_sum:
     return selectWaveReduceSum(ResVReg, ResType, I);
   case Intrinsic::spv_wave_readlane: