[NFC][HLSL] Move emitter out of AMDGPU.cpp

[farzonl]

• Move all HLSL code out of AMDGPU.cpp to CGHLSLBuiltins.cpp
• Fixes accidental reorganization of HLSL code into AMDGPU caused by ([NFC][clang] Split clang/lib/CodeGen/CGBuiltin.cpp into target-specific files #132252, 7f920e2e5f70b)

[llvmbot]

@llvm/pr-subscribers-clang
@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-hlsl

Author: Farzon Lotfi (farzonl)

Changes

• Create a home for LangBuiltins
• Move all HLSL code out of AMDGPU.cpp to CGHLSLBuiltins.cpp

---

Patch is 63.61 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/133251.diff

4 Files Affected:

• (modified) .github/new-prs-labeler.yml (+1)
• (modified) clang/lib/CodeGen/CMakeLists.txt (+1)
• (added) clang/lib/CodeGen/LangBuiltins/CGHLSLBuiltins.cpp (+740)
• (modified) clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp (-724)

diff --git a/.github/new-prs-labeler.yml b/.github/new-prs-labeler.yml
index c375fa5dc7516..50a01a8e2908f 100644
--- a/.github/new-prs-labeler.yml
+++ b/.github/new-prs-labeler.yml
@@ -637,6 +637,7 @@ hlsl:
   - clang/lib/Sema/HLSLExternalSemaSource.cpp
   - clang/lib/Sema/SemaHLSL.cpp
   - clang/lib/CodeGen/CGHLSLRuntime.*
+  - clang/lib/CodeGen/LangBuiltins/CGHLSLBuiltins.cpp
   - llvm/include/llvm/Frontend/HLSL/**
   - llvm/lib/Frontend/HLSL/**
 
diff --git a/clang/lib/CodeGen/CMakeLists.txt b/clang/lib/CodeGen/CMakeLists.txt
index 94a908197d795..10407a33932b1 100644
--- a/clang/lib/CodeGen/CMakeLists.txt
+++ b/clang/lib/CodeGen/CMakeLists.txt
@@ -115,6 +115,7 @@ add_clang_library(clangCodeGen
   PatternInit.cpp
   SanitizerMetadata.cpp
   SwiftCallingConv.cpp
+  LangBuiltins/CGHLSLBuiltins.cpp
   TargetBuiltins/ARM.cpp
   TargetBuiltins/AMDGPU.cpp
   TargetBuiltins/Hexagon.cpp
diff --git a/clang/lib/CodeGen/LangBuiltins/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/LangBuiltins/CGHLSLBuiltins.cpp
new file mode 100644
index 0000000000000..5709594a34826
--- /dev/null
+++ b/clang/lib/CodeGen/LangBuiltins/CGHLSLBuiltins.cpp
@@ -0,0 +1,740 @@
+//===------- CGHLSLBuiltins.cpp - Emit LLVM Code for HLSL builtins --------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This contains code to emit HLSL Builtin calls as LLVM code.
+//
+//===----------------------------------------------------------------------===//
+
+#include "CGBuiltin.h"
+#include "CGHLSLRuntime.h"
+
+using namespace clang;
+using namespace CodeGen;
+using namespace llvm;
+
+static Value *handleAsDoubleBuiltin(CodeGenFunction &CGF, const CallExpr *E) {
+  assert((E->getArg(0)->getType()->hasUnsignedIntegerRepresentation() &&
+          E->getArg(1)->getType()->hasUnsignedIntegerRepresentation()) &&
+         "asdouble operands types mismatch");
+  Value *OpLowBits = CGF.EmitScalarExpr(E->getArg(0));
+  Value *OpHighBits = CGF.EmitScalarExpr(E->getArg(1));
+
+  llvm::Type *ResultType = CGF.DoubleTy;
+  int N = 1;
+  if (auto *VTy = E->getArg(0)->getType()->getAs<clang::VectorType>()) {
+    N = VTy->getNumElements();
+    ResultType = llvm::FixedVectorType::get(CGF.DoubleTy, N);
+  }
+
+  if (CGF.CGM.getTarget().getTriple().isDXIL())
+    return CGF.Builder.CreateIntrinsic(
+        /*ReturnType=*/ResultType, Intrinsic::dx_asdouble,
+        {OpLowBits, OpHighBits}, nullptr, "hlsl.asdouble");
+
+  if (!E->getArg(0)->getType()->isVectorType()) {
+    OpLowBits = CGF.Builder.CreateVectorSplat(1, OpLowBits);
+    OpHighBits = CGF.Builder.CreateVectorSplat(1, OpHighBits);
+  }
+
+  llvm::SmallVector<int> Mask;
+  for (int i = 0; i < N; i++) {
+    Mask.push_back(i);
+    Mask.push_back(i + N);
+  }
+
+  Value *BitVec = CGF.Builder.CreateShuffleVector(OpLowBits, OpHighBits, Mask);
+
+  return CGF.Builder.CreateBitCast(BitVec, ResultType);
+}
+
+static Value *handleHlslClip(const CallExpr *E, CodeGenFunction *CGF) {
+  Value *Op0 = CGF->EmitScalarExpr(E->getArg(0));
+
+  Constant *FZeroConst = ConstantFP::getZero(CGF->FloatTy);
+  Value *CMP;
+  Value *LastInstr;
+
+  if (const auto *VecTy = E->getArg(0)->getType()->getAs<clang::VectorType>()) {
+    FZeroConst = ConstantVector::getSplat(
+        ElementCount::getFixed(VecTy->getNumElements()), FZeroConst);
+    auto *FCompInst = CGF->Builder.CreateFCmpOLT(Op0, FZeroConst);
+    CMP = CGF->Builder.CreateIntrinsic(
+        CGF->Builder.getInt1Ty(), CGF->CGM.getHLSLRuntime().getAnyIntrinsic(),
+        {FCompInst});
+  } else
+    CMP = CGF->Builder.CreateFCmpOLT(Op0, FZeroConst);
+
+  if (CGF->CGM.getTarget().getTriple().isDXIL())
+    LastInstr =
+        CGF->Builder.CreateIntrinsic(CGF->VoidTy, Intrinsic::dx_discard, {CMP});
+  else if (CGF->CGM.getTarget().getTriple().isSPIRV()) {
+    BasicBlock *LT0 = CGF->createBasicBlock("lt0", CGF->CurFn);
+    BasicBlock *End = CGF->createBasicBlock("end", CGF->CurFn);
+
+    CGF->Builder.CreateCondBr(CMP, LT0, End);
+
+    CGF->Builder.SetInsertPoint(LT0);
+
+    CGF->Builder.CreateIntrinsic(CGF->VoidTy, Intrinsic::spv_discard, {});
+
+    LastInstr = CGF->Builder.CreateBr(End);
+    CGF->Builder.SetInsertPoint(End);
+  } else {
+    llvm_unreachable("Backend Codegen not supported.");
+  }
+
+  return LastInstr;
+}
+
+static Value *handleHlslSplitdouble(const CallExpr *E, CodeGenFunction *CGF) {
+  Value *Op0 = CGF->EmitScalarExpr(E->getArg(0));
+  const auto *OutArg1 = dyn_cast<HLSLOutArgExpr>(E->getArg(1));
+  const auto *OutArg2 = dyn_cast<HLSLOutArgExpr>(E->getArg(2));
+
+  CallArgList Args;
+  LValue Op1TmpLValue =
+      CGF->EmitHLSLOutArgExpr(OutArg1, Args, OutArg1->getType());
+  LValue Op2TmpLValue =
+      CGF->EmitHLSLOutArgExpr(OutArg2, Args, OutArg2->getType());
+
+  if (CGF->getTarget().getCXXABI().areArgsDestroyedLeftToRightInCallee())
+    Args.reverseWritebacks();
+
+  Value *LowBits = nullptr;
+  Value *HighBits = nullptr;
+
+  if (CGF->CGM.getTarget().getTriple().isDXIL()) {
+
+    llvm::Type *RetElementTy = CGF->Int32Ty;
+    if (auto *Op0VecTy = E->getArg(0)->getType()->getAs<clang::VectorType>())
+      RetElementTy = llvm::VectorType::get(
+          CGF->Int32Ty, ElementCount::getFixed(Op0VecTy->getNumElements()));
+    auto *RetTy = llvm::StructType::get(RetElementTy, RetElementTy);
+
+    CallInst *CI = CGF->Builder.CreateIntrinsic(
+        RetTy, Intrinsic::dx_splitdouble, {Op0}, nullptr, "hlsl.splitdouble");
+
+    LowBits = CGF->Builder.CreateExtractValue(CI, 0);
+    HighBits = CGF->Builder.CreateExtractValue(CI, 1);
+
+  } else {
+    // For Non DXIL targets we generate the instructions.
+
+    if (!Op0->getType()->isVectorTy()) {
+      FixedVectorType *DestTy = FixedVectorType::get(CGF->Int32Ty, 2);
+      Value *Bitcast = CGF->Builder.CreateBitCast(Op0, DestTy);
+
+      LowBits = CGF->Builder.CreateExtractElement(Bitcast, (uint64_t)0);
+      HighBits = CGF->Builder.CreateExtractElement(Bitcast, 1);
+    } else {
+      int NumElements = 1;
+      if (const auto *VecTy =
+              E->getArg(0)->getType()->getAs<clang::VectorType>())
+        NumElements = VecTy->getNumElements();
+
+      FixedVectorType *Uint32VecTy =
+          FixedVectorType::get(CGF->Int32Ty, NumElements * 2);
+      Value *Uint32Vec = CGF->Builder.CreateBitCast(Op0, Uint32VecTy);
+      if (NumElements == 1) {
+        LowBits = CGF->Builder.CreateExtractElement(Uint32Vec, (uint64_t)0);
+        HighBits = CGF->Builder.CreateExtractElement(Uint32Vec, 1);
+      } else {
+        SmallVector<int> EvenMask, OddMask;
+        for (int I = 0, E = NumElements; I != E; ++I) {
+          EvenMask.push_back(I * 2);
+          OddMask.push_back(I * 2 + 1);
+        }
+        LowBits = CGF->Builder.CreateShuffleVector(Uint32Vec, EvenMask);
+        HighBits = CGF->Builder.CreateShuffleVector(Uint32Vec, OddMask);
+      }
+    }
+  }
+  CGF->Builder.CreateStore(LowBits, Op1TmpLValue.getAddress());
+  auto *LastInst =
+      CGF->Builder.CreateStore(HighBits, Op2TmpLValue.getAddress());
+  CGF->EmitWritebacks(Args);
+  return LastInst;
+}
+
+// Return dot product intrinsic that corresponds to the QT scalar type
+static Intrinsic::ID getDotProductIntrinsic(CGHLSLRuntime &RT, QualType QT) {
+  if (QT->isFloatingType())
+    return RT.getFDotIntrinsic();
+  if (QT->isSignedIntegerType())
+    return RT.getSDotIntrinsic();
+  assert(QT->isUnsignedIntegerType());
+  return RT.getUDotIntrinsic();
+}
+
+static Intrinsic::ID getFirstBitHighIntrinsic(CGHLSLRuntime &RT, QualType QT) {
+  if (QT->hasSignedIntegerRepresentation()) {
+    return RT.getFirstBitSHighIntrinsic();
+  }
+
+  assert(QT->hasUnsignedIntegerRepresentation());
+  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 Intrinsic::spv_wave_reduce_sum;
+  case llvm::Triple::dxil: {
+    if (QT->isUnsignedIntegerType())
+      return Intrinsic::dx_wave_reduce_usum;
+    return Intrinsic::dx_wave_reduce_sum;
+  }
+  default:
+    llvm_unreachable("Intrinsic WaveActiveSum"
+                     " not supported by target architecture");
+  }
+}
+
+// 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 Intrinsic::spv_wave_reduce_umax;
+    return Intrinsic::spv_wave_reduce_max;
+  case llvm::Triple::dxil: {
+    if (QT->isUnsignedIntegerType())
+      return Intrinsic::dx_wave_reduce_umax;
+    return 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) {
+  if (!getLangOpts().HLSL)
+    return nullptr;
+
+  switch (BuiltinID) {
+  case Builtin::BI__builtin_hlsl_adduint64: {
+    Value *OpA = EmitScalarExpr(E->getArg(0));
+    Value *OpB = EmitScalarExpr(E->getArg(1));
+    QualType Arg0Ty = E->getArg(0)->getType();
+    uint64_t NumElements = Arg0Ty->castAs<VectorType>()->getNumElements();
+    assert(Arg0Ty == E->getArg(1)->getType() &&
+           "AddUint64 operand types must match");
+    assert(Arg0Ty->hasIntegerRepresentation() &&
+           "AddUint64 operands must have an integer representation");
+    assert((NumElements == 2 || NumElements == 4) &&
+           "AddUint64 operands must have 2 or 4 elements");
+
+    llvm::Value *LowA;
+    llvm::Value *HighA;
+    llvm::Value *LowB;
+    llvm::Value *HighB;
+
+    // Obtain low and high words of inputs A and B
+    if (NumElements == 2) {
+      LowA = Builder.CreateExtractElement(OpA, (uint64_t)0, "LowA");
+      HighA = Builder.CreateExtractElement(OpA, (uint64_t)1, "HighA");
+      LowB = Builder.CreateExtractElement(OpB, (uint64_t)0, "LowB");
+      HighB = Builder.CreateExtractElement(OpB, (uint64_t)1, "HighB");
+    } else {
+      LowA = Builder.CreateShuffleVector(OpA, {0, 2}, "LowA");
+      HighA = Builder.CreateShuffleVector(OpA, {1, 3}, "HighA");
+      LowB = Builder.CreateShuffleVector(OpB, {0, 2}, "LowB");
+      HighB = Builder.CreateShuffleVector(OpB, {1, 3}, "HighB");
+    }
+
+    // Use an uadd_with_overflow to compute the sum of low words and obtain a
+    // carry value
+    llvm::Value *Carry;
+    llvm::Value *LowSum = EmitOverflowIntrinsic(
+        *this, Intrinsic::uadd_with_overflow, LowA, LowB, Carry);
+    llvm::Value *ZExtCarry =
+        Builder.CreateZExt(Carry, HighA->getType(), "CarryZExt");
+
+    // Sum the high words and the carry
+    llvm::Value *HighSum = Builder.CreateAdd(HighA, HighB, "HighSum");
+    llvm::Value *HighSumPlusCarry =
+        Builder.CreateAdd(HighSum, ZExtCarry, "HighSumPlusCarry");
+
+    if (NumElements == 4) {
+      return Builder.CreateShuffleVector(LowSum, HighSumPlusCarry, {0, 2, 1, 3},
+                                         "hlsl.AddUint64");
+    }
+
+    llvm::Value *Result = PoisonValue::get(OpA->getType());
+    Result = Builder.CreateInsertElement(Result, LowSum, (uint64_t)0,
+                                         "hlsl.AddUint64.upto0");
+    Result = Builder.CreateInsertElement(Result, HighSumPlusCarry, (uint64_t)1,
+                                         "hlsl.AddUint64");
+    return Result;
+  }
+  case Builtin::BI__builtin_hlsl_resource_getpointer: {
+    Value *HandleOp = EmitScalarExpr(E->getArg(0));
+    Value *IndexOp = EmitScalarExpr(E->getArg(1));
+
+    // TODO: Map to an hlsl_device address space.
+    llvm::Type *RetTy = llvm::PointerType::getUnqual(getLLVMContext());
+
+    return Builder.CreateIntrinsic(
+        RetTy, CGM.getHLSLRuntime().getCreateResourceGetPointerIntrinsic(),
+        ArrayRef<Value *>{HandleOp, IndexOp});
+  }
+  case Builtin::BI__builtin_hlsl_all: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/llvm::Type::getInt1Ty(getLLVMContext()),
+        CGM.getHLSLRuntime().getAllIntrinsic(), ArrayRef<Value *>{Op0}, nullptr,
+        "hlsl.all");
+  }
+  case Builtin::BI__builtin_hlsl_and: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    Value *Op1 = EmitScalarExpr(E->getArg(1));
+    return Builder.CreateAnd(Op0, Op1, "hlsl.and");
+  }
+  case Builtin::BI__builtin_hlsl_or: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    Value *Op1 = EmitScalarExpr(E->getArg(1));
+    return Builder.CreateOr(Op0, Op1, "hlsl.or");
+  }
+  case Builtin::BI__builtin_hlsl_any: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/llvm::Type::getInt1Ty(getLLVMContext()),
+        CGM.getHLSLRuntime().getAnyIntrinsic(), ArrayRef<Value *>{Op0}, nullptr,
+        "hlsl.any");
+  }
+  case Builtin::BI__builtin_hlsl_asdouble:
+    return handleAsDoubleBuiltin(*this, E);
+  case Builtin::BI__builtin_hlsl_elementwise_clamp: {
+    Value *OpX = EmitScalarExpr(E->getArg(0));
+    Value *OpMin = EmitScalarExpr(E->getArg(1));
+    Value *OpMax = EmitScalarExpr(E->getArg(2));
+
+    QualType Ty = E->getArg(0)->getType();
+    if (auto *VecTy = Ty->getAs<VectorType>())
+      Ty = VecTy->getElementType();
+
+    Intrinsic::ID Intr;
+    if (Ty->isFloatingType()) {
+      Intr = CGM.getHLSLRuntime().getNClampIntrinsic();
+    } else if (Ty->isUnsignedIntegerType()) {
+      Intr = CGM.getHLSLRuntime().getUClampIntrinsic();
+    } else {
+      assert(Ty->isSignedIntegerType());
+      Intr = CGM.getHLSLRuntime().getSClampIntrinsic();
+    }
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/OpX->getType(), Intr,
+        ArrayRef<Value *>{OpX, OpMin, OpMax}, nullptr, "hlsl.clamp");
+  }
+  case Builtin::BI__builtin_hlsl_cross: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    Value *Op1 = EmitScalarExpr(E->getArg(1));
+    assert(E->getArg(0)->getType()->hasFloatingRepresentation() &&
+           E->getArg(1)->getType()->hasFloatingRepresentation() &&
+           "cross operands must have a float representation");
+    // make sure each vector has exactly 3 elements
+    assert(
+        E->getArg(0)->getType()->castAs<VectorType>()->getNumElements() == 3 &&
+        E->getArg(1)->getType()->castAs<VectorType>()->getNumElements() == 3 &&
+        "input vectors must have 3 elements each");
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/Op0->getType(), CGM.getHLSLRuntime().getCrossIntrinsic(),
+        ArrayRef<Value *>{Op0, Op1}, nullptr, "hlsl.cross");
+  }
+  case Builtin::BI__builtin_hlsl_dot: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    Value *Op1 = EmitScalarExpr(E->getArg(1));
+    llvm::Type *T0 = Op0->getType();
+    llvm::Type *T1 = Op1->getType();
+
+    // If the arguments are scalars, just emit a multiply
+    if (!T0->isVectorTy() && !T1->isVectorTy()) {
+      if (T0->isFloatingPointTy())
+        return Builder.CreateFMul(Op0, Op1, "hlsl.dot");
+
+      if (T0->isIntegerTy())
+        return Builder.CreateMul(Op0, Op1, "hlsl.dot");
+
+      llvm_unreachable(
+          "Scalar dot product is only supported on ints and floats.");
+    }
+    // For vectors, validate types and emit the appropriate intrinsic
+
+    // A VectorSplat should have happened
+    assert(T0->isVectorTy() && T1->isVectorTy() &&
+           "Dot product of vector and scalar is not supported.");
+
+    auto *VecTy0 = E->getArg(0)->getType()->castAs<VectorType>();
+    [[maybe_unused]] auto *VecTy1 =
+        E->getArg(1)->getType()->castAs<VectorType>();
+
+    assert(VecTy0->getElementType() == VecTy1->getElementType() &&
+           "Dot product of vectors need the same element types.");
+
+    assert(VecTy0->getNumElements() == VecTy1->getNumElements() &&
+           "Dot product requires vectors to be of the same size.");
+
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/T0->getScalarType(),
+        getDotProductIntrinsic(CGM.getHLSLRuntime(), VecTy0->getElementType()),
+        ArrayRef<Value *>{Op0, Op1}, nullptr, "hlsl.dot");
+  }
+  case Builtin::BI__builtin_hlsl_dot4add_i8packed: {
+    Value *A = EmitScalarExpr(E->getArg(0));
+    Value *B = EmitScalarExpr(E->getArg(1));
+    Value *C = EmitScalarExpr(E->getArg(2));
+
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDot4AddI8PackedIntrinsic();
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/C->getType(), ID, ArrayRef<Value *>{A, B, C}, nullptr,
+        "hlsl.dot4add.i8packed");
+  }
+  case Builtin::BI__builtin_hlsl_dot4add_u8packed: {
+    Value *A = EmitScalarExpr(E->getArg(0));
+    Value *B = EmitScalarExpr(E->getArg(1));
+    Value *C = EmitScalarExpr(E->getArg(2));
+
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDot4AddU8PackedIntrinsic();
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/C->getType(), ID, ArrayRef<Value *>{A, B, C}, nullptr,
+        "hlsl.dot4add.u8packed");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_firstbithigh: {
+    Value *X = EmitScalarExpr(E->getArg(0));
+
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/ConvertType(E->getType()),
+        getFirstBitHighIntrinsic(CGM.getHLSLRuntime(), E->getArg(0)->getType()),
+        ArrayRef<Value *>{X}, nullptr, "hlsl.firstbithigh");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_firstbitlow: {
+    Value *X = EmitScalarExpr(E->getArg(0));
+
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/ConvertType(E->getType()),
+        CGM.getHLSLRuntime().getFirstBitLowIntrinsic(), ArrayRef<Value *>{X},
+        nullptr, "hlsl.firstbitlow");
+  }
+  case Builtin::BI__builtin_hlsl_lerp: {
+    Value *X = EmitScalarExpr(E->getArg(0));
+    Value *Y = EmitScalarExpr(E->getArg(1));
+    Value *S = EmitScalarExpr(E->getArg(2));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("lerp operand must have a float representation");
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/X->getType(), CGM.getHLSLRuntime().getLerpIntrinsic(),
+        ArrayRef<Value *>{X, Y, S}, nullptr, "hlsl.lerp");
+  }
+  case Builtin::BI__builtin_hlsl_normalize: {
+    Value *X = EmitScalarExpr(E->getArg(0));
+
+    assert(E->getArg(0)->getType()->hasFloatingRepresentation() &&
+           "normalize operand must have a float representation");
+
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/X->getType(),
+        CGM.getHLSLRuntime().getNormalizeIntrinsic(), ArrayRef<Value *>{X},
+        nullptr, "hlsl.normalize");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_degrees: {
+    Value *X = EmitScalarExpr(E->getArg(0));
+
+    assert(E->getArg(0)->getType()->hasFloatingRepresentation() &&
+           "degree operand must have a float representation");
+
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/X->getType(), CGM.getHLSLRuntime().getDegreesIntrinsic(),
+        ArrayRef<Value *>{X}, nullptr, "hlsl.degrees");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_frac: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("frac operand must have a float representation");
+    return Builder.CreateIntrinsic(
+        /*ReturnType=*/Op0->getType(), CGM.getHLSLRuntime().getFracIntrinsic(),
+        ArrayRef<Value *>{Op0}, nullptr, "hlsl.frac");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_isinf: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    llvm::Type *Xty = Op0->getType();
+    llvm::Type *retType = llvm::Type::getInt1Ty(this->getLLVMContext());
+    if (Xty->isVectorTy()) {
+      auto *XVecTy = E->getArg(0)->getType()->castAs<VectorType>();
+      retType = llvm::VectorType::get(
+          retType, ElementCount::getFixed(XVecTy->getNumElements()));
+    }
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("isinf operand m...
[truncated]

[arsenm]

I wouldn't bother with the LangBuiltins subdirectory

[farzonl]

done!

[jthackray]

Thanks for creating this PR, LGTM (I wasn't able to do it earlier as one of my kids was ill, so away from laptop)

[farzonl]

Thanks for creating this PR, LGTM (I wasn't able to do it earlier as one of my kids was ill, so away from laptop)

No worries. Thanks for the approval!