[clang] [llvm] adding clang codegen (PR #109331)
via cfe-commits
cfe-commits at lists.llvm.org
Fri Sep 20 17:01:48 PDT 2024
https://github.com/joaosaffran updated https://github.com/llvm/llvm-project/pull/109331
>From 7493fc50beae8223e6ba83a1b2fcc8e14d31b352 Mon Sep 17 00:00:00 2001
From: Joao Saffran <jderezende at microsoft.com>
Date: Thu, 19 Sep 2024 00:13:51 +0000
Subject: [PATCH] adding clang codegen
---
clang/include/clang/Basic/Builtins.td | 6 ++
clang/lib/CodeGen/CGBuiltin.cpp | 40 +++++++++++++
clang/lib/CodeGen/CGCall.cpp | 6 ++
clang/lib/CodeGen/CGExpr.cpp | 13 ++++-
clang/lib/CodeGen/CGHLSLRuntime.h | 2 +-
clang/lib/CodeGen/CodeGenFunction.h | 6 ++
clang/lib/Headers/hlsl/hlsl_intrinsics.h | 20 +++++++
clang/lib/Sema/SemaHLSL.cpp | 56 ++++++++++++++++---
.../builtins/asuint-splitdouble.hlsl | 10 ++++
llvm/include/llvm/IR/IntrinsicsDirectX.td | 5 ++
llvm/lib/Target/DirectX/DXIL.td | 1 +
.../Target/DirectX/DXILIntrinsicExpansion.cpp | 13 +++++
12 files changed, 166 insertions(+), 12 deletions(-)
create mode 100644 clang/test/CodeGenHLSL/builtins/asuint-splitdouble.hlsl
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 8c5d7ad763bf97..b38957f6e3f15d 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -4788,6 +4788,12 @@ def HLSLStep: LangBuiltin<"HLSL_LANG"> {
let Prototype = "void(...)";
}
+def HLSLAsUintSplitDouble: LangBuiltin<"HLSL_LANG"> {
+ let Spellings = ["__builtin_hlsl_asuint_splitdouble"];
+ let Attributes = [NoThrow, Const];
+ let Prototype = "void(...)";
+}
+
// Builtins for XRay.
def XRayCustomEvent : Builtin {
let Spellings = ["__xray_customevent"];
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index e2711f1ba70239..5993d49ddcecd4 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -11,6 +11,7 @@
//===----------------------------------------------------------------------===//
#include "ABIInfo.h"
+#include "Address.h"
#include "CGCUDARuntime.h"
#include "CGCXXABI.h"
#include "CGHLSLRuntime.h"
@@ -25,6 +26,7 @@
#include "clang/AST/ASTContext.h"
#include "clang/AST/Attr.h"
#include "clang/AST/Decl.h"
+#include "clang/AST/Expr.h"
#include "clang/AST/OSLog.h"
#include "clang/AST/OperationKinds.h"
#include "clang/Basic/TargetBuiltins.h"
@@ -59,6 +61,7 @@
#include "llvm/IR/MatrixBuilder.h"
#include "llvm/IR/MemoryModelRelaxationAnnotations.h"
#include "llvm/Support/AMDGPUAddrSpace.h"
+#include "llvm/Support/Casting.h"
#include "llvm/Support/ConvertUTF.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/ScopedPrinter.h"
@@ -18824,6 +18827,43 @@ case Builtin::BI__builtin_hlsl_elementwise_isinf: {
retType, CGM.getHLSLRuntime().getSignIntrinsic(),
ArrayRef<Value *>{Op0}, nullptr, "hlsl.sign");
}
+ // This should only be called when targeting DXIL
+ case Builtin::BI__builtin_hlsl_asuint_splitdouble: {
+
+ assert((E->getArg(0)->getType()->isDoubleType() &&
+ E->getArg(1)->getType()->isUnsignedIntegerType() &&
+ E->getArg(2)->getType()->isUnsignedIntegerType()) &&
+ "asuint operands types mismatch");
+
+ Value *Op0 = EmitScalarExpr(E->getArg(0));
+ auto *Op1 = dyn_cast<HLSLOutArgExpr>(E->getArg(1));
+ auto *Op2 = dyn_cast<HLSLOutArgExpr>(E->getArg(2));
+
+ auto [Op1BaseLValue, Op1TmpLValue] =
+ EmitHLSLOutArgLValues(Op1, E->getArg(1)->getType());
+ auto [Op2BaseLValue, Op2TmpLValue] =
+ EmitHLSLOutArgLValues(Op2, E->getArg(2)->getType());
+
+ llvm::Type *retType = llvm::StructType::get(Int32Ty, Int32Ty);
+ if (Op0->getType()->isVectorTy()) {
+ auto *XVecTy = E->getArg(0)->getType()->getAs<VectorType>();
+
+ llvm::VectorType *i32VecTy = llvm::VectorType::get(
+ Int32Ty, ElementCount::getFixed(XVecTy->getNumElements()));
+
+ retType = llvm::StructType::get(i32VecTy, i32VecTy);
+ }
+
+ CallInst *CI =
+ Builder.CreateIntrinsic(retType, llvm::Intrinsic::dx_asuint_splitdouble,
+ {Op0}, nullptr, "hlsl.asuint");
+
+ Value *arg0 = Builder.CreateExtractValue(CI, 0);
+ Value *arg1 = Builder.CreateExtractValue(CI, 1);
+
+ Builder.CreateStore(arg0, Op1TmpLValue.getAddress());
+ return Builder.CreateStore(arg1, Op2TmpLValue.getAddress());
+ }
}
return nullptr;
}
diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 4ae981e4013e9c..4581d06c7c12e5 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -4681,6 +4681,11 @@ void CallArg::copyInto(CodeGenFunction &CGF, Address Addr) const {
IsUsed = true;
}
+void CodeGenFunction::EmitWritebacks(CodeGenFunction &CGF,
+ const CallArgList &args) {
+ emitWritebacks(CGF, args);
+}
+
void CodeGenFunction::EmitCallArg(CallArgList &args, const Expr *E,
QualType type) {
DisableDebugLocationUpdates Dis(*this, E);
@@ -4696,6 +4701,7 @@ void CodeGenFunction::EmitCallArg(CallArgList &args, const Expr *E,
// Add writeback for HLSLOutParamExpr.
if (const HLSLOutArgExpr *OE = dyn_cast<HLSLOutArgExpr>(E)) {
EmitHLSLOutArgExpr(OE, args, type);
+ emitWritebacks(*this, args);
return;
}
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 35b5daaf6d4b55..79b4c898fcdf92 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -28,6 +28,7 @@
#include "clang/AST/DeclObjC.h"
#include "clang/AST/NSAPI.h"
#include "clang/AST/StmtVisitor.h"
+#include "clang/AST/Type.h"
#include "clang/Basic/Builtins.h"
#include "clang/Basic/CodeGenOptions.h"
#include "clang/Basic/SourceManager.h"
@@ -5458,9 +5459,8 @@ LValue CodeGenFunction::EmitOpaqueValueLValue(const OpaqueValueExpr *e) {
return getOrCreateOpaqueLValueMapping(e);
}
-void CodeGenFunction::EmitHLSLOutArgExpr(const HLSLOutArgExpr *E,
- CallArgList &Args, QualType Ty) {
-
+std::pair<LValue, LValue>
+CodeGenFunction::EmitHLSLOutArgLValues(const HLSLOutArgExpr *E, QualType Ty) {
// Emitting the casted temporary through an opaque value.
LValue BaseLV = EmitLValue(E->getArgLValue());
OpaqueValueMappingData::bind(*this, E->getOpaqueArgLValue(), BaseLV);
@@ -5474,6 +5474,13 @@ void CodeGenFunction::EmitHLSLOutArgExpr(const HLSLOutArgExpr *E,
TempLV);
OpaqueValueMappingData::bind(*this, E->getCastedTemporary(), TempLV);
+ return std::make_pair(BaseLV, TempLV);
+}
+
+void CodeGenFunction::EmitHLSLOutArgExpr(const HLSLOutArgExpr *E,
+ CallArgList &Args, QualType Ty) {
+
+ auto [BaseLV, TempLV] = EmitHLSLOutArgLValues(E, Ty);
llvm::Value *Addr = TempLV.getAddress().getBasePointer();
llvm::Type *ElTy = ConvertTypeForMem(TempLV.getType());
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index a8aabca7348ffb..bdb47835e6d1b8 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -87,7 +87,7 @@ class CGHLSLRuntime {
GENERATE_HLSL_INTRINSIC_FUNCTION(SDot, sdot)
GENERATE_HLSL_INTRINSIC_FUNCTION(UDot, udot)
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveIsFirstLane, wave_is_first_lane)
-
+
//===----------------------------------------------------------------------===//
// End of reserved area for HLSL intrinsic getters.
//===----------------------------------------------------------------------===//
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index 2df17e83bae2ee..8c19b2e0fec040 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -4296,6 +4296,9 @@ class CodeGenFunction : public CodeGenTypeCache {
LValue EmitCastLValue(const CastExpr *E);
LValue EmitMaterializeTemporaryExpr(const MaterializeTemporaryExpr *E);
LValue EmitOpaqueValueLValue(const OpaqueValueExpr *e);
+
+ std::pair<LValue, LValue> EmitHLSLOutArgLValues(const HLSLOutArgExpr *E,
+ QualType Ty);
void EmitHLSLOutArgExpr(const HLSLOutArgExpr *E, CallArgList &Args,
QualType Ty);
@@ -5145,6 +5148,9 @@ class CodeGenFunction : public CodeGenTypeCache {
SourceLocation ArgLoc, AbstractCallee AC,
unsigned ParmNum);
+ /// EmitWriteback - Emit callbacks for function.
+ void EmitWritebacks(CodeGenFunction &CGF, const CallArgList &args);
+
/// EmitCallArg - Emit a single call argument.
void EmitCallArg(CallArgList &args, const Expr *E, QualType ArgType);
diff --git a/clang/lib/Headers/hlsl/hlsl_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
index 6cd6a2caf19994..d48b60bab16d51 100644
--- a/clang/lib/Headers/hlsl/hlsl_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
@@ -423,6 +423,26 @@ template <typename T> _HLSL_INLINE uint asuint(T F) {
return __detail::bit_cast<uint, T>(F);
}
+//===----------------------------------------------------------------------===//
+// asuint splitdouble builtins
+//===----------------------------------------------------------------------===//
+
+/// \fn void asuint(double D, out uint lowbits, out int highbits)
+/// \brief Split and interprets the lowbits and highbits of double D into uints.
+/// \param D The input double.
+/// \param lowbits The output lowbits of D.
+/// \param highbits The highbits lowbits D.
+#if __is_target_arch(dxil)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_asuint_splitdouble)
+void asuint(double, out uint, out uint);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_asuint_splitdouble)
+void asuint(double2, out uint2, out uint2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_asuint_splitdouble)
+void asuint(double3, out uint3, out uint3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_asuint_splitdouble)
+void asuint(double4, out uint4, out uint4);
+#endif
+
//===----------------------------------------------------------------------===//
// atan builtins
//===----------------------------------------------------------------------===//
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index 03b7c2edb605fe..f8c388dcc1df81 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -1467,18 +1467,27 @@ bool CheckVectorElementCallArgs(Sema *S, CallExpr *TheCall) {
return true;
}
+bool CheckArgTypeIsCorrect(
+ Sema *S, Expr *Arg, QualType ExpectedType,
+ llvm::function_ref<bool(clang::QualType PassedType)> Check) {
+ QualType PassedType = Arg->getType();
+ if (Check(PassedType)) {
+ if (auto *VecTyA = PassedType->getAs<VectorType>())
+ ExpectedType = S->Context.getVectorType(
+ ExpectedType, VecTyA->getNumElements(), VecTyA->getVectorKind());
+ S->Diag(Arg->getBeginLoc(), diag::err_typecheck_convert_incompatible)
+ << PassedType << ExpectedType << 1 << 0 << 0;
+ return true;
+ }
+ return false;
+}
+
bool CheckArgsTypesAreCorrect(
Sema *S, CallExpr *TheCall, QualType ExpectedType,
llvm::function_ref<bool(clang::QualType PassedType)> Check) {
for (unsigned i = 0; i < TheCall->getNumArgs(); ++i) {
- QualType PassedType = TheCall->getArg(i)->getType();
- if (Check(PassedType)) {
- if (auto *VecTyA = PassedType->getAs<VectorType>())
- ExpectedType = S->Context.getVectorType(
- ExpectedType, VecTyA->getNumElements(), VecTyA->getVectorKind());
- S->Diag(TheCall->getArg(0)->getBeginLoc(),
- diag::err_typecheck_convert_incompatible)
- << PassedType << ExpectedType << 1 << 0 << 0;
+ Expr *Arg = TheCall->getArg(i);
+ if (CheckArgTypeIsCorrect(S, Arg, ExpectedType, Check)) {
return true;
}
}
@@ -1762,6 +1771,37 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
return true;
break;
}
+ case Builtin::BI__builtin_hlsl_asuint_splitdouble: {
+ if (SemaRef.checkArgCount(TheCall, 3))
+ return true;
+
+ Expr *Op0 = TheCall->getArg(0);
+
+ auto CheckIsNotDouble = [](clang::QualType PassedType) -> bool {
+ return !PassedType->isDoubleType();
+ };
+
+ if (CheckArgTypeIsCorrect(&SemaRef, Op0, SemaRef.Context.DoubleTy,
+ CheckIsNotDouble)) {
+ return true;
+ }
+
+ Expr *Op1 = TheCall->getArg(1);
+ Expr *Op2 = TheCall->getArg(2);
+
+ auto CheckIsNotUint = [](clang::QualType PassedType) -> bool {
+ return !PassedType->isUnsignedIntegerType();
+ };
+
+ if (CheckArgTypeIsCorrect(&SemaRef, Op1, SemaRef.Context.UnsignedIntTy,
+ CheckIsNotUint) ||
+ CheckArgTypeIsCorrect(&SemaRef, Op2, SemaRef.Context.UnsignedIntTy,
+ CheckIsNotUint)) {
+ return true;
+ }
+
+ break;
+ }
case Builtin::BI__builtin_elementwise_acos:
case Builtin::BI__builtin_elementwise_asin:
case Builtin::BI__builtin_elementwise_atan:
diff --git a/clang/test/CodeGenHLSL/builtins/asuint-splitdouble.hlsl b/clang/test/CodeGenHLSL/builtins/asuint-splitdouble.hlsl
new file mode 100644
index 00000000000000..de410b50bc8f5b
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/asuint-splitdouble.hlsl
@@ -0,0 +1,10 @@
+// RUN: %clang_cc1 -finclude-default-header -x hlsl -triple dxil-pc-shadermodel6.3-library %s -fnative-half-type -emit-llvm -O0 -o - | FileCheck %s
+
+// CHECK: define {{.*}}test_scalar{{.*}}(double {{.*}} [[VAL1:%.*]], i32 {{.*}} [[VAL2:%.*]], i32 {{.*}} [[VAL3:%.*]]){{.*}}
+// CHECK: [[VALD:%.*]] = load double, ptr [[VAL1]].addr{{.*}}
+// CHECK: call { i32, i32 } @llvm.dx.asuint.splitdouble.{{.*}}(double [[VALD]])
+float fn(double D) {
+ float A, B;
+ asuint(D, A, B);
+ return A + B;
+}
diff --git a/llvm/include/llvm/IR/IntrinsicsDirectX.td b/llvm/include/llvm/IR/IntrinsicsDirectX.td
index 3ce7b8b987ef86..d8092397881550 100644
--- a/llvm/include/llvm/IR/IntrinsicsDirectX.td
+++ b/llvm/include/llvm/IR/IntrinsicsDirectX.td
@@ -88,4 +88,9 @@ def int_dx_rsqrt : DefaultAttrsIntrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>]
def int_dx_wave_is_first_lane : DefaultAttrsIntrinsic<[llvm_i1_ty], [], [IntrConvergent]>;
def int_dx_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_dx_step : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty, LLVMMatchType<0>], [IntrNoMem]>;
+
+def int_dx_asuint_splitdouble : DefaultAttrsIntrinsic<
+ [llvm_anyint_ty, LLVMMatchType<0>],
+ [LLVMScalarOrSameVectorWidth<0, llvm_double_ty>],
+ [IntrNoMem, IntrWillReturn]>;
}
diff --git a/llvm/lib/Target/DirectX/DXIL.td b/llvm/lib/Target/DirectX/DXIL.td
index 9aa0af3e3a6b17..06c52da5fc07c8 100644
--- a/llvm/lib/Target/DirectX/DXIL.td
+++ b/llvm/lib/Target/DirectX/DXIL.td
@@ -778,6 +778,7 @@ def FlattenedThreadIdInGroup : DXILOp<96, flattenedThreadIdInGroup> {
let stages = [Stages<DXIL1_0, [compute, mesh, amplification, node]>];
let attributes = [Attributes<DXIL1_0, [ReadNone]>];
}
+//
def AnnotateHandle : DXILOp<217, annotateHandle> {
let Doc = "annotate handle with resource properties";
diff --git a/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp b/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
index dd73b895b14d37..2bc9ebc962e71a 100644
--- a/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
+++ b/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
@@ -12,6 +12,7 @@
#include "DXILIntrinsicExpansion.h"
#include "DirectX.h"
+#include "llvm-c/Core.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/Analysis/DXILResource.h"
@@ -346,6 +347,15 @@ static Value *expandStepIntrinsic(CallInst *Orig) {
return Builder.CreateSelect(Cond, Zero, One);
}
+// static Value *expandSplitdoubleIntrinsic(CallInst *Orig) {
+// Value *X = Orig->getOperand(0);
+// Type *Ty = X->getType();
+// IRBuilder<> Builder(Orig);
+
+// Builder.CreateIntrinsic()
+
+// }
+
static Intrinsic::ID getMaxForClamp(Type *ElemTy,
Intrinsic::ID ClampIntrinsic) {
if (ClampIntrinsic == Intrinsic::dx_uclamp)
@@ -459,6 +469,9 @@ static bool expandIntrinsic(Function &F, CallInst *Orig) {
break;
case Intrinsic::dx_step:
Result = expandStepIntrinsic(Orig);
+ break;
+ // case Intrinsic::dx_asuint_splitdouble:
+ // Result = expandSplitdoubleIntrinsic(Orig);
}
if (Result) {
Orig->replaceAllUsesWith(Result);
More information about the cfe-commits
mailing list